Development and Functionalization of Visible-Light-Driven Water-Splitting Photocatalysts

With global warming and the depletion of fossil resources, our fossil fuel-dependent society is expected to shift to one that instead uses hydrogen (H2) as a clean and renewable energy. To realize this, the photocatalytic water-splitting reaction, which produces H2 from water and solar energy through photocatalysis, has attracted much attention. However, for practical use, the functionality of water-splitting photocatalysts must be further improved to efficiently absorb visible (Vis) light, which accounts for the majority of sunlight. Considering the mechanism of water-splitting photocatalysis, researchers in the various fields must be employed in this type of study to achieve this. However, for researchers in fields other than catalytic chemistry, ceramic (semiconductor) materials chemistry, and electrochemistry to participate in this field, new reviews that summarize previous reports on water-splitting photocatalysis seem to be needed. Therefore, in this review, we summarize recent studies on the development and functionalization of Vis-light-driven water-splitting photocatalysts. Through this summary, we aim to share current technology and future challenges with readers in the various fields and help expedite the practical application of Vis-light-driven water-splitting photocatalysts.


Water-Splitting Photocatalysts
With the increasing threat of global warming and the depletion of fossil resources, society is expected to shift to using clean and renewable energy instead of fossil fuels. Hydrogen (H 2 ) does not emit carbon dioxide or other harmful materials when used in energy generation. In addition to producing energy in combustion engines, H 2 can also be directly converted into electric power by fuel cells ( Figure 1A). If H 2 can be produced via a photocatalytic water-splitting reaction , renewable energy can be produced from sunlight and water, which are abundant on Earth ( Figure 1B). Therefore, watersplitting photocatalysts have attracted huge attention as a means to address energy and environmental problems.
However, to realize the practical use of water-splitting photocatalysts, solar-to-hydrogen conversion efficiency (STH) should be improved to about 10%. As shown in Figure 2, it is difficult to achieve STH = 10% using only ultraviolet (UV) light [44]. Therefore, the use of visible (Vis)-light-driven water-splitting photocatalysts that conduct water splitting under Vis light (400 nm ≤ λ ≤ 800 nm), which accounts for most sunlight, is essential to achieve this goal [45][46][47][48].  Reproduced with permission from reference [44]. Copyright 2017 Springer Nature Limited.

Development of Vis-Light-Driven Water-Splitting Photocatalysts
When a water-splitting reaction is conducted using a semiconductor photocatalyst, the reaction consists of three main steps ( Figure 3A), as follows: (1) The semiconductor photocatalyst absorbs light energy, resulting in electronic excitation from the valence band (VB) to the conduction band (CB); (2) the excited electrons and remaining holes in the VB migrate to the photocatalyst surface or cocatalyst nanoparticles (NPs), respectively; (3) the H 2 evolution reaction (HER; Equation (1)) and oxygen (O 2 ) evolution reaction (OER; Equation (2)) proceed on the cocatalyst or photocatalyst surface. 2H + + 2e − → H 2 (1) Theoretically, H 2 O reduction and H 2 evolution proceed when the CB minimum edge (CBM) of the semiconductor photocatalyst is more negative than the reduction potential of H 2 O (0 V vs. normal H 2 electrode (NHE); pH = 0). The H 2 O oxidation reaction proceeds, and O 2 is evolved when the VB maximum edge (VBM) of the semiconductor photocatalyst is more positive than the oxidation potential of H 2 O (1.23 V vs. NHE; pH = 0) ( Figure 3B). When the band gap (BG) of the semiconductor photocatalyst is sufficiently wide (UV-lightdriven photocatalyst), such control of the CBM and VBM positions is relatively easy. In contrast, when the BG of the semiconductor photocatalyst is narrow (Vis-light-driven watersplitting photocatalyst), achieving appropriate CBM and VBM positions simultaneously becomes difficult. Furthermore, a semiconductor photocatalyst with a CBM and VBM that satisfy the above conditions is not guaranteed to achieve an overall water-splitting reaction (OWSR). This is due to the following factors: (i) The high activation energy of the water-splitting reaction makes it difficult for the reaction to proceed; (ii) recombination of electrons and holes (excitons) causes the reaction to be deactivated; and (iii) generated H 2 and O 2 cause a reverse reaction. Therefore, only a few studies have reported one-step photocatalytic materials that can achieve an OWSR under Vis light ( Figure 4A) [49]. The OWSR can also be achieved by combining two semiconductor photocatalysts, which can conduct the half-reactions of water splitting (HER and OER), and a redox couple (mediator), which can transfer excitons between them ( Figure 4B) [50,51]. This two-step reaction system, which imitates plant photosynthesis, is called the Z-scheme water-splitting reaction. In this reaction, any semiconductor that can cause a half-reaction can be used. Therefore, the number of available photocatalysts in this reaction is much larger than in the one-step water-splitting reaction, and longer-wavelength light can be used. As the HER and OER occur on two separate photocatalysts, the reverse reaction of H 2 and O 2 evolution can be suppressed in the Z-scheme system using a two-port H-type electrolytic cell with an ion-exchange membrane. Furthermore, separation of the evolved gases is not necessary using this method. However, the reverse reaction involving the redox couples can occur in this system, which does not occur in the one-step water-splitting reaction [51]. Furthermore, as a disadvantage, the theoretical STH of the Z-scheme water-splitting reaction is lower than that of the one-step water-splitting reaction, as two photons are required in a single reaction.
Unfortunately, at present, there is no Vis-light-driven water-splitting photocatalyst that can realize practical application in either the one-step or Z-scheme reaction. To generate sufficient H 2 to withstand market competition using water-splitting photocatalysts, and thereby realize a H 2 -energy society, it is essential to greatly improve the functionality of Vis-light-driven photocatalysts in the future.

Structure of This Review
The outline of this review is as follows. Section 2 describes the methods used to develop Vis-light-driven water-splitting photocatalysts. Specifically, Section 2.1 describes the fabrication of Vis-light-driven water-splitting photocatalysts by modifying the BG of metal oxide water-splitting photocatalysts (UV-light-driven water-splitting photocatalyst) ( Figure 5A), and Section 2.2 describes the development of Vis-light-driven water-splitting photocatalysts using materials other than metal oxides ( Figure 5B). Section 3 describes controlling the cocatalyst ( Figure 3A), which operates as the active site ( Figure 5C). Section 4 describes means for separating each reaction site ( Figure 5D) to enhance the efficiency of the water-splitting reaction. Section 5 provides a short summary and, finally, Section 6 describes our future outlook. In this review, we have categorized methods for the fabrication and functionalization of Vis-light-driven water-splitting photocatalysts according to our viewpoint. Therefore, this categorization might differ from previous reviews [44,49,51,[80][81][82][83][84][85][86][87][88][89][90]. Furthermore, as this review was written for readers not previously involved in water-splitting photocatalysis, methods used to prepare each photocatalyst and analyze their properties are not described. Therefore, readers interested in details of these experiments are referred to the original papers cited.

Modification of BG of Metal Oxide Semiconductor Photocatalysts (UV-Light-Driven Water-Splitting Photocatalysts)
As metals form strong bonds with O, metal oxides are highly stable. Therefore, early studies on water-splitting semiconductor photocatalysts have mainly used metal oxides as photocatalytic materials. Such research has shown that metal oxides with an electronic structure of d 0 (transition metal ions, such as Ti 4+ , zirconium ion (Zr 4+ ), niobium ion (Nb 5+ ), tantalum ion (Ta 5+ ), vanadium ion (V 5+ ), tungsten ion (W 6+ ), and cerium ion (Ce 4+ )) and d 10 (typical metal ions, such as zinc ion (Zn 2+ ), indium ion (In 3+ ), gallium ion (Ga 3+ ), germanium ion (Ge 4+ ), tin ion (Sn 4+ ), and antimony ion (Sb 5+ )), possess functions as watersplitting photocatalysts. However, as most of these metal oxides have a BG of more than 3 eV, they can cause an OWSR only when irradiated with UV light. Therefore, to cause a water-splitting reaction under Vis light, the BG of the semiconductor photocatalyst must be narrowed to match the energy of the Vis-light region (BG < 3.0 eV).
Furthermore, in recent years, several examples of the VB being shifted to the negativepotential side by substituting O 2− with halide ions have been reported (Table 1) [108][109][110][111]. In such cases, the VBM shifts continuously to the negative-potential side with increasing electronegativity of the halide anion. For example, in bismuth oxyhalide (BiOX), the BGs of BiOCl, BiOBr, and BiOI are 3.42, 2.78, and 1.84 eV, respectively ( Figure 8C) [108,109]. In these metal halide photocatalysts, the Bi 6p orbital forms a CBM at a more positive position compared with the reduction potential of H 2 O. Therefore, this photocatalyst alone cannot conduct water-splitting reactions under Vis-light irradiation. Accordingly, metal halide photocatalysts are expected to be OER photocatalysts in the Z-scheme. Regarding such metal halide photocatalysts, in 2016, Kageyama, Abe, and co-workers reported that the formation of Bi 4 NbO 8 Cl with Nb oxide enhanced its stability under Vis-light irradiation [109,110].

Substitution of Metal Cation
The BG of the semiconductor can also be narrowed by introducing metal ions with a d 10 s 2 -type electron configuration, such as copper ion (Cu + ), silver ion (Ag + ), lead ion (Pb 2+ ), and Bi 3+ , into the metal oxide photocatalyst, resulting in a widened VB through hybridization of their orbitals with O 2p orbitals [112][113][114][115][116][117][118]. For example, in 1999, Kudo and co-workers developed monoclinic BiVO 4 (m-BiVO 4 ) by adding Bi 3+ to V oxide and found that the BG of this photocatalyst was narrowed to 2.4 eV owing to Bi 6s and O 2pπ hybridization ( Figure 8D) [112,113]. These authors also confirmed that this photocatalyst promoted O 2 evolution under Vis-light irradiation in the presence of a sacrificial agent (Table 1). They also succeeded in increasing the VBM by substituting alkali metal ions with Ag + or Cu + at the near surface of bulky materials, such as NaTaO 3 (Table 1) [114,115]. Furthermore, there have been several reports of Vis-light-driven water-splitting photocatalysts created by replacing alkali metal ions in the interlayer of layered oxide photocatalysts with Ag + or Cu + (Table 1) [116][117][118].

Formation of Impurity Levels by Doping
Metal oxide photocatalysts become Vis-light responsive by forming donor levels (impurity levels) when 0.1% to several percent of the metal ions are substituted with other transition metal ions (such as chromium ion (Cr 3+ ), rhodium ion (Rh 3+ ), and iridium ion (Ir 3+ )) while maintaining the lattice structure, known as doping. Using such a method, Kudo and co-workers succeeded in developing a Rh-doped strontium titanium oxide (SrTiO 3 :Rh) photocatalyst in 2004 ( Figure 7) [119]. For SrTiO 3 :Rh, a donor level (impurity level) was formed at the negative side of VBM (namely, in the forbidden band) due to doping of some Ti 4+ sites with Rh 3+ ( Figure 9A). For SrTiO 3 :Rh, optical absorption occurred at both 580 and 420 nm immediately after preparation ( Figure 9B) [119]. These absorptions were attributed to impurity levels based on Rh 4+ and Rh 3+ , respectively. However, the acceptor level (impurity level) formed by Rh 4+ became an exciton recombination center [120]. Therefore, to proceed with the water-splitting reaction efficiently, Rh 4+ needs to be photoreduced to Rh 3+ by light irradiation [121]. Furthermore, the formation of such acceptor levels can be avoided by co-doping other metal cations to maintain the charge balance [122][123][124]. For example, Onishi and co-workers succeeded in suppressing the formation of Rh 4+ by co-doping some Ti 4+ sites with two types of ion, Sb 5+ and Rh 3+ ( Figure 9B) [125]. In contrast, Domen and co-workers succeeded in reducing the ratio of Rh 4+ and increasing the ratio of Rh 3+ by doping some Ti 4+ sites with Rh 3+ and some Sr 2+ sites with lanthanum ion (La 3+ ) (co-doping) [126].
Doping is also effective at shifting the positions of the CB and VB [127]. In 2015, Lee and co-workers simultaneously substituted Bi 3+ and V 5+ in m-BiVO 4 with In 3+ and molybdenum ion (Mo 6+ ), respectively, to create BiVO 4 :In,Mo. In this BiVO 4 :In, Mo photocatalyst, BiVO 4 was a mixture of m-BiVO 4 and tetragonal BiVO 4 (t-BiVO 4 ), which induced an increase in the compressive lattice strain ( Figure 9C). This caused an increase in the CB and allowed BiVO 4 :In,Mo to promote the HER ( Figure 9D) [128].

Narrowing of BG by Solid-Solution Formation
Semiconductor materials with the same crystal structure can easily form a solid solution [136]. The BG and energy levels of the solid solution continuously change depending on the ratio of the two types of semiconductor material [129][130][131][132][133]. Domen and co-workers succeeded in creating GaN:ZnO, which provided the first example of efficient and stable Vis-light-driven OWSR ( Figure 10A) by solid-solution formation from GaN (3.4 eV) and ZnO (3.2 eV), which are UV-light-driven photocatalysts [137]. In this solid solution, both Zn 3d and N 2p electrons were present in the VBM, and p-d repulsion occurred between them [137,138], resulting in a narrower BG for this solid solution (~2.6 eV) compared with those of GaN (3.4 eV) and ZnO (3.2 eV) ( Figure 10A). In 2017, Domen and co-workers also succeeded in creating La 5 Ti 2 Cu(S 1−x Table 1) [132]. The absorption edge of this solid solution shifted monotonically toward longer wavelengths with increasing Se content and, at x = 1, the solid solution absorbed light up to 820 nm ( Figure 10B). At x = 0.2, HER activity was observed under Vis-light irradiation of up to 720 nm ( Figure 10B).

Shift of Band Structure by Reducing Particle Size
In 2014, Bao and co-workers showed that reducing the size of cobalt (II) oxide (CoO) NPs to~10 nm shifted their band to a more suitable position for water splitting ( Figure 11A). The CoO NPs with a particle size of~10 nm had a BG (2.6 eV) capable of absorbing Vis light ( Figure 7 and Table 1) and induced overall water splitting with a very high efficiency of STH = 5% [134]. Unfortunately, CoO NPs have low stability, like most other Vis-light-driven photocatalysts, and the reaction was deactivated in about 1 h. To address this problem, in 2017, Mao, Liu, Kang, and co-workers created a CoO single crystal with a submicrometer-sized octahedral structure and the active (111) face exposed ( Figure 11B) [135]. In this photocatalyst, exposure of the (111) surface suppressed hydrogen peroxide (H 2 O 2 ) poisoning, which led to a decrease in photocatalytic activity ( Figure 11C). However, even CoO single crystals obtained in this manner did not show sufficient stability, with thermal oxidation of CoO to Co 3 O 4 observed. Therefore, these authors combined CoO single crystals with graphene, a thermal conductor, to create a highly active and stable Vis-light-driven water-splitting photocatalyst ( Figure 11D). Furthermore, in 2017, Kang, Liu, Huang, and co-workers succeeded in improving stability during the degradation reaction of antimicrobial agents by forming a heterojunction of CoO NPs, a p-type semiconductor photocatalyst, and graphitic carbon nitride (g-C 3 N 4 ), an n-type semiconductor photocatalyst, to promote charge separation [139]. Lu, Shi, and co-workers have shown that such high stability due to heterojunctions can also be induced using BiVO 4 , an n-type semiconductor photocatalyst [140]. Unfortunately, there have been no reports on the water-splitting activity of such heterojunction photocatalysts using CoO NPs and n-type semiconductor photocatalysts, but water splitting is expected to be studied on such heterojunction photocatalysts in the future.

Connection of Indium Gallium Nitride with Different BGs
For indium gallium nitride (InGaN), the BG varies continuously from the UV to nearinfrared regions depending on the composition of In and Ga. In 2018, Vayssieres, Mi, and co-workers developed a photocatalyst with a dual-band structure ( Figure 12A) by combining this material with GaN (3.4 eV) [155]. In 2019, these authors also succeeded in creating a nanowire photocatalyst with a quadruple-band structure ( Figure 12B [141]. The multiband InGaN nanowires were directly grown on a non-planar silicon wafer, had a large surface area, and absorbed light with high efficiency. Furthermore, the combination of four types of photocatalysts with different BGs allowed this photocatalyst to absorb sunlight across almost the entire Vis-light spectrum ( Figure 12C). The obtained photocatalyst maintained its water-splitting activity for a long time under Vis-light irradiation ( Figure 12D).
In addition, several studies have shown that OWSR can be achieved by changing the morphology of g-C 3 N 4 [172,173]. For example, in 2021 Bao, Pu, and Wang successfully exfoliated g-C 3 N 4 by irradiating bulk g-C 3 N 4 with a femtosecond pulsed laser and thereby synthesized g-C 3 N 4 ultra-thin nanosheets (UTN) [152]. The deposition of platinum (Pt) single atoms on g-C 3 N 4 UTN resulted in OWSR with a H 2 production rate of 42.6 µmol g −1 h −1 and an O 2 production rate of 18.7 µmol g −1 h −1 (Figure 13C(a)). They attributed this phenomenon to the formation of C≡N defects caused by laser stripping, which induced the two following things: (1) aiding the loading of Pt single atoms and thereby increasing the number of active sites, and (2) downshifting the CBM and VBM to promote the OER ( Figure 13C(b)).

Connection of Covalent Organic Frameworks with Metal-Organic Frameworks or Covalent Organic Frameworks
Covalent organic frameworks (COFs) and metal-organic frameworks (MOFs) are crystalline and porous materials that consist of pure organic molecules and metals bonded by covalent bonds or coordination bonds. They are used in various fields, such as gas storage, catalysis, and sensing. COFs have long-range ordered structure, excellent surface area, and adjustable BG and have recently attracted much attention as water-splitting photocatalysts [174]. Most two-dimensional COFs reported to date form layered structures through π-π stacking, which allows them to transport charge carriers not only within but also between covalent sheets [175][176][177]. Furthermore, most COFs (especially Schiff-based COFs) have colors ranging from orange to dark red and provide excellent light harvesting in the Vis-light range. However, COFs not only exhibit low hydrophilicity and stability but also easily cause the recombination of photoexcited carriers. Therefore, the watersplitting activity of COFs has generally been lower than those of metal oxide (sulfide) Vis-light-driven water-splitting photocatalysts.
In 2018, Lan and co-workers reported the development of a hybrid photocatalyst of a COF (TpPa-1-COF) and a MOF (NH 2 -UiO-66) [153]. The hybrid photocatalyst had a band structure suitable for charge separation, and the COF and MOF were strongly connected by covalent bonds ( Figure 14A). In these hybrid photocatalysts, the transfer of excited electrons occurred efficiently ( Figure 14B), resulting in high HER activity ( Table 2). Other groups have also been working on the development of such hybrid photocatalysts. For example, using a MOF (NH 2 -UiO-66) and a COF (1,3,5-tris-(4-formylphenyl) triazine (TFPT)-2,5-diethoxybenzene-1,4-dicarbohydrazide (DETH)), Jiang and co-workers successfully created octahedral core-shell hetero-framework photocatalysts (TFPT-DETH/NH 2 -UiO-66) [154]. These photocatalysts were formed by epitaxial growth of a TFPT-DETH shell on a NH 2 -UiO-66 core ( Figure 15A). The authors obtained a series of TFPT-DETH/NH 2 -UiO-66 samples (TFPT-DETH/NH 2 -UiO-66 (n) (n = 1, 2, 4, or 6)) with different shell thicknesses by varying the amounts of TFPT and DETH. As shown in Figure 15B, the HER rate increased with increasing TFPT shell thickness, reaching a maximum at DETH/NH 2 -UiO-66 (4), and then steadily decreased. These phenomena were due to some of the bulk material, including the MOF core, not being able to absorb sufficient light when the COF shell was too thick, which reduced the amount of photogenerated excitons. Furthermore, for organic semiconductors, a thicker shell results in a faster exciton recombination rate inside the photocatalyst owing to the shorter diffusion length and lower mobility of excitons. Owing to these factors, the photocatalytic activity of DETH/NH 2 -UiO-66 (n) samples gave the highest HER rate when the thickness of the COF shell was at the optimum value (n = 4). The DETH/NH 2 -UiO-66 (4) photocatalyst showed excellent HER activity with an apparent quantum yield (AQY) of 1.11% (at 420 nm). The authors suggested that synergistic effects in the hetero-framework were responsible for this high AQY ( Figure 15C).

Control of Cocatalysts
As described in Section 2, the development of Vis-light-driven water-splitting photocatalysts has seen rapid progress in recent years (Figure 7, Tables 1 and 2). However, water splitting was only achieved using a semiconductor photocatalyst alone in a few reported cases. This is largely due to two factors, as follows: (i) Most of the photoexcited charges recombine in a short time and relax energetically, producing heat (non-radiative deactivation) and luminescence (radiative deactivation); and (ii) the surface of the semiconductor photocatalyst has few active sites for reaction. Therefore, for the photocatalytic reaction to proceed with a high quantum yield (QY), electrons and holes must be spatially separated and transferred to active sites on the surface before recombination.

Particle Size Control
The size of cocatalyst NPs has a significant effect on the photocatalytic activity. For example, Teranishi and co-workers synthesized monodisperse Rh NPs of different sizes (1.6 ± 0.5, 2.7 ± 0.3, or 5.1 ± 0.5 nm) using a polyol synthesis method, and then this HER cocatalyst was loaded on GaN:ZnO (Rh NPs/GaN:ZnO) by removing the ligands via calcination [218]. Transmission electron microscopy (TEM) images ( Figure 18A) showed that the Rh NPs cocatalyst was slightly aggregated on GaN:ZnO but still highly monodisperse (1.5 ± 0.3, 3.8 ± 0.8, or 6.6 ± 1.1 nm). A Cr 2 O 3 shell (see Section 3.5.1) was then formed on the Rh NPs (Cr 2 O 3 /Rh NPs/GaN:ZnO), which acted as a reverse-reaction suppressor layer. The water-splitting activity of Cr 2 O 3 /Rh NPs/GaN:ZnO was enhanced with a reduction in size of the Rh NPs cocatalyst ( Figure 18B and Table 3). These size effects were attributed to an increase in the proportion of surface atoms (HER active sites) in the Rh NPs cocatalyst, and enhanced charge separation as the Rh NPs cocatalyst decreased in size. Such an enhancement of the water-splitting activity caused by a reduction in size of the cocatalyst has also been observed in our study on gold (Au)-NCs or Rh 2−x Cr x O 3 -NCs-loaded BaLa 4 Ti 4 O 15 photocatalysts (UV-light-driven photocatalysts; Figure 19) [73][74][75]77]. filter to cut UV light; reaction vessel, Pyrex inner-irradiation type. Reproduced with permission from reference [218]. Copyright 2013 American Chemical Society.

Chemical Composition Control
Alloying enables the creation of materials with physical properties and functions different from those of single metals. In 2015, Qin, Kang, and co-workers showed that the HER activity of SrTiO 3 was significantly enhanced when Cu-Pt alloy NPs were used as the cocatalyst compared with using Cu NPs or Pt NPs as the cocatalyst [219]. Specifically, Cu 0.95 Pt 0.05 NPs/SrTiO 3 showed an HER rate about 2.79, 1.76 times higher than those of Cu NPs/SrTiO 3 and Pt NPs/SrTiO 3 , respectively ( Figure 20 and Table 3). The authors' interpretation of these results was that using Cu-Pt alloy NPs as the cocatalyst led to high HER rates because excited electrons were efficiently transferred to the cocatalyst, which suppressed charge recombination.  In 2015, Ge and co-workers reported using an alloy consisting of Pt and Co as a cocatalyst [220]. In this study, Pt 0.5 Co 0.5 NPs/g-C 3 N 4 nanosheets were found to have HER activity 1.34 times higher than that of Pt NPs/g-C 3 N 4 nanosheets with the same Pt loading ( Figure 21A and Table 3). Alloying with Co was interpreted to increase the driving force for photoexcited electron transfer from the CB of g-C 3 N 4 to the cocatalyst, resulting in enhanced HER activity ( Figure 21B). However, when the Co content exceeded the Pt content, the HER activity decreased ( Figure 21A). This was attributed to the reduction in HER reaction sites on the cocatalyst surface when the Co content became too large. As described above, alloying is extremely effective at improving the cocatalyst function. Recently, the synthesis of alloy NCs with precisely controlled chemical compositions by liquid-phase synthesis has become possible . The use of these fine alloy NCs as precursors will allow precise control of the chemical composition of loaded metal NCs (Figure 22), which is expected to provide an improved understanding of the factors that contribute to the enhanced activity induced by alloying [86].

Morphology Control
When metal NPs are used as a cocatalyst, the water-splitting activity changes depending on the exposed crystal plane. This is due to the electronic structure and surface energy of the metal NPs' surface differing depending on the crystal plane, resulting in different adsorption properties with the substrate. The optimal shape and crystal planes are different depending on the metal species.
In 2016, Yu and co-workers synthesized Pt NPs with different shapes (cubic, octahedral, or spherical;~10 nm; Figure 23A) and successfully loaded them on g-C 3 N 4 [221]. Studies on the as-obtained Pt NPs/g-C 3 N 4 photocatalysts showed that their HER activity increased in the order of cubic-Pt NPs/g-C 3 N 4 < octahedral-Pt NPs/g-C 3 N 4 < spherical-Pt NPs/g-C 3 N 4 ( Figure 23B and Table 3). The cubic-Pt NPs consist of six (100) planes and do not have many active sites, consisting of sharp edges and corners, which are necessary for HER evolution. In contrast, octahedral-Pt NPs are composed of eight (111) planes and have more active sites, consisting of sharp edges and corners, than cubic-Pt NPs. Meanwhile, spherical-Pt NPs are composed of a large number of (100) and (111) planes, and these structures contain many active sites consisting of sharp edges and corners. These factors were interpreted to be related to the HER activity of Pt NPs/g-C 3 N 4 with Pt NPs cocatalysts of different geometries, in the order of cubic < octahedral < spherical. Similar shape dependence has also been observed for TiO 2 loaded with Pd NPs. In 2018, Yu and co-workers prepared Pd NPs/TiO 2 photocatalysts loaded with cubic or tetrahedral Pd nanocrystals (cubic-Pd NPs/TiO 2 and tetrahedral-Pd NPs/TiO 2 , respectively; Figure 24A) and investigated their HER activity [222]. The results demonstrated that tetrahedral-Pd NPs/TiO 2 showed HER activity 1.5-2.0 times higher than that of cubic-Pd NPs/TiO 2 ( Table 3). The tetrahedral-Pd NPs consisted of four (111) planes, to which photoexcited electron transfer from the CB of TiO 2 proceeded efficiently ( Figure 24B). On the tetrahedral-Pd NPs surface, adsorption of H, H-to-H 2 conversion of molecules, and desorption of H 2 molecules readily occurred ( Figure 24C). Furthermore, theoretical calculations indicated that the (111) planes had a large work function, and that not only the edge/corner atoms but also the uncoordinated surface atoms, could be reaction sites. Accordingly, tetrahedral-Pd NPs/TiO 2 exhibited higher HER activity compared with cubic-Pd NPs/TiO 2 ( Figure 24D).  (111) surface. (C) Energy variation in the H 2 evolution process on Pd (100) and Pd (111) facets. The total energy of the initial Pd surface and two isolated H atoms is set to zero. Parenthesized numbers indicate the three reaction steps. TS represents the transition state. E ads , E bar , and E des are the adsorption energy of H atoms, energy barrier, and desorption energy of the H 2 molecule, respectively. The calculated E abs , E bar , and E des of Pd (100) facets are −6.57, 0.034, and 0.29 eV, respectively, and those of Pd (111) facets are −6.77, 0.018 and 0.16 eV, respectively. (D) Photocatalytic H 2 evolution activities of tetrahedral-Pd NPs/TiO 2 and cubic-Pd NPs/TiO 2 from 10 vol% lactic acid solution under Vis-light irradiation (λ > 420 nm). Reproduced with permission from reference [222]. Copyright 2018 American Chemical Society.
In contrast, in a 2015 study on Pd NPs/cadmium sulfur (CdS), Yao and co-workers reported a different shape dependence than the above two reports [223]. In this study, the authors synthesized cubic-Pd NPs (~8.9 nm) surrounded by six (100) planes and octahedral-Pd NPs (~6.0 nm) surrounded by eight (111) planes ( Figure 25A) and loaded them on CdS photocatalysts. From photocatalytic activity measurements, it was found that cubic-Pd NPs/CdS shows higher HER activity compared with octahedral-Pd NPs/CdS ( Figure 25B). The photocurrent generation efficiency of cubic-Pd NPs/CdS was higher than that of octahedral-Pd NPs/CdS ( Figure 25C), which indicated that electron transfer from the cocatalyst to the reactants was more efficient in the former. Estimation of the electrochemical surface area (ECSA) ( Figure 25D) showed that the cubic-Pd NPs/CdS (21.7 m 2 g −1 Pd) had an ECSA 1.49 times higher than that of octahedral-Pd NPs/CdS (14.6 m 2 g −1 Pd), meaning that more proton adsorption/desorption sites were present in the former. These were attributed as the two main factors causing cubic-Pd NPs/CdS to show higher HER activity compared with octahedral-Pd NPs/CdS. In the above studies, one metal element was used in the cocatalyst. However, in 2016, Yao, Xu, and co-workers reported using two elements in the cocatalyst. The authors found that the HER activity of Pt-Pd alloy NPs/CdS depended on both the shape and composition of the alloy NPs cocatalyst [224]. Regarding this shape dependence, cubic-PtPd NPs/CdS had a much higher HER activity ( Figure 26A) and a 3.4 times higher photocatalytic turnover frequency (TOF) compared with octahedral-PtPd NPs/CdS. Electrochemical experiments ( Figure 26B) showed that the interfacial electron transfer rate in cubic-PtPd NPs/CdS was higher than that in octahedral-PtPd NPs/CdS. Regarding the composition dependence, H 2 evolution and the TOF were enhanced when the atomic ratio of Pt to Pd was changed from 1:0 to about 2:1. Reproduced with permission from reference [224]. Copyright 2016 American Chemical Society.

Interfacial Structure Control
Improving bonding between the cocatalyst and photocatalyst surface is also effective in enhancing the photocatalytic activity. In 2021, Domen, Teshima, and co-workers established a method for loading highly dispersed and uniformly sized HER cocatalysts. In this method, Pt NPs were first loaded on BaTaO 2 N (Figure 7 and Table 3) by impregnation, followed by additional loading of Pt on Pt NPs by photodeposition ( Figure 27A) [225]. This sequential loading strongly immobilized Pt NPs on BaTaO 2 N, which facilitated the transfer of photoexcited electrons from the semiconductor to the cocatalyst, and the resulting photocatalyst showed high HER activity (AQY = 6.8 ± 0.5% at 420 nm; Figure 27B and Table 3). The Z-scheme water-splitting reaction using Pt NPs/BaTaO 2 N as the HER photocatalyst and tungsten oxide (WO 3 ) as the OER photocatalyst ( Figure 4B) showed an AQY = 4.0% (at 420 nm) and STH = 0.24%.

Surface Structure Control
The HER proceeds on the noble metal (such as Pt and Rh) NPs cocatalyst. However, when O 2 is present in the system, the reverse reaction also proceeds in parallel (O 2 photoreduction and reverse reaction; see Figure 28) on the noble metal NPs cocatalyst. Therefore, to efficiently produce H 2 , the reverse reaction on the noble metal NPs cocatalyst must be suppressed. For this purpose, effective methods are as follows: (i) formation of a reverse-reaction-suppressing layer on the cocatalyst, (ii) formation of a reverse-reactionsuppressing layer on the entire photocatalyst surface, and (iii) giving the cocatalyst itself a reverse-reaction-suppressing function. Such methods are described in the following section. Pt NP/BaTaO 2 N photocatalyst, 0.1 g; surface-treated WO 3 , 0.15 g; 1 mM of aqueous NaI solution, 150 mL; light source, 300 W Xe lamp (λ ≥ 420 nm) or solar simulator (AM1.5G); irradiation area for solar simulator, 7.6 cm 2 ; reaction system, Pyrex top-illuminated vessel connected to the closed gas-circulation system with the periodical evacuation of gas products. Reproduced with permission from reference [225]. Copyright 2021 Springer Nature Limited.

Formation of Cr 2 O 3 Shell on Cocatalysts
In 2006, Maeda, Domen, and co-workers found that formation of a Cr 2 O 3 layer on the Rh NPs' surface ( Figure 29) suppressed one of the reverse reactions, namely, the O 2 photoreduction reaction ( Figure 28C) [226][227][228][229][230]. Various experiments on the reaction mechanism showed that H 2 O and H + ions, which are polar molecules, penetrated the Cr 2 O 3 layer because this layer was hydrated in water, resulting in H 2 being generated on the Rh surface. On the other hand, O 2 , which is a nonpolar molecule, cannot penetrate the Cr 2 O 3 layer from the outside, meaning that the formation of this layer prevented the reverse reaction. This inhibition of the reverse reaction has also been observed when other noble NPs, such as Cu, Pd, Pt, and Au, were used as cocatalysts [75,232,233]. Figure 29. Schematic model of HER on core-shell Cr 2 O 3 /noble-metal NPs system as a cocatalyst for photocatalytic overall water splitting. Reproduced with permission from reference [232]. Copyright 2009 American Chemical Society.

Formation of Other Amorphous Metal (Oxy) Hydroxide Layers on the Photocatalyst Surface
The reverse reactions can also be suppressed by loading amorphous oxyhydroxides, such as TiO 2 , Nb 2 O 5 , and Ta 2 O 5 , on the entire surface of the photocatalyst particles [231]. These layers were created by loading peroxide complexes onto the photocatalytic surface by photodeposition. This reaction is a downhill reaction that proceeds more easily than the formation of Cr 2 O 3 layers described above using an uphill reaction (in this case, photodeposition using CrO 4 2− ). Therefore, this method can be applied to water-splitting photocatalysts that have weak reducing and oxidizing power owing to their narrow BG.
For example, in 2015, Takata, Domen, and co-workers mixed TiO 2 and RhCrO x NPs/LaMg 1/3 Ta 2/3 O 2 N (Mg = magnesium) in hydrogen peroxide (H 2 O 2 ) solution and irradiated them with light to form a core-shell structure in which TiOXH (OXH = oxyhydroxide) covered the entire surface of the semiconductor photocatalyst (LaMg 1/3 Ta 2/3 O 2 N) and the cocatalyst particles (RhCrO x NPs) ( Figure 30A). The authors found that this suppressed the O 2 reduction reaction in the as-obtained TiOXH/RhCrO x NPs/LaMg 1/3 Ta 2/3 O 2 N [234]. The authors also succeeded in forming a double-coating layer consisting of SiOXH and TiOXH on the surface of LaMg 1/3 Ta 2/3 O 2 N using a similar preparation method ( Figure 30A). In this case, the amount of H 2 and O 2 evolution increased linearly with irradiation time ( Figure 30B). Double coating resulted in the formation of a more uniform layer. Additionally, when SiOXH was mixed into the layer, the hydrophilicity of the layer increased, further inhibiting the permeation of O 2 molecules. These factors resulted in more effective suppression of the reverse reaction. In 2016, these authors also reported the effect of the type of precursor on the function of the photocatalyst. Changing the precursor from TiO 2 to titanium tetraisopropoxide (TTIP) was found to enhance the water-splitting activity of the resulting photocatalyst 1.4 times ( Figure 30C) [235]. This paper also reported that the formation of TiOXH layers enhanced the stability of LaMg 1/3 Ta 2/3 O 2 N during photocatalytic reactions ( Figure 30D).

Formation of Solid Solution (Rh-Cr, Rh-Zr) Oxide Cocatalysts
In 2006, Maeda and Domen showed that suppression of the reverse reaction due to Cr 2 O 3 also occurred when the solid-solution structure was formed instead of the core-shell structure [236][237][238][239][240]. Such a suppression effect of Rh 2−x Cr x O 3 NPs on the reverse reaction was also observed in our study on UV-light-driven photocatalysis ( Figure 19) [77].
Furthermore, long-term light irradiation of the obtained photocatalyst led to the dissolution of Cr 6+ ions [284]. This results in decreased catalytic activity and is considered to have an impact on environmental destruction and health hazards. Therefore, Saruyama, Teranishi, and co-workers worked to create cocatalysts with elements other than Cr that have similar effects. As a result, in 2020, these authors found that the RhZrO x solid-solution NPs cocatalyst, which contained Zr and Rh, also had the ability to suppress the reverse reaction [241]. Although UV-light-driven aluminum (Al)-doped SrTiO 3 (SrTiO 3 :Al) [285] was used as a photocatalyst in this study, future studies are expected to be conducted on Vis-light-driven water-splitting photocatalysts using the same NPs as a cocatalyst.

Charge-State Control
The function of the cocatalyst also depends on the charge state of the metal contained in the cocatalyst. Cobalt oxide (CoO x ) is often used as an OER cocatalyst owing to its high efficiency, low cost, and earth abundance. Yamakata, Maeda, and co-workers reported in 2020 that g-C 3 N 4 loaded with Co 2+ -based spinel-type CoAl 2 O 4 NPs cocatalysts (5-20 nm) showed higher OER activity compared with g-C 3 N 4 loaded with spinel-type Co 3 O 4 NPs cocatalysts (5-20 nm) consisting of a mixture of Co 2+ and Co 3+ ( Figure 31A) [242]. The results of transient absorption spectroscopy showed that the stronger hole-trapping effect in CoAl 2 O 4 NPs compared with Co 3 O 4 cocatalysts caused this phenomenon ( Figure 31B). However, such a slow decay was not clear at 6000 cm −1 (b), and not observed at 9000 cm −1 (c). These results strongly suggest that photogenerated holes in g-C 3 N 4 could move to the loaded CoAl 2 O 4 NPs, thereby increasing the electron population. Reproduced with permission from reference [242]. Copyright 2020 American Chemical Society.
Similar results have been reported by Liu, Yang, and co-workers. In 2021, these authors successfully loaded Co 2+ -based CoO x cocatalysts (~2.5 nm) or Co 3+ -based CoO x cocatalysts (~2.6 nm) on TaON photocatalysts using a photochemical metal-organic deposition (PMOD) method [243]. Photocatalytic studies showed that Co 2+ -based CoO x cocatalysts were 1.6 times more effective for OER than Co 3+ -based CoO x cocatalysts and that TaON loaded with Co 2+ -based CoO x cocatalysts had an AQY of 21.2% (at 420 ± 15 nm; Figure 32A and Table 3). Photoelectrochemical reactions and photoluminescence (PL) measurements indicated that Co 2+ species played an important role in accelerating charge separation and transport ( Figure 32B).

Separation of Each Reaction Site
To create highly active water-splitting photocatalysts, exciton recombination must be suppressed [286]. Therefore, in recent years, the development of water-splitting photocatalysts in which the excited electrons and holes are spatially separated has been promoted, such that the reduction and oxidation reactions can proceed on different surfaces. One approach is to load both the HER and OER cocatalysts on the photocatalyst (dualcocatalyst loading). With such a dual-cocatalyst loading, the excited electrons and holes transfer toward the respective cocatalysts. As a result, they can be spatially separated, and the water-oxidation and water-reduction reactions can proceed on different surfaces of the photocatalyst. However, when loading of the dual-cocatalyst is conducted randomly, the two types of cocatalysts cannot be spatially separated, thereby causing the recombination of excitons and reverse reactions to proceed, which might decrease the photocatalytic activity ( Figure 33A). Therefore, loading cocatalysts onto suitable sites for each reaction is essential. The following methods can somewhat avoid the exciton recombination and reverse reactions proceeding: (i) forming facets that both excited electrons and holes can easily reach (Section 4.1; Figure 33B) [287,288]; (ii) forming onedimensional structures (Section 4.2; Figure 33C) [289,290]; (iii) forming a yolk-shell structure (Section 4.3; Figure 33D) [291][292][293]; and (iv) forming a built-in electric field in the band structure (Section 4.4; Figure 33E) [141,155]. For some of the semiconductor photocatalysts described in this section, the appropriate cocatalysts, available reactions (OWSR, HER, or OER; Figure 4), activities, and references are summarized in Table 4 [287][288][289][290][291][292][293].   [291] a For example, Pt(PD) NPs represent Pt NPs loaded by photodeposition. b OER, HER, and OWSR represent the oxygen evolution reaction, hydrogen evolution reaction, and overall water-splitting reaction, respectively. c AQY, and ER, represent the apparent quantum yield, and evolution rate, respectively.

Forming Facets That Both Excited Electrons and Holes Can Easily Reach
Some crystallized semiconductor photocatalysts might have facets that are more suitable for reduction or oxidation reactions [294]. For example, m-BiVO 4 has a decahedral structure consisting of (010) and (110) facets, where reduction reactions selectively occur on the (010) facets and oxidation reactions on the (110) facets. In 2013, Li, Zhang, and co-workers succeeded in selectively loading Pt NPs as an HER cocatalyst on the (010) facet and MnO x (PD) NPs as an OER cocatalyst on the (110) facet using the photodeposition method ( Figure 34A) [287]. As-obtained Pt(PD) NPs-MnO x (PD) NPs/BiVO 4 showed a much higher OER rate compared with the photocatalyst with cocatalysts randomly loaded using the impregnation method ( Figure 34B). The high OER rate of Pt(PD) NPs-MnO x (PD) NPs/BiVO 4 obtained using the photodeposition method was attributed to the excited electrons and holes being transferred to different crystal facets and separated efficiently ( Figure 34C) [295]. In 2017, Fan, Dittrich, Li, and co-workers also showed that the migration direction of electrons and holes strongly depended on the built-in electric field present in the space charge region of each facet; therefore, controlling the shape of photocatalytic particles and increasing the difference in the internal electric field between each facet can enhance the separation of electrons and holes ( Figure 34D) [296]. These authors also found that an asymmetrical array of cocatalysts further enhances the difference in the built-in electric field at the surface ( Figure 34D).
The flux method is extremely effective for the synthesis of photocatalysts with such specific facets. In the flux method, raw powder is heated and dissolved in the flux, and then crystals are precipitated by the increase in supersaturation caused by cooling and evaporation of the flux. This method has been used in the synthesis of metal oxides such as K 4 Nb 6 O 17 (K = potassium) [297][298][299][300], KNb 3 O 8 [301,302], Na 2 Ti 6 O 13 [303][304][305], and SnNb 2 O 6 [306,307]. In 2020, Takata, Domen, and co-workers controlled the particle shape of SrTiO 3 :Al using a flux method and subsequently successfully loaded HER and OER cocatalysts selectively onto specific crystal surfaces of SrTiO 3 :Al particles ( Figure 35A) [288]. In the structure, the photoexcited electrons and holes were selectively transferred to the HER and OER cocatalysts, respectively, and the recombination of electrons and holes, which caused the QY to decrease in conventional photocatalysts, was almost completely suppressed. Therefore, this photocatalyst showed an AQY of 96% (at 350-360 nm) ( Figure 35B). In this study, the UV-light-driven semiconductor SrTiO 3 :Al was used as a water-splitting photocatalyst. In future studies, such a method is expected to be applied to create highly functional Vis-light-driven water-splitting photocatalysts.

Formation of One-Dimensional (1D) Nanostructures
In some cases, the formation of 1D nanostructures can efficiently separate electrons and holes spatially. In 2019, Li and co-workers successfully loaded Pt NPs (HER cocatalyst) on the tip and PdS NPs (OER cocatalyst) on the side of CdSe 1D nanorods (CdSe(1D-NRs)) by photodeposition (Pt NPs-PdS NPs/CdSe(1D-NRs); Figure 36A) [289]. The asobtained Pt NPs-PdS NPs/CdSe(1D-NRs) photocatalyst showed a HER activity more than 20 times higher than that of the conventional Pt NPs-PdS NPs/CdSe NPs photocatalyst (AQY =~45% (at 420 nm); Figure 36B). PL measurements showed that these excellent photocatalytic performances were attributed to the decrease in exciton recombination in the CdSe(1D-NRs) caused by the spatial separation of the cocatalysts ( Figure 36C). In this photocatalyst, the excited electrons migrate along the long axis to the tip, while the holes migrate to the sides, causing the effective separation of electrons and holes. Furthermore, in 2020, Li and co-workers reported that P-doping the surface of the Cd 0.5 Zn 0.5 S(1D-NRs) photocatalyst (Cd 0.5 Zn 0.5 S(1D-NRs):P) enhances electron and hole transfer, producing a Vis-light-driven water-splitting photocatalyst with a high QY [290]. These authors successfully improved the HER activity of the Cd 0.5 Zn 0.5 S(1D-NRs) photocatalyst under Vis-light irradiation by more than two orders of magnitude. They also succeeded in improving the AQY to 89% (at 420 nm) by loading Pt NPs and PdS NPs cocatalysts on the tips and sides of the obtained Cd 0.5 Zn 0.5 S:P(1D-NRs), respectively ( Figure 37).   1 M of Na 2 S-Na 2 SO 3 aqueous solution, 100 mL; Xe lamp light source (300 W) with an optical filter (λ > 420 nm). (B) Proposed photogenerated electron-hole transfer mechanism for P-doped CZS NRs to induce a localized intrinsic electric field for spatial separation of redox cocatalysts. Reproduced with permission from reference [290]. Copyright 2020 Wiley-VCH.

Formation of Yolk-Shell (Hollow) Nanostructure
Photocatalysts with suppressed recombination and reverse reactions can be developed by loading HER and OER cocatalysts on the inside and outside of the yolk-shell nanostructure, respectively ( Figure 33D). Indeed, in 2013, Domen and co-workers succeeded in significantly enhancing the HER activity by forming Ta 3 N 5 yolk-shell nanostructures using SiO 2 as a template and selectively loading Pt NPs (HER cocatalyst) and IrO 2 NPs (OER cocatalyst) onto the inside and outside surfaces, respectively ( Figure 38) [291]. In 2016, Wang and co-workers also successfully formed g-C 3 N 4 hollow spheres using aminated SiO 2 as a template and loaded HER cocatalysts (Pt NPs) and OER cocatalysts (Co 3 O 4 NPs) onto the inside and outside surfaces, respectively ( Figure 39A,B). The obtained photocatalyst showed overall water splitting with the molar ratio of H 2 to O 2 of 2:1 under UV-light irradiation ( Figure 39C(a)) [292]. When both Pt NPs and Co 3 O 4 NPs were loaded onto the outside surface of g-C 3 N 4 , a decrease in water-splitting activity due to decreased O 2 evolution was observed ( Figure 39C(b)). This indicated that separately loading each cocatalyst onto the inside and outside of the yolk-shell nanostructure is extremely important to obtain high activity. In 2021, Wang and co-workers also succeeded in creating photocatalysts with HER cocatalysts (MoS 2 NPs) and OER cocatalysts (PdS NPs) loaded onto the outside and inside of CdS hollow spheres, respectively ( Figure 40A) [293]. As exciton recombination was greatly suppressed, the obtained photocatalyst showed 115 times higher HER activity compared with CdS NPs ( Figure 40B).

Formation of Built-in Electric Field in Band Structure
In 2018, Mi and co-workers proposed a photochemical diode structure in which the excitons generated by photoexcitation were directed to different active sites ( Figure 41A) [155]. The proposed structure consisted of vertically aligned InGaN nanowires, with the active sites of HER and OER clearly defined. Photogenerated electrons and holes were instantly separated by the electric field formed perpendicularly to the nanosheet ( Figure 41A), which suppressed recombination and reverse reactions at the surface and in the bulk. Specifically, the authors introduced an in-built electric field along the lateral direction of the nanowires by varying the amount of Mg doping ( Figure 41B). Photoexcited electrons transferred to the less Mg-doped side and proceeded with HER, while holes transferred to the more Mgdoped side and proceeded with OER ( Figure 41C). When the HER cocatalyst (Cr 2 O 3 /Rh NPs) and OER cocatalyst (CoO x NPs) were loaded on this photocatalyst, the STH reached 5.2% (Table 2) [141]. Energy-band representation of the proposed photochemical diode (PCD) with radial thickness "d" showing the built-in electric field (band-bending) that separates the excitons (electron and hole) and drives them towards the opposite cathode and anode surfaces. In contrast to a conventional p-n PCD, only a single photon absorption is required to generate one active electron-hole pair to participate in the redox reaction (such as a Schottky-type photochemical diode). (B) Schematic of the quadruple-band InGaN nanowire. The p-type dopant originating from the tilted Mg effusion cell (relative to the nanowire orientation) leads to the Mg-doping gradient profile in the lateral direction of the nanowire. (C) The InGaN nanowire can introduce a built-in electric field for efficient charge carrier separation and extraction for water redox reactions. Reproduced with permission from references [141,155]. Copyright 2018 Springer Nature Limited and 2019 The Royal Society of Chemistry.

Summary
This review summarized representative studies of Vis-light-driven water-splitting photocatalysts. This summary clarified the following points regarding the fabrication and functionalization of Vis-light-driven water-splitting photocatalysts.
(1) To develop Vis-light-driven water-splitting photocatalysts, modifying the band structure of stable metal oxide photocatalysts is effective. The main methods are (i) shifting the energy position of the VBM to the negative-potential side by anion or metal cation substitution, (ii) forming impurity levels in the BG by doping, (iii) narrowing the BG by solid solution, and (iv) shifting the entire band structure to the negative-potential side by a reduction in the size of semiconductor particles. (2) Vis-light-driven water-splitting photocatalysts can also be created by (i) forming multiband-InGaN nanowires, (ii) utilizing g-C 3 N 4 , and (iii) forming MOF/COF connections. (3) Controlling the particle size, chemical composition, morphology, interfacial structure, surface structure, and charge state of the cocatalyst is extremely effective at enhancing the functionality of the photocatalyst. (4) Forming crystal facets on the photocatalyst surface that excited electrons and holes can easily reach, one-dimensional NR structures, yolk-shell structures, and a built-in electric field in the band structure effectively promote charge separation and suppress recombination, resulting in a high QY.
Sharing these findings with readers is expected to further accelerate the development and practical application of Vis-light-driven water-splitting photocatalysts.

Outlook
For the practical application of Vis-light-driven water-splitting photocatalysts, much effort is expected to be devoted to the following research areas: (1) Identifying a simple synthesis method for Vis-light-driven water-splitting photocatalysts. Although several materials have been developed for Vis-light-driven watersplitting photocatalysts, metal (oxy)nitrides and metal (oxy)sulfides are the most attractive materials for OWSR in terms of QY. However, synthesis methods for these photocatalysts have been established only under specific atmospheres. Therefore, only a few research groups with synthesis experience have been able to study these photocatalysts. In the future, simpler methods for synthesizing metal(oxy)nitrides and metal(oxy)sulfide water-splitting photocatalysts are expected to be developed. If realized, more research groups will be able to participate in research on the functionalization of these materials, which is expected to bring their practical application closer to fruition. (2) Enhancement of exciton separation efficiency. To obtain highly active water-splitting photocatalysts, the exciton separation efficiency must be enhanced [288]. In future, fluorescence lifetime and transient absorption spectroscopic measurements [308][309][310][311][312][313][314] are expected to be conducted for a number of photocatalysts to gain a deeper understanding of the influence of the photocatalyst substrate, cocatalyst, and interfacial structure between them on the charge separation efficiency. (3) Structural analysis of loaded cocatalysts. In order to understand structure-property relationships, it is essential to gain a deeper understanding of the geometric structure of the loaded cocatalysts, especially during the reaction. Therefore, in the future, the geometric structure of loaded cocatalysts is expected to be directly observed using aberration-corrected TEM [315] or scanning TEM. Furthermore, the geometric structure observed under electron irradiation in a vacuum is not necessarily the same as the geometric structure during the water-splitting reaction. Therefore, operando measurements using X-ray absorption fine-structure analysis and other techniques [316] are expected to be applied in photocatalyst studies, which will provide a deeper understanding of the geometric structure during the water-splitting reaction than available at present. (4) Theoretical calculation for real system. Theoretical calculations are also useful for developing highly functional water-splitting photocatalysts. Indeed, previous theoretical calculations have clarified the adsorption state of water molecules [317][318][319], the rate-limiting step of the reaction [320,321], and the exciton transfer process at the interface between water molecules and photocatalysts [322,323]. However, in most of these studies, theoretical calculations have been performed on simplified models of real systems, and simulations have been performed on a timescale (<1 µs) shorter than the actual reaction time (10-900 µs). In future, calculations are expected to be performed on longer timescales for real systems, which will provide a deeper understanding of the photocatalytic reaction process. (5) Construction of practical application system. Photocatalysts with an STH exceeding 10% need to be developed for practical use in water splitting. Furthermore, a system for the social implementation of water-splitting photocatalysts must be constructed simultaneously. Accordingly, Domen and co-workers have recently succeeded in constructing a H 2 -production system using photocatalytic panels. However, at present, remaining challenges include the fabrication cost of photocatalytic panels and the performance of the module that separates the evolved gases (H 2 and O 2 ) [324,325]. In future, industry-academia collaborative research is anticipated to be conducted more extensively, which will lead to these challenges being overcome.

Conflicts of Interest:
There are no conflict to declare.