Mimicking Natural Photosynthesis: Solar to Renewable H2 Fuel Synthesis by Z-Scheme Water Splitting Systems

Visible light-driven water splitting using cheap and robust photocatalysts is one of the most exciting ways to produce clean and renewable energy for future generations. Cutting edge research within the field focuses on so-called “Z-scheme” systems, which are inspired by the photosystem II–photosystem I (PSII/PSI) coupling from natural photosynthesis. A Z-scheme system comprises two photocatalysts and generates two sets of charge carriers, splitting water into its constituent parts, hydrogen and oxygen, at separate locations. This is not only more efficient than using a single photocatalyst, but practically it could also be safer. Researchers within the field are constantly aiming to bring systems toward industrial level efficiencies by maximizing light absorption of the materials, engineering more stable redox couples, and also searching for new hydrogen and oxygen evolution cocatalysts. This review provides an in-depth survey of relevant Z-schemes from past to present, with particular focus on mechanistic breakthroughs, and highlights current state of the art systems which are at the forefront of the field.


INTRODUCTION
Solar irradiation hitting the Earth amounts to over 120 000 terawatts of power per year, while human energy consumption will be no more than 30 terawatts per year by 2050 according to nonconservative estimates. 1,2 Understandably then, solar hydrogen fuel synthesis has attracted substantial interest over the past half century as it has the potential to meet the growing global energy demand, mitigate greenhouse gas emission, and provide a solution to the concerns about sustainable energy supplies. Furthermore, both sunlight and water are abundant and essentially inexhaustible, hence there is an opportunity to achieve solar-driven hydrogen synthesis on a large scale via photocatalytic water splitting under solar irradiation. 3 Recently, interest has peaked in renewable solar hydrogen production from water via a robust two photocatalyst system, in which the two half reactions, proton reduction and water oxidation, should both be efficient. However, crucial technological advances are required in order to achieve this potential. Most importantly, large scale hydrogen fuel production demands a solar-to-hydrogen (STH) efficiency exceeding 10%, and considering almost half of the sunlight on the earth's surface lies in the visible region (400−700 nm), harvesting visible light photons efficiently is among the most crucial of all challenges. 1,4−6 The earliest demonstration of photoassisted water splitting dates back to 1921, reported by Baur and Rebmann, using UV light and a AgCl/TlCl system in pure water. 7 Since the discovery of Honda−Fujishima effect in 1972, many efforts have been made in search for a single photocatalyst, where both proton reduction and water oxidation reactions should take place concurrently at different sites. 8,9 Such a single photocatalyst system is shown in Scheme 1, where a semiconductor material can absorb photons with energy greater or equal to the band gap, thus exciting electrons from the valence band (VB) into the conduction band (CB), leaving positively charged holes in the valence band. The overall water splitting process requires a positive Gibbs free energy (H 2 O → H 2 + 0.5 O 2 , ΔG 0 = 237.13 kJ/mol) and thus cannot occur spontaneously. Photogenerated electrons and holes can have strong reduction and oxidation potentials and can be used for catalytic surface redox reactions. 3 Water splitting requires two electrons or four holes to evolve one molecule of H 2 or O 2 , respectively. Hence, the suitable size of band gaps and the positions of conduction band minimum (CBM) and valence band maximum (VBM) are the most important aspects for overall water splitting. The CBM of semiconductors must be more negative than the reduction potential of H + /H 2 (0 V vs standard hydrogen electrode (SHE) at pH = 0), and the VBM must be more positive than the oxidation potential of O 2 /H 2 O (+1.23 V vs SHE at pH = 0). Therefore, the theoretical minimum requirement of photon energy is 1.23 eV, which corresponds to a wavelength in the range of approximately 1000 nm and is in the near-infrared range. However, activation energy (overpotential) is necessary to overcome the energy barrier of redox reactions between charge carriers and water molecules. Therefore, the band gap of a photocatalyst should be larger than the minimum requirement of 1.23 eV. In terms of the solar spectrum, 52−55% sunlight at Earth's surface is infrared light (above 700 nm), 42−43% visible light (400−700 nm), and 3−5% ultraviolet light (below 400 nm). 10 Thus, to achieve a reasonable efficiency of the overall water splitting reaction, the band gap of photocatalysts should be narrower than 3.1 eV to utilize more visible light from the solar spectrum but also possess a large enough overpotential. As for some sulfides and nitrides with suitable band gap potentials for overall water splitting, many of them are not stable in an overall water splitting reaction due to self-oxidation. 11,12 For example, most (oxy)nitrides with the introduction of N 2p orbital in the VB formation, undergo self-oxidative deactivation to some degree, in which photogenerated holes oxidize N 3− to N 2 (2N 3− + 6h + → N 2 ). Self-oxidative deactivation proceeds competitively, and in some cases even when appropriate cocatalysts are used, the water splitting rate will gradually decrease as the compound decomposes.
As for some materials with narrower band gaps, such as the ground-breaking example of GaN-ZnO solid solution (2.5 eV), that could split water into H 2 and O 2 under visible light, the

Chemical Reviews
Review efficiency was still unsatisfactory (average quantum yield of 0.14% in the range of 300−480 nm). 13 Up to now, only a few examples of overall water splitting by a single photocatalyst system under visible light irradiation exist in literature, such as GaN:ZnO solid solutions, 13 nitrides, 14 sulfides, 5 and g-C 3 N 4. 15 The majority of reported visible light-driven photocatalysts are only efficient for half reactions of either proton reduction or water oxidation, as the combination of both band gap magnitude and positioning are difficult to engineer. 16,17 Rather than split water using a single photocatalyst, nature uses a two-stage, double-excitation process (known as a "Zscheme") in PSI and PSII. 18 Inspired by natural photosynthesis of green plants, inorganic analogues have been continually researched over the past 20 years. Artificial photosynthesis is achieved by coupling two narrow band gap semiconductors: an oxygen evolution photocatalyst (OEP) and a hydrogen evolution photocatalyst (HEP), connected via an electron mediator. 19 −21 Briefly, photoexcited electrons in the CB of a HEP reduce protons while the holes in the VB of the OEP oxidize water, evolving H 2 and O 2 , respectively, on separate photocatalysts. The CB electrons of the OEP pass through a relay/shuttle to combine with photoholes in the VB of the HEP, thus completing the cycle and maintaining charge neutrality. Overall water splitting based on the Z-scheme mechanism requires three conditions: (a) an OEP which can inject electrons to hole acceptors, (b) a HEP that can accept electrons from donors, and, crucially, (c) an electron relay system with suitable potentials where subsequent back/side reactions are suppressed. In such a dual photocatalyst system, both OEP and HEP can be driven by visible light because the band gap conditions for overall water splitting are less stringent, theoretically enabling a maximum solar energy conversion efficiency of 40% compared with a ca. 30% maximum solar conversion efficiency on a single photocatalyst. 22 For this reason, this review only focuses on visible driven Z-scheme water splitting systems and the main factors influencing performance, which are hereby divided into the following sections: • Mimicking natural photosynthesis: rationale and mechanism • Photocatalyst development for Z-scheme water splitting • Redox couples in Z-scheme water splitting systems • Strategies for improving efficiency of Z-scheme systems • Conclusions and future outlook 2. MIMICKING NATURAL PHOTOSYNTHESIS: RATIONALE AND MECHANISM Scheme 2 shows the double excitation system in natural photosynthesis along with the electron cascade steps. At certain wavelengths, this system can operate at a quantum yield over 95%. 23,24 Reaction center chlorophylls in photosystem II (PSII) absorb photons (λ ≤ 680 nm), leading to the generation of electron−hole pairs. The photoinduced holes oxidize water to produce oxygen, protons, and electrons. The electrons are transferred via the electron transport chain, whereas the protons released are accumulated in one side of the membrane to build an electronic potential to support the conversion of ADP (adenosine diphosphate) into ATP (adenosine triphosphate). In photosystem I (PS I), electrons are excited by photons (λ ≤ 700 nm), however, now the holes are quenched by the electrons produced by PSII, simultaneously elongating the electron lifetime so more reduction reactions can occur. The electrons from PSI are finally used to reduce NADP + (nicotinamide adenine dinucleotide phosphate) to NADPH, which, together with ATP, provides the reduction power and energy for the conversion of CO 2 to sugars. 11 Inspired by natural photosynthesis, an inorganic double excitation system as shown in Scheme 3 was proposed first by Sayama and Abe et al. 19 In such a system, the water-splitting reaction is separated into two reactions: O 2 evolution and H 2 evolution. When photons are absorbed by an OEP, electrons in the VB will be excited to CB, leaving holes to oxidize water to O 2 . The electrons on the CB first reduce the corresponding electron acceptors (i.e., oxidant, Ox) to form the electron donors (i.e., reductant, Red). Then, the electron donors (Red) are converted back to its oxidized form (Ox) by the oxidation of photoholes on the surface of a HEP. Hence, the electrons are transferred between two photocatalysts by the appropriate shuttle redox couple in solution continuously. Meanwhile, protons are reduced to hydrogen by the photoexcited electrons on the CB of a HEP. As such, the overall water splitting reaction is complete, with regeneration of redox mediators (ions or molecules) and light being the only energy input. However, Red can be oxidized by holes from the VB of OEP, and similarly the electrons on the CB of HEP can react with Ox, i.e., back reactions take place. Thus, for an efficient Zscheme system, back reactions should be inhibited (dotted curves shown in Scheme 3).
A Z-scheme system can utilize visible light more efficiently than a single photocatalyst water splitting system (Scheme 1) due to the lower energy requirement to drive each half reaction Scheme 2. A Schematic Diagram of a Double Excitation Process on P680 and P700 in Natural Photosynthesis Scheme 3. A Double Excitation System Based on a HEP and an OEP with a Shuttle of Ox, oxidant, and Red, reductant, molecules a in comparison to the higher energy requirement to drive an overall water splitting reaction using one photocatalyst. For instance, visible-light-responsive oxides such as WO 3 can carry out O 2 evolution if a mediator Ox can be reduced to a Red. Similarly, non-oxide photocatalysts such as, sulphides, nitrides, and dyes can be applied in H 2 evolution if they are able to oxidize mediator Red to Ox. Therefore, much more semiconductor materials can be applied in the Z-scheme system even if they are not suitable for the harsh requirements for overall water splitting as a single photocatalyst.

PHOTOCATALYST DEVELOPMENT FOR Z-SCHEME WATER SPLITTING
Several well-known photocatalysts for half reactions are listed in Figure 1. In constructing a Z-scheme system, photocatalysts are first examined in half (sacrificial) reactions to test their activity before coupling with a counterpart. As a test for the oxygen evolving half reaction under irradiation of visible light, the photoexcited electrons in a suitable photocatalyst CB would move to react with the sacrificial reagents which are easily reduced, such as Ag + , leaving the photogenerated holes in VB to oxidize water to O 2 . For the H 2 production reaction, the photogenerated holes on the photocatalysts are consumed by a hole scavenger such as methanol or triethanolamine, hence the remaining photoexcited electrons could be used in water reduction efficiently. 3,25 There are much more visible driven photocatalysts working for half reactions (especially with sacrificial reagents) than those so far reported in overall water splitting using Z-scheme, and the construction of Z-scheme systems based on these visible driven photocatalysts for half reactions is one of the targets of solar energy to hydrogen fuel development. 26 The reported Z-scheme systems are listed below in Table 1.
There are a few reasons why certain photocatalysts are either unsuitable or less efficient for the process, including incorrect band positions of the photocatalysts, insufficient efficiency, and incompatible redox charge mediator choice. These factors will be addressed in detail later in the review. In some cases, redox activity can take place at the solid interface of two photocatalysts and thus an aqueous mediator is not required. These cases are detailed in section 3.3. The majority of the reported Z-scheme systems use dissolved ions as redox mediators, which will be reviewed immediately in sections 3.1 and 3.2.

Oxidation Photocatalyst Dominated Systems with Dissolved Ions as Redox Mediators
The O 2 /H 2 O potential lies at +1.23 V (vs SHE at pH = 0). However, due to the overpotential to the chemical reaction, it usually requires a more positive VBM position (vs SHE at pH = 0) to achieve gaseous oxygen production. 27 Oxygen generation from water splitting is a relatively slow procedure because it requires four holes to form one molecule of O 2 by water oxidation, while per H 2 production only needs two electrons. Furthermore, electrons transfer much faster than holes in most oxide photocatalysts which are n-type semiconductors, so the process is much more facile. 28 Hence, a large driving force has to be provided by a deep VBM of a semiconductor. A consequence of this deep lying VBM, a problem in the majority of metal oxides, is that this makes the band gap very large, so a large portion of the visible spectrum from sunlight cannot not be sufficiently harvested. 29 Only a few metal oxides have been reported to work under visible irradiation for efficient half reactions (O 2 evolution), for example monoclinic BiVO 4 (∼2.4 eV), 30 WO 3 (∼2.7 eV), 25 and Ag 3 PO 4 (2.6 eV). 31 3.1.1. WO 3 . Tungsten oxide (WO 3 ) is a well-documented photocatalyst in water oxidation, which has an appreciably more positive CBM than water reduction potential and a much more positive position of VBM than the water oxidation potential. 32 The band gap energy of WO 3 is around 2.7−2.8 eV with relatively positive CBM of +0.5 V and VBM of +3.2 V (vs SHE at pH = 0). The redox potentials of the two most commonly used mediator pairs, Fe 3+ /Fe 2+ and IO 3 − /I − , are around +0.77 V and +1.09 V (vs SHE at pH = 0), respectively, which are thermodynamically able to be reduced by CB electrons in WO 3 . 33 With decent stability under low pH and suitable band positions, WO 3 samples are widely used with Fe 3+ /Fe 2+ redox couples under acidic conditions. WO 3 is among the most widely used oxygen production photocatalyst in half reactions. WO 3 was the first material utilized in an UV-driven Z-scheme water splitting system as a water oxidation photocatalyst to achieve a stoichiometric H 2 and O 2 gas generation by Sayama et al. in 1997. 19 In a Fe 3+ /Fe 2+ mediated solution, WO 3 loaded with RuO 2 produced O 2 along with the other half reaction of proton reduction by Fe 2+ . The activity was promoted by RuO 2 loaded on WO 3 as a cocatalyst, which is crucial to reduce the overpotential needed for oxidation. 34 The pathway is accessible because the Fe 3+ /Fe 2+ potential is +0.77 V (vs SHE at pH = 0), which is positive enough for the CB electrons of WO 3 (+0.5 V vs SHE at pH = 0) to reduce.  WO 3 has been widely reported for water oxidation in Z-Scheme system coupled with either inorganic semiconductors such as doped SrTiO 3 or polymeric photocatalyst g-C 3 N 4 and so on as the counter HEP. 35−38 Still, tailoring cocatalysts and electron mediator conditions is just as important as photocatalyst selection. Interestingly, it is found that IO 3 − /I − is the most suitable redox shuttle in the WO 3 related systems. This is probably due to the suitable pH of the mediator solution (between pH 7−10), as some papers claim 39 that the acidic media (pH < 4) would change the WO 3 to H x WO y because WO 3 is not stable in very acidic (pH < 3) or basic (pH > 9) conditions. Admittedly, some reports studied WO 3 in acidic media where stability is less studied. WO 3 is also the first semiconductor applied in visible-driven Z-scheme water splitting system by Sayama and Abe et al. 37 (Pt/SrTiO 3 :Cr,Ta)−(IO 3 − /I − )−(PtO x /WO 3 ) system was the first example of visible light-driven overall water splitting reported in 2002, and an apparent quantum yield (AQY) of 1% was achieved at 420 nm under optimized conditions in 2005 ( Figure 2). 37,40 For the first time, selective O 2 evolution was achieved on PtO x /WO 3 even in the presence of a considerable amount of I − which was thermodynamically more favorable to be oxidized by holes. This is the key property for the construction of a Z-scheme; the adsorption of IO 3 − anions, instead of I − , to the photocatalyst surface is thermodynamically preferable under pH of 7, thus ensuring a close contact between shuttle species and charges. H 2 and O 2 could be generated at a ratio of 2:1 when the two half reactions were combined. In this work, the pH of the solution was also important. Alkaline conditions were shown to be more favorable for water splitting because acidic conditions generate side products such as I 3 − which can be reduced to I − in place of IO 3 − , breaking the redox cycle. The pH value was also optimized to avoid the formation of an oxyanionic tungstate compound or a hydrogen tungstate ion. WO 3 was also involved in the first visible light-driven organic photocatalyst based Z-scheme system. It was via both (Pt/g-C 3 N 4 )−(IO 3 − /I − )−(Pt/WO 3 ) and (Pt/g-C 3 N 4 )−(Fe 3+ / Fe 2+ )−(BiVO 4 ) systems recently reported by Martin et al, and the former one containing WO 3 achieved the most efficient performance of H 2 and O 2 evolution at rates of 2.12 and 1.10 μmol h −1 , respectively, from water for over 24 h. 35 The organicderived g-C 3 N 4 was synthesized from a facile and extremely cheap precursor, urea. In this work, the optimum pH of 8.3 pointed out not only the inefficiency due to the formation I 3 − at lower pH but also the instability in alkaline media (pH > 9). The mass ratio of WO 3 to g-C 3 N 4 was also shown to be significant, as the difference of light absorption between the two photocatalysts influences the overall water splitting rate. Pt species act as an electron trap for photogenerated electrons in WO 3 so that the mediator could accept them efficiently and also elongate electron−hole lifetimes. IrO 2 −Pt-coloaded WO 3 for O 2 evolution was also successfully coupled with a coumarin dye sensitized Pt intercalated H 2 K 2 Nb 6 O 17 for H 2 evolution in the presence of I 3 − /I − redox couple. 41 3.1.2. BiVO 4 and Bi 2 MoO 6 . As a polymorphic compound with various crystal structures including, tetragonal zircon, monoclinic scheelite, and tetragonal scheelite phases, the monoclinic scheelite phase of BiVO 4 is one of the most efficient visible light-driven water oxidation photocatalysts ( Figure 3I). 42 Monoclinic BiVO 4 has a band gap of ca. 2.4 eV, with CBM locating at +0.46 V (vs SHE at pH = 0) and VBM at +2.86 V (vs SHE at pH = 0). 43 The more negative VBM of BiVO 4 than those of conventional oxides is basically attributed to the hybridization between the O 2p and Bi 6s orbitals. 30 In the presence of sacrificial agent Ag + , the performance of BiVO 4 outperforms WO 3 . The narrower band gap of BiVO 4 facilitates a larger range of light absorption in comparison to WO 3 , and the more negative band position allows more mediators with different redox potentials to be used. Other Bi-based materials such as Bi 2 MoO 6 also come into sight, while the major attention has been paid on BiVO 4 probably due to a higher quantum yield achieved on it. 36 The first BiVO 4 based Z-scheme water splitting system was reported in 2004 in a (Pt/SrTiO 3 :Rh)−(Fe 3+ /Fe 2+ )−(BiVO 4 or Bi 2 MoO 6 ) system ( Figure 3II). 36 Using Fe 3+ (FeCl 3 ) as an electron scavenger, BiVO 4 could evolve oxygen steadily, while separately, Fe 2+ (FeCl 2 ) was found to be sufficient as a hole scavenger for hydrogen evolution over Pt/SrTiO 3 :Rh. Combining the two sides, H 2 and O 2 from overall water splitting were observed steadily even after 158 h. This system worked under wavelengths of up to 530 nm, consistent with the absorption    3 based system showed a higher AQY of 0.5% than 0.4% of the BiVO 4 system at 420 nm. 36 Co complex electron mediators such as [Co(bpy) 3 ] 3+/2+ and [Co(phen) 3 ] 3+/2+ (bpy = 2,2′-bipyridine, phen = 1,10phenanthroline) redox couples are also suitable for use in Ru/SrTiO 3 :Rh and BiVO 4 systems for visible light-driven water splitting. The maximum AQY was 2.1% at 420 nm. It is suggested by the authors that the doped Rh species in SrTiO 3 was very important because Rh not only assisted long wavelength photon absorption but also acted as active sites for the oxidation of these mediator ions. A lack of Rh as active oxidation sites on other photocatalysts failed to form oxidation product of Co 3+ , hence resulting in low activity. 45 The choice of OEP is also crucial in achieving stable and stoichiometric water splitting using these complexes. The use of other oxides, such as TiO 2 and WO 3 , failed to evolve O 2 at a stable rate in the presence of such a Co complex, suggesting the occurrence of undesirable decomposition of such Co complexes. The moderate oxidation potentials of holes generated in the VB of BiVO 4 , which is more negative than those oxides due to significant contribution of Bi 6s orbitals, is probably crucial to suppress such undesirable decomposition. This system also emphasizes the influence of pH of the reactant solution on water splitting. The optimum pH was 7 because under acidic conditions, the oxidation of Co(II) complex ions on BiVO 4 suppressed oxygen evolution. Conversely, the reduction of the Co(III) complex was not influenced dramatically by pH and remained unchanged. Further discussion will be presented later in section 4.
3.1.3. Oxynitrides. Oxynitrides of early transition metals with d 0 electronic configurations have been utilized in visible light induced water splitting systems due to properties of suitable band structure, good stability, and facile synthesis. 5 Oxynitrides can be synthesized by annealing the corresponding metal oxide powder under an ammonia atmosphere at temperatures of ca. 1123−1273 K. 5 Due to the more negative potential of the N 2p orbital in comparison to O 2p, the partially or fully replacement of O 2p orbitals could lead the VBM of the oxynitrides to shift upward without affecting the CBM, thus tuning the material's band gap to respond to visible light ( Figure 4I). Oxynitrides could be partially decomposed to N 2 by photogenerated holes (2N 3− + 6h + → 3N 2 ), and this could be suppressed as the reaction processes. 5 Tantalum oxynitride (TaON) is a typical photocatalyst herein and its CBM is −0.3 V (vs SHE at pH = 0) and the VBM is +2.1 V (vs SHE at pH = 0), thus the band gap is 2.4 eV corresponding to 520 nm. 46 When suitable cocatalysts were loaded on TaON, H 2 or O 2 could be evolved depending on the nature of the cocatalyst. For an oxygen evolution reaction in the presence of Ag + as an electron scavenger, the initial quantum yield of TaON was nearly 34% (λ ∼ 420 nm). 47 The first example of an oxynitride-based water splitting system was demonstrated in 2008. Although unmodified TaON could not generate O 2 from IO 3 − aq, loading of either IrO 2 or RuO 2 cocatalyst afforded appreciable O 2 evolution. It was reported that (Pt/TaON)−(IO 3 − /I − )−(RuO 2 /TaON) was an efficient system which produced H 2 and O 2 with an AQY of 0.1−0.2% (λ ∼ 420 nm). This pioneering work then led to the discovery of a plethora of oxynitride materials as H 2 /O 2 evolution photocatalysts in Z-schemes ( Figure 4II). 48 By focusing on different cocatalysts on the photocatalyst, this study demonstrated the effective role in controlling the selectivity of mediators and the flexibility of the system. 49 The most effective oxynitride-based Z-scheme system was found to be Pt/ZrO 2 /TaON for H 2 and RuO 2 /TaON for O 2 evolution. 50 RuO 2 was found to promote the IO 3 − reduction as well as the water oxidation, which was supported by the cathodic response in an absence and presence of IO 3 − . The enhancement of gas evolution rate was observed when the loading amount of RuO 2 was increased from 0 to 0.5 wt % due to the increased density of active sites. However, an excess amount of Ru loading also overcovered the TaON surface so that light absorption decreased along with activity. The concentration of I − ions was also an important factor which will be discussed in next section. 50 The shortcomings in this work were mainly on the back reaction of water splitting on RuO 2 and TaON as proved by cathodic photocurrent due to O 2 reduction on both RuO 2 /TaON and TaON in a PEC setup. The intrinsic property of the back reaction on RuO 2 /TaON and the competitive oxidation of I − were the main reasons for a lower performance. 50 TaON is an intermediate phase between Ta 2 O 5 and Ta 3 N 5 . Ta 3 N 5 has a longer absorption edge than TaON (active to 591 nm, CBM at −0.4 V and a VBM at +1.7 V), therefore Ta 3 N 5 has been applied in visible driven Z-scheme overall water splitting as an O 2 evolution photocatalyst as well in the presence of cocatalysts such as Ru, Rh, and Ir. 51 The half reaction of Ta 3 N 5 showed a high AQY of 10% for O 2 evolution in the existence of Ag + , then a Z-scheme system was constructed by (Pt/ZrO 2 /TaON)−(IO 3 − /I − )−(Ir/R-TiO 2 / Ta 3 N 5 ) (anatase and rutile TiO 2 noted as A-and R-TiO 2 ). 52,53 Ir loaded on the surface acted as active sites, while R-TiO 2 effectively controlled the IO 3 − /I − adsorbed on the surface, thus suppressing the backward reaction. As mentioned before, competitive reactions in a Z-scheme are often observed and must be taken into consideration. R-TiO 2 selectively adsorbed IO 3 − and hindered adsorption of I − , which is needed for O 2 evolution. It is reasonable that A-TiO 2 did not perform as well as R-TiO 2 due to the lack of unique adsorption properties. TiO 2 loading increased the activity 10-fold by the enhanced absorbance, which indicates that the suitable surface modification should not be neglected.
Oxynitrides, as mentioned above, were reported to undergo self-oxidative reaction where holes competitively oxidize nitrogen anions (2N 3− + 6h + = N 2 ) during water oxidation, causing the deactivation to the system. To solve this issue, cocatalysts including IrO x and CoO x nanoparticles were developed to scavenge holes and thus act as active sites for water oxidation. Considering the short diffusion lengths of photoholes in bulk n-type semiconductors, the dispersion of such cocatalysts should be as homogeneous and high as possible without compromising light harvesting ability. 54 3.  Figure 5I shows the initial rate of O 2 evolution from Fe(ClO 4 ) 3 aqueous solution over H 2 WO 4 and WO 3 under the irradiation conditions with different cutoff filters, along with their UV−vis absorption spectra. It should be noted that the H 2 WO 4 sample showed appreciable activity for O 2 generation under the irradiation longer than 460 nm, where the WO 3 photocatalyst showed almost no activity. H 2 WO 4 is often used as a precursor of WO 3 ; there has been no report on  its application to photocatalytic water splitting. One of the reasons why the H 2 WO 4 so far received little attention as an OEP is the lack of O 2 evolution ability of H 2 WO 4 with Ag + electron acceptor, which is generally used for the first test of O 2 evolution ability of photocatalysts. Thus, this result implies that careful attention should be paid to electron acceptor choice when searching for new photocatalytic materials for water splitting. The working diagram of H 2 WO 4 based Z-scheme is shown in Figure 5II Figure 6I-a). It was revealed that the VBM of Bi 4 NbO 8 Cl consisting of mainly O 2p orbitals (not Cl 3p) is much more negative (∼2.2 V vs SHE) than those of conventional oxides (∼3.0 V vs SHE in most cases) or oxychlorides (e.g., BiOCl, ∼3.0 V vs SHE). Consequently, Bi 4 NbO 8 Cl possesses both a narrow bandgap (∼2.4 eV) for visible light absorption and a more negative CBM than the water reduction potential ( Figure 6III). It was presumed that the photogenerated holes populated on O 2− anions near VBM (or O − ions) were rather stable and just brought about O 2 evolution, instead of oxidative decomposition, which often occurred in conventional anion-doped narrow band gap materials such as (oxy)nitrides. The authors also demonstrated that a Silleń−Aurivillius related oxychloride Bi 6 NbWO 14 Cl (see Figure 6I-b) worked as a stable OEP in the Z-scheme water splitting system. These findings would provide new strategies for developing durable photocatalytic materials for water splitting under visible light by manipulating the perovskite block as well as the (Bi 2 O 2 ) 2 Cl block in the Silleń−Aurivillius phases, with a general formula of [(Bi 2 O 2 ) 2 X] 3+ [A n−1 B n O 3n+1 ] 3− (X = halogens, n = 1, 2, 3,...).
Maeda et al. recently demonstrated that a rutile TiO 2 powder doped with tantalum and nitrogen (TiO 2 :Ta/N) could function as a stable OEP for Z-scheme water splitting with both Fe 3+ / Fe 2+ and IO 3 − /I − redox couples under visible light. The preintroduction of Ta cations into TiO 2 precursor via microwave-assisted solvothermal synthesis was found to suppress the production of Ti 3+ and/or anionic vacancies during nitridation process with NH 3 stream. The decreased formation of such unfavorable sites that enhance charge recombination probably afford TiO 2 :Ta/N to oxidize water in a stable manor. Although the formation of such reduced cation species (e.g., Ti 3+ , Nb 4+ ), as well as anion defects, during nitriding has so far limited the active oxynitride photocatalyst to Ta-based one (e.g., TaON), the present new strategy probably enables us to develop new oxynitride (or nitrogen-dope) materials as stable OEPs with high photocatalytic activity in Zscheme water splitting systems.

Reduction Photocatalyst Dominated Systems with Dissolved Redox Mediators
As mentioned above, the development of the other half reaction (H 2 evolution) is naturally indispensable for an efficient Z-scheme system. The up to date visible light-driven HEP are next discussed.
3.2.1. Cation Doped SrTiO 3 . Doped SrTiO 3 is one of the most well documented H 2 production photocatalysts used in the construction of Z-schemes. Non-doped SrTiO 3 has a relatively large band gap (3.1 eV), which prohibits its use in visible light driven water splitting system. 37 Hence, cationic doping of SrTiO 3 has often been used to introduce a new energy level above the original VB to act as an electron donor level, which allows visible light absorption. However, besides the electron donor level, the dopant can also act as a recombination center for charge carriers so the activity can decrease in certain circumstances. Therefore, the dopant must be selected carefully to ensure it is efficient for H 2 evolution in overall water splitting, striking a balance between light absorption, redox functionality, and electron trapping capabilities.
The visible light response of Rh-doped SrTiO 3 is thought to be due to a new donor level induced by Rh (Figure 7). Pt/ SrTiO 3 :Rh can achieve a quantum yield of 5.2% at 420 nm for the H 2 evolution half reaction in an aqueous methanol solution. 16 There are two onsets of the absorption band of SrTiO 3 :Rh including a shoulder one at 420 nm and a band from 580 to 1000 nm due to the oxidation state of the rhodium. Rh dopants existed in two forms, Rh 3+ and another higher oxidation state (e.g., Rh 4+ ), according to the diffuse reflectance spectrum. The Rh 4+ could be easily reduced to Rh 3+ by electrons excited by photons. The photogenerated electrons from the donor level formed with Rh 3+ were transferred to the CB of SrTiO 3 to reduce protons, while the holes (positive charge) left on the electron donor level oxidized methanol. This Pt/SrTiO 3 :Rh system was also examined for the O 2 generation, but was relatively inefficient, likely due to the discrete Rh 3+ sites. It appears that such discrete oxidation site of Rh 3+ cannot catalyze water oxidation efficiently because it requires fourelectron abstraction while it can catalyze one or two-electron oxidation processes (e.g., methanol or Fe 2+ ).
As mentioned above, Sayama, Abe et al. combined Pt/ SrTiO 3 :Cr,Ta with PtO x /WO 3 in the presence of an IO 3 − /I − mediator in an aqueous solution and first reported visible responsive overall water splitting system via a Z-scheme ( Figure  2). 37 With the extremely low overpotential for H 2 production in water electrolysis, Pt was selected as the cocatalyst for proton reduction to hydrogen. It is worth noting that Pt/SrTiO 3 worked well for O 2 evolution with IO 3 − as the electron scavenger. However, after doping of Cr and Ta, O 2 could hardly be detected under the same conditions, therefore encouraging active site selectivity which is favorable to construct a Z-scheme system. This work pointed out only oxide semiconductors containing Cr 3+ such as doped SrTiO 3 (Cr−Ta-doped), doped CaTiO 3 (Cr−Ta-doped), and CrTaO 4 showed H 2 production activities if loaded by Pt as a cocatalyst, notably Pt/ SrTiO 3 :Cr,Ta exhibited the highest H 2 production rate. Considering similar ionic radii of Cr 3+ and Ta 5+ in comparison to Ti 4+ , it was more likely that the dopants substituted the titanium ions in the lattice. The codoping of Ta and Cr makes the photocatalyst more efficient than that doped by Cr only for water splitting due to the neutralized total charge on the lattice, which likely decreased the probability of charge recombination and also increased stability for the reaction.
3.2.2. Oxynitrides. As mentioned above, TaON was investigated as both OEP and HEP. Pt-loaded TaON as a HEP was coupled with PtO x /WO 3 in an IO 3 − /I − aqueous solution to construct a Z-scheme system. 57 H 2 could be produced in the presence of NaI as hole scavenger on Pt/ TaON, and the evolution rate decreased due to the accumulation of IO 3 − ions, which were competitively reduced back to I − in place of water reduction. Additionally, TaON was shown to be resistant to photocorrosion in this half reaction process, as no O 2 or N 2 was detected (by gas chromatography), probably due to efficient scavenge of holes by I − adsorbed on the TaON surface. 57 To some extent, O 2 was also evolved, which was due to a thin Ta-oxide overlayer on TaON because the surface of TaON was oxidized under visible light irradiation as confirmed by the in situ multiple internal reflection IR (MIR-IR) results. 58 Overall water splitting was achieved when PtO x / WO 3 was combined and stably produced the stoichiometric H 2 and O 2 for 60 h at the pH of 5.3, reaching 0.5% of AQY at 420 nm. 59 Mixed tantalum oxynitride perovskites of ATaO 2 N (A = Ca, Sr, Ba) have shown to be active for visible light driven H 2 evolution. 60,61 These materials were synthesized by annealing amorphous A 2 Ta 2 O 7 under NH 3 . The absorption edges of ATaO 2 N (A = Ca, Sr, Ba) were at 520, 600, and 660 nm, respectively, showing a cationic effect that absorption increased along with the ionic radii ( Figure 8I). The highest H 2 evolution rate among them was from BaTaO 2 N (50 μ mol/h), followed by SrTaO 2 N (42 μ mol/h) and then CaTaO 2 N (24 μ mol/h). It was speculated that Ta 3 N 5 might exist in these materials, but Pt/Ta 3 N 5 yielded a nondetectable amount of H 2 in the presence of I − , therefore it could be confirmed that the activity came from the unique perovskite−oxynitride crystal structure. 47 The first Z-scheme for an overall water splitting system was reported to respond to wavelengths longer than 600 nm using the (Pt/BaTaO 2 N)−(IO 3  Figure 8II). 60 The quantum yield (based on Pt/BaTaO 2 N) for overall water splitting was determined to be 0.1% at 420−440 nm. Crucially, self-oxidative photocorrosion was not significantly detected in this system. However, as for Pt/SrTaO 2 N, self-oxidative photocorrosion was detected in the form of gaseous N 2 , and ultimately the system failed to work with PtO x /WO 3 . The reason why only Pt/SrTaO 2 N suffered from that was not clear yet but could likely be due to the formation of strontium iodate/iodide. However, photocorrosion could be decreased if calcination temperature was increased, suggesting that a higher crystallinity helped anchor the strontium ions in place. It was noticed though, that the rate of gas evolution eventually decreased in this system due to I 3 − formation. An accumulation of I 3 − ions that could not be reduced by PtO x /WO 3 effectively could be avoided by increasing pH of solution. Unfortunately, WO 3 is not stable in alkaline solution with pH over 9. Therefore, other visible light-driven photocatalysts with satisfying stability in high pH are still needed in order to be coupled with ATaO 2 N.
The modification of a TaON surface with ZrO 2 was reported to be useful as a passivation layer. 47 In the (Pt/ZrO 2 /TaON)− (IO 3 − /I − )−(PtO x /WO 3 ) system, 6.3% AQY at 420 nm was achieved. When annealed in ammonia, the formation of reduced species (Ta 3+ or Ta 4+ ) along with anion defects that act as recombination sites are basically inevitable. However, the formations of such unexpected reduced species and defects were effectively suppressed by ZrO 2 modification, resulting in higher efficiency. Various zirconium precursors were also investigated, with the best precursor being ZrO(NO 3 ) 2 ·H 2 O, which dispersed well and could modify the anionic vacancies during the nitridation process, leaving fewer impurities and reducing the recombination sites.
So far, the highest AQY of a Z-scheme in the presence of a dissolved mediator for overall water splitting was achieved using a (MgTa 2 O 6−x 47 The AQY of 6.8% at 420 nm under optimum conditions has been reported in 2015. The band locations ensured an absorption of visible light up to 570 nm as a type II heterojunction structure. 47 As expected, electron transfer occurs from the CB of MgTa 2 O 6−x N y to the CB of TaON, and holes transfer from the VB of TaON to the VB of MgTa 2 O 6−x N y , thus the recombination of charge carriers was effectively suppressed by the heterostructure, as confirmed by the prolonged charges' lifetime from time-resolved infrared spectra

Chemical Reviews
Review (TRIR). When irradiated by light with wavelength above 520 nm, the junction still produced H 2 while TaON alone did not. The gas evolution rate was even higher than that of Pt/ MgTa 2 O 6−x N y only with wavelength >420 nm, revealing a spatially separated charge transfer mechanism that was responsible for the enhancement of photocatalytic activity at longer wavelength. Pt metal was observed to be photoreduced on the surface of TaON instead of MgTa 2 O 6−x N y , which confirmed photoexcited electrons were transferred from MgTa 2 O 6−x N y to TaON. The one-pot synthesis method of ammonia treating MgTa 2 O 6 /Ta 2 O 5 provided a more intimate interfacial contact and fewer defects than a simple mechanical mixture of MgTa 2 O 6−x N y with TaON. Although heterojunction structures were widely used in photocatalysts, this was the first time for heterojunctions to be applied as a HEP in a Z-scheme system. This work demonstrated the possibility of using heterostructured photocatalysts for half reactions in overall water splitting.
3.2.3. Dye-Sensitized Systems. Although various semiconductors, not only conventional metal oxides, have been successfully employed as HEPs in Z-scheme water splitting systems as introduced above, there are still a limited number of efficient photocatalysts. This is primarily due to the difficulty in tuning the band levels of inorganic semiconductors. On the other hand, it is much easier to tune the energy levels (HOMO and LUMO) of organic dyes, as proven in the extensive studies on dye-sensitized solar cells with redox couples such as I 3 Thus, using dye-sensitized n-type semiconductors as a HEP with such a redox couple appears to be one of the promising ways to achieve efficient water splitting by harvesting wide range of visible light. However, the most serious issue of such organic dyes is their instability, specifically the instability of the oxidized state (reactive radical species in most cases) of simple dye molecules in aqueous medium. Actually, there are few reliable dye-sensitized photocatalysts that are active for H 2 production using a reversible electron donor such as I − .
Abe et al. demonstrated that the introduction of an oligothiophene moiety between the donor and acceptor parts dyes (see the structures in Figure 9) drastically improved the stability during their redox cycles in aqueous solution. 62 Both the electrochemical measurements and photocatalytic reactions indicated that the oxidized states of dyes having oligothiophene moieties possessed sufficiently long lifetimes even in aqueous solution. The long lifetime of oxidized species enabled dye regeneration by accepting an electron from the I − anion, achieving sustained and stoichiometric production of H 2 and I 3 − under visible light irradiation when combined with a n-type semiconductor, internally platinized layered niobate (Pt(in)/ H 4 Nb 6 O 17 ). Then water splitting into H 2 and O 2 under visible light was achieved using such coumarin or carbazole dyesensitized Pt(in)/H 4 Nb 6 O 17 as HEP, coupled with an OEP (IrO 2 −Pt-coloaded WO 3 ) in the presence of I 3 − /I − redox couple (see Figure 10), while the use of dyes without  oligothiophene moiety failed to split water stably due to the occurrence of oxidative decomposition of dyes. 62 3.2.4. g-C 3 N 4 . Apart from crystalline inorganic semiconductors, which have been well studied since the discovery of the Fujishima−Honda effect in early 1970s, polymeric semiconductors have attracted increasing attention since graphitic carbon nitride (g-C 3 N 4 ) was reported for visible light-driven photocatalytic hydrogen production ( Figure  11I). 63,64 Composed of ordered heptazine units, g-C 3 N 4 has been reported to have a relatively small band gap of 2.7−2.9 eV with a negative enough CBM position (at −1.2 V vs SHE at pH = 0) to drive electrons to reduce protons to hydrogen. 29 Composed of extremely abundant elements, graphitic carbon nitride is low-cost, robust, and nontoxic. However, as g-C 3 N 4 has only been used as a photocatalyst since 2008, efficiencies are still relatively low for H 2 production and the wavelength absorption limit is only 460 nm. Consequently, many strategies have since been employed to improve the compound in terms of performance and durability. Precursors including urea, melamine, dicyandiamide, and thiourea have been found to be able to produce g-C 3 N 4 by simple pyrolysis in various atmospheres, including air. 64 Elemental doping of P, B, F, I, S, O, and N was also examined. 65−72 Other methods such as nanostructure design and molecular incorporation were also applied herein to enhance the moderate performance. 73,74 However, only a few groups have succeeded in coupling the g-C 3 N 4 in Z-scheme water splitting.
Liu et al. reported a route to prepare g-C 3 N 4 from human waste (urea), making g-C 3 N 4 extremely cheap. 75 The first g-C 3 N 4 based overall water splitting system was reported by Martin and Tang et al. in 2014. 35 Further, it was found that the degree of polymerization influenced the performance of hydrogen evolution and the optimum urea-derived g-C 3 N 4 showed an internal quantum yield of 26.5% under visible light (400 nm) with a TON of 641, as one of the highest efficient visible light photocatalysts for water reduction reaction. 64 For this half reaction, various other cocatalysts such as Au and Ru were also examined, yet Pt was found to be the most efficient.
In the urea-derived g-C 3 N 4 based Z-scheme system, both (Pt/g-C 3 N 4 )−(IO 3 − /I − )−(PtO x /WO 3 ) and (Pt/g-C 3 N 4 )− (Fe 3+ /Fe 2+ )−(BiVO 4 ) represented splitting water continuously, with an initial gas evolution rate of H 2 and O 2 at 7.4 and 3.7, 1.5, and 0.8 μmol h −1 , respectively. The systems worked both under both full arc and visible, yielding reproducible H 2 and O 2 for over 24 h. The VBM of g-C 3 N 4 lied as deep as +1.5 V (vs SHE at pH 0), which is suitable for the both redox couples (IO 3 − /I − , +0.67 V at pH 7; Fe 3+ /Fe 2+ , +0.77 V) ( Figure 11II). One thing to note was that only urea-derived g-C 3 N 4 was active in these systems. Thiourea and dicyandiamide did not work here although they could photocatalytically produce H 2 from water. 35 Their unsatisfying band positions and the surface conditions could be the reason of the phenomena, but it could also be that they are simply not efficient enough to generate a detectable amount of hydrogen. As for urea-derived g-C 3 N 4 , the lower degree of surface protonation was more favorable for the adsorption of I − or Fe 2+ ions on the photocatalyst surface, thus making the cycle reaction easy to facilitate.
3.2.5. (Oxy)sulfides. (Oxy)sulfide materials have been studied as HEPs because many of them have both the narrow bandgaps for visible light absorption and very negative CBMs for water reduction, due to the contribution of S 3p orbitals to VBM formation. Zhao et al. applied an oxysulfide Sm 2 Ti 2 S 2 O 5 with a band gap of ca. 2 eV as a HEP in Z-scheme water splitting coupled with rutile TiO 2 as an OEP in the presence of IO 3 − /I − redox couple, while this system still required UV light for the excitation of the OEP. 76 Then, Kato et al. demonstrated for the first time a metal sulfide based Z-scheme water splitting with a [Co(terpy) 3 ] 3+/2+ redox in 2015 by utilizing Ru-loaded (CuGa) 0.8 Zn 0.4 S 2 as a HEP and BiVO 4 as an OEP. 77 The combination resulted in stable generation of H 2 and O 2 with a stoichiometric ratio under visible light irradiation (λ > 420 nm), as well as under a solar simulator, without any reduced activity. It was proposed that the predominant contribution of Cu 3d orbital to the VBM of the (CuGa) 1−x Zn 2x S 2 photocatalyst is crucial in achieving such stable reaction without suffering from self-oxidative deactivation (photocorrosion). This efficient molecular shuttle will be discussed in more detail in a later section (section 4.3.3).

Water Splitting with Solid Redox Mediators
As mentioned above, the redox mediator pairs existing in a Zscheme more or less compete with reduction on the HEP and oxidation on the OEP, although these redox couples relayed electrons well in many reported systems. A simple solid electron shuttle instead of the redox pairs should be more favorable as it rules out the redox diffusion process, 48 which could also be facile and appropriate to practical use. 99 Dissolved ions redox-mediated systems are easily investigated by half reaction tests for either H 2 or O 2 with appropriate sacrificial reagent, following pure water splitting by the coupled pairs. However, in a solid shuttle system, one has to synthesis the HEP/shuttle/OEP double junction first before testing the gas evolution activities. Another obstacle to solid junctions is the charge recombination between layers. 96 The following section describes the successful examples of Z-schemes for pure water splitting using a solid mediator, which in principle should be more facile than systems that use a liquid mediator. The primary requirement for the solid mediator is high conductivity, which includes such materials as graphene oxide, gold nanoparticles, and carbon nanoparticle/dots.
3.3.1. Photoreduced Graphene Oxide. The first visible driven solid mediator Z-scheme for overall water splitting by Iwase et al. is illustrated in Figure 12, where photoreduced graphene oxide (PRGO) was used as the electron shuttle in a system composed of BiVO 4 and Ru/SrTiO 3 :Rh. Doped SrTiO 3 acted as the HEP and BiVO 4 as the OEP, which were the most commonly used two photocatalysts in a dissolved mediate system. 99 The highest rates achieved were 11 and 5.5 μmol h −1 for H 2 and O 2 production, respectively, indicating pure water splitting. When irradiated by light, due to the staggering of band positions, electrons in BiVO 4 migrated through the PRGO layer to combine with the positive charges (i.e., Rh 4+ species) in SrTiO 3 :Rh. Meanwhile, the photogenerated holes in BiVO 4 and the photoexcited electrons in Ru/SrTiO 3 :Rh were used to oxidize and reduce water, respectively, thus forming an overall water spitting. As a conductor, PRGO worked as the mediator. Different from other "dissolved redox mediators", PRGO was a conductor and transferred electrons freely between semiconductors, but the back pathway was suppressed because the photoexcited electrons in BiVO 4 cannot reduce proton and the positive charges (i.e., Rh 4+ species) in Ru/ SrTiO 3 :Rh cannot oxidize water at the same time. 96 PRGO was photo reduced on BiVO 4 instead of on Ru/SrTiO 3 :Rh, which could help electron relay. Both the reduced GO on Ru/ SrTiO 3 :Rh and hydrazine-reduced GO failed to perform overall water splitting. The reason was that in those two occasions, the hydrophobicity was so high that the powders turned to be immiscible in water. According to the authors, this phenomenon could be seen by eye and the carbon state shown in XPS confirmed the low portion of C−O bond which was favorable for good dispersion. So, to maintain the enhanced activity in the Z-scheme, the hydrophobicity at the surface had to be controlled. 99 These authors have also applied this strategy (i.e., the use of PRGO as solid electron mediator) to metal sulfide-based Zscheme systems, in which various Cu-based sulfides such as CuGaS 2 , CuInS 2 , and Cu 2 ZnGeS 4 were employed as HEPs combined with rutile TiO 2 (OEP) via PRGO. 100 Although these early systems required UV light irradiation for the excitation of TiO 2 , these authors later combined CuGaS 2 with CoO x -loaded BiVO 4 to demonstrate water splitting under visible light. 100 The CoO x cocatalyst was revealed to play a crucial role in the Z-scheme system to promote water oxidation on BiVO 4 , thereby enhancing the electron injection from CB of BiVO 4 to CuGaS 2 through PRGO. It should be noted that there is a strong relation between the activity of present Zscheme systems and the photoelectrochemical properties of photocatalyst materials employed. Only the combination of ptype and n-type photocatalysts for H 2 evolution and O 2 evolution, respectively, was effective at present even if the sufficient overlap between the band potentials for electron transfer was satisfied between the HEP and OEP.
3.3.2. Au and Ag. The first all-solid-state Z-scheme system reported was composed of Au@CdS/TiO 2 for methyl viologen reduction reaction instead of water splitting. 110 By the deposition−precipitation method, a core−shell-type nanoparticle was obtained, which was specified to be CdS-coated Au nanocrystals on TiO 2 surface. The Au@CdS/TiO 2 photocatalytic activity was much higher than any of two components. The reasons could be due to the higher chargeseparation efficiency by Z-scheme mechanism, for electrons in CB of TiO 2 transfer to Au and then to VB of CdS, influenced by the large work function of gold. 110 Although this work was not visible light-driven Z-scheme water splitting, it opened a possible window for all-solid-state Z-scheme system development. Far from satisfactory, many efforts have been made in the past decade but the majority mainly focused on the dye degradation or half reaction of H 2 evolution, while only several systems worked for water splitting. One of the significant challenges is the interfacial contact between the solid electron mediator, HEP and OEP. Another challenge is that the system should remain stable during reaction. Thus electron mediators are limited to conductive/electron rich materials such as Au and reduced graphene oxide (RGO). 98,111 Very recently, an efficient Z-scheme system composed of SrTiO 3 :La,Rh/Au nanoparticle/BiVO 4 :Mo photocatalyst sheets was reported, which could be prepared by two methods. 98,112 On the photocatalyst sheet prepared by the first method, HEP and OEP particles were physically embedded into the gold layer (see Figure 13a). The STH energy conversions efficiency reached 1.1% with an AQY of 33% at 419 nm, which were both nearly 1 order of magnitude higher than that reported in the liquid mediator system and became a bench mark result in the Z-scheme systems. 98 Unlike PRGO used in the previous section, the underlying gold layer provided a much better charge transfer for the embedded photocatalysts via a particle transfer method. The contact resistance of a Schottky barrier, which is the most common problem in solid state Z-scheme systems and also the strongest dependency in this work, between the metallic gold layer and semiconductors, was reduced by controlling the annealing time and temperature. The extended annealing temperature and duration enhanced the contact between them, probably due to crystallinity, while too high temperature or too long durations decreased the activity considerably, which was attributed to the volatilisation of V 2 O 5 . 30 The back reaction, which was a crucial factor in other Z-scheme systems, was considered in this work as well, as H 2 and O 2 formed at sites in close proximity to each other. Protection layers of Cr 2 O 3 and amorphous TiO 2 were applied for effectively capping the surface and suppressing back reactions. 113 The second method was a screen-printing technology. 98 The sheet design could be screen printed by an ink, which contained photocatalysts loaded with Au. The water splitting activity of a printed sheet went up according to an increasing amount of Au in the colloid, indicating the Au nanoparticles acted as an electron conductor between particles ( Figure 14). The STH in the second system was 0.1%, which was less than the first example that was due to difficulty in the appropriate thickness of cocatalyst loading. The activity was also found to decay with time, which was due to aggregation of Au that lowered electron relay between HEP and OEP particles. However, the authors showed how screen printing technology could be a viable design for facile large-scale production in the future.
Apart from Rh-doped SrTiO 3 as the effective HEP, an oxynitride LaMg 1/3 Ta 2/3 O 2 N was recently demonstrated to function as visible light responsive HEP in such systems. Pan et al. reported that a photocatalytic sheet consisting of RhCrO x / LaMg 1/3 Ta 2/3 O 2 N (HEP) and rutile-TiO 2 (OEP) embedded in the Au layer with an amorphous TiO 2 protective layer exhibited activity for simultaneous evolution of H 2 and O 2 , while this sheet required UV light. 105 The authors then demonstrated water splitting under visible light by combining the RhCrO x / LaMg 1/3 Ta 2/3 O 2 N and BiVO 4 :Mo coated with an amorphous TiO 2 layer. 98 A reduced graphene oxide (RGO) was also applied as an additional solid mediator to the conductive layer to bridge particulate photocatalysts. 106 Other metals including Ag were also reported to work as shuttles in Z-scheme water splitting systems, such as ZnRh 2 O 4 / Ag/Ag 1−x SbO 3−y by Kobayashi et al., 108 where electrons are transferred from the CB of Ag 1−x SbO 3−y , via the solid-state mediator Ag, to the VB of ZnRh 2 O 4 . The size of Ag is believed to be crucial, which is still in the nanometer regime after heating to 1173 K. Such small Ag nanoparticles ensure maximum use of surface area and thus more sites for charge   transfer. It should be noted that even without Ag, the physical mixture of Ag 1−x SbO 3−y and ZnRh 2 O 4 was also able to split water but only 33% of the activity. This also confirms the important role of silver mediator in the charge transfer of Zscheme water splitting.
3.3.3. Conductive Carbon. As introduced above, quite efficient Z-scheme system composed of SrTiO 3 :La,Rh/Au nanoparticle/BiVO 4 :Mo photocatalyst sheets was reported to exhibit the highest STH (1.1%) of an artificial Z-scheme systems based on particulate photocatalysts. However, there were some issues in such a Au-mediator based system. One is the non-negligible cost of Au mediator, while it depends on the amount of use. However, the scaling up potential could be more practical if low-cost conducting media, as well as narrower bandgap semiconductors, are developed in this system. Another issue is the occurrence of backward reaction (i.e., water formation from evolved H 2 and O 2 ) on Au, while the rate of this reaction is much lower than that on Pt. Although the occurrence of backward reaction was considerably suppressed by the protection layer of Cr 2 O 3 and amorphous TiO 2 as described above, it was still appreciable when the reactions were carried out at high background pressures. Thus, the reactions were carried out at low background pressures of 5 KPa in most cases in the report. 107 Very recently, the same research group successfully introduced a conductive carbon as an efficient electron mediator to a particulate photocatalyst sheet. The unfavorable occurrence of backward reaction was effectively suppressed (see Figure 15), thereby affording efficient Z-scheme water splitting even at ambient pressure. 107 The SrTiO 3 :La,Rh/C/BiVO 4 :Mo sheet was demonstrated to split pure water (pH 6.8) with an STH energy conversion efficiency of 1.2% at 331 and 10 KPa while retaining 80% of this efficiency at 91 KPa. Importantly, the photocatalyst sheet showed higher efficiency than conventional powder suspension systems and photoelectrochemical parallel cells, mainly because of suppressed overpotential. The simple and inexpensive preparation of a carbon-based photocatalyst sheet, as well as the quite high STH efficiency, enables the possibility of large-scale application of such photocatalyst panels for the practical H 2 production from water under solar light.

Water Splitting without Mediators
Further, to simplify a Z-scheme structure for water splitting, a system based on interparticle electron transfer could be developed as mediator-free Z-scheme systems. Compared with Z-scheme containing dissolved redox mediator and solid mediator systems, mediator-free Z-scheme system is very challenging and was reported very recently, e.g., (Ru/ SrTiO 3 :Rh)−(BiVO 4 ) or (Ru/SrTiO 3 :Rh)−(Ir/CoOx/Ta 3 N 5 ) systems as illustrated in Figure 16. 94,97 The reversibility of an oxidation state between Rh 3+ and Rh 4+ is the benefit side in this Z-scheme charge transfer process. Ru/ SrTiO 3 :Rh has two absorption onsets, one at 520 nm due to the donor level formed by Rh 3+ to CB and the other at 580 nm assigned to the VB to acceptor level by Rh 4+ . The 520 nm band was responsible to the H 2 evolution, while the band at 580 nm was inactive. To clarify the details of Rh oxidation reversibility, H 2 reduction under different temperatures was tested. The Rh 3+ of SrTiO 3 :Rh under 473 K and below was able to be reversibly oxidized, but the one reduced at 573 K was quite stable. The characterizations of the samples treated at lower temperatures were similar to the nonreduced compounds, while the high temperature appeared to significantly reduce the reversible oxidation state. It was concluded that the surface Rh ions, with reversible oxidation state, served for the electron transfer process as the role of a shuttle. 107 In the (Ru/SrTiO 3 :Rh)−(BiVO 4 ) system, the AQY was reported to be 1.7% at 420 nm and the STH conversion efficiency reached 0.12%. To prove there was electron relay between two photocatalysts, two systems denoted as a separation system and a mixture system composed of the same materials were compared. When the two photocatalysts located at different regions in the reactor (separation system), overall water splitting was not observed because there was no physical contact or conductive/redox species in this system so that the charge transfer could not proceed through aqueous media. Water splitting only followed if the powders were mixed (mixture system), indicating the electron transfer between particles happened to realize the overall water splitting in this dual excitation system. The pH of the system influenced the activity mainly due to a zeta potential effect whereby the overall charge upon Ru/SrTiO 3 :Rh caused aggregation, the highest degree when pH was 3.5, while BiVO 4 was found not to be influenced by pH value.
An (Ru/SrTiO 3 :Rh)−(Ir/CoOx/Ta 3 N 5 ) system applied a similar concept. Ta 3 N 5 worked as an OEP and Ru/SrTiO 3 :Rh as a HEP. Electrons from the CB of Ta 3 N 5 transfer to the Rh    3d orbitals which were the donor in the HEP with the help of metallic Ir served as an electron sink. One highlight of this work lay on the CoO x cocatalysts which suppressed the degradation of the nitride compound, thus enhancing the water oxidation activity. 97,98 4. REDOX COUPLES IN Z-SCHEME WATER SPLITTING SYSTEMS As reviewed in the previous sections, most Z-scheme water splitting systems rely on an effective redox couple such as IO 3 Some new redox couples (e.g., metal complexes) and solid electron mediators (e.g., Au particles or reduced graphene oxides) have recently been demonstrated as well. Because of their significance in Z-scheme systems, this section concentrates on the fundamental properties of such redox couples dissolved in aqueous solution (except for the solid electron mediators), their development over the decades, and the features of Z-scheme systems associated with each redox couple. Table 1 summarizes the redox shuttles used so far in a Z-scheme system.
Redox couples used to achieve efficient Z-scheme water splitting should have the following properties: (1) an appropriate redox potential (between the reduction and oxidation potentials of water), (2) sufficient reversibility (i.e., redox cyclability) under reaction conditions (e.g., solution, pH) without deposition or precipitation, and (3) good transparency to UV and visible light. However, Z-scheme water splitting systems are difficult to construct, even if all of the intrinsic properties described above are present. This is because of the preferential occurrence of backward reactions on each photocatalyst. Reduction of Ox to Red (indicated by the broken lines in Figure 17 which is a redraw of Scheme 3 for discussion in this section), a thermodynamically advantageous backward reaction is favored over reduction of water to H 2 on a HEP. In contrast, oxidation of Red to Ox proceeds preferentially over oxidation of water to O 2 on an OEP. Furthermore, it is more  Figure 17). The key is to clarify the selectivity of redox couples based on their ad/ desorption onto the relevant photocatalysts. Thus, this section will mainly discuss the adsorption properties of each redox couple on specific photocatalysts and their intrinsic properties as both are crucial to construct an efficient Z-scheme water splitting system. Although iodine can possess wide range of valences from −1 to +7, the following redox cycles can proceed stably in aqueous solution: It should be noted that disproportional reactions of I 3 − can occur under basic conditions, for example: Thus, the pH of the reaction solution has a significant influence on the redox properties and thereby on the overall efficiency of Z-scheme water splitting. It appears that the redox cycle of I 3 − /I − dominates under acidic conditions, whereas the contribution of IO 3 − /I − increases with the increasing pH. However, it is fundamentally difficult to distinguish between the IO 3 − directly produced from I − via the six-electron process and that produced via disproportion of the I 3 − oxidatively generated from I − . I 3 − exhibits photoabsorption up to ∼500 nm, whereas neither I − nor IO 3 − shows photoabsorption of wavelengths longer than 300 nm, as shown in Figure 18. This strong, broad absorption by I 3 − undoubtedly causes a significant light shielding effect that decreases the number of photons absorbed by photocatalyst particles and consequently lowers the photocatalytic efficiency. Although photochemical reactions of these anionic species rarely occur under irradiation with wavelengths longer than 300 nm, a photochemical reaction of IO 3 − can occur under irradiation with UV light with λ < 300 nm and is accompanied by O 2 release. The reaction can be expressed by the following equation: This reaction can occur when a combination of high-pressure Hg lamp and quartz water jacket is employed by which strong UV light centered at 254 and 265 nm is emitted. Although the quantum yield of this reaction is not high, we should use a Pyrex water jacket (or a Pyrex window cell), which blocks UV light with wavelengths shorter than 300 nm, to distinguish photocatalytic O 2 evolution from photochemical O 2 evolution. 4.1.2. Photocatalytic Water Reduction Using I − as a Hole Acceptor. In 1996, Ohno et al. reported photocatalytic H 2 evolution using I − as a hole scavenger on Pt-loaded TiO 2 photocatalysts. 114 As seen in Figure 19, UV irradiation (λ > 300 nm) to Pt/TiO 2 (rutile, TIO-5) suspended in an aqueous KI solution (1 M, pH 2.4) resulted in simultaneous, stoichiometric (1:1) generation of H 2 and I 3

−
. The reaction was readily terminated, most likely due to a backward reaction (i.e., reduction of I 3

−
). The authors also revealed that pH had a

Chemical Reviews
Review significant influence on activity. As seen in Figure 20, the initial reaction rates decreased drastically when the pH increased to ∼4.
This remarkable change in activity was explained by the change in the surface charge of TiO 2 from positive to negative near the isoelectric point (∼6). Indeed, I − adsorption on the TiO 2 surface in aqueous solutions decreased significantly due to the increased pH, undoubtedly due to electrical repulsion between the negatively charged surface and I − anions. 114 However, the authors might underestimated the H 2 evolution activity of these Pt/TiO 2 photocatalysts under basic conditions because the activity was mainly evaluated based on I 3 − generation rather than H 2 evolution (see Figure 20).
In 2001, Abe et al. revealed that Pt-loaded TiO 2 photocatalysts efficiently produced H 2 and generate IO 3 − from basic aqueous solutions containing I − . 38,115 IO 3 − was the dominant product at pH > 9, whereas I 3 − was the main product when the pH was below 3. Mixtures were generated when the pH was between 3 and 9. The initial H 2 evolution rates from NaI (aq) on Pt/TiO 2 were almost same regardless of the pH (3 or 11) ( Figure 21). The rates decreased gradually in both cases, undoubtedly due to backward reactions (i.e., reduction of I 3 − and/or IO 3 − ). Fast decrease of H 2 evolution in the basic solution strongly suggested that the IO 3 − anions were reduced more efficiently than I 3 − , which was confirmed via ion chromatography instead of UV−vis. absorption. The amount of IO 3 − anions measured was smaller than the stoichiometric value derived from H 2 evolution amount, which was likely due to the facile adsorption of IO 3 − anions on the photocatalyst surfaces. In addition, the study showed that anatase TiO 2 particles could generate H 2 from NaI (aq) much more efficiently than rutile TiO 2 particles under basic conditions. 4.1.3. Photocatalytic Water Oxidation Using IO 3 − as an Electron Acceptor. The construction of Z-scheme water splitting systems using I 3 − /I − redox reactions has likely been disrupted by the low reactivity of I 3 − as an electron acceptor on conventional OEP. In contrast, IO 3 − as an electron acceptor enabled efficient and selective O 2 evolution on some photocatalysts such as rutile TiO 2 . 11,115 Figure 22 shows, for example, the time sequence of O 2 evolution on four different TiO 2 samples (A, anatase; R, rutile) in an aqueous NaIO 3 solution under UV light irradiation (λ > 300 nm). When the reaction took place in the presence of IO 3 − , both the rutile (R1 and R2) and anatase TiO 2 samples (A1 and A2) generated O 2 . However, the rutile samples showed a higher rate of O 2 evolution than their anatase counterparts. The AQY on rutile TiO 2 was ∼10% at 350 nm. Interestingly, O 2 evolution over rutile TiO 2 photocatalysts proceeded at an almost constant rate and continued until the total amount of O 2 reached 1.5 mmol     in the absence of NaI (Figure 22a). This occurred even when a considerable amount of I − was present (Figure 22b). An O 2 amount of 1.5 mmol was in line with the amount stoichiometrically anticipated from the IO 3 − (1 mmol) added to the solution before irradiation. The rate of O 2 evolution over anatase TiO 2 samples gradually decreased with time. This difference between anatase and rutile was observed more clearly when the reaction took place with excess I − , where O 2 evolution was completely suppressed on both of the anatase TiO 2 samples as shown in Figure 22b. This phenomenon on rutile TiO 2 was very interesting because the thermodynamically less favorable water oxidation proceeds preferentially even in the presence of I − . In contrast, such selective water oxidation did not take place on anatase TiO 2 samples.
The difference was due to adsorption characteristics regarding of IO 3 − and I − anions on TiO 2 samples. IO 3 − ions could readily adsorb on the surfaces of rutile TiO 2 particles (Figure 23c,d), whereas I − ions could not. Thus, one could expect IO 3 − anions to readily adsorb onto rutile TiO 2 surfaces, even at low concentrations, thereby reacting efficiently with photoexcited electrons whereas the produced I − could not be oxidized easily on rutile TiO 2 surface, leading to efficient water oxidation. In contrast, both IO 3 − and I − ions relatively strongly adsorbed onto the surface of anatase TiO 2 (Figure 23a,b), enhancing the unexpected back reaction (i.e., oxidation of I − ).
Such interesting adsorption of IO 3 − /I − couple and selective water oxidation were also observed on visible light-responsive WO 3 photocatalysts with PtO x species as cocatalysts for IO 3 − reduction. 11,40,59,79,116 As shown in Figure 24, O 2 generation from NaIO 3 (aq) proceeded consistently on a PtO x -loaded WO 3 11,40,115 Simultaneous evolution of H 2 and O 2 in a stoichiometric ratio was observed when a combination of Pt/ anatase TiO 2 and rutile TiO 2 was used but not observed on one of these photocatalysts or in other combinations, indicating that overall water splitting proceeded via a two-step photoexcitation mechanism combined with a redox cycle between anions IO 3 − and I − . As proved in the Figure 23, different adsorption behavior of IO 3 − to I − on rutile TiO 2 resulted in a very low IO 3 − concentration in solution, effectively suppressing the undesirable IO 3 − reduction (i.e., backward reaction) on the Pt/ anatase TiO 2 catalyst, thus producing a higher H 2 evolution rate. Thus the key to achieve water splitting is that preferential oxidation of I − to IO 3 − over the HEP and preferential oxidation of water to O 2 over the OEP must occur simultaneously.
In such IO 3 − /I − redox based Z-scheme systems, the pH of the solution significantly affected the total efficiency. 11,40,115 Figure 25a shows, for example, the rates of gas evolution over a mixture of Pt/anarase-TiO 2 with bare rutile-TiO 2 , suspended in 0.1 M NaI aqueous solutions with different pH values. The rates increased substantially as the pH increased from 3 to 9 and then decreased when the pH was above 11. The O 2 evolution was negligible at pH 3 because the main oxidation product produced on Pt/anatase-TiO 2 was I 3 − , which is not an efficient electron acceptor for O 2 evolution over bare rutile-TiO 2 . Although O 2 evolution was observed after an induction period at pH 5 to 7, the quantity produced was less than stoichiometric amount with respect to H 2 evolved. 11 This was certainly due to the accumulation of I 3 − in the solution. The induction period suggested that some IO 3 − anions could be produced via the disproportional reaction of I 3 − in the aqueous solution during the photocatalytic reaction as mentioned in the section 4.1.1. The accumulation of I 3 − also caused loss of light because of strong light absorption by the anion. In basic solutions (pH ≥ 9), the stoichiometric evolution of H 2 and O 2   was observed without induction period. Thus, conditions above pH of 9 resulted into high overall water splitting efficiencies due to efficient redox cycling between IO 3 − and I − at these conditions. The concentration of the redox mediator also significantly affected the overall efficiency, mainly due to the occurrence of backward reactions on each photocatalyst. Figure 25b shows, for example, the rates of gas evolution over the above mixture suspended in aqueous solutions with different NaI concentrations. The rates increased with increasing NaI concentration up to 40 mM, undoubtedly due to the enhanced reaction on HEP (Pt/anarase-TiO 2 ) with electron donor (I − ), as also supported by the increased rate of H 2 evolution on HEP in these concentrations (closed triangles). However, the higher concentration of I − resulted in the decrease in the rate of O 2 evolution over OEP (rutile TiO 2 ) due to the increased rate of the competitive oxidation of I − .
A visible light-driven water-splitting system was then constructed. SrTiO 3 co-doped with Cr and Ta (SrTiO 3 :Cr/ Ta) was found to be an active photocatalyst for both H 2 and IO 3 − production from an aqueous NaI solution under visible light irradiation (λ > 410 nm). 11,40,115 The photocatalyst was found to have a reasonably high activity for H 2 evolution from water under visible light irradiation in the presence of methanol as a sacrificial electron donor. 117 Although the rate of H 2 evolution over the Pt-loaded SrTiO 3 :Cr/Ta photocatalyst decreased remarkably with irradiation time due to the backward reaction (reduction of IO 3 − to I − by photoexcited electrons), using both Pt-loaded SrTiO 3 :Cr/Ta and Pt-loaded WO 3 photocatalysts for the first time resulted in simultaneous evolution of H 2 and O 2 (initial H 2 rate, 16 μmol h −1 ; initial O 2 rate, 8 μmol h −1 ) from an aqueous NaI solution under visible light irradiation (λ > 410 nm), as shown in Figure 26. It represents one of the most important technical breakthroughs in the field of photocatalytic water splitting. Since this demonstration, various photocatalyst materials, including metal oxides, oxynitrides, nitrides, oxysulfides, and carbon nitrides, have been proved to form effective H 2 -evolving photocatalyst couples with WO 3 for Z-scheme water splitting under visible light. there are significant issues that should be addressed in order to improve total efficiency. The first is the rapid backward reduction of IO 3 − on a Pt cocatalyst loaded HEP, which significantly lowers the efficiency of H 2 evolution. The second is the requirement of a more efficient cocatalyst (e.g., Pt) for reduction of IO 3 − on an OEP due to the nature of six-electron reduction process. An in-depth understanding of the adsorptivity and reactivity of IO 3 − /I − will provide clue to help control the selectivity of each step and developing new cocatalysts that exhibit high selectivity for forward reactions. The third issue is the limited choice of OEP that can function under visible light other than WO 3 . Although some metal (oxy)nitirdes such as TaON and Ta 3 N 5 can be used as OEP after appropriate surface modifications, 51 their long-term stabilities are insufficient due to the occurrence of self-oxidative deactivation. As shown above, the efficiencies of IO 3 − /I − redoxbased Z-scheme systems are significantly affected by the pH of the solution. Basic conditions (pH ∼ 9) resulted in the highest efficiencies over a combination of Pt/anatase TiO 2 and rutile TiO 2 under UV light. This is due to the efficient IO 3 − /I − redox cycle and absence of unfavorable intermediates such as I 3

−
. Unfortunately, such conditions cannot be applied to visible light-responsive systems that use WO 3 as an OEP due to the instability of WO 3 particles in basic solutions. Therefore, the development of a visible light-driven photocatalyst with good durability in basic solutions and good selectivity for water oxidation is strongly desired in order to improve the overall efficiency of IO 3 − /I − redox-based Z-scheme water splitting under visible light irradiation. 4 E°= +0.77 V) has been employed as a useful redox couple since the early stages of research on water splitting. In particular, the oxidant (Fe 3+ ) has been used as an effective electron acceptor for photocatalytic water oxidation (half reaction). The most advantageous characteristic of the Fe 3+ / Fe 2+ redox system, especially compared to the above IO 3 − /I − redox system, is the facile reduction of Fe 3+ to Fe 2+ due to an one-electron transfer process, which eliminates the need for a    19,60,90,118 The loading of an appropriate cocatalyst such as RuO 2 enhanced O 2 evolution in some cases. 19,119 Another advantage is the thermodynamic stability of the reduced Fe 2+ due to the significantly negative potential of further Fe 2+ reduction (Fe 2+ + 2e − → Fe; E°= −0.44 V). On the other hand, Fe 3+ /Fe 2+ redox has one disadvantage related to their chemical stability in high pH value solution. The Fe 3+ / Fe 2+ redox cycle is stable and reversible only at pH < 2.5, while the prompt precipitation of Fe(OH) 3 from Fe 3+ cationic hydrolysis takes place at higher pH when water is used as solvent. This irreversibility of Fe 3+ /Fe 2+ redox consequently limits of the option to photocatalysts who have sufficient stability under highly acidic conditions.
With regard to photoabsorption of the Fe 3+ /Fe 2+ redox couple itself, careful attention must be paid to aqueous solutions containing Fe 2+ species during UV light irradiation. As shown in Figure 27, aqueous solutions of FeCl 2 exhibit obvious absorption of wavelengths shorter than ∼280 nm. Aqueous FeCl 3 solutions exhibit stronger absorption across the UV spectrum and even in the visible light region up to ∼550 nm. This is especially noticeable when the concentration is high (e.g., 16 mM). It has been reported that photochemical oxidation of Fe 2+ to Fe 3+ proceeded under irradiation with UV light with wavelengths <300 nm, accompanied by reduction of water to H 2 as indicated by the following equation This photochemical reaction readily occurs if one uses a combination of high-pressure mercury (Hg) lamp and quartz water jacket, 120 leading to a misunderstanding that H 2 gas is generated on a HEP. To avoid this confusing photochemical reaction, it is suggested to use a Pyrex water jacket (or at least a Pyrex window) to block UV light with wavelengths shorter than 300 nm. In contrast, such photochemical reaction rarely occurs with Fe 3+ in spite of its strong, broad photoabsorption up to the visible light region, while such strong absorption up to the visible light region will involve a significant light shielding effect, lowering the photocatalytic efficiency. Thus, the Fe 3+ /Fe 2+ redox concentration must be carefully optimized during development of a Z-scheme system, as is the case with IO 3 − /I − .

Photocatalytic
Water Oxidation Using Fe 3+ as an Electron Acceptor. Nevertheless, the Fe 3+ /Fe 2+ redox system, especially with Fe 3+ as an electron acceptor, has been employed extensively. This is most likely due to its easy handling, understood reaction mechanism and easy quantitative determination of product (i.e., Fe 2+ ) concentrations. In 1962, Krasnovskii et al. presented a brief description of water oxidation with a Fe 3+ electron acceptor using WO 3 particles as a photocatalyst. 118,121 Unfortunately, this pioneering work attracted little attention for long time. In 1982, Dawent et al. reported a valuable study on photocatalytic O 2 evolution from aqueous FeCl 3 solutions using unmodified or cocatalyst-loaded WO 3 particles under visible light (λ > 400 nm). 119 In this detailed study, a reasonable correlation between the O 2 evolution rate and the WO 3 photoabsorption spectrum was confirmed using a high-radiance monochromator. The typical quantum yield of O 2 evolution was estimated to be ∼0.3% at 405 nm. Furthermore, the significant influence of redox concentration was revealed. As shown in Figure 28, the rate of photocatalytic O 2 evolution on WO 3 increased dramatically with the initial Fe 3+ concentration in aqueous solutions of up to 5 mM. This was attributed to the enhanced rate of consumption of photoexcited electrons by Fe 3+ . The rate decreased significantly as the concentration increased beyond 5 mM. This decreased O 2 evolution could be justified via light shielding by Fe 3+ , as explained above. However, the authors did not identify the origin at that time (they proposed several possibilities, including light shielding). On the other hand, increasing the Fe 2+ concentration, even up to 1 mM as seen in Figure 29, significantly decreased the rate of O 2 evolution, most likely due to the preferential occurrence of backward oxidation of Fe 2+ by photogenerated holes (Fe 2+ + h + → Fe 3+ ) in place of the oxidation of water. Such a significant backward reaction, i.e., reoxidation of Fe 2+ by holes, was observed with WO 3 as well as with other visible light-responsive photocatalysts such as BiVO 4 . 86 Therefore, it is one of the biggest issues that must be addressed to improve the overall efficiency of water splitting.
4.2.3. Unique Adsorption Properties of Fe 3+ /Fe 2+ that Enable Highly Selective Water Oxidation. Although recent progress on strategies for addressing backward reactions will be   reviewed in the next section (Section 5), Ohno et al. have performed a critical study that has shed light upon how to address this issue. Their study revealed the unique adsorption properties of the Fe 3+ /Fe 2+ redox system on specific photocatalysts such as rutile TiO 2 . They demonstrated that the photocatalytic oxidation of water with a Fe 3+ electron acceptor proceeded with reasonably good quantum yield (∼9% at 365 nm) on rutile TiO 2 particles. More importantly, O 2 evolution continued until almost all of the Fe 3+ ions in the solution were reduced into Fe 2+ ions. 122,123 As seen in Figure 30, rutile TiO 2 particles preferentially adsorb Fe 3+ ions. Particularly, Fe 3+ ions adsorb exclusively when both Fe 3+ and Fe 2+ ions are present. Thus, the photoexcited electrons generated on rutile TiO 2 particles could react efficiently with significant amounts of Fe 3+ adsorbed on the surface. Simultaneously, photogenerated holes preferentially oxidized water to produce O 2 even in the presence of Fe 2+ , in spite of a thermodynamic disadvantage, because of the negligible adsorption of Fe 2+ ions on the rutile surface. Although such unique Fe 3+ /Fe 2+ redox couple adsorption properties were found only in a few specific photocatalysts, such as rutile TiO 2 , this pioneering work provided better understanding of photocatalytic reactions with redox couples, as well as strong suggestion on how to address these backward reactions to improve total efficiency.

Half Z-Scheme
Coupled with Photochemical or Electrochemical Reaction for Water Splitting. As described above, photocatalytic O 2 evolution using Fe 3+ electron acceptor has been studied extensively since the 1980s and demonstrated on several photocatalysts, including some that were responsive to visible light such as WO 3 . However, reliable overall water splitting (i.e., simultaneous generation of H 2 and O 2 ) through Fe 3+ /Fe 2+ redox-based Zscheme photocatalysis was not achieved until 2004. This was undoubtedly due to the lack of effective H 2 -evolving photocatalysts that could reduce water to H 2 during oxidation of Fe 2+ ions to Fe 3+ , even under UV light irradiation, despite extensive exploration of the subject.
Thus, several complex systems were developed to demonstrate overall water splitting with a Fe 3+ /Fe 2+ redox system. For example, Sayama et al. demonstrated a Z-scheme type water splitting system in 1997 by taking advantage of the abovementioned "photochemical" H 2 evolution from aqueous Fe 2+ ions (see Figure 31). 19 In this system, O 2 evolution proceeded "photocatalytically" on RuO 2 -loaded WO 3 particles suspended in an aqueous solution containing a Fe 3+ acceptor. This was accompanied by the generation of Fe 2+ . H 2 evolution occurred "photochemically" via the absorption of photons by the aqueous Fe 2+ ions with wavelengths shorter than 280 nm, resulting in the simultaneous, stoichiometric evolution of H 2 and O 2 . 23 Although this system seems distant from artificial photosynthesis because the natural sunlight on Earth's surface contains little short-wavelength UV (λ < 280 nm), the authors demonstrated the interesting concept of achieving separate production of H 2 and O 2 by applying alternating photoirradiation, as shown in Figure 32. The first run was carried out via irradiation of FeCl 2 (aq) with short wavelength UV, without a photocatalyst, which generated only H 2 . The second run was performed by using a Pyrex jacket to irradiate the system with UV wavelengths longer than 300 nm, in the presence of a WO 3 photocatalyst, resulting only in generation of O 2 . Thus, alternating photoirradiation with different wavelengths of UV light through two different water jackets demonstrates alternating production of H 2 and O 2 .
Another example of a complex Fe 3+ /Fe 2+ redox system was demonstrated by Ohno et al. 124 Two kinds of rutile TiO 2 photocatalysts were suspended in aqueous solution, and two    types of redox couples (Fe 3+ /Fe 2+ and Br 2 /Br − ) in each solution were "electrochemically" coupled via two Pt electrodes and an external circuit (Figure 33a). In compartment I, a Ptloaded TiO 2 photocatalyst reduces water (H + ) to H 2 and oxidizes Br − to Br 2 (and/or probably Br 3 − ) under UV light irradiation. Another bare TiO 2 photocatalyst is suspended in compartment II to reduce Fe 3+ to Fe 2+ and oxidize water to O 2 under UV light. Simultaneously, the Fe 2+ ions accumulated at the electrodes in compartment II were "electrochemically" oxidized back to Fe 3+ by the stronger oxidation potential of Br 2 (or Br 3 − ), which was reduced back to Br − (see the energy diagram shown in Figure 33b). Consequently, water could continuously split into H 2 and O 2 under photoirradiation. This was accompanied by the transportation of H + through a Nafion membrane. Although this system is more complicated than other Z-scheme systems consisting of suspension of two different photocatalysts and one redox couple in each singlecompartment cell, conditions such as concentrations of reagents or quantities of photocatalyst powders could be independently optimized in each compartment and thus might allow us to improve the total water splitting efficiency in some cases.
4.2.5. Fe 3+ /Fe 2+ Redox-Based Z-Scheme Water Splitting. In 2004, a simple Z-scheme water splitting system consisting of two different photocatalysts and a Fe 3+ /Fe 2+ redox couple was demonstrated by Kudo et al. The system used SrTiO 3 doped with Rh cations (SrTiO 3 :Rh) as the HEP. 36 Kudo et al. explored new class of visible light-responsive photocatalysts based on the doping of cations such as Cr 3+ or Rh 3+ into a conventional SrTiO 3 semiconductor and found that Pt-loaded SrTiO 3 :Rh photocatalysts showed reasonable activity for H 2 evolution from water as well as simultaneous oxidation of Fe 2+ ions via harvesting of visible light with wavelengths of up to ∼550 nm. 16 They then combined the Pt/SrTiO 3 :Rh with BiVO 4 , an OEP that was also developed by the same group in the presence of a Fe 3+ /Fe 2+ redox system. 3,36,85,86,96,125−127 One important feature of this system is its unexpectedly high selectivity for H 2 evolution on the Pt/SrTiO 3 :Rh photocatalyst, even in the presence of Fe 3+ ions.
As shown in Figure 34a, excess O 2 was produced during the initial stage of the first run in 2 mM FeCl 3 (aq); due to the more efficient reaction with the Fe 3+ acceptor on BiVO 4 , the rate of O 2 generation gradually decreased and became stoichiometric with H 2 evolution. H 2 and O 2 were produced in a stoichiometric ratio (2:1) from the start of the second run after evacuation of the gas phase. The reaction in 2 mM FeCl 2 (aq) also exhibited nearly stoichiometric evolution of H 2 and O 2 at the end of the second run, while H 2 evolution predominated during the initial stage of the first run (see Figure  34b). This was attributed to a more efficient reaction on Pt/ SrTiO 3 :Rh with a higher concentration of the Fe 2+ donor. In both cases, Fe 3+ and Fe 2+ ions reaching a steady state have  Careful analysis of the two redox couples, one can find that H 2 production in IO 3 − /I − redox system was significantly suppressed by the coexistence of a small amount of IO 3 − due to a preference for reduction of IO 3 − anions by photoexcited electrons, 11,40,116 whereas H 2 evolution in Fe 3+ /Fe 2+ redox system was enhanced by an increase in the Fe 3+ concentration when concentration of FeCl 2 was fixed at 2 mM (see Table 2). Enhanced H 2 production with increased Fe 3+ concentration is unusual in view of the electron acceptability of Fe 3+ . However, the results of reactions in various aqueous solutions containing Fe 3+ suggest that the adsorption of [Fe(SO 4 )(H 2 O) 5 ] + and/or [Fe(OH)(H 2 O) 5 ] 2+ on the Pt surface efficiently suppresses both the undesirable backward reactions on the Pt surface, such as water formation from H 2 and O 2 and the reduction of Fe 3+ by H 2 (see Figure 35). The authors did not mention the possibility of suppressing another key backward reaction, the photocatalytic reduction of Fe 3+ by photoexcited electrons on Pt. However, the presence of highly selective H 2 evolution implies that this backward reaction is also suppressed. The rate of O 2 evolution over the BiVO 4 photocatalyst decreased as the Fe 2+ concentration increased due to backward reactions, in a manner similar to that observed with WO 3 . However, the BiVO 4 photocatalyst can generate O 2 even in the presence of a high (e.g., 5 mM) Fe 2+ ion concentration (see Figure 36).
Because both the HEP and OEP possess sufficiently high selectivity, the combination of SrTiO 3 :Rh and BiVO 4 resulted in efficient water splitting through the Fe 3+ /Fe 2+ redox cycle.
The unusually high selectivity of the Pt/SrTiO 3 :Rh photocatalyst for both reduction of H + and oxidation of Fe 2+ enabled the simultaneous, stoichiometric evolution of H 2 and O 2 under visible light via combination with various narrow-bandgap metal oxides such as WO 3 and BiMoO 6 , which acted as O 2 -evolving photocatalysts, even if their selectivity for water oxidation was not high. Not only such conventional metal oxides, but various      56,88 For long time after the first demonstration of Fe 3+ /Fe 2+ redox-based Z-scheme water splitting in 2004, there was only one available HEP (SrTiO 3 :Rh) that worked with the system. In 2014, Tang et al. demonstrated that an organic semiconductor g-C 3 N 4 could function as an effective HEP in Fe 3+ / Fe 2+ redox-based Z-scheme systems, 35 as described in the previous section.
4.2.6. Issues and Future, Prospective Fe 3+ /Fe 2+ Redox-Based Z-Scheme Systems. One of the biggest issues with the Fe 3+ /Fe 2+ redox system, along with the strict limitation on pH conditions below 2.5, is the limited choice of active HEP. Only SrTiO 3 :Rh and g-C 3 N 4 have been demonstrated so far. Thus, only wavelengths of up to ∼550 nm can be used for H 2 evolution. Such a troublesome lack of photocatalyst materials is attributed to (1) high rate of reduction of Fe 3+ on the cocatalyst or photocatalyst surface and (2) slow Fe 2+ oxidation on the photocatalyst surface. The former is reasonable considering that Fe 3+ cations work as efficient electron acceptors on various semiconductor photocatalysts even without any cocatalysts. Thus, it appears that a considerable fraction of photoexcited electrons readily react with Fe 3+ cations adsorbed on the surfaces of conventional metal oxides before they reach the cocatalyst (e.g., Pt) where water reduction should occur. The development of effective oxidation sites for Fe 2+ oxidation will be the key to addressing the latter cause (i.e., slow Fe 2+ oxidation). Photocatalytic H 2 evolution on SrTiO 3 :Rh photocatalysts proceeds via photoexcitation of electrons from the Rh 3+ donor level to the conduction band under irradiation, generating oxidized Rh 4+ species. It is thus reasonable to consider that the Rh 4+ species work as effective and efficient oxidation sites for Fe 2+ cations, especially when they are near the photocatalyst particle surface. The deeper understanding of the unique reactivity of Pt-(or Ru-loaded) SrTiO 3 :Rh photocatalysts may provide clues which aid in the development of other active photocatalysts for H 2 production from aqueous Fe 2+ solution. The activity of the g-C 3 N 4 photocatalyst for this extremely difficult reaction also implies that surface properties which contrast with those of conventional metal oxides can be used for effective adsorption and reactivity in the Fe 3+ /Fe 2+ redox system. Further investigation and understanding of the surface properties of the newly developed g-C 3 N 4 photocatalyst will also support the development of alternative photocatalysts that can harvest a wider range of visible light.

Metal Complex Redox System
Transition metal complexes have some advantages as redox mediator compared to simple ion couples, such as tunable redox potential by changing the ligands and good reversibility over a relatively wide pH range, and thus they can undoubtedly be alternative redox candidates in Z-scheme water splitting. In 2001, Graẗzel et al. have proved that cobalt complexes could rival the conventional I 3 − /I − redox couple for dye-sensitized solar cells (DSSCs) because of their favorable kinetic behaviors, 128,129 and then demonstrated highly efficient DSSCs with a light-to-electric power conversion efficiency of ∼8% by using cobalt complexes such as [Co(dbbip) 2 ](ClO 4 ) 2 (dbbip = 2,6-bis(1′-butylbenzimidazol-2′-yl)pyridine) in organic solvents (e.g., a mixture of acetonitrile and ethylene carbonate). 128 In 2013, Spiccia et al. reported the first application of a cobalt tris(2,2′-bipyridine)-based aqueous electrolyte for the fabrication of highly efficient and stable DSSCs with a maximum energy conversion efficiency of ∼5%, 130 suggesting that such complexes show high potential to applications to the Z-scheme water splitting. However, the greatest concern in such applications is the possibility of oxidative decomposition of the metal complexes with organic ligands, specifically on OEPs.  92 One of the most important findings in this study was that the appropriate choice of photocatalyst materials, especially from the standpoint of oxidation power, is crucial in establishing such metalcomplex-based Z-scheme water splitting. As seen in Table 3, only SrTiO 3 :Rh photocatalysts (loaded with a Pt or Ru cocatalyst for water reduction) generate H 2 in the presence of reduced species of a Co complex (i.e., [Co(bpy) 3 ] 2+ or [Co(phen) 3 ] 2+ ) as the electron donor. On the other hand, conventional photocatalysts such as Pt/TiO 2 and Pt/SrTiO 3 exhibit no activity for H 2 evolution even under UV light irradiation. These findings indicated that the Rh species doped in SrTiO 3 :Rh played a crucial role in oxidizing the Co complex based redox, as in the case described for the Fe 3+ /Fe 2+ redox system in the previous sections. Additionally, the negligible activity of nondoped materials such as SrTiO 3 implies that the oxidative decomposition occurred via the holes generated in the valence bands of such metal oxides, which possess a strong oxidation power. The choice of OEP is also crucial in achieving stable and stoichiometric water splitting using these complexes. Table 4 summarizes the results of the overall water splitting by using combinations of Ru/SrTiO 3 :Rh (HEP) and various OEPs in the presence of a Co complex redox, along with the results of comparative experiments (e.g., in the presence of Co 2+ or ligands instead of the Co complex). The use of conventional WO 3 as an OEP resulted in the generation of a much smaller amount of O 2 gas than the stoichiometrically expected values. On the other hand, the use of BiVO 4 afforded stoichiometric generation of H 2 and O 2 with steady rates (see Figure 37 for the time courses of gas generation).
The use of TiO 2 co-doped with Cr and Sb cations also resulted in almost stoichiometric gas evolution. This can be explained by the oxidation potentials of photogenerated holes on each OEP. The oxidation power of conventional metal oxides such as TiO 2 and WO 3 is generally strong and sufficient for the decomposition of various organic molecules and metal complexes, attributed to their deeply positive levels of valence band maximum at around +3.0 V vs SHE formed by the O 2p orbitals. On the other hand, the valence band maximum of BiVO 4 with a monoclinic scheelite structure is located at a more negative potential (∼2.5 V vs SHE) because of the significant contribution of the Bi 6s orbitals to the VBM. Thus, it appears that the undesirable oxidative decomposition of Co complexes is suppressed because of the moderate oxidation power of BiVO 4 and TiO 2 :Cr,Sb photocatalysts.
The authors also found that the use of excess Sr in the preparation of SrTiO 3 :Rh via a solid-state reaction, and the calcination temperature changes, significantly affected the activity for H 2 production using Co complex redox mediator. The SrTiO 3 :Rh sample prepared at 1372 K with 3 mol % excess Sr afforded the highest rates (H 2 , 100 μmol h −1 ; O 2 , 47 μmol h −1 ) in the Z-scheme reaction combined with BiVO 4 under visible light (λ > 420 nm, from 300 W Xe lamp with cutoff filter). As seen in Figure 37, the water splitting reaction using such an optimized combination proceeds steadily for 17 h of irradiation, with a 2.1% quantum yield at 420 nm. The turnover number of reacted electrons with the [Co(bpy) 3 ] 3+/2+ redox was calculated to be 56, indicating sufficient stability of the metal complex as an electron mediator in the Z-scheme water splitting. The combination can also split water not only under a solar simulator (AM-1.5, 100 mW cm −2 ) but also under natural sunlight, with about 0.06% solar energy conversion efficiency for H 2 production.
For the [Co(bpy) 3 ] 3+/2+ redox, its availability over a relatively wide pH range from 3.5 to 9 was confirmed, while almost neutral conditions (pH 6.9) shown the highest efficiency in water splitting with the (Ru/SrTiO 3 :Rh)−(BiVO 4 ) combination. The influence of pH on the total efficiency could be explained by the change in the driving forces of electron transfer between each photocatalyst and the [Co(bpy) 3 ] 3+/2+ redox, as illustrated in Figure 38. The band levels of metal oxide semiconductors generally shift with pH, which is attributed to surface protonation and deprotonation, whereas the redox potential of [Co(bpy) 3 ] 3+/2+ remains unchanged. Thus, the decreasing pH value increases the driving force of electron injection from [Co(bpy) 3 ] 2+ to HEP but simultaneously decreases the driving force from the OEP to [Co(bpy) 3 ] 3+ . Consequently, the total efficiency for water splitting shows a volcano-type dependence on pH, resulting in the highest gas evolution rates at around neutral pH with this combination of HEP and OEP.

Separate Evolution of H 2 and O 2 by [Co-(bpy) 3 ] 3+/2+
Redox-Based Z-Scheme. One of the advantages of a Z-scheme water splitting system compared to a conventional one-step system is its capability of separate evolution of H 2 and O 2 by employing a membrane that divides the HEP and OEP. Such a separate gas production is useful in practical applications to the storage of H 2 separately from O 2 (i.e., air) to avoid explosion. However, demonstration of such a separate production has likely been inhibited by the facile occurrence of backward reactions on each photocatalyst. Kudo et al. demonstrated for the first time such separate evolution by using the above-mentioned (Ru/SrTiO 3 :Rh)−(BiVO 4 ) combination with [Co(bpy) 3 ] 3+/2+ redox in a two-compartment reactor as shown in Figure 39. Such a successful demonstration   (e.g., 0.25 mM), thus affording the successful separate production of H 2 and O 2 in the two-compartment reactor equipped with a simple membrane. It should also be noted that the mass transport resistance of redox couples across a membrane would limit the Z-scheme water splitting activity in such design, and thus engineering an adequate membrane should be on a case-by-case basis. A significant challenge would be the overall design of a two-compartment reactor as well, however, that is an engineering-based problem and would have to be addressed separately.

Application of Metal Sulfide
Semiconductor with [Co(terpy) 3 ] 3+/2+ Redox Couple. Various metal sulfides have been extensively studied as photocatalysts for visible-lightinduced H 2 production because many sulfides possess the narrow bandgaps required for visible light absorption and the appropriate band levels for water reduction. Another fascinating feature of such metal sulfides is the potential for band tuning via the formation of solid solutions, as demonstrated in binary CuGaS 2 −ZnS (i.e., (CuGa) 1−x Zn 2x S 2 ) and ternary ZnS-CuInS 2 −AgInS 2 (i.e., (CuAg) x In 2x Zn 2(1−2x) S 2 ) systems. However, the biggest drawback of such metal sulfide photocatalysts is their instability, which is attributed to the occurrence of selfoxidative deactivation (e.g., CdS + 2h + → Cd 2+ + S). Thus, one of the most reasonable and attractive strategies to achieve overall water splitting using such metal sulfides under visible light involves a Z-scheme water splitting system.
In 2015, Kudo et al. demonstrated metal sulfide based Zscheme water splitting with a [Co(terpy) 3 ] 3+/2+ redox. 77 The combination of Ru (0.5 wt %)-loaded (CuGa) 0.8 Zn 0.4 S 2 as HEP and BiVO 4 with [Co(terpy) 3 ] 3+/2+ redox resulted in stable generation of H 2 and O 2 with a stoichiometric ratio under visible light irradiation (λ > 420 nm), as well as under a solar simulator, without any reduced activity. The turnover number of the reacted electrons to the number of S atoms at the surface of (CuGa) 0.8 Zn 0.4 S 2 exceeded 70 after 17 h of irradiation, indicating that this sulfide semiconductor could function stably as a HEP with a [Co(terpy) 3 ] 3+/2+ redox with negligible photocorrosion. The VBM of the (CuGa) 0.8 Zn 0.4 S 2 was confirmed to predominantly consist of the Cu 3d orbital at a potential higher than that of the S 3p orbital. It is therefore speculated that the possibility of photocorrosion was decreased in this material compared to that in other conventional metal

Chemical Reviews
Review sulfides such as CdS or ZnS in which the VBMs are mainly formed by S 3p orbitals. Although such a successful combination of metal sulfides and Co complex redox is limited to this combination at present, the results may provide a new and promising strategy to achieve efficient water splitting under visible light using metal sulfide photocatalysts.

Polyoxometalate Redox System
Although some Co complexes have been employed as electron mediators in Z-scheme water splitting as descried above, the available OEP is still limited to BiVO 4 that possess moderate oxidation potential than conventional metal oxides. This is probably associated with oxidative decomposition of organic ligands in such metal complexes by photogenerated holes. It has been reported that many transition metal-incorporating polyoxometalates (POMs), which basically contain no organic components, exhibit reversible redox behavior derived from valence differences in the incorporated transition metals and/or in the constituent elements of the framework. Additionally, most POMs are stable in mildly acidic conditions, while the available ranges of pH are depended on their structure and constituent elements. Thus, such POMs will be alternative redox candidates with high stability in Z-scheme water splitting.
In  Figure 41). Interestingly, the initial rates were almost independent of the initial concentration of SiW 11 Mn II (50 or 100 μmol). Although the rate of H 2 evolution gradually decreased with irradiation time, finally reached the stoichiometric values expected from the amount of SiW 11 Mn II present in the initial solutions. This behavior can be explained by the adsorption property of SiW 11 Mn III /SiW 11 Mn II on the photocatalyst. As shown in Figure 42, the adsorption of SiW 11 Mn II on the SrTiO 3 :Rh was almost saturated at around 1 mM (corresponding to 50 μmol in the present reaction solutions). Thus, the rate of electron injection from SiW 11 Mn II was certainly limited even when the reaction was started with higher amount. More importantly, SiW 11 Mn II preferentially adsorbed on the surface of SrTiO 3 :Rh over SiW 11 Mn III ; the adsorption amount of SiW 11 Mn III is much less than that of SiW 11 Mn II when compared at a same concentration (e.g., 1 mM). It is thus reasonable to consider that the rate of undesirable backward reaction (reduction of SiW 11 Mn III ) is low even if the concentration of SiW 11 Mn III increases with the progress of reactions, resulting in relatively steady rate of H 2 evolution in the present system.
The oxidized SiW 11 Mn III was also proven to function as an electron acceptor on PtO x /WO 3 photocatalyst for O 2 evolution (see Figure 43). When the reaction was initiated with ca. 96 μmol of SiW 11 Mn III , O 2 evolution was observed at almost steady rates saturating at about 23 μmol, which was very close to the stoichiometrically expected value (24 μmol). When the

Chemical Reviews
Review reaction was initiated with a larger amount of SiW 11 Mn III (197 μmol), the rate of O 2 evolution gradually decreased with irradiation time, undoubtedly due to the occurrence of backward reaction to some extent. Then, the combination of Ru/SrTiO 3 :Rh and PtO x /WO 3 with SiW 11 Mn III /SiW 11 Mn II redox at pH = 4.5 resulted in stable generation of H 2 and O 2 with a stoichiometric ratio under visible light irradiation (λ > 420 nm).
Although the present SiW 11 Mn III/II POM redox possesses an advantage over conventional Fe 3+ /Fe 2+ redox, i.e., the availability in mildly acidic conditions (e.g., pH = 4.5), it has some disadvantages. One is the necessity of effective cocatalyst on an OEP, even though this redox cycle takes place via oneelectron transfer similar to Fe 3+ /Fe 2+ . Although bare WO 3 could generate O 2 in the presence of SiW 11 Mn III , the loading of PtO x or Pt significantly enhanced the O 2 evolution, indicating that the reduction of SiW 11 Mn III requires an appropriate cocatalyst. The second one is the light absorption in visible light region. Although the reduced SiW 11 Mn II possesses absorption mainly at UV light region along with weaker one in visible region, the oxidized SiW 11 Mn III shows strong and relatively broad absorption peak (at 500 nm) ranging from 450 to 600 nm. Thus, the accumulation of SiW 11 Mn III undoubtedly causes a significant light shielding effect that decreases the number of photons absorbed by photocatalyst particles and consequently lowers the photocatalytic efficiency. Therefore, the searching and/or developing POMs with efficient redox cycle and good transparency to visible light is strongly desired for improving the total efficiency of POM-based Z-scheme water splitting.

Vanadate Redox System
A vanadate redox mediator has been reported to function in a Z-scheme system of (Ru/SrTiO 3 :Rh)− (VO 2 + /VO 2+ )−(Fe− H−Cs−WO 3 ) by Miseki et al. 109 The vanadate ion shuttle is expected to be redox active according to the following equation: It should be noted that the V 5+ structure varies with the pH of solution (i.e., pH 0−3 = VO 2+ , pH 3−9 = H 2 VO 4− , and pH 9−13 = HVO 4 2− ), 131 and the redox potential of the V 5+ /V 4+ is also highly dependent on the complex structure (VO 2 + /VO 2+ = 1.0 V and H 2 VO 4 − /HV 2 O 5 − = 0.7 V, vs NHE, pH = 0). 109 Therefore, the authors carefully adjusted the pH of the system to 2.4, ensuring the desired pathway occurred, and the pH did not fluctuate. Prior to this, the authors measured the O 2 and H 2 half reactions in the presence of VO 2+ and VO 2 + , respectively. The intermediate species were confirmed by UV−vis spectra as the VO 2+ peak at 766 nm submerged in O 2 production and disappeared during H 2 production. Water is selectively oxidized due to H 3 O + species on the surface of Fe−H−Cs−WO 3 and that the selective water oxidation may be much faster than unfavorable back-reactions. Ru/SrTiO 3 :Rh as a HEP contains

Chemical Reviews
Review active VO 2+ oxidation sites but no VO 2 + reduction sites, thus suppressing the undesirable reactions and resulting in ideally selective performance. This work also demonstrated the context of separated stepwise Z-scheme H 2 and O 2 production. Briefly, a HEP was added at the beginning with VO 2+ under visible irradiation. After the total H 2 reached a stable value, indicating the VO 2+ were all converted to VO 2 + , the HEP was filtered, and OEP was added to perform the O 2 evolution consequently. The activity was reported to be H 2 , 50 μmol h −1, and O 2 , 25 μmol h −1 , in one-pot reaction, and stepwise reaction was about 1.4 times higher. This might infer that the physical mixing of HEP and OEP caused negative interactions and a light scattering/ shielding effect of Ru/SrTiO 3 :Rh also influenced the activity to some extent.
To summarize, there are still some limitations for Z-scheme systems apart from advantages. For example, proper working conditions (e.g., pH) of different types of mediators and mediator-free Z-scheme for water splitting are very important for their performance. These are summarized in Table 5.

STRATEGIES FOR IMPROVING EFFICIENCY OF Z-SCHEME SYSTEMS
The previous sections detailed various Z-scheme systems with the appropriate redox couples. However, their efficiencies, except for the recently developed photocatalyst panels, are still moderate and must be further improved to demonstrate the feasibility of practical H 2 production. Various methods such as cocatalyst loading, surface modifications, morphology controls, and introduction of a heterojunction have been attempted to improve the efficiency of existing visible light responsive photocatalysts. This section will focus on the cocatalyst loading and the surface modifications, both of which have been proven as effective ways to improve the efficiency of Z-scheme systems.

Cocatalyst Loading
Cocatalyst loading onto photocatalyst particles has been widely studied as an effective way to improve the performance of semiconductor photocatalysts. The main role of such cocatalyst has been generally regarded as (1) enhancing the charge separation in the photocatalyst and (2) facilitating the surface chemical reactions by the carriers. For most photocatalysts used for conventional one-step water splitting, the cocatalysts that enhance water reduction but not catalyze the water formation from H 2 and O 2 , (e.g., NiO x , Cr 2 O 3 /Rh) are often employed to improve the efficiency of water splitting. Some recent studies also investigated coloading of an effective O 2 -evolving cocatalyst (e.g., Mn 3 O 4 , IrO 2 ) with a H 2 -evolving cocatalyst. 132 Cocatalyst loading also influences the efficiencies of Z-scheme water splitting systems significantly. In the Z-scheme systems with a dissolved redox couple, loading of the cocatalysts that facilitate reduction or oxidation of the redox mediator, rather than water, often dominates the total efficiency. For example, the reduction of IO 3 − to I − via six electrons often becomes the rate-determining step in O 2 evolving part, thus the loading of effective cocatalysts (e.g., PtO x , RuO 2 ) can drastically increase the rate of O 2 evolution and thereby improve the total efficiency of Z-scheme water splitting.
5.1.1. Cocatalysts for H 2 Evolution Photocatalysts with IO 3 − /I − Redox Couple. Most semiconductor photocatalysts that have been developed for water splitting require suitable cocatalyst particles to promote H 2 production. Noble metals such as Pt that have low overpotentials for water reduction are often used as effective cocatalysts. However, it is well recognized that Pt-loaded photocatalysts such as Pt/TiO 2 cannot split pure water into H 2 and O 2 in a simple aqueous suspension system. This is basically attributed to backward reactions (i.e., catalytic H 2 O formation from the evolved H 2 and O 2 ) on Pt surface before the produced gases escape from the suspension to gas phase.
On the other hand, many of the Z-scheme water splitting systems based on IO 3 − /I − or Fe 3+ /Fe 2+ redox couple can generate H 2 and O 2 simultaneously even in the presence of Pt cocatalyst on HEP, as seen in Table 1. One can notice that most of the Z-scheme systems with IO 3 − /I − (or I 3 − /I − ) redox employ Pt cocatalyst for the HEPs, with a few exceptions. The spontaneous formation of iodine layer on Pt surface is probably one of the reasons why H 2 and O 2 can generate stably even in the presence of Pt. 133 It has been reported that I − anions in aqueous solution were spontaneously and oxidatively chemisorbed as zerovalent I atoms on Pt polycrystalline electrode and formed stable monolayer on the Pt surface. 134 As seen in Figure 44, a Pt/TiO 2 photocatalyst after photoreaction in an aqueous solution of NaI (1 mM) exhibits strong peaks at 618.5 and 630 eV in the I 3d XPS spectra, which indicate the presence of zerovalent iodine. These peaks were also observed over the same sample that was stirred in the NaI aqueous solution for 12 h in dark but not over the unloaded TiO 2 under the same conditions. As shown in Figure 45, the pressure of gas mixture of H 2 and O 2 rapidly decreased on a fresh Pt/TiO 2 sample due to the water formation from H 2 and O 2 on naked Pt surface. On the other hand, the pressure of the mixture decreased much more slowly on an I-adsorbed Pt/TiO 2 sample. These findings indicate that the iodine monolayers formed on Pt surface effectively suppress the backward reaction from H 2 and O 2 . It was reported that electron transfer could occur through the iodine layer formed on the Pt surface and also that the iodine layers could be reductively eliminated from the surface either by exposure to electrochemically generated H 2 or by application of sufficiently negative potentials. 133 Therefore, the following two mechanisms can be proposed for the H 2 production over Iadsorbed Pt cocatalyst surfaces: (1) water reduction proceed directly on the iodine layer by photoexcited electrons migrated from semiconductor bulk via Pt, and (2) a part of the iodine layers is reductively eliminated by photogenerated electrons or

Chemical Reviews
Review H 2 and then water reduction takes place on the exposed Pt surface.  126 On the other hand, Kudo et al. demonstrated that Ru cocatalyst exhibited better performance than Pt. As seen in Figure 46, the possible backward reactions (e.g., water formation from H 2 and O 2 and reduction of Fe 3+ ions by H 2 ) are much more effectively suppressed on Ru cocatalyst. 85  85 The gas evolutions proceeded with almost steady rates on Ru cocatalyst-based system even at relatively high pressures, whereas the rate of gas evolution over the Pt cocatalyst-based system gradually decreased after 10 h of irradiation, undoubtedly due to the occurrence of backward reactions accompanied by an increase in the pressures of evolved H 2 and O 2 . It should be noted that the X-ray photoelectron spectroscopy (XPS) analysis on the samples revealed that the surface of Ru cocatalyst loaded by photodeposition is oxidized by air or water. Thus, it appears that the backward reactions are effectively suppressed by the presence of such thin oxide layer on the surface, which is different from what happened on Pt cocatalyst. 5.1.3. Nanostructured Cocatalysts for Selective H 2 Evolution. Suppressing backward reactions of redox couples is undoubtedly one of the most effective ways to improve the total efficiency of Z-scheme water splitting systems. For example, Pt nanoparticles that are selectively deposited in the interlayer space of layered semiconductor can work as selective reduction sites for H 2 production with I 3 − /I − or IO 3 − /I − redox couple. 62,135 As seen in Figure 48, H 2 evolution over dyeadsorbed Pt/TiO 2 photocatalyst with reversible I − electron donor rapidly terminates with quite small amount of H 2 evolved, while it can generate H 2 stably from an aqueous triethanolamine solution. This is due to the significant backward reaction (reduction of I 3 − to I − ) on the Pt particles as illustrated in Figure 49a. On the other hand, H 2 evolution proceeded at a steady rate on the dye-adsorbed Pt(in)/ H 4 Nb 6 O 17 sample, in which Pt nanoparticles were selectively loaded into the interlayer spaces of layered niobate. This is a

Chemical Reviews
Review result of the reduced backward reaction caused by I 3 − anions on Pt particles because of the electrostatic repulsion between I 3 − and the negatively charged niobate layers (see Figure 49b).
Suzuki et al. also demonstrated that the selective deposition of Pt cocatalyst into interlayer spaces of a nitrogen-doped tantalate (H + /RbLaTa 2 O 7−x N y ) was effective to achieve nonsacrificial and endergonic H 2 evolution from water using I − as a reversible electron donor under visible light. 135 5.1.4. Cocatalysts for O 2 Evolution Photocatalysts. The loading of cocatalyst, such as PtO x or RuO 2 , is indispensable for efficient O 2 evolution over OEPs such as WO 3 using the IO 3 − anion as an electron acceptor, because the reduction of IO 3 − to I − via six-electrons often becomes the ratedetermining step. On the other hand, even unmodified photocatalysts such as WO 3 and BiVO 4 can generate O 2 with Fe 3+ electron acceptors. Suzuki et al. revealed a reasonable relation between the overpotential of each cocatalyst material for IO 3 − reduction and the rate of O 2 evolution on the cocatalyst-loaded photocatalyst with IO 3 − as electron acceptor. 136 As shown in Figure 50, the overpotentials of MO x catalysts for IO 3 − reduction, which were determined by electrochemical measurements, decrease in the following order: CoO x > Rh 2 O 3 > IrO 2 > RuO 2 > PtO. Clearly, a good correlation was drawn that the rate of O 2 evolution on ex-Ca 2 Nb 3 O 10 /K + samples (prepared via exfoliation and subsequent restacking of layered KCa 2 Nb 3 O 10 material) loaded with these MO x cocatalysts increased with the decreased overpotentials for IO 3 − reduction on the MO x . 136 The surface chemical states also significantly influence the performance of cocatalysts. It was revealed that WO 3 photocatalysts loaded with platinum oxide (PtO) showed a much higher activity for O 2 evolution in the presence of the electron acceptor IO 3 − compared to those loaded with Pt metal. 59 One can see in Figure 51 that an interesting dependence of the O 2 evolution rate from aqueous IO 3 − over the WO 3 photocatalyst loaded with Pt species upon calcination temperature after impregnation of H 2 PtCl 6 . 59 The rate increases significantly with increasing calcination temperature from 723 to 823 K, and then decreases drastically above 873 K. The XPS and STEM analysis revealed that the WO 3 samples loaded with Pt species at low temperatures (723−823 K) mainly contained highly dispersed PtO species, while that prepared at 873 K contained mixture of PtO and Pt species. On the other hand, the samples prepared at higher temperatures (>873 K) dominantly had relatively large particles of metallic Pt. Considering the fact that the photodeposited Pt/WO 3 samples, which were loaded with highly dispersed Pt metal particles, showed a much lower activity for O 2 evolution (ca. 5 μmol h −1 ), it appears that PtO is a more favorable cocatalyst for O 2 evolution with IO 3 − acceptor than Pt metal. It is well-known that Pt nanoparticles act as highly efficient catalyst for O 2 reduction. Thus, one possible reason for the low activity of WO 3 samples loaded with a Pt metal cocatalyst is competitive reduction of O 2 molecules on the Pt metal surface alongside the reduction of IO 3 − . The increase in the O 2 evolution rate with temperature from 723 to 823 K is also due to the improved interface between WO 3 and PtO particles, which enables efficient electron transport from WO 3 to PtO.
The loading of the RuO 2 cocatalyst on TaON enhanced the O 2 evolution in the presence of IO 3 − , which enabled the RuO 2 / TaON photocatalyst to be used as an OEP in combination with Pt/TaON. Structural analyses and (photo)electrochemical measurements revealed that the activity of RuO 2 /TaON photocatalyst was strongly dependent on the generation of optimally dispersed RuO 2 nanoparticles, which simultaneously promote both the reduction of IO 3 − and oxidation of water. 33 It was recently revealed that the Ru(OH) x Cl y species obtained by the calcination of RuCl 3 ·nH 2 O at or below 200°C showed a much higher activity for IO 3 − reduction than conventional RuO 2 , while the Ru(OH) x Cl y species rarely catalyzed water

Chemical Reviews
Review oxidation. 137 Figure 52 shows current−potential curves for the TaON electrodes loaded with Ru species at various temperatures (x) in a Na 2 SO 4 aqueous solution containing IO 3 − (1.0 mM) under dark conditions. XPS analysis indicated that the partially hydroxylated Ru species, e.g., Ru(OH) x Cl y , were loaded onto TaON by the wet-impregnation process followed by calcination at relatively low temperatures (∼200°C), whereas the calcination at higher temperatures resulted in the formation of conventional RuO 2 . Surprisingly, the onset potentials on the Ru-150 and Ru-200 electrodes, both of which mainly contain Ru(OH) x Cl y species, were observed around much more positive potentials of +0.1 V vs Ag/AgCl, along with astonishingly higher current densities. Co loading the Ru(OH) x Cl y and Co cation species (by simple impregnation of Co(NO 3 ) 2 at 100°C) as water oxidation cocatalysts significantly improved O 2 evolution on the TaON photocatalyst by more than a factor of 2 compared to that loaded with only the conventional RuO 2 cocatalyst (see Figure 53). These findings strongly suggest that the coloading of two different cocatalysts independently optimized for reduction and oxidation can be an effective strategy to enhance the efficiency of OEPs (see Figure 54) in Z-scheme water splitting systems.

Surface Modification
Sayama et al. demonstrated that the surface modification of WO 3 photocatalysts with Cs + species drastically improved the efficiency in O 2 evolution with Fe 3+ /Fe 2+ redox couple. 138 As seen in Table 6, the rate of O 2 evolution on WO 3 from aqueous Fe 2 (SO 4 ) 3 solution is significantly increased after the hydrothermal reaction of WO 3 particles in aqueous Cs 2 CO 3 solution at 393 K. The rate of O 2 evolution was further improved by the impregnation of aqueous Cs 2 CO 3 solution followed by calcination in air at 773 K. Interestingly, the rate of O 2 evolution over the Cs-impregnated sample was much enhanced in the second run (196 μmol h −1 ) compared to that in the first run (117 μmol h −1 ).
As seen in Figure 55, the total amount of O 2 evolved on Csmodified WO 3 reached 315 μmol, agreeing with the stoichiometric amount expected from the amount of Fe 3+ (1260 μmol) in the initial solution. Such stoichiometric O 2 generation indicates high selectivity of Cs-modified WO 3 photocatalyst toward water oxidation by holes (i.e., low selectivity toward backward oxidation of Fe 2+ ). Results about material characterization on such Cs-modified WO 3 photocatalysts and photoanodes revealed that "ion-exchangeable sites" were formed on the WO 3 surface by incorporating Cs cations via hydrothermal or impregnation treatment with Cs 2 CO 3 . The Cs cations incorporated could be easily replaced by H + via acid-treatment as well as during the reaction in aqueous Fe 2 (SO 4 ) 3 solution with low pH (∼2). Then, the H +exchanged sites strongly absorbed water molecules as H 3 O + , which were more readily oxidized to O 2 by the photogenerated holes. It was also suggested that Fe 2+ cations occupy a part of "ion-exchangeable sites", which enhances the electron transfer     80 As seen in Table 7, the rate of O 2 evolution on PtO x -loaded WO 3 photocatalyst was much increased by Cs-modification. The activity of Cs-modified WO 3 was drastically lowered by the subsequent calcination in air at 573 K, implying that the ionexchangeable sites formed via Cs modification absorb water molecules, similar to the case of Fe 3+ /Fe 2+ redox system, and facilitate water oxidation to O 2 .
Interestingly, the Cs-modification enables the PtO x /WO 3 photocatalyst to generate O 2 even in the presence of I 3 − electron acceptor. As seen in Figure 56, unmodified PtO x / WO 3 photocatalyst shows negligibly low activity for O 2 evolution with I 3 − electron acceptor, as previously reported. Such low activity of conventional PtO x /WO 3 photocatalyst with I 3 − is one of the biggest issues in Z-scheme water splitting with IO 3 − /I − redox couple because I 3 − anions readily generate and accumulate under acidic (or even neutral) conditions. However, the Cs-modified WO 3 photocatalysts exhibit surprisingly high activity for O 2 evolution as seen in Figure  56; the O 2 evolution continues with an almost steady rate up to the expected amounts. Consequently, Z-scheme water splitting proceeds steadily with much higher rates of gas evolution by employing such Cs-modified WO 3 as an OEP coupled with Pt/ SrTiO 3 :Cr,Ta as HEP in aqueous NaI solutions with wider range of pH, as seen in Figure 57 for example.
Developing adsorption-controlled photocatalyst particles will be another effective strategy for improving the efficiency of Zscheme water splitting with dissolved redox couple. Tomita et al. recently demonstrated that the WO 3 particles with selectively exposed faces exhibited better selectivity for O 2 evolution with Fe 3+ /Fe 2+ redox couple. 91 As shown in Figure  58, the WO 3 particles prepared via hydrothermal reaction at greater than 255°C (HT-255 and HT-315) have rectangularlike shapes with considerably flat surfaces. It was found that the HT-315 sample, which was prepared at 315°C, exhibited favorable adsorption property to Fe 3+ /Fe 2+ redox couple compared to commercial samples (K-WO 3 and S-WO 3 ). For example, the HT-315 sample adsorbed much higher amount of Fe 3+ (∼7.0 μmol g −1 ) than commercial WO 3 sample (e.g., ∼1.1 μmol g −1 for K-WO 3 ) when compared at low concentrations of

Review
Fe 3+ (1.6 mM), in spite of the smaller specific surface area, as seen in Figure 59. On the other hand, the amount of Fe 2+ adsorbed on the HT-315 sample was slightly but appreciably smaller than that on K-WO 3 .
Consequently, the HT-315 showed higher rates and selectivity for O 2 evolution than commercial ones even with decreased Fe 3+ and increased Fe 2+ concentrations, as clearly seen in Figure 60. The efficiency of Z-scheme water splitting was indeed improved by employing this HT-315 sample as OEP couples with Ru/SrTiO 3 :Rh as a HEP in the presence of Fe 3+ /Fe 2+ redox couple. 91

Morphology Control and Heterojunctions
The morphology of photocatalysts also influences the properties, and generally a smaller particle size or higher crystallinity are believed to benefit the performance. 139 For example, the surface structure and particle size of NaTaO 3 can be tuned from 4 to 0.1 μm by doping with lanthanum and alkaline earth metal ions, resulting in an order of magnitude enhancement of water splitting activity. 140,141 A proper control of morphology also results in the enhancement of Z-scheme water splitting systems.
One interesting example is the "nature-inspired" architecture CdS/Au/TiO 2 system. 142 Ding et al. investigated the black butterfly wing scale and realized that such a quasi-honeycomb (QH) morphology could promote a significant increase in light absorbance. The authors managed to synthesize TiO 2 with a QH structure using the wing of the black butterfly as an organic template, creating three different varieties of templates (wing, plate, and scale). CdS/Au/TiO 2 with wing architecture morphology exhibited a 230% increase in comparison to plate architecture based TiO 2 , indicating that morphology control is of importance.
The construction of heterojunction is another commonly used strategies to control the properties of photocatalysts for superior performance including suppressed charge recombination. 29 An example is (MgTa 2 O 6 83 Compared to the mechanically mixed powder of both, the heterojunction structure exhibited a 2-fold increase in activity due to a much more intimate interfacial contact. Both the mixed and junction samples were found to have a prolonged lifetime of charge carries according to time-resolved IR spectra, but the latter had a smaller defect density according to UV/vis spectra. SEM results also proved that the photodeposited Pt was on the surface of TaON, indicating the electron transfer from MgTa 2 O 6−x N y to TaON enhanced charge separation. The advantages of such junctions compared to single HEP results in a benchmark AQY of 6.8% in Z-scheme water splitting in presence of liquid mediators.

CONCLUSIONS AND OUTLOOK
In this review article, following a short introduction of the conception of Z-scheme systems, we have thoroughly reviewed the recent progress on visible light-driven Z-scheme water splitting systems, with specific focuses on the recent developments of semiconductor materials, redox mediators, and the function of cocatalysts. Since the turn of the 21st century, various semiconductors, including metal oxides, (oxy)nitrides, (oxy)sulfides, oxyhalides, organic dyes, and carbon based nitrides, have been successfully incorporated into Z-scheme systems with a variety of different redox couples. The introduction of nonoxide semiconductors into Z-scheme water splitting systems has extended the available wavelengths

Chemical Reviews
Review up to 700 nm in comparison to conventional water splitting with one photocatalyst.
Among the three types of Z-scheme systems, the highest AQY in a liquid mediator system of 6.8% at 420 nm was achieved in a MgTa 2 O 6−x N y /TaON)−(IO 3 − /I − )−(PtO x / WO 3 ) system. An AQY of 1.7% at 420 nm was reported in the BiVO 4 −Ru/SrTiO 3 :Rh system without a need of shuttle mediator, resulting in an STH conversion efficiency of 0.12%. The most recent solid mediator system (Cr 2 O 3 /Ru-modified SrTiO 3 :La,Rh/C/BiVO 4 :Mo sheets) represented the record STH energy conversion efficiency of 1.2% at 331 K under 10 KPa. Although the three systems have individual advantages and disadvantages, they could be developed to meet the required solar to fuel conversion efficiency of 10% if all components could be optimized. Taking the recent solid mediator system (SrTiO 3 :La,Rh/Au nanoparticle/BiVO 4 :Mo sheets) as an example, one of its unsatisfactory is that the STH efficiency of 1.1% under 10 KPa dramatically declines as the pressure goes up due to back reactions, while later on the same group replaced Au layer by carbon film and maintained an STH of 1.0% under higher pressure of 90 KPa, showing the importance of a proper mediator to suppress the back reactions. 107 The solid shuttle mediator system at present represents the best efficiency to date and therefore should be further researched with impetus. To develop efficient OEPs and HEPs, new strategies for achieving efficient separation of electrons and holes in powdered semiconductor photocatalysts, such as p−n junctions or specifically exposed crystal facets, should be pursued continuously, along with developing new semiconductor materials with light absorption to longer wavelengths in the visible region. Polymer photocatalysts with readily tunable bandgaps and band positions could be another option to investigated, provided stability was on par with that of for example carbon nitride.
Although new redox couples such as metal complexes and polyoxometalates have been proven to work in the Z-scheme water splitting, most of these systems still require an IO 3 − /I − or Fe 3+ /Fe 2+ redox couple. However, both redox couples possess intrinsic problems in improving the total efficiency of water splitting, for example, the requirement of an efficient cocatalyst (e.g., Pt, PtOx) for reduction of IO 3 − on O 2 -evolving photocatalysts, attributed to the nature of the six-electron reduction. The limited choice of OEPs that can function with IO 3 − /I − under visible light and proper pH range, other than WO 3 , is also a problem. Although some metal (oxy)nitrides such as TaON and Ta 3 N 5 can be used as OEPs after appropriate surface modifications, their long-term stabilities are insufficient because of the occurrence of self-oxidative deactivation. One of the most critical issues with the Fe 3+ /Fe 2+ redox system, along with a strict pH operation window (i.e., below 2.5), is the limited choice of active H 2 -evolving photocatalysts. Only SrTiO 3 :Rh and g-C 3 N 4 have been demonstrated so far. Thus, only wavelengths of up to ∼550 nm can be used for H 2 evolution. Hence, the development of versatile and efficient redox couples will be a key step in achieving highly efficient water splitting. In such a development of redox couples, the control over reaction selectivity (forward or backward) by the intrinsic adsorption property of the redox couple, designing the surface properties of photocatalysts, and introducing reaction sites with high selectivity, will be other key issues. In parallel, solid mediators can be discovered by not only considering the electron conductivity but also their work functions to match with two separate photocatalysts. Clearly conductive and low cost carbon based materials such as graphene could be a good option.
So far, the charge transfer dynamics in the entire Z-scheme system has been less investigated although there are a few studies focusing on the so-called half Z-scheme. This poor understanding is one factor limiting the progress of the Zscheme system for solar fuel synthesis, which should be put more efforts in future. Nevertheless, a Z-scheme promises a higher maximum solar to fuel conversion efficiency (∼40%) than a single photocatalyst system (∼30%), 22 which is a major driving force for development. Furthermore, such Z-scheme systems have the potential to spatially separate H 2 and O 2 if a suitable reactor design is conceived. However, the current solar to fuel conversion efficiency is still very moderate (e.g., only μmol/hour gas evolution rates) for a Z-scheme and a scale-up process could be considered when a very efficient system is developed. Thus, construction of tandem systems that enable the utilization of different wavelengths on each photocatalyst (e.g., short wavelength on OEP and longer wavelengths on HEP), as well as separate production of H 2 and O 2 , will demonstrate a practical H 2 production system based on the mechanism of a double excitation Z-scheme.
water splitting and photocatalytic selective oxidation of organic compounds.
David James Martin studied Physics at the University of Liverpool (M.Phys.), and then completed a Ph.D. at UCL under the supervision of Prof. Junwang Tang, focusing on developing new photocatalysts for water splitting. He then accepted a postdoctoral position under Prof. Andy Beale and used X-ray scattering and absorption techniques to study heterogeneous catalysts under operando conditions. He currently works as a senior researcher in the group of Prof. Moniek Tromp at the University of Amsterdam, applying X-ray absorption and emission techniques to study homogeneous and heterogeneous catalysts, for applications such as selective oligomerization of ethylene, partial oxidation of propene, the low temperature WGS reaction, and electrochemical water oxidation. After that, he was appointed as a JSPS Fellow at NIMS, Japan, and a senior researcher in the Department of Chemistry at Imperial College London, working on solar fuels synthesis and mechanistic studies by time-resolved spectroscopies. After that, he took a faculty position at UCL. His current research interests lie in photocatalytic small molecules activation (e.g., CH 4 , N 2 , H 2 O, CO 2 ) and mechanistic aspects of solar energy conversion and water treatment, alongside microwave intensified fluidic chemistry. He is also a Fellow of the RSC.