Bridging electrocatalyst and cocatalyst studies for solar hydrogen production via water splitting

Solar-driven water-splitting has been considered as a promising technology for large-scale generation of sustainable energy for succeeding generations. Recent intensive efforts have led to the discovery of advanced multi-element-compound water-splitting electrocatalysts with very small overpotentials in anticipation of their application to solar cell-assisted water electrolysis. Although photocatalytic and photoelectrochemical water-splitting systems are more attractive approaches for scaling up without much technical complexity and high investment costs, improving their efficiencies remains a huge challenge. Hybridizing photocatalysts or photoelectrodes with cocatalysts has been an effective scheme to enhance their overall solar energy conversion efficiencies. However, direct integration of highly-active electrocatalysts as cocatalysts introduces critical factors that require careful consideration. These additional requirements limit the design principle for cocatalysts compared with electrocatalysts, decelerating development of cocatalyst materials. This perspective first summarizes the recent advances in electrocatalyst materials and the effective strategies to assemble cocatalyst/photoactive semiconductor composites, and further discusses the core principles and tools that hold the key in designing advanced cocatalysts and generating a deeper understanding on how to further push the limits of water-splitting efficiency.


Introduction
Reducing the amount of CO 2 emitted from burning fossil fuels is essential to mitigate global warming. 1 To meet this challenge while addressing the growing global energy demand, it is becoming increasingly important to develop sustainable, carbon-neutral energy sources. 2 Molecular hydrogen (H 2 ) is regarded as an ideal green fuel because it releases zero emissions and only produces water upon combustion. 3 Therefore, H 2 is expected to hold prime signicance as a high-energy-density fuel in future energy systems. In fact, H 2 fuel cells have been implemented to power vehicles with advanced H 2 transportation technologies. 4 Although hydrogen is the most abundant element on earth, the dihydrogen molecule rarely exists in nature; therefore, H 2 must be produced articially. Currently, steam reforming is the preferable method for producing commercial H 2 ; however, Masaki Saruyama is a Program-specic assistant professor in Institute for Chemical Research, Kyoto University. He received his PhD in Science from Tsukuba University in 2011 under the supervision of Prof. Toshiharu Teranishi. He joined Kyoto University in 2015 aer working for Mitsui Chemicals, Inc. His research interests involve development of inorganic nanoparticles for photocatalytic water splitting, ion exchange reaction of ionic nanocrystals, and self-assembly of nanoparticles.

Christian
Mark Pelicano received his PhD in 2019 from the Nara Institute of Science and Technology in Japan. He is currently working as a postdoctoral fellow in Kyoto University under the supervision of Prof. Toshiharu Teranishi to develop multi-metallic cocatalysts for photocatalytic water splitting. His research interests are centered on the synthesis and design of functional nanomaterials for solar energy conversion especially in the area of hybrid solar cells and photo/electrocatalytic hydrogen production. a substantial amount of energy is required to drive this process and CO 2 is emitted as a byproduct. 5 Recently, sunlight-driven water splitting has emerged as an attractive approach for green and sustainable H 2 production. 6 Among solar-hydrogen technologies, water electrolysis using electricity generated from solar cells is considered to be the most advanced pathway. 7 Combining a solar cell with an electrocatalyst (EC)-loaded electrode can promote water splitting, reaching >30% solar-tohydrogen efficiency (STH) using an InGaP/GaAs/GaInNAsSb triple-junction solar cell. 8 Thermodynamically, the reduction and oxidation of water start at 0 V RHE and +1.23 V RHE , respectively, meaning that the overall water-splitting reaction can start from 1.23 V. However, the activation energy for each half-reaction contributes to an overpotential, which requires an additional voltage (hundreds of millivolts) to overcome and drive the water-splitting process. 9 In practical applications, ECs are employed to reduce this overpotential, thereby improving the STH.
To date, the high cost of device manufacturing has hindered the practical implementation of solar cell-assisted H 2 -production devices. However, the direct decomposition of water through photocatalysis represents a promising alternative approach owing to its technical simplicity and low associated investment costs. 10,11 Scaling up a photocatalytic system is also relatively much easier because water splitting can proceed by simply immersing semiconductor photocatalyst (PC) powder in water under light irradiation. 12 Recently, a large-scale experiment (100 m 2 scale) demonstrated that immobilizing powdered PCs on panels could lead to the production of about 600 L of H 2 on a sunny day. 13 Such photoactive semiconductors can also be used as light-absorbing layer of photoelectrodes (PEs) for photoelectrochemical (PEC) water-splitting cells. 14 These electrodes are usually placed in separate compartments which are electrically connected through an external circuit. Although this conguration makes PEC systems more complex in terms of design, higher STH efficiencies have been achieved due to easier product collection and elimination of potential back-reactions. 15 For example, a tandem-type PEC cell reached an STH of >7%, which far exceeds current STH values for PC systems. 16 However, both PC and PEC systems still have large rooms to reach a target STH of 10% with a long-term durability for practical application. 12 To realize a higher water-splitting performance, substantial efforts have been mainly devoted to the development of semiconductors as efficient photoactive materials. 10 Cocatalysts (CCs), which are water-splitting catalysts coupled with PCs or PEs, also have been systematically studied as equally-important components in both PC and PEC systems. 17 Because the semiconductor surface tends to have a weak driving force for redox reactions involving water, coupling with CCs can facilitate these redox reactions by lowering the activation energy barrier. 17 Specically, charged-up CCs retain suitable steady-state potentials to drive the surface reactions by serving as trapping sites for photogenerated carriers (i.e., electrons and holes). In turn, this process not only promotes charge separation but also suppresses adverse charge recombination and photocorrosion. 18,19 While the integration of CCs with PCs and PEs is indispensable for achieving excellent efficiencies, the CC materials have been explored much less than the EC materials. The straightforward application of previously-reported ECs materials as CCs is challenging because it introduces various factors that must each be critically considered. 18 First, CCs should be small or thin enough so that they do not limit the amount of light absorbed by the PC and PE. 20 Second, the change in potential shi at the CC/semiconductor junction is critical for evaluating charge transfer efficiency. 21 These additional requirements have not yet been sufficiently addressed, delaying advancement related to CC materials.
In this perspective, we discuss the recent advances in the design of EC materials and the strategies for constructing CC/ PC and CC/PE composites, including our group's nanoparticle Fig. 1 Application of materials active for solar-driven water splitting via the hydrogen evolution reaction (HER) and oxygen evolution reaction (OER).

Toshiharu
Teranishi is a professor at Institute for Chemical Research, Kyoto University. He received his PhD from The University of Tokyo under the direction of Prof. Naoki Toshima in 1994, and spent seven and a half years at Japan Advanced Institute of Science and Technology as an Assistant Professor and an Associate Professor. In 2004, He moved to University of Tsukuba as a Full Professor, and moved to Kyoto University in 2011. Current research interests include precise structural control of inorganic nanomaterials and structure-specic functions for high-performance devices and photo-energy conversion.
(NP)-adsorption approach. A special emphasis is put on the key principles governing the rational assembly of an efficient CC/PC and CC/PE system. We hope this perspective will inspire researchers in pursuit of bridging the gap between the developments of highly-active EC and CC materials, which have only been studied separately to date (Fig. 1).

Development of electrocatalyst materials
Water splitting involves two concurrent catalytic half-reactions: the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER). Materials that can lower the activation energies of these reactions are utilized as ECs. Although noble metals such as Pt, Ru(O x ) and Ir(O x ) are still representative efficient ECs for water splitting, diverse earth-abundant compounds with comparable catalytic activities have also been developed based on comprehensive theoretical and experimental studies. 22 Herein, we briey introduce recently-developed EC materials for efficient water splitting and discuss their typical preparation methods. A comprehensive introduction of EC materials is beyond the scope of this perspective; thus, we encourage the interested reader to refer to recent reviews summarizing the development of EC materials. 23 22 It is believed that ECs with DG H* ¼ 0 are most suitable for efficient HER, and Pt has the DG H* closest to 0 among all elements. 25 From this perspective, materials based on the combination of other elements can be a promising replacement for Pt (a precious metal) if their DG H* is close to 0. 22 Signicant research efforts have uncovered new combinations of earthabundant elements that exhibit good HER activities (Fig. 2a). 23 MoS 2 was a pioneering EC material 26 that launched successive explorations into HER-active 2D metal dichalcogenides. With support from theoretical studies, considerable enhancements of the HER activities of MoS 2 compounds have been demonstrated by introducing defects and elemental dopants (Fig. 2b). 27,28 Metal (Ni, Co, Fe, W, Mo, .) phosphides, nitrides, and carbides are also subjects of extensive studies owing to their advantageous electronic structures. 27,[29][30][31][32][33][34][35][36][37][38] Tuning the metal composition has proven to be pivotal in realizing exceptional HER performance. 29 Such rationally designed EC compounds have realized comparable HER activities over a wide range of pH for practical applications.
2.1.2 OER electrocatalysts. Although noble metal-based IrO x and RuO x materials remain as the state-of-the-art ECs for OER, 39,40 earth-abundant 3d transition metal-based materials have attracted attention owing to their excellent OER activities (i.e., NiFeOOH). 41,42 Computational and experimental evidence have shown that the difference between the adsorption energies of the O* and OH* intermediates (i.e., DG O* À DG OH* ) is a reliable indicator for predicting the OER activity of an EC (Fig. 2c). 43 For example, Sargent et al. used theory and experiments to establish that an exceptional OER activity can be attained by tailoring the material's composition in terms of three metals (Co, Fe, and W), which enables an optimization of DG OH* (Fig. 2d). 44 In addition, researchers have discovered that chalcogenides and phosphides containing two or more types of metals can also function as excellent OER ECs. [44][45][46][47][48][49][50][51][52][53][54] 2.1.3 Bifunctional compounds as electrocatalysts. While exploring new EC materials, researchers have identied various compounds that are highly-active for both HER and OER. For example, multi-metallic compounds, such as oxides, phosphides, and chalcogenides, achieve efficient overall water splitting in a two-electrode conguration driven by a cell voltage of $1.6 V in alkaline electrolyte (Table 1). 48,[55][56][57][58][59][60] Higher pH conditions (alkaline media) are usually more appropriate for such bifunctional materials because they facilitate the ratedetermining OER, which has a larger overpotential than the HER. 61 Bifunctional EC materials represent promising CCs for photocatalytic water splitting owing to their inherent abilities to promote HER and OER simultaneously.

Synthesis and use of electrocatalysts
To date, innovative EC materials have been synthesized through various chemical and physical routes. Because the overall size of the ECs does not require ne-tuning (as long as they are conductive enough to work on electrodes), they have mainly been fabricated using facile strategies, including solvothermal syntheses, solid-state reactions, and electrochemical deposition (Fig. 3a). This enables rapid and widespread exploration of novel EC materials via screening experiments. 62 Simple oxides and chalcogenides can be directly synthesized by liquid-phase methods under mild conditions, 62,63 whereas hightemperature annealing is oen required to obtain phosphides, nitrides, and carbides from their corresponding preformed oxides. [34][35][36]51,52 Specically, they are annealed with precursors (P ¼ NaH 2 PO 2 ; N ¼ NH 3 ; C ¼ organic molecule) in a furnace at >300 C for complete conversion. To assess the resulting material's electrocatalytic activity, powdered ECs are embedded on a conductive carbon substrate, and the composite  is typically glued to an electrode surface using a Naon binder. 64 ECs can also be grown directly on porous conductive supports, such as metal foam or carbon paper/cloth, which increase their conductivity and loading per geometrical unit area ( Fig. 3c and d). 37,38,65 Such EC-grown substrates can be directly connected to a potentiostat as free-standing electrodes. Alternatively, nanosizing is a practical approach for designing active ECs with large surface areas because it provides unsaturated surface atoms and promotes greater atomic usage efficiency. 23 Effective techniques for fabricating EC nanoparticles (NPs) include colloidal synthesis and precursor decomposition on a support material. 51,66 Annealing a metal-organic framework (MOF) is a facile technique to produce multi-metallic NPs supported on a porous conductive carbonaceous substrate (Fig. 3b). 49 Moreover, recent advances have revealed that single-atom (SA) catalysts are emerging as new ECs with exceptional properties, such as maximum atomic usage efficiency, high activity of the metal complex, and high durability of the bulk catalyst. 67 Similar to powdered ECs, EC NPs and SA catalysts can be supported on conductive substrates and xed to an electrode. These EC electrodes operate as working electrodes in a three (working, reference, and counter)-electrode system to evaluate their catalytic performance towards HER and/or OER. 68 Overall water splitting is investigated in a two (working + counter)-electrode conguration using two identical EC-immobilized electrodes (Fig. 3h). 55 It is worth noting that ECs can undergo some changes during catalysis, especially in a OER environment. 69 According to the Pourbaix diagram, the surface or entire EC can transform into its corresponding [(oxy)hydr]oxide under alkaline conditions under a positive applied potential ( Fig. 4a-e). 70 Wu et al. reported that the surface of Co 4 N was progressively oxidized during OER cyclic voltammetry (CV) experiments in 1 M KOH electrolyte. 71 Even if the composition is carefully controlled during synthesis, the real active species may deviate from the as-prepared material. Such transformed oxides and (oxy)hydroxides frequently exhibit superior catalytic activities because their amorphous surfaces offer more active sites and enhanced electronic interactions relative to their pure oxide analogs ( Fig. 4f and g). 72 In calculations of the surface state, a lack of understanding of the real active species may mislead research conclusions.

Cocatalysts for PC and PEC systems
In the case of solar cell-assisted water splitting, EC-loaded electrodes can be employed in the same conguration as two electrode system. Therefore, improving intrinsic catalytic activity of EC directly increases the STH, aside from the solar cell efficiency. In terms of PC and PEC systems, the latest advancements in semiconductor design (e.g., defect engineering and morphology control) have paved the way for advanced PCs and PEs with high light-energy conversion efficiencies. 73 Nevertheless, additional considerations concerning the successful integration of efficient EC materials as CCs onto the PC and PEC systems must be addressed. First, it is essential for the CCs to have suitable NP dimensions and homogenous distribution on the PC and PE to ensure that they do not block the incident light. Second, designing an optimal interface between the semiconductor and CC is necessary because the stability and performance of the entire system largely depend on this junction. Such requirements signicantly limit the range of material dimensions (e.g. size, shape, and light absorbance.) that can be viewed as valuable for CCs, thereby hindering further development of CC research compared with EC research.
Despite these difficulties, a signicant amount of effort has been dedicated to the hybridization of water-splitting catalyst materials with semiconductors to realize high-performance photocatalytic and PEC systems. 74 Unlike in the synthesis of ECs, the stability of PCs and PEs must be considered when constructing CC/semiconductor composites. PCs based on (oxy) nitride and (oxy)chalcogenide decompose easier than pure oxides, while other PCs become inactive because of a change in the metal valence. 75 As introduced in Section 2, some methods for efficient EC formation require high-temperature annealing, and such a harsh condition cannot be applied to unstable semiconductors. In this section, we summarize representative strategies for loading CCs onto PCs and PEs (Fig. 5a, Tables 2  and 3).

Cocatalysts loading on powdery photocatalysts
In photocatalysis using powdery PCs, both reduction and oxidation reactions take place on an individual PC particle, requiring two suitable surfaces on an PC particle to drive each particular half-reaction. Generally, CCs for a specic halfreaction are rst loaded onto the PC surface leaving the other regions available to facilitate the other half-reaction, which is especially advantageous for examining individual half-reactions including the cases using sacricial reagents. For overall watersplitting PC, the loading of CCs for both HER and OER can further accelerate the total reaction to attain higher photocatalytic performance. 76 3.1.1 Hybridizing photocatalysts with cocatalysts. Mechanical mixing is the simplest way to modify a PC with an CC. Ultrasonication and (ball-)milling (a physical mixing method) are useful for preparing a well-mixed powdery composite. [77][78][79][80][81][82] Thermal treatment application is oen necessary to strengthen the interfacial contact aer mixing. [83][84][85][86] The main advantage of this technique is that it does not impose any restrictions on the type of PC or CC that can be used; thus,   (Fig. 5b). 80 Chemical methods are also applicable to hybridize PC with synthetic CC materials. Adding a CC to a PC synthetic system can grow typical PCs, such as CdS, on the CC surface under mild hydrothermal conditions (Fig. 5c). [87][88][89][90][91][92][93][94] However, this technique is not viable for bulk oxide-based PCs (e.g., SrTiO 3 ) which require high-temperature (>800 C) solid-state synthetic reactions that can lead to CC decomposition.
3.1.2 Cocatalyst growth and transformation on photocatalysts. A simple route to deposit a nanoparticulate metal oxide CC on a PC is through an impregnation method. [95][96][97][98][99] Typically, the CC precursor solution and the PC are successively mixed and dried, followed by an annealing step to completely convert the precursors into active CC components. This strategy also enables homogeneous connement of certain elements in small domains, thus allowing for facile manipulation of the metal composition. Although the majority of reported examples only describe a single type of metal, the deposition of multimetallic oxide CCs has also been achieved. For example, RhCrO x (a popular CC for photocatalytic water splitting) is commonly fabricated using this process. 100 Tsuji et al. also performed a simple impregnation to load ternary-metallic Brownmillerite Ca 2 FeCoO 5 NPs onto a TiO 2 PC (Fig. 5d). 101 However, this route requires high-temperature annealing to decompose the precursor salts, and this limits its application for thermally-unstable PCs. Moreover, ne-tuning of the CC size is difficult, and aggregation may adversely affect the catalytic performance of the CC. 102 A solvothermal method can also be applied to grow metal (hydr)oxide or sulphide CCs directly on a PC. [103][104][105][106][107] The straightforward addition of the PC into the EC synthesis solution can generate a CC/PC composite.
As mentioned in Section 2, a gas phase reaction can be employed to modify the composition of preformed oxide CCs. To promote the formation of high-quality chalcogenide, phosphide, nitride, and carbide CCs, which hardly form via impregnation or solvothermal methods, supplemental annealing treatments must be implemented. As a result, stable PCs with excellent thermal resistance (e.g., C 3 N 4 ) are mostly used for this approach. [108][109][110][111] 3.1.3 Photodeposition. Irradiating the PC with light is a clever way to create an internal potential through the excitation of carriers, which can reduce or oxidize the precursor to form nanoparticulate CCs. 112 Because photodeposition occurs at sites where photogenerated carriers preferentially migrate, CCs tend to nucleate on specic crystal planes of a PC, thus enabling location-controlled CC deposition. 14 Photodeposition can also be performed at RT, and therefore, unstable materials [e.g., (oxy)nitrides and (oxy)sulphides] can be used as the PC. 113,114 Researchers have demonstrated that metal or metal oxide CCs, such as Rh, Pt, MnO x , and CoO x can be loaded using this strategy. 115,116 Recently, S-or P-containing precursors have been used for photodepositing metal sulphides and phosphates, respectively. In order to shed some light on the dependence of photocatalytic performance on the type of CC-loading method, apparent quantum yield (AQY) values (at 420 nm) for wellstudied CdS/CC composites in the presence of hole scavengers are listed in Table 2. Although fair comparison between works under different conditions is difficult, rough tendency of AQY depending on loading method can be seen. CdS PCs loaded with CCs through physical mixing tend to show relatively lower AQY of around 10%. Meanwhile, CdS/CC composites assembled using hydrothermal and photodeposition methods can reach much higher AQY of up to $70%. These results suggest that the creation of an atomic-level PC/CC interface is paramount for a smooth carrier migration and reduced bulk recombination of photogenerated charge carriers in PCs.

Cocatalysts loading on photoelectrodes
In PEC systems, the photogenerated minority carriers diffuse to the semiconductor surface and the majority carriers are collected by the conductive substrate. For example, in photoanode containing n-type semiconductor, the electrons travel to the counter electrode, and the holes on photoanode surface and the electrons on the counter electrode surface are used to drive oxidation and reduction half-reactions, respectively. Therefore, unlike powdery PC systems, only one half-reaction (oxidation on n-type and reduction on p-type semiconductor) takes place on the PE surface, which requires the loading of appropriate CCs for each reaction. Accordingly, efficient collection of majority carriers (holes in p-type and electrons in n-type semiconductors) requires a small resistance at the semiconductor/electrode interface and within the grain boundaries of semiconductor layer. 123 Hence, CCs are typically loaded aer fabricating high-quality PEs, wherein a mild CC-loading condition is necessary not to damage those prebuilt PEs. Impregnation by drop casting the CC precursor and a subsequent mild annealing is oen applied when constructing CC/PE composites. [124][125][126] Hydrothermal method can also be employed by immersing PEs in a reaction vessel containing CC precursors (Fig. 5e). [127][128][129] These methods enable the formation of multi-metallic oxide CC layer on relatively unstable semiconductors like Ta 3 N 5 (Table 3). 130 An effective technique to homogeneously coat a CC layer on PEs is an electrodeposition. By applying a potential using a potentiostat, CCs can be deposited as reduced or oxidized species on the PE surface. Light-irradiation is sometimes applied during electrodeposition to assist adequate CC-loading under small applied potential. A representative example of an electrodeposited CC is a CoPi thin lm, which is formed by oxidizing a Co precursor with a phosphate electrolyte. 131 In addition, FeOOH, NiOOH, and CoOOH are popular OER CCs that can be anodically deposited in a similar way. 132-138 These catalysts, coated as homogeneous thin lms or NPs, have low absorption coefficients, which is a desirable feature for a CC. Considering that electrodeposition can be carried out at room temperature (RT), both robust BiVO 4 and fragile Ta 3 N 5 PCs can be used in this method (Fig. 5f).

Nanoparticle-adsorption approach
\Recent milestones in colloidal synthesis allow for the preparation of advanced NPs with controlled size and composition. 139 Colloidal NPs represent ideal CC candidates because they can offer numerous active sites and enable uninterrupted light absorption by the semiconductor. Selective binding of organic ligands on the NP surface grants the opportunity to precisely engineer their surface. 140 Specically, organic molecules, which have polar functional groups at both ends, can link the NP and semiconductor through hydrogen bonding between the functional groups and OH units on the PC surface (Fig. 7a). However, such ligands behave as an insulating, carrier-blocking layer at the CC/semiconductor interface, and therefore, they should be eliminated (Fig. 6). The standard approach to remove these protective ligands and activate the NP catalysts involves applying an annealing treatment. Previous reports have shown that the required annealing temperature for complete ligand removal changes depending on the type of ligand and the PC support. Our group successfully deposited monodisperse Rh NPs as HER CCs on a GaN:ZnO PC. [141][142][143][144][145] Specically, ultrasmall polyvinylpyrrolidone (PVP) coated-Rh NPs were rst adsorbed onto the PC using a linker molecule and then annealed in air at 400 C (Fig. 7a). Interestingly, the size of the NPs remained $1 nm without aggregation aer annealing. Negishi et al. also veried the crucial role of annealing to eliminate the ligands on Au clusters without changing their size. 146 Such ultrane NP deposition can only be realized through an NP-adsorption strategy. Depositing ultrasmall CCs can drastically enhance the photocatalytic performance of the composite because the NPs introduce a larger surface area and a greater number of active sites. 141 However, larger NPs could reduce the number of cocatalytic sites at the same degree of loading, and this might shield the incident light needed for carrier generation. In this context, ne-tuning the NPs can contribute to a deeper fundamental understanding of the CC effect (e.g., size and structure effects) on photocatalysis (Fig. 7b). 144 Although a thermal treatment is favourable for ligand removal, elevated temperatures and a harsh environment may cause irreversible damage to the PC, PE and the CC NPs. Alternatively, replacing long-chain ligands with small surface-coordinating molecules is an effective way to increase the interfacial contact between the NPs and semiconductor under mild conditions. For example, our group treated oleylamine (OAm)-capped Ni 3 S 4 NPs with S 2À at RT to replace the OAm and yield S 2À -stabilized Ni 3 S 4 NPs (Fig. 7c-e), 147 which were then directly deposited on a CdS/Cu(In,Ga)Se 2 PE through a layer-by-layer assembly process. Remarkably, the PE coated with S 2À -Ni 3 S 4 NPs exhibited an increase in cathodic photocurrent compared with the OAm-Ni 3 S 4 -NPs spin-coated PE, which highlighted the importance of establishing good contact between the CC and PE (Fig. 7f). Additionally, heating treatment to remove OAm from OAm-Ni 3 S 4 layers at 300 C resulted in deterioration of CdS/Cu(In,Ga)Se 2 PE. These results demonstrate that the chemical ligandremoval process can enable the application of NP CCs to unstable PEs.
By exploiting the transformation phenomenon of ECs during catalysis, we demonstrated that applying potential for OER to a NiP x @FeP y O z NP-loaded electrode transformed the NPs into highly OER active NiFeOOH lm while simultaneously removing their organic ligands (Fig. 8). 148 Notably, the colour of the NPs changed from black (phosphide) to colourless (hydroxide), with enhanced transparency for incident visible light. Such a characteristic transformation is highly advantageous for coupling NPs with PCs and PEs under mild conditions. Evidently, the NiP x @FeP y O z NP deposition substantially increased the photocurrent of the BiVO 4 PE without any posttreatment ( Fig. 8c-g), showing that homogeneous NP dispersion can be used as ready-to-use CC ink.

Application of ECs to CCs
activity could be identied. For example, what are the core principles relevant for unlocking high-performance CC/PC and CC/PE systems and bypassing rigorous material screening experiments? In this section, we introduce fundamental concepts that play a signicant role in designing advanced ECs as CCs for photocatalytic and PEC water splitting.

Energy level matching at the CC/semiconductor interface
When an EC comes in contact with a semiconductor, the band structure of the semiconductor is modulated. 149 In general, an electronic contact between a metallic CC and a semiconductor initiates an electron transfer between the semiconductor and the CC until their Fermi levels (E F ; the energy level with a 50% probability of electron occupation) equilibrate, which forms a space charge layer. In the space charge layer, the band edge energies of the semiconductor are continuously shied by the electric eld between the semiconductor and the metallic CC. As a result, the energy bands of the semiconductor bend toward the CC. The band bends upward (downward) when the E F of PC is higher (lower) than the E F of CC. Upward band bending creates an energetic upslope (Schottky barrier) for electron transfer but facilitates hole transfer from PC to CC, which encourages photo-oxidation reaction on CCs. 149,152 On the contrary, downward band bending promotes electron transfer (ohmic contact) but hinders hole transfer from PC to CC, which is desirable for photo-reduction reaction. Therefore, proper energy level matching to minimize the barrier height or create an ohmic contact for specic carriers is vital for a maximum utilization of photogenerated carriers. Although nanosized CCs have limited capabilities for donating or accepting electrons, loading sufficient amount of CC NPs can still cause band bending in the semiconductor, which has been veried experimentally using Kelvin probe force microscopy and surface photovoltage measurements. 150,151 For a CC with semiconductor-like properties, its band alignment with the PC and PE should also be considered to assess the charge transfer efficiency at their interface (Fig. 9b). 149  Chen et al. clearly demonstrated the benecial effect of an appropriate CC/PC junction on the photocatalytic activity by loading Pt or MoS 2 CCs on the surface of CdS PCs. 153 Although Pt has a higher HER activity than MoS 2 as an EC (Fig. 9c), the photocatalytic HER activity of the MoS 2 -loaded CdS was superior to that of the Pt-loaded CdS (Fig. 9d). This observation arose from the differences in their electron transfer rates. Specically, the deeper E F of Pt caused a greater upward band bending effect in CdS, which simultaneously disturbed the photoexcited electron transfer from CdS to Pt. This suggests that a lower activation potential for electron transfer from CdS makes MoS 2 more suitable as a CC than Pt (Fig. 9e). Therefore, the relative performances of ECs do not necessarily reect their relative applicability as CCs.  To evaluate the carrier transfer efficiency at CC/ semiconductor junctions, key parameters (e.g., work function of the metal CC, E F of the semiconductor) deserve deeper investigations. In the case of metals, standard work function data can be used to estimate the bending direction and Schottky barrier height at CC/PC junctions; in contrast, E F and bandgap values for complex heterogeneous CCs are not readily available, which makes evaluations of their band alignment difficult. Nevertheless, DFT and rst-principle calculations can provide critical insights on the fundamental electronic structure of EC materials (including band structure, density of states and E F ), enabling rapid prediction of suitable PC/CC combinations.
Even if the interfacial band alignments are unknown, the interfacial carrier dynamics can be measured through optical experiments, which can be used to assess the PC/CC interface. 154 Transient absorption spectroscopy (TAS) monitors the photogenerated carrier dynamics in a PC by recording the temporal absorption evolution aer a pulse excitation. 155 For example, TAS reveals (i) how electrons at the conduction band minimum behave in a PC/CC composite aer photoexcitation and (ii) the subsequent ultrafast intraband relaxation of hot electrons by probing the absorption in the infrared region derived from free electrons. 155,156 Although TAS has been conducted on clear quantum-dot dispersions, TAS measurements for solid-state bulk materials in reectance mode are still uncommon. Yamakata et al. showed how Pt and CoO x CCs affected the carrier dynamics when deposited on an LaTiO 2 N PC in the solid state. 157 Upon CoO x loading, visible (17 000 cm À1 ) and IR (2000 cm À1 ) probes traced a temporal decrease and increase in the population of holes and free electrons in LaTiO 2 N, respectively, aer a 500 nm laser pulse excitation (Fig. 10). These results indicated that CoO x can rapidly extract photogenerated holes in LaTiO 2 N to dramatically extend the lifetime of electrons. In contrast, Pt could not effectively extract photogenerated electrons because of the trapping mechanism occurring at mid-gap states in the LaTiO 2 N. These TAS ndings help pinpoint which specic process is the bottleneck in attaining excellent activity. Therefore, applying ultrafast spectroscopy to CC/PC(PE) systems could be very useful in providing guidelines for designing CC materials.

Assessment of CC activity
In addition to the interfacial carrier dynamics, the intrinsic electrocatalytic activity of the CC material is still a signicant descriptor of photocatalytic and PEC performance. Since the CCs serve as highly-active sites to drive redox reactions (like ECs), an understanding and evaluation of their intrinsic HER and OER activities are requisite for enhancing the catalytic performance. Notably, their catalytic activities can be measured with a CCloaded electrode using an electrochemical system. In EC studies, the catalytic activities are typically monitored under strong acidic or alkaline conditions, which are advantageous in showing their maximum activities. However, such extreme pH conditions cannot be applied in photocatalytic and PEC systems because they would likely cause chemical damage to the semiconductor. 158 Since the stability of the photoactive semiconductor is the top priority for water splitting, it is important to conduct assessments at a practically-applicable pH. Even if an EC exhibits excellent performance  under specic pH conditions, that same material might not be equally effective as CCs under different pH conditions. 159 Therefore, it is necessary to examine the CC performance in an environment similar to that optimized for the photocatalytic and PEC systems. For example, based on the synergistic improvement of OER activity by a bimetallic Co-Mn oxide EC, 160 our group investigated the effect of Co doping on the catalytic OER activity of Mn 3 O 4 NPs on the SrTiO 3 PC, and found that increasing the Co doping content improved the water-splitting activity (Fig. 11). 145 By measuring the OER activity of a Co x Mn 3Àx O 4 NP-loaded electrode in neutral pH [0.1 M phosphate buffer solution (PBS)], we found that the Co content monotonically enhanced the OER activity (Fig. 11d). The calculated band structure of the Co x Mn 3Àx O 4 NPs had similar valence band levels, regardless of the Co content, which indicated that the enhanced OER kinetics achieved by the Co-doping in Mn 3 O 4 CC primarily contributed to boosting the overall photocatalytic water-splitting activity by accelerating slow OER process (Fig. 11e). Electrochemical assessments under similar conditions as those used for the photocatalytic and PEC systems can offer more accurate predictions of an EC's suitability as a CC. Recently reported ECs that are active over universal pH represent promising candidates as CCs in various photocatalytic and PEC environments. [161][162][163]

Prevention of backward reaction
The oxygen reduction reaction (ORR) is usually not an issue to consider for electrochemical systems because they employ a divided cell conguration, wherein the HER and OER electrodes are placed in separated cells. However, the ORR must be considered in overall water splitting using powdered PCs because both half-reactions take place in a single reactor. Outstanding HER CCs, such as Pt and Rh metals, tend to actively catalyse the ORR, 22 which consumes the H 2 and O 2 evolved during the photocatalysis, thus reducing the H 2 uptake. 164 Strategies to effectively suppress this backward reaction and increase the quantity of evolved H 2 include the application of protective coatings on the CCs (e.g., amorphous oxides such as CrO x , 165 TaO x , 166 SiO x , 167 and ZrO x (ref. 168 and 169)). Our electrochemical measurements demonstrated that the ZrO x matrix could suppress the ORR of Rh 3+ active sites, which conrmed the O 2 -blocking ability of ZrO x (Fig. 12a) and prevented the loss of H 2 -uptake in photocatalytic overall water splitting by ZrRhO x / SrTiO 3 :Al (Fig. 12b). Evaluating the ORR activities of CC materials in addition to their HER and OER activities can provide insights relevant for improving the total photocatalytic activity.

Structural and chemical transformations of CCs during photocatalysis
Similar to ECs, nanosized CCs can undergo structural and chemical transformations during photocatalytic and PEC water-  splitting. In fact, our group's study revealed that the NiP x @-FeP y O z CCs on BiVO 4 PE transformed into NiFeOOH based on post-OER characterizations with X-ray photoelectron spectroscopy (XPS) and Fourier transform infrared spectroscopy (FTIR). 148 We also observed a compositional change in the novel Rh-Cl-Zr-O CCs deposited on SrTiO 3 :Al PCs under operating conditions. Aer the photocatalysis, XPS analyses detected the removal of Cl from the initial Rh-Cl-Zr-O CC to form RhZrO x during photocatalysis. 169 These results indicated the importance of post-catalysis characterization in determining the actual active species. Additionally, in situ/operando characterization techniques are powerful tools for monitoring the temporal evolution of CCs under the relevant working conditions, and they can help us to deeply understand the transformation mechanism of CCs. 170 Recently, synchrotron-based characterizations have provided valuable data with high temporal resolution, even for a small concentration of CCs. Altomare and co-workers conducted operando X-ray absorption spectroscopy (XAS) experiments for CuNi CCs loaded onto TiO 2 during photocatalysis (Fig. 13a). 171 At the early stage of the reaction, the preformed CuO x and NiO species were reduced to metallic Cu and Ni, thus forming CuNi alloy CCs (Fig. 13b-e). Subsequently, the H 2 evolution rate increased as the transformation progressed, which identied the metallic CuNi as the actual active sites. Operando XAS can also be used to conrm the stability of CCs in a sustained catalytic setting. Fundamentally, the integrated utilization of the methods discussed here is highly desirable to accelerate progress in designing highlyactive CC materials.

Conclusions and outlook
In this perspective, we have discussed recent developments in EC materials and their application as CCs in photocatalytic and PEC water-splitting systems. Although signicant efforts have already been directed toward the construction of high-performance CC/ PC and CC/PEC composites, exploration of CC materials still lags its EC counterpart. If methodologies to properly hybridize active EC materials with a variety of photoactive semiconductors (including unstable chalcogenides and nitrides) are developed, the accumulated knowledge regarding ECs will dramatically accelerate advances in photocatalytic and PEC systems. We believe that the colloidal approach using NPs of active EC materials with controllable sizes and compositions represents a promising solution to bridge the studies of ECs and CCs.
However, there are urgent issues that must be overcome in order to further enhance their photocatalytic performance. For particulate PC systems, spatial separation of HER and OER sites is highly desirable to efficiently promote both half-reactions. Facet-selective photodeposition of HER and OER CCs has been realized on PCs (e.g., SrTiO 3 :Al and BiVO 4 ) by utilizing the intrinsic built-in potential generated from the work function difference between the PC crystal planes. 76,115 To realize such site-specic CC deposition in other loading methods, exploiting the preferential adsorption of surfactants on different crystal planes of PCs could be useful. 172 The surfactants on CC precursor complex or NP surface have functional groups that could interact with coordinatively unsaturated atoms on the PC surface via dynamic adsorption and desorption. Again, the distinct atomic arrangement on the crystal facets of PCs might affect the binding affinities of incoming functional groups. For example, the (111) surface of Cu 2 O PCs contains more Cu dangling bonds that can be passivated by polar functional ligands. 173 In turn, CCs could be preferentially loaded on the exposed (100) surface of Cu 2 O through proper linker molecules. Likewise, PtCl 6 2À interacted strongly with OH groups on the anatase TiO 2 (101) surface, whereas Pt(NH 3 ) 4 2+ does not. 174 As such, this strategy can only be implemented on limited PCs and CCs, thus expanding the library of applicable materials is of primary importance to improve the poor STH efficiencies in PC systems. Another crucial factor that needs to be taken into account is the formation of an intimate interfacial contact between CCs and semiconductors. Although CCs are usually loaded on asprepared PC powder and PEs, the presence of pre-existing defective and/or insulative oxide layers on the semiconductor surface might cause charge recombination and/or carrier transfer blocking, and therefore should be removed before CC deposition. 175 Acid treatment is a simple and effective way to dissolve such surface oxide layers. 175 Ensuring long-term durability of CCs is also critical in maintaining excellent photocatalytic performance over extended operation periods. Since small amount of CCs is usually loaded on NPs or thin lms, CC degradation can potentially occur during photocatalysis through dissolution and physical removal. 125 The loss of carrier extraction capability of CCs could lead to self-corrosion of PCs due to uncompensated total charge. 19 Employing newly developed efficient and durable multi-metallic compound EC materials as CCs is a straightforward but promising way in accelerating the discovery of high-performance PC/CC combinations. Aside from improving the chemical and physical durability, CC regeneration by feeding CC precursor during operation could be a logical solution in prolonging the lifetime of practical photocatalytic systems. 176 As stated from Section 3, the need to control the CC morphology, internal electronic structure, and intrinsic surface catalytic activity, in addition to the CC's interfacial structure with the semiconductors, make advancements in this eld challenging. However, recent advances in characterization techniques such as ultrafast spectroscopy and synchrotronbased in situ analyses have gradually revealed the crucial role of such factors in determining the bottlenecks of photocatalytic systems. In addition to developing these technologies, acquiring fundamental properties from basic electrocatalytic analysis and theoretical DFT and ab initio calculation can potentially contribute to the understanding and construction of a sustainable and highly-efficient solar-driven photocatalytic water-splitting system.

Conflicts of interest
There are no conicts to declare.