Strategies of Anode Materials Design towards Improved Photoelectrochemical Water Splitting Efficiency

Abstract

This review presents the latest processes for designing anode materials to improve the efficiency of water photolysis. Based on different contributions towards the solar-to-hydrogen efficiency, we mainly review the strategies to enhance the light absorption, facilitate the charge separation, and enhance the surface charge injection. Although great achievements have been obtained, the challenges faced in the development of anode materials for solar energy to make water splitting remain significant. In this review, the major challenges to improve the conversion efficiency of photoelectrochemical water splitting reactions are presented. We hope that this review helps researchers in or coming to the field to better appreciate the state-of-the-art, and to make a better choice when they embark on new research in photocatalytic water splitting.

1. Introduction

Nowadays, it is becoming the world’s biggest challenge to meet the increasing energy demand by developing new sustainable substitution energy and reducing carbon dioxide emission. Overall, water splitting using sunlight is identified as a promising technology for production of renewable fuels and chemicals without causing environmental pollution.

In a photoelectrochemical cell for solar water-splitting, there are three main elements; namely, an anode, a cathode, and an electrolyte. At the anode, the oxygen evolution reaction (OER) will happen, while through the hydrogen evolution reaction (HER), hydrogen forms at the cathode, as shown in reactions R1 and R2 [1].

$$\mathrm{R}1: 2{\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)\to 4{\mathrm{e}}^{-}+4{\mathrm{H}}^{+}+{\mathrm{O}}_2\left(\mathrm{g}\right)$$

(1)

$$\mathrm{R}2: 4{\mathrm{H}}^{+}\left(\mathrm{aq}\right)+4{\mathrm{e}}^{-}\to 2{\mathrm{H}}_2$$

(2)

The two half-reactions constitute the overall water splitting process (2H₂O → 2H₂ + O₂), with the thermodynamic threshold energy E₀ = 1.23 eV. Although both reactions are important for the overall water splitting efficiency, the process is mainly hindered by the hypokinetic four-electron water oxidation. Based on electron transfer process, OER can be divided into four reaction paths with 1.23 eV in the surface site (denoted by the * symbol) of anode materials for thermodynamic threshold energy, as shown in reactions R3 to R6 [2].

$$\mathrm{R}3: {\mathrm{H}}_2\mathrm{O}+ \ast \leftrightarrow {\mathrm{OH}}^{\ast }+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$

(3)

$$\mathrm{R}4: {\mathrm{OH}}^{\ast }\leftrightarrow {\mathrm{O}}^{\ast }+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$

(4)

$$\mathrm{R}5:{\mathrm{O}}^{\ast }+{\mathrm{H}}_2\mathrm{O}\leftrightarrow {\mathrm{OOH}}^{\ast }+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$

(5)

$$\mathrm{R}6: {\mathrm{OOH}}^{\ast }\leftrightarrow {\mathrm{O}}_2+{\mathrm{H}}^{+}+{\mathrm{e}}^{-}$$

(6)

After the Honda–Fujishima effect was discovered in the early 1970s [3], intensive studies were carried out for photoelectrochemical (PEC) water splitting on semiconducting materials. The semiconducting anode materials will greatly determine the efficiency of absorbing light, suitable over-potential, and stability. It is challenging to develop efficient new anode materials or improve the ability of existing anode materials. Consequently, extensive efforts were implemented to improve the efficiency of anode materials; for example, TiO₂, α-Fe₂O₃, WO₃, and BiVO₄ [4]. However, there is still lack of cheap and stable photoanode materials with sufficient efficiency for water oxidation. Although significant progress has been made by developing new materials, nanostructures, and other novel concepts over the past decades, efficiencies are still below our expectations. Therefore, to assist future development, we wish to summarize the strategies to develop anode materials for improved efficiency in PEC water splitting in a concise manner. It is up to the readers to decide how different strategies could be applied in their specific applications, or to apply them synergistically in some situations.

The solar-to-hydrogen (STH) efficiency is an important metric for benchmarking and performance evaluation. The STH for a PEC system constructed by photoelectrodes can be expressed by [5]:

$$\mathrm{STH}=\frac{J_P\times 1.23}{I_0}\times 100\%$$

(7)

where $J_P$ is the photocurrent observed from the experiment, $I_0$ is the power intensity of the sunlight. The number 1.23 in the equation is the approximated value of the threshold energy for water splitting which requires the bandgap energy of any possible photoanode lager than 1.23 eV. For a photoelectrode, $J_P$ is determined by the photogeneration absorption efficiency (η_abs), the product of the injection efficiency (η_sep), and the injection efficiency (η_inj) of the photogenerated carrier to the reactant [6], which can be described by:

$$J_P=J_0\times {\mathsf{\eta }}_{abs}\times {\mathsf{\eta }}_{sep}\times {\mathsf{\eta }}_{inj}$$

(8)

where $J_0$ is the theoretical photocurrent when 100% of the photons in the solar spectrum with energies exceeding the bandgap are absorbed and converted. Theoretical STH and solar photocurrent of a photoelectrode under AM 1.5 G irradiation (100 mW·cm⁻²) are displayed in Figure 1 [6].

Thus, to achieve efficient PEC water splitting, several key criteria must be met, as shown in Figure 2: (1) wide range of light absorption of a material that can utilize more energy from sunlight; (2) effective generation of the charges and their transportation from the internal electrode to the photoelectric surface; (3) fast reaction of photogenerated carriers on the surface. Therefore, we will summarize recent progress of improving those efficiencies separately.

In this review, we summarize recent progress in the development of anode materials for photoelectrochemical water splitting, which are mainly based on our research work. This is carried out through a review of the strategies to enhance the light absorption, facilitate the charge separation, and enhance the surface charge injection. Finally, we share our critical views on the remaining challenges in this field.

2. Strategies to Enhance the Light Absorption

Generally, light absorption of a photoelectrode depends on its bandgap. For a known photoelectrode with a fixed bandgap, the film thickness and structure would affect its light absorption efficiency. Some reported strategies used to enhance the light absorption are summarized below.

2.1. Nanostructure Formation

Nanostructure is a good way to capture the scattered light if the light is not absorbed immediately after it reaches the surface. Nanostructure formation is a well-known strategy for silicon solar cells to reduce reflection loss and increase the energy conversion efficiency [7]. Similarly, nanostructure construction is proven effective in improving the light capture in PEC electrodes [8]. It has been reported that multiple light scattering in engineered tapered nanostructures enhances light absorption. (Figure 3) [9]. With a nanoporous Mo:BiVO₄ layer on the engineering conical nanophotonic structure combined with a solar cell, a STH efficiency of 6.2% was realized. Zhou et al. reported a hierarchical BiVO₄ photoelectrode, which consists of small nanoparticles and voids, could induce efficient light harvesting due to multiple light scattering [6]. Kim reported high-efficiency light absorption of GaN truncated nanocones. The truncated nanocones can trap the light within the nanostructure; thus, light loss, which is caused by surface reflection, is reduced. The TiO₂ solar cell prepared by using a 1D nanowire array can have a light conversion efficiency of up to 5.02%, which is much higher than that of a simple TiO₂ powder [10].

2.2. Band Engineering

To reduce the band gap of semiconductors, one commonly adopted and promising strategy is to add a foreign element (doping). Doping of foreign elements may include adding cation(s), anion(s), or adding cation(s) and anion(s) at same time.

Addition of cations usually induces a shallow valence band in the wide band gap through the hybrid valence orbital [11]. In metal oxide photocatalysts, incorporation of metal ions will introduce a donor level by hybridizing the O 2p valence orbitals with a metal ion having high-lying valence orbitals, such as Bi(III), Ag(I), Sn(II), Pb(II), or Cu(I). The crystal structure and the amount of metal incorporated greatly affect the contribution of the valence band. For example, in a MoO₃ electrode, Li⁺, Na⁺, and Mg²⁺ can greatly reduce the gap of the band [12,13]. Thus, modulating the band structure by cation doping is considered to be effective at the atomic level.

Anion doping has been employed to shift the absorption edge from the UV to the visible region; indeed, it is one of the strategies to enhance the light absorption. A lot of attention has been paid to nonmetal doping into TiO₂ to tune the light absorption, such as N doping as well as C, F, S, B doping [14]. For example, N and B doping greatly expands the light absorption to 700 nm [15]. Also, O incorporation in MoS₂ is able to improve the hybridization between Mo d-orbital and S p-orbital, resulting in a much smaller band gap [16]. Similarly, replacement of O by N in Ta₂O₅ induces a narrowed bandgap; for example, it was observed that the absorption was extended from 320 to 500 nm by N-doping (Figure 4) [17]. Simultaneously adding cations and anions can greatly change the band structure for enhancing the light absorption; e.g., the band gap of GaN is about 3.4 eV and the band gap of ZnO is 3.2 eV (Figure 5). However, the solid solution of GaN and ZnO narrowed the band gap to around 2.58 eV with visible light response [18]. This strategy provides us with a powerful method for adjusting the bandgap of photoelectrode materials through band engineering.

Another strategy for enlarging the light absorption is by intrinsic doping, such as introducing oxygen vacancy in oxides. The typical example is black TiO₂, in which lots of Ti³⁺ species are generated accompanied with the oxygen vacancy, being self-doping [19,20]. Doping and presence of lattice vacancies introduce midgap states which form the band tails and narrow the band gap. The light absorption of TiO₂ was enlarged to around 1000 nm from the 400 nm of white TiO₂. However, it is noted that the enlarged light absorption has little contribution to the enhanced performance, because the there is little visible light response for the black TiO₂ (Figure 6) [21]. The enhanced performance is possibly from the better conductivity.

2.3. Dual Absorber

Wang et al. reported extraordinary PEC performance by using dual BiVO₄ photoanodes made of two BiVO₄ which has the same transparency (Figure 7) [22]. The second electrode can use the transmitted light unabsorbed by the first electrode. It was found that the enhanced performance was mainly because of the higher light utilization rate in the airway scope of band edge (400–520 nm). A similar strategy was employed by Kim et al. with a different absorber; BiVO₄ in the front, and Fe₂O₃ behind BiVO₄ (Figure 8) [23]. This can further utilize the solar light, because Fe₂O₃ has a smaller bandgap than BiVO₄; thus, the transmitted light can be utilized by Fe₂O₃. This has increased the efficiency of hydrogen conversion to 7.7%. Wang et al. reported PEC water splitting using a hematite/Si nanowire dual-absorber system, which yielded a lower onset potential [24]. In their work, hematite was coated on the surface of Si nanowire, and the transmitted light through hematite was utilized by Si. The generated electrons of hematite combine with the holes generated on Si. Thus, holes generated on hematite are used for water oxidation, and at the same time, electrons generated on Si are used for reduction reaction. The match of the band edge positions of the two absorbers is important. A double absorber series battery was reported by Sivula et al., in which the light transmitted through the photoelectrode material is utilized by the dye-sensitized solar cells (DSSC) that provides the bias needed by the photoelectrode [25]. This concept is very useful in tandem PEC cells, which do not need additional bias and can increase the solar conversion efficiency. Besides, Chang and Ye also added Co₃O₄ and carbon quantum dot (CQDs) components on the surface of BiVO₄ to enlarger light absorption range [26,27]. For instance, CQDs decorated BiVO₄ photoanode cocatalyzed by a bi-layered Ni-FeOOH oxygen evolution catalyst have obtained an outstanding photocurrent density of 5.99 mA·cm⁻² at 1.23 V vs. reversible hydrogen electrode (RHE) with an enlarged light absorption range. Noble metal nanoparticles can also enhance the absorption light scattering/trapping, such as Au nanoparticles coated BiVO₄ and hematite [28,29]. However, the plasmonic effect also involves direct electron transfer mechanism (DET) and plasmon induced resonance electron transfer (PIRET) to enhance the performance except the light absorption. Currently, the full understanding of the involved physical mechanisms remains elusive, and requires more work to be clarifies [30].

3. Strategies to Improve the Charge Separation

3.1. Doping

Doping is proved to be an effective way to improve the conductivity of semiconducting compounds and thus increase the charge separation efficiency [5]. Zhao et al. reported that extrinsic Ti doping and oxygen vacancy generation (intrinsic doping) have greatly enhanced the PEC performance of hematite (Figure 9) [31]. In another example, incorporation of Mo⁶⁺ into a partial site of V⁵⁺ in BiVO₄ not only has changed the crystal symmetry of BiVO₄ and introduced some polarons, but also provided more free carriers (electrons) which facilitate the charge transport [32,33].

Exploration of new materials with suitable band position and narrow bandgap for water splitting is still a vital issue. Since doping is proven as an effective method to ameliorate the efficiency through improved conductivity, this method is usually combined with the design and synthesis of new PEC electrode materials. Recently, the performance of a new promising material, Zn₂FeO₄, for PEC water splitting was improved by Ti doping (Figure 10) [34]. Lumley et al. reported that the performance of Cu₁₁V₆O₂₆ was enhanced after doping by Mo or W [35]. Jo et al. found the 0.5% PO₄-doped BiVO₄ can lower the charge transfer resistance of BiVO₄ remarkably, about 30 times higher than that of BiVO₄ before doping [36].

One special doping is gradient doping, which means that the dopant concentration varies from surface to the bulk. In such a case, the band bending not only occurs on the surface, but also in the bulk, which provides an electric field throughout the bulk material. It is easier for the carriers to separate and transport to the surface in a gradient electrode. Yang et al. investigated how gradient doping affects a GaAs photocathode and found that both the escape probability and the diffusion length increased when compared with uniform doping [37]. Abdi et al. synthesized a doped W gradually decreased BiVO₄ from 1% to 0% between the interface of the FTO/semiconductor and semiconductor/electrolyte (Figure 11) [38]. Based on their experiment, the efficiency of charge separation increases to ~60%, while the efficiency of charge separation is ~38% for equally doped BiVO₄ (Figure 12).

3.2. Nanostructure to Shorten the Diffusion Length

Besides conductivity improvement, the minority carrier’s diffusivity is another key factor influencing the charge separation efficiency. Nanostructure provides a shorter diffusion distance for the minority carrier; therefore, it has been widely applied to improve the charge separation as shown in Figure 12 [39]. Kim et al. found nanoporous morphology of BiVO₄ can effectively suppress recombination, which yields an electron-hole separation efficiency as high as 90% at 1.23 V vs. RHE [40]. Zhao et al. also found enhancement of the performance by nanostructure compared with dense structure (Figure 13) and proved that the main contribution of the observed enhancement is the improved charge separation efficiency, which is caused by the easy diffusion of minority carriers to the interface [41]. Zhu et al. reported a nanorod-structured ZnFe₂O₄ with a new benchmark solar photocurrent of 1.0 mA·cm⁻² at 1.23 V and 1.7 mA·cm⁻² at 1.6 V vs. RHE (Figure 14) [42]. In general, if the average particle diameter of the nanopores is shorter than holes diffusion length, carrier recombination in bulk can be effectively inhibited.

Table 1. The hole diffusion length of different materials.

Materials	Hole Diffusion Length	Ref.
BiVO4	100 nm	[43]
α-Fe2O3	2–4 nm	[44]
TiO2	70 nm	[45]
WO3	150 nm	[46]

shows holes diffusion length of some typical materials for OER.

3.3. Heterojunction

A heterojunction comprises, but is not limited to, two materials, whereby electron–hole pairs are produced by one or more semiconductor photocatalysts and then transfer to the other material, which facilitates the space charge separation [47]. The two materials should meet some requirements. Taking a photoanode as an example, the valence band of A is more positive than B; at the same time, the conduction band of photocatalyst B should be more negative than A. This would facilitate the electron and hole transfer from one to the other (Figure 15). A lot of work has reported the heterojunction for enhanced PEC performance, such as WO₃/BiVO₄ [48,49], Cu₂O/g-C₃N₄ [50], Cu₂O/TiO₂, p-Cu₂O/n-TaON [51]. It is also found that heterojunction can be formed between different facets of the same crystal, e.g., BiVO₄ (010) and BiVO₄ (110) surfaces. The electrons and holes can accumulate on different surfaces of BiVO₄ (Figure 16) [52], which causes the space separation of electron–hole. Wang et al. found improved hydrogen production in a TiO₂ photocatalyst by forming {001}-{010} “quasi” heterojunctions [53]. Another case of heterojunctions is α-Fe₂O₃-TiO₂ heterojunction. Holes generated in TiO₂ will be transferred to α-Fe₂O₃ with electrons preferentially accumulating in TiO₂ at the same time. Under specific conditions, hybrid composites such as Fe₂TiO₅, Fe₃TiO₄, and FeTiO₃ form at the interface between hematite and titania that, despite some controversial results, exhibit remarkable performances as photoanode materials in PEC devices [54].

3.4. Internal Electric Field to Improve the Charge Separation

Recent studies indicate that the internal electric field in a photocatalyst can improve the charge separation [55]. The internal electric field provides a driving force for the separation of photoinduced charge carriers. Ferroelectric materials are a big family for having internal electric field due to their spontaneous electric polarization property among various photocatalytic materials. The spontaneous polarization in ferroelectric materials are due to the noncentrosymmetric property and the polarization occurs because the positive and negative charges have different centers of symmetry [56]. Perovskite oxides (ABO₃) are among the most extensively studied ferroelectrics, such as BaTiO₃ [57] and BiFeO₃ [58]. Rohrer et al. found that Ba_xSr_1−xTiO₃ solid solution by replacing Ba with Sr can enlarge the space-charge region and improve the activity [59]. Zhang et al. also found that incorporation of carbon into the Bi₃O₄Cl lattice can increase internal electric field to boost bulk charge separation [60]. Hao et al. found that double perovskite ferroelectric Bi₂FeMo_xNi_1−xO₆ thin films also possess strong ferroelectric self-polarization, which has played a crucial part in improving the photovoltaic effect [61]. Another case of charge separation is α-Fe₂O₃-TiO₂ heterojunction. Some research works found that the Fe₂O₃-TiO₂ heterojunction showed enhancement in the photocurrent [62,63]. However, somehow it is not consistent with our general understanding on the heterojunction, because hematite has a more positive conduction band potential than TiO₂, which prohibits the electron transfer from hematite to TiO₂. Further study shows that this kind of heterojunction is through iron titanates (e.g., Fe₂TiO₅, Fe₃TiO₄, and FeTiO₃) formed at the interface [54].

4. Strategies to Enhance the Surface Charge Injection

4.1. Catalyst Loading

In PEC water splitting, the catalyst loading play three vital roles for improving both the activity and reliability: (1) to reduce the overpotential or activation energy; (2) to assist charge separation at the semiconductor/electrolyte interface; (3) to suppress the photocorrosion and improve the durability of the working device [64]. Transition metal or transition metal oxides are usually used as catalyst loading to achieve a higher activity and decent reaction rates [65]. Recently, many (Fe, Co, Ni)-based catalyst loading have been widely investigated. For example, cobalt phosphate (CoPi) is the star cocatalyst to lower the overpotential and enhance the performance of water oxidation (Figure 17) [66]. Hydro-oxides are also one kind of effective catalyst loading, such as FeOOH, NiOOH, and CoOOH [67,68]. Kim et al. employed FeOOH/NiOOH dual-layer catalyst loading and achieved maximum ABPE of 1.75% due to reduction of the surface recombination [40]. Zhang et al. extended the catalyst loading to metal organic framework (MOF) materials, through applying in situ synthesized poly[-Co₂(benzimidazole)₄] nanoparticles on the surface of BiVO₄ to boost surface reaction kinetics (Figure 18) [69]. In this respect, the bilayered Ni-FeOOH oxygen evolution catalyst layer has been reported as the best catalysts for BiVO₄ to date [70].

4.2. Surface Treatment and Surface Passivation

Point defects are usually found on the surface of photoelectrode. In some cases, defects play a positive role in the surface catalysis. However, in some other cases, defects serve as recombination centers. Thus, it is important to suppress the recombination on these surface centers. Recently, it was found that the surface treatment can improve the PEC performance. The mechanism for surface treatment and surface passivation is illustrated in Figure 19 [71].

Surface defect states will lead to high charge recombination rate and water oxidation with low efficiency, caused by the photogenerated holes. Surface passivation and treatment can passivate defect states, forcefully suppressing surface recombination for improving water oxidation efficiency. For example, Luo et al. found that electrochemical cyclic voltammetry surface pretreatment on Mo-doped BiVO₄ can greatly enhances the photocurrent of Mo-doped BiVO₄ [72]. This is mainly because the pretreatment that can remove the surface recombination center MoO_x which is segregated during the preparation of the photoelectrode. After the preprocessing, MoO_x on the surface is dissolved into the electrolyte (Figure 20).

Further, Wang et al. found that the electrochemical treatment is not only suitable for doped BiVO₄, but also applicable to pristine BiVO₄ and other materials, such as TiO₂, Fe₂O₃ [73]. Besides the surface treatment, a thin passivation layer is also proved to be effective to enhance the surface catalysis. Originally, passivation layers were applied to semiconductor photoelectrodes to improve their photochemical or chemical stability. It was reported that even an extremely thin layer on the photoelectrodes can change the surface properties [74]. Hisatomi investigated the oxide surface layer on the onset potential of hematite photoanode, such as Al₂O₃, Ga₂O₃, and In₂O₃, which can greatly weaken the original potential for oxidation of water [75]. Formal et al. covered an extremely thin layer (0.1 to 2 nm) of alumina on nanostructured hematite photoanodes. The photocurrent was shifted cathodically by as much as 100 mV. Further investigation found that the cathodic shift is owing to the passivation of surface trapping states caused by the Al₂O₃ overlayer [76]. Besides, IrO₂ and CoO_x thin layers on TaON photoanodes were found to improve the durability by reducing the self-oxidation of the electrode [77,78]. In general, a suitable structure of the passivation layer for photoelectrodes can either improve the reaction kinetics, or increase the chemical or physical protection against corrosion, or both.

4.3. Active Sites

The chemical reactions for water splitting occur on the electrode surface through its active sites. The active sites, therefore, affect the performance of the surface catalysis. The low catalytic activities are mainly attributed to the relatively lower activity and smaller number of reactive sites. The activity of active sites can be reflected by the overpotential at the anode. Experimentally, it might be extremely challenging to identify the atomic surface active sites. Therefore, calculations using density functional theory (DFT) become important to reveal the overpotential at different sites, as well as to understand the atomic/molecular mechanism. There are some methods to calculate the overpotential based on different reaction process [79,80]. The well acceptable reaction process is the four intermediate steps reaction path, as shown in R3 to R6.

The overpotential tends to zero if $∆G_{R3}^{\mathsf{\theta }}$ = $∆G_{R4}^{\mathsf{\theta }} = ∆G_{R5}^{\mathsf{\theta }}$ = $∆G_{R6}^{\mathsf{\theta }}$ = 1.23 eV and this would define the ideal catalyst, as shown in Figure 21 [81].

Until now, the atomic level interaction between catalytic activity and active sites is still an outstanding problem, owing to the large difference between idealized models and real catalysts. Thus, it is necessary to identify the pivotal factors that influence the activity of reactive sites, in order to promote the catalytic properties. It is generally considered that the low coordinated steps, edges, terraces, kinks, and/or corner atoms are often the favorable catalytic reaction sites [82,83]. For example, Sun et al. suggested that the low-coordinated Co³⁺ atoms of porous atomically-thick Co₃O₄ sheets serve as the catalytically active sites [84]. Currently, there are no quantitative models to describe the relationship between the coordinated number and activity. The main reason for this difficulty is that there might be other key parameters that have not been separately verified. For example, some results show that the surface state of the atom and position of neighbor atom will greatly affect the activity of the reactive sites. Hu et al. found that oxygen vacancy in BiVO₄ will induce more active sites on a BiVO₄ surface, which makes the vacancy sites active for water oxidation (Figure 22) [85]. Similarly, a Ti 3d defect state in the band gap of titania also plays an important role in the surface catalysis of defect contained TiO₂ [86]. Zhao et al. also revealed that W doping and W–Ti codoping into BiVO₄ have led to more active sites on the surface and reduced charge transfer resistance (Figure 23) [87,88]. Theoretical calculation revealed that Ti sites become active after the codoping.

5. Conclusions and Outlook

In this review, we have highlighted the latest studies directed at developing metal oxide-based photoanodes, including BiVO₄, α-Fe₂O₃, TiO₂, and WO₃ nanostructures. We have summarized the strategies to enhance the light absorption, charge separation, and surface charge injection of this class of n-type semiconductor oxide materials. Although great achievements have been obtained, significant challenges remain before commercially viable solar-driven water splitting can be realized. Among them, one challenge is to balance the contributions from factors affecting light absorption, charge separation, and charge injection. Sometimes, their contributions contradict each other; e.g., a thicker layer increases absorption efficiency, but reduces the charge separation efficiency due to a longer distance to be travelled by the charge carrier. Another challenge lies with the lack of thorough understanding of the effect of the surface-state, due to the complex roles it plays in the water splitting process. Surface states may affect the amount of stored charges, act as surface recombination centers, and pin the Fermi-level. Most of the existing models can only explain one of these potential effects and there is no commonly used model to describe the full impact of surface state. A big challenge is to combine all the advantages and improve the overall efficiency for water splitting. Usually, it is inevitable to introduce some other problems when improving one factor. For example, nanostructuring is mostly employed to enhance the charge separation and light absorption; however, more surface states will be formed which sometimes are unfavorable to the interfacial charge injection. Thus, catalyst loading is required to passivate the surface states. Then, the challenge becomes finding a suitable catalyst loading to passivate and to control its thickness through a proper deposition technology.

Recently, researchers are focusing on the formation of ultrathin layers of cocatalyst on the surface, which will provide very efficient charge injection while avoiding the recombination in the surface layer and at the interface. However, it is very difficult to precisely control the formation of a uniform cover layer of the cocatalyst on the surface with reasonable cost of production. New low-cost processing technology and materials need to be developed. Another related issue is the stability and durability from the point of view of practical application. Newly developed materials should be coated with good adhesion while having good chemical and photochemical stability. More work is needed to build a more comprehensive model to guide the design and optimization of the electrode materials for solar water splitting.