Recent Advances on Small Band Gap Semiconductor Materials (≤2.1 eV) for Solar Water Splitting

Abstract

The conversion of solar energy into renewable H₂ fuel via photoelectrochemical and photocatalytic water splitting approaches has attracted considerable attention due to its potential to solve significant energy and environmental issues. To achieve reasonable energy conversion efficiency of 10%, which is amenable to the economic feasibility of this technology, semiconductor materials, which are the main carrier for solar H₂ production, must fulfill several important criteria. One of the most important criteria is the band gap of the semiconductor material, which should be low enough to allow the efficient absorption of incident light. This is because the band gap of semiconductor material fundamentally determines the upper limit of the theoretical energy conversion efficiency of the solar conversion system. In this work, recent advances in utilizing semiconductor material with a band gap smaller than 2.1 eV instead of the simply-defined visible-light-responsive semiconductor materials toward solar H₂ production from water splitting was reviewed.

1. Introduction

1.1. Solar Hydrogen Production from Water Splitting with Semiconductor Materials

The increasing concern over the depletion of traditional fossil fuels and the serious environmental problems has triggered interest in the utilization of solar energy. The main features afforded by solar energy are its abundance and cleanness. It is estimated that the magnitude of the available solar energy striking the earth’s surface in 1 h (4.3 × 10²⁰ J) is even higher than that consumed on the planet in 1 year (4.1 × 10²⁰ J) [1]. Therefore, harvesting energy in a sustainable, environment-benign, and cost-effective manner from sunlight offers an ideal approach toward fulfilling the human need for renewable energy [2,3]. However, the large disparity between the practical utilization of solar energy and its huge theoretical capacity creates a compelling need to develop credible ways to convert solar energy.

Up to now, several strategies have been proposed and developed for the conversion of solar energy into different energy forms, such as thermal energy, chemical fuels, and electricity. Among all the strategies under investigation, solar water splitting to produce hydrogen fuel on semiconductor materials has been an important research topic in the past decades [4]. Water splitting to produce H₂ and O₂ is an uphill reaction and involves a large positive Gibbs free energy change (ΔG⁰) of 237 kJ/mol [5]. Therefore, it defines the unfavorable nature of the reaction under standard conditions. However, water splitting can be realized by the well-known Honda–Fujishima effect [6], which involves illuminating an n-type TiO₂ photoanode coupled with a Pt cathode at an applied bias much lower than the voltage needed for the electrolysis of water. This finding, reported in 1972, has inspired the subsequent intense research on the conversion of solar energy to H₂ fuel via water splitting.

Two approaches, photocatalytic water splitting on semiconductor photocatalyst powder and photoelectrochemical (PEC) water splitting on semiconductor film electrodes, are typically used to produce H₂, both of which can be regarded as artificial photosynthesis [7]. Figure 1 shows the general steps required for achieving water splitting via photocatalytic reactions on photocatalyst powder suspended in an aqueous solution. In the first step, when the semiconductor powder harvests incident photons with enough energy, the electrons in the valence band are excited to the conduction band, while the holes stay in the valence band of the photocatalyst. The energy difference between the valence band and the conduction band is defined as the band gap (E_g), and the energy of the incident photon must surpass the band gap of the semiconductors to allow for light absorption. In the second step, the photogenerated charges will move to the surface of the semiconductor photocatalysts before recombination. In the third step, the photogenerated electrons will reduce protons to generate H₂, and the holes will oxidize H₂O to generate O₂ on the different active sites of the photocatalyst. As the charges photogenerated on wide band gap semiconductors have high energy for the redox reactions, they are more likely to initiate photocatalytic reactions. The three steps described above merely represent the necessary steps involved in a complete photocatalytic water-splitting reaction. There are several underlying criteria for the material that has the potential to achieve water splitting. One of the most important criteria is the valence and conduction band positions of semiconductors relative to those of the water reduction and oxidation potentials. As shown in Figure 1, the bottom level of the conduction band must be more negative than the redox potential of H⁺/H₂ (0 V vs. normal hydrogen electrode (NHE)), and the top level of the valence band must be more positive than the redox potential of O₂/H₂O (1.23 V vs. NHE). Therefore, finding a material with conduction and valence band positions that can match the redox potential of water splitting is crucial, which makes this research topic extremely challenging.

For PEC water splitting, the general steps include creating photogenerated charges, migration of charges, and subsequent charge-induced water splitting reactions, which are quite similar to those of photocatalytic water splitting (Figure 2). However, there is one distinct difference between the two approaches. For PEC water splitting, an external potential could be applied between the working electrode made of semiconductor materials and the counter electrodes, typically made of active materials capable of catalyzing water oxidation or reduction. This applied potential will provide additional energy for the photogenerated charges and straddle the limitations that existed in photocatalytic water splitting due to the inappropriate band positions of semiconductors relative to water redox potentials. Moreover, the applied potential will provide the driving force for the directional migration of photogenerated charges, consequently decreasing their recombination and enhancing the overall efficiency of the PEC system. For example, several visible-light-responsive oxide photocatalysts, such as n-type Fe₂O₃ or WO₃ semiconductors, cannot reduce protons to produce H₂ due to their highly positive conduction band edge positions relative to the water reduction potential. However, under applied potential, photoelectrodes made of these materials can efficiently split water under visible light. Therefore, materials that are inappropriate for photocatalytic water splitting may still be qualified for PEC water splitting with the aid of applied bias. For n-type semiconductor materials, water is oxidized on the semiconductor surface and reduced on the counter electrode. Contrary to n-type semiconductor materials, water is reduced on the surface of p-type semiconductor materials and oxidized on the counter electrode.

1.2. Band-Gap: The Fundamental Factor Determining the Theoretical Solar-to-H₂ (STH) Efficiency

Semiconductor materials are the main carriers for the PEC and photocatalytic reactions. Considering the general steps involved in the PEC and photocatalytic water splitting, it is evident that the semiconductors must first absorb incident photons to initiate the subsequent reactions as described by the first law of photochemistry (the Grotthuss–Draper law). The more photons semiconductors could capture, the higher efficiency the semiconductor-based reaction system could obtain. Therefore, for any semiconductor material to be potentially feasible for solar H₂ generation, it must be an excellent light harvester.

The magnitude of the band gap of a semiconductor determines the photons that could be utilized. Small band gap semiconductors harvest more photons, while wide band gap semiconductors harvest less. If we presume that all the photons with energy larger than the semiconductor band gap can be absorbed by the semiconductor and transformed to the H₂, the maximal theoretical STH efficiency of a material can be estimated from its band gap [8]. The theoretical maximal photocurrent (J_max) and STH efficiency of a semiconductor were fundamentally determined by its band gap. It was found that the theoretical STH obtained on anatase TiO₂ (3.2 eV), monoclinic WO₃ (2.6 eV), and α-Fe₂O₃ (2.1 eV) were 1%, 6%, and 15%, respectively [8]. If a semiconductor can absorb light of up to 800 nm (corresponds to 1.55 eV), it could achieve a theoretical STH efficiency of 32% [8]. Therefore, semiconductors with smaller band gaps will be favorable in the actual PEC water splitting, considering its possibly higher conversion efficiency of solar energy.

Service demonstrated that the energy conversion efficiency must be ≥10% if the economic viability of the PEC hydrogen production can be met (the energy conversion efficiency is defined as the ratio of the amount of the energy output to the energy input) [9]. This corresponds to a band gap of ca. 2.4 eV and a threshold wavelength of ca. 520 nm [8]. Therefore, even wide-band semiconductor materials are more likely to convert absorbed photons into H₂ fuel due to the high kinetic energy of the photogenerated charges. Their application as photoanode materials is impractical due to the limited theoretical energy conversion efficiency that is incompatible with what is demanded by the practical applications. It is worth noting that the band gap of semiconductor materials should also satisfy the thermodynamic energy requirement for water splitting [10]. From the standard Gibbs free energy change of 237 kJ/mol for water splitting, the theoretical band gap of the semiconductor must be at least 1.23 eV for unassisted photocatalytic water splitting [11,12]. Moreover, considering the kinetic overpotentials for the oxygen and hydrogen evolution reactions on the semiconductor surface, the actual band gap required should be more than 1.5 eV for unassisted photocatalytic water splitting [13]. Therefore, materials with a band gap in the range of 1.5 to 2.4 eV will be potentially viable candidates for photocatalytic water splitting [14]. However, in PEC water splitting, due to the assistance of applied potential, semiconductor materials with a band gap of less than 1.23 eV are still applicable [15,16].

In terms of photocatalytic water splitting efficiency, Domen et al. suggested a tentative goal as a starting point for the research on hydrogen production using a particulate photocatalyst [17]. According to their estimations, material that can split pure water with a quantum efficiency of 30% at 600 nm is required, corresponding to a theoretical STH conversion efficiency of ca. 5% as shown in Figure 3. Even though this proposed energy conversion efficiency in photocatalytic water splitting is only half of that proposed for PEC water splitting, it is evident that the development of semiconductors with a band gap as small as 2 eV (corresponding to 600 nm) is also indispensable. Therefore, compared with PEC water splitting, the photocatalytic water splitting is a more challenging topic in terms of the energy conversion efficiency.

The aim of PEC and photocatalytic materials development is to identify a semiconductor system that has the ability to meet most, if not all, of the requirements for cost-effective solar hydrogen generation [8]. Various semiconductor nanomaterials such as oxide, sulfide and oxysulfide, nitride and oxynitride, phosphide, selenide, silicon, and silicide have been developed for solar water splitting applications. Special attention has been paid to the development of visible-light-responsive semiconductor materials (λ > 420 nm) due to their capability to utilize the larger fraction of solar electromagnetic irradiation. However, strictly speaking, from the above analysis of the maximum theoretical STH efficiency as a function of the band gap and the efficiency required in terms of economic feasibility, the development of semiconductor materials with a band gap of lower than 2.4 eV instead of simply defined visible-light-responsive materials will be highly desirable. Moreover, we should also keep in mind that when we calculate the energy conversion efficiency as a function of the band gap, we ideally assume that all the photons with energy larger than the band gap of materials can be harvested by the semiconductors to produce photogenerated charges and the photogenerated charges can be collected and converted to the H₂ fuel. Due to the unavoidable recombination of photogenerated charges and the energy loss during the conversion processes, this estimation is only ideal. If we tentatively assume 1/3 of the input energy is lost in this whole process, a theoretical energy conversion efficiency of 15% is therefore needed to achieve an actual energy conversion efficiency of 10%, which is amenable to the economic feasibility of this process. In this case, semiconductor materials with a band gap of less than 2.1 eV will be more viable candidates.

1.3. Scope of this Review Article

In recent years, owing to the rising interest in solar H₂ production from PEC and photocatalytic water splitting, significant progress has been made in this field, and several excellent reviews have been published [7,10,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36]. However, there are still no reviews focusing on semiconductor materials with band gaps smaller than 2.1 eV. Considering the high theoretical energy conversion efficiency and correspondingly higher economic feasibility that could be obtained on small band gap semiconductor materials, it is highly desirable to give a summary of these materials. In this review article, we will provide an overview of the recent advances in the development of emerging small band gap (E_g ≤ 2.1 eV) semiconductor materials or the modification of existing ones for solar H₂ production. Only the representative results will be discussed, and therefore, this review is not meant to be wholly thorough. Readers interested in the profile of photocatalytic and PEC water splitting could refer to other review articles and the references herein [10,11,12,13,14,15,16].

2. Small Band Gap (E_g ≤ 2.1 eV) Semiconductor Materials for Solar H₂ Production

Generally speaking, the conduction band of a semiconductor material is composed of the metal element’s orbitals, while the valence band is composed of the non-metal element’s orbitals, except for a few cases that the metal element’s orbitals can also form the valence bands. Therefore, the band gap of semiconductor material is fundamentally determined by the elements that constitute the materials. Kudo et al. gave an excellent summary of these elements in terms of their functionalities [37].

At present, oxide-based semiconductor materials account for the main part of the heterogeneous photocatalysts or photoelectrode materials under investigation. For oxide-based semiconductor materials composed of fixed metal cations, in most cases, the band gap of the materials will be decreased by replacing the oxygen elements with other non-metal elements such as nitrogen, sulfur, phosphorus, selenium, and iodine, among others. This is because the valence band positions of the resulting materials will be elevated relative to that of the oxides, where the valence band is usually composed of O 2p orbitals. At the same time, the conduction band position of the resulting materials composed of metal orbitals will remain relatively stable. Therefore, the valence band modification of semiconductor materials with various non-metal elements will bring a pronounced effect on the amplitude of the band gaps and will lead to the formation of semiconductor materials with much-reduced band gaps. In the following section, we will summarize small band gap semiconductor materials according to the main non-metal elements in the semiconductor materials.

2.1. Oxide-Based Semiconductor Materials

2.1.1. α-Fe₂O₃ (Hematite)

Hematite has a band gap of ca. 2.1 eV and theoretical STH efficiency of 15%. The conduction band position of hematite is lower than the water reduction potential; therefore, hematite cannot split pure water photocatalytically. However, hematite is one of the most promising semiconductor materials for PEC water splitting due to its attractive advantages, such as wide light absorption, low cost, excellent chemical durability, and environmental compatibility. Fe is among the most available elements that can be relied upon for the large-scale fabrication of devices [38]. However, hematite has some inherent disadvantages, such as a short minority carrier lifetime (ca. 10 ps), unsuitable conduction band energy level for H₂ production, poor conductivity, short hole diffusion distance (ca. 2 to 4 nm), and poor surface oxygen evolution reaction kinetics, which leads to much lower STH efficiency compared to the theoretical value [39,40]. Major efforts have been devoted to addressing these problems. Several methods, including elemental doping, nanostructuring, and loading of oxygen evolution catalysts, have been used to tune the electronic and structural characteristics of hematite film, and significant progress has been made [41,42,43,44,45,46,47,48,49,50,51]. As the synthesis, characterization, charge carrier dynamics, and photoelectrochemical properties of hematite were comprehensively reviewed in papers [52,53,54,55,56,57,58], we will only introduce some representative results for highly efficient hematite photoelectrode.

Gratzel and Sivula et al. extensively investigated the preparation of hematite photoanodes for PEC water splitting. They reported a highly efficient nanostructured Si-doped hematite electrode. The hematite film was deposited on a conductive FTO substrate by atmospheric pressure chemical vapor deposition from Fe (CO)₅ and tetraethoxysilane at 688 K [59,60]. The as-deposited film consisted of a perpendicularly oriented dendritic structure of 500 nm thickness with a feature size of only 10–20 nm at the surface. Under AM 1.5G illumination (100 mW cm⁻²), the hematite film delivered a photocurrent density of 2.3 mA cm⁻² in 1 M NaOH electrolyte, much higher than without Si doping (10 μA cm⁻²). Moreover, the treatment of the hematite film with cobalt further increased the current density to 2.7 mA cm⁻² due to enhanced water oxidation dynamics by the cobalt cocatalyst. The high efficiency achieved in this hematite film is supposed to be due to the beneficial advantages afforded by the dendritic nanostructure, which permits efficient light absorption while reducing the length that photogenerated holes have to diffuse to reach the hematite/electrolyte interface. In the following work, by optimizing the preparation conditions, hematite film with similar nanostructures while higher thickness was obtained [59]. IrO₂ nanoparticles were then loaded on the hematite surface by electrophoresis method. The loading of IrO₂ cocatalyst was found to lead to a dramatic shift of the onset potential from +1.0 V to 0.8 V vs. reversible hydrogen electrode (RHE). Moreover, a high photocurrent of over 3 mA cm⁻² was achieved with this hematite electrode at 1.23 V vs. RHE under AM 1.5G 100 mW cm⁻² simulated sunlight illuminations due to the improved nanostructure and catalytic properties of the hematite film. The important role of cocatalysts in improving the PEC performance of hematite has also been investigated by other groups from both experimental and fundamental aspects [45,61,62]. Gratzel and Sivula et al. also developed several strategies for the low-cost preparation of hematite film, the confinement of the feature size of hematite film nanoparticles, and surface modification with oxide layers to improve the onset potential for water oxidation [44,63,64].

Li et al. developed a series of low-cost and facile approaches for fabricating highly efficient hematite photoanodes such as Sn-doped, Ti-doped, and oxygen-deficient hematite [41,65,66]. Hydrothermal and subsequent high-temperature sintering in air was used to prepare Sn-doped hematite nanowires or nanocorals films on FTO substrate [41]. Photocurrents of 1.24 and 1.86 mA cm⁻² were achieved at 1.23 V vs. RHE on the nanowire film and nanocoral films, respectively. Compared with undoped hematite nanostructures, the Sn-doped ones showed remarkable photocurrent densities due to the enhanced conductivity and surface area by Sn doping. To prepare a Ti-doped hematite photoelectrode, a facile deposition-annealing (DA) process was used, during which iron (III) chloride and titanium butoxide were used as the Fe and Ti precursors, respectively [65]. An annealing at a temperature as low as 823 K was used. Under optimized conditions, the Ti-doped hematite film achieved a photocurrent density of 2.8 mA cm⁻² at 1.23 V vs. RHE (Figure 4). Impedance measurements exhibit that the Ti-dopant acts as an electron donor and improves the donor density by two orders of magnitude. The increased photocurrent is ascribed to the increased donor density and decreased electron-hole recombination due to the Ti doping. Moreover, Li’s group prepared a highly efficient oxygen-deficient hematite photoelectrode by thermal treatment of akaganeite nanowire arrays in an oxygen-deficient environment at a quite low activation temperature of 823 K [66]. The as-prepared hematite electrode delivered a photocurrent density of 3.37 mA cm⁻² at 1.50 V vs. RHE. The drastic enhancement of the PEC performance of the hematite film is due to the enhanced donor density originating from the formation of an oxygen vacancy (Fe²⁺). This work showed a simple and effective method for the preparation of highly photoactive undoped hematite at a relatively low activation temperature for PEC water splitting.

2.1.2. Cu₂O

Cuprous oxide, Cu₂O, has a direct band gap of 2.0 eV and a corresponding theoretical STH conversion efficiency of 18% [67]. The conduction band of Cu₂O is located at 0.7 V, negative of the water reduction potential, and the valence band is just positive of the water oxidation potential [68,69]. Therefore, from the thermodynamic point, Cu₂O can achieve the hydrogen evolution half reaction while it is ineffective for the water oxidation reaction. Furthermore, considering that the redox potentials for the reduction and oxidation of Cu₂O lie within its band gap, Cu₂O is not stable when illuminating in aqueous solutions, which severely limits its application as a photocathode material for PEC water splitting. Therefore, finding a proper coating layer that can protect a Cu₂O photoelectrode against photocathodic decomposition while allowing an efficient electron tunnel is important for this material [70,71,72,73,74].

Paracchino et al. developed a highly efficient and relatively stable Cu₂O photocathode for PEC water splitting [75,76]. Cu₂O film was first prepared onto FTO substrates with an electrodeposition method. Ultrathin protective layers of Al-doped ZnO and TiO₂ were then deposited onto the Cu₂O photocathode layer by layer with an atomic layer deposition technique. Finally, Pt nanoparticles were electrodeposited on the top layer of the photoelectrode to enhance the kinetics of the water reduction reaction. It was found that the cathodic photocurrent density observed on the bare Cu₂O electrode reached a plateau of 2.4 mA cm⁻² and decreased dramatically with short-term irradiation due to the reductive decomposition of Cu₂O. However, the photocurrent density of Cu₂O with a protective layer reached 7.8 mA cm⁻² at 0 V vs. NHE. Moreover, the stability of the Cu₂O electrode was greatly improved. The authors pointed out that the protective layer must meet several criteria. First, the protective layer must have a staggered type-II offset to allow the flow of photogenerated charges to the desirable direction. Second, the conduction band of the protective layer should be higher than the water reduction potential, and the protective layer should be stable. Third, the surface of the protective layer in contact with the electrolyte should be favorable for the water reduction reactions. By tuning the deposition conditions of the protective layer, a highly efficient Cu₂O photocathode was obtained in this study. Even the Cu₂O photocathode degraded with prolonged irradiation; the authors expected that its stability could be further improved by optimizing the building block material for the protective layer.

To address the photocorrosion problem on the Cu₂O photocathode, Zhang and Dua et al. developed a facile strategy [77]. In their approach, a glucose solution was first coated onto the Cu₂O nanowire array electrode. A subsequent carbonization treatment will decompose glucose to form a protective carbon coating onto the Cu₂O electrode. The thickness of the carbon layer was tuned by changing the concentration of the glucose solution. After depositing the carbon layer on the Cu₂O, the as-obtained electrode exhibited remarkably improved photostability as well as considerably enhanced photocurrent density. Under optimum conditions, carbon-layer protected Cu₂O nanowire arrays delivered a photocurrent density of 3.95 mA cm⁻² and an energy conversion efficiency of 0.56% under AM 1.5 G illumination (100 mW cm⁻²). Compared with the atomic layer deposition strategy mentioned above, this strategy is much more facile and inexpensive and will provide a general approach that can mitigate the stability problem of many nonstable photoelectrodes.

To improve the stability of the Cu₂O electrode, Lin deposited NiO_x cocatalyst film onto the Cu₂O nanowire electrode [78]. In this work, a NiO_x precursor solution was first spin-coated onto a Cu₂O electrode. Subsequent annealing of the Cu₂O electrode in air will decompose the NiO_x precursor to NiO_x cocatalyst. After loading the NiO_x cocatalyst, a photocurrent density of −0.56 mA cm⁻² at 0.1 V vs. NHE was obtained under light irradiation (26 mW cm⁻²). Moreover, the NiO_x-modified Cu₂O electrode showed greatly improved stability compared with the bare ones. The authors suppose the photogenerated electrons will transfer from the Cu₂O to NiO_x cocatalyst and reduce NiO_x to metallic Ni. Metallic Ni will act as the H₂ evolution cocatalyst and is re-oxidized to Ni²⁺. Therefore, the deposition of NiO_x on Cu₂O can significantly improve the water reduction reaction dynamics and simultaneously improve the stability of Cu₂O.

2.1.3. CuNbO₃ and CuNb₃O₈

Maggard’s group developed a series of novel Cu⁺-based semiconductor materials such as CuNbO₃, CuNb₃O₈, CuNb₁₃O₃₃, Cu₅Ta₁₁O₃₀, and Cu₃Ta₇O₁₉ and systematically investigated their PEC performance for water-splitting [79,80,81,82]. All these materials are visible light responsive. However, only CuNbO₃ and CuNb₃O₈ have band gaps of less than 2.1 eV. CuNbO₃ and CuNb₃O₈ powder were synthesized by heating a stoichiometric mixture of Cu₂O and Nb₂O₅ in an evacuated fused silica tube at the proper temperature and duration. The photocathodes were then prepared by spreading the as-synthesized powder onto FTO conductive substrate with post-heat treatment under a vacuum. CuNbO₃ is a red color p-type semiconductor with a band gap of ca. 2.0 eV [79]. Density functional theory calculations show the visible light absorption results from a nearly direct band gap transition owing mainly to copper-to-niobium excitations. Figure 5 shows the current-potential curve of the CuNbO₃ photocathode under chopped visible light illumination (λ > 420 nm). It was found that the CuNbO₃ photocathode exhibited an onset potential of +0.10 V vs. saturated calomel electrode (SCE) and achieved a cathodic photocurrent of −0.10 mA cm⁻² at −0.6 V vs. SCE. The incident photon-to-current conversion efficiency (IPCE) of ca. 1–2% is attained at the lowest visible-light energies and rises sharply to >5% at higher incident photon energies. Moreover, the p-type CuNbO₃ film exhibited a stable photocurrent and notable resistance to photocorrosion.

Similar to CuNbO₃, CuNb₃O₈ has a p-type property [80]. However, because of the different crystal structures, CuNb₃O₈ was found to have indirect and direct band gap values of 1.26 and 1.47 eV and consequently can utilize a larger fraction of the solar irradiation. Furthermore, the conduction and valence bands of CuNb₃O₈ are higher than the water reduction and oxidation potentials, respectively. As a result, CuNb₃O₈ can thermodynamically reduce water to H₂ but not oxidize it to O₂. Under visible light irradiation (λ > 420 nm), the CuNb₃O₈ photocathode showed an onset potential of +0.30 V vs. RHE and a cathodic photocurrent of −0.40 mA cm⁻² at −0.35 V vs. SCE. Moreover, this p-type CuNb₃O₈ film also exhibited a stable photocurrent. The above studies indicated that CuNbO₃ and CuNb₃O₈ are potentially good candidates for fabricating photocathodes for PEC water splitting due to their excellent light absorption properties, suitable conduction band edge position, and relatively high stability against photocorrosion.

2.1.4. CaFe₂O₄

CaFe₂O₄, a p-type semiconductor material, has a band gap of 1.9 eV. The conduction and valence band edges of CaFe₂O₄ are located at −0.6 and +1.3 V vs. NHE, respectively, fulfilling the thermodynamic requirement for the hydrogen and oxygen evolution reactions [83,84,85]. Ida et al. reported the fabrication of a p-type CaFe₂O₄ photocathode for water splitting [86]. The CaFe₂O₄ photocathode was prepared by spreading CaFe₂O₄ powder onto a Pt plate followed by annealing at 1373–1473 K for 2 h. The CaFe₂O₄ photocathode prepared at 1373 K consisted of powder with particle sizes of 2–4 μm, while annealing at 1473 K will melt the powder to form a (hk0)-oriented CaFe₂O₄ planar film. The electrode prepared at 1473 K presented a photocurrent that was ca. 10–50 times higher than the electrode prepared at 1373 K due to the improvements in film quality and the better contact between the CaFe₂O₄ film and Pt substrate (Figure 6). IPCE of 5–15% and less than 5% was obtained in the UV and visible regions, respectively. A small photocurrent was also observed up to 650 nm, corresponding to the band edge of CaFe₂O₄. However, the photocurrent was quite low in the visible region, possibly due to the faster recombination of the as-generated electron-hole pairs. When a bias of 1 V was applied between the CaFe₂O₄ cathode and Pt anode in a two-electrode system under irradiation, H₂ and O₂ can be produced with a nearly stoichiometric ratio. Moreover, a photocell short-circuited by connecting the p-type CaFe₂O₄ and n-type TiO₂ can also split water with no applied bias under illumination. However, the ratio of H₂ to O₂ is not stoichiometric, probably attributed to the oxidation or oxygen adsorption on the TiO₂/Ti electrode. Moreover, the photocurrent of this photocell gradually decreased, possibly due to the decomposition of CaFe₂O₄ upon long-time illumination. This instability problem is anticipated to be solved by loading the proper cocatalyst. Following this work, Cao et al. prepared p-type CaFe₂O₄ nanofilms on an FTO glass substrate using a pulsed laser deposition (PLD) method. The as-prepared film showed high PEC performance for the H₂ production even without applied bias [87].

2.1.5. Sr_1-xNbO₃ (0.1 ≤ x ≤ 0.2)

Xu et al. reported the synthesis of a series of red oxide semiconductors with a chemical formula of Sr_1-xNbO₃ (0.1 ≤ x ≤ 0.2) and their properties for the photocatalytic water splitting reaction under visible irradiation [88]. The absorption edges of the Sr_1-xNbO₃ (0.1 ≤ x ≤ 0.2) catalysts are located at ca. 700 nm, corresponding to a band gap of 1.9 eV. Furthermore, the band gaps of Sr_1-xNbO₃ (0.1 ≤ x ≤ 0.2) catalysts reduce gradually with reducing Sr content. Under visible light illumination, the Sr_1-xNbO₃ catalysts displayed activity for the oxidation and reduction of water in the presence of suitable sacrificial reagents, indicating the proper band edge positions of Sr_1-xNbO₃ relative to both the hydrogen and oxygen evolution potentials. Furthermore, the photocatalytic activity of Sr_1-xNbO₃ can be further enhanced by ball milling treatment due to the resulting increased surface area. One interesting property of the Sr_1-xNbO₃ materials is their high electronic conductivity of ca. 3 × 10³ S cm⁻¹ (x = 0.2), which indicates their metallic nature instead of that observed in traditional semiconductor photocatalysts. The authors suggested that the absorption of the photons was due to the excitation from the partially filled conduction band to a higher-level unoccupied band. Although the exact origin of the light absorption needs to be further investigated, the wide absorption up to 700 nm, the appropriate energy level, and the high conductivity may entail Sr_1-xNbO₃ a promising candidate for PEC water splitting.

2.2. Nitride and Oxynitride-Based Semiconductor Materials

Domen’s group developed a series of Ta, Nb, Ti, Ga, and Ge-based nitride and oxynitride semiconductor materials for photocatalytic and PEC water splitting. The tops of the valence bands of these materials consist primarily of N 2p orbitals or O 2p hybridized N 2p orbitals, and the bottoms of the conduction band consist of metal orbitals. Compared with oxide materials, the corresponding nitride or oxynitride materials will have much-reduced band gaps due to the contribution of N 2p orbitals to the valence band. Moreover, compared with sulfide materials, nitride or oxynitride materials exhibited relatively high stability against photo corrosion, forming a unique type of visible-light-responsive semiconductor material.

2.2.1. Ta₃N₅

Ta₃N₅ is one of the most investigated nitride semiconductor materials for photocatalytic and PEC water-splitting applications [89,90,91,92]. Ta₃N₅ has several important features that are compatible with solar water-splitting reactions. For example, Ta₃N₅ has a small band gap of around 2.1 eV and can achieve a maximum theoretical energy conversion efficiency of 15.9%. This property is comparable with that of Fe₂O₃. Moreover, the conduction and valence band levels of Ta₃N₅ are located at ca. +1.6 V and −0.4 V vs. NHE at pH = 0, respectively, and therefore can fulfill the thermodynamic requirement for both water reduction and oxidation reactions.

Ta₃N₅ was initially investigated as a photocatalyst for photocatalytic water-splitting reactions [89,90,93,94,95,96,97,98,99,100]. It was prepared by heating Ta₂O₅ powder in an NH₃ flow at 1123 K. Under visible light irradiation, Ta₃N₅ can reduce water to produce H₂ in an aqueous methanol solution. Moreover, Ta₃N₅ can produce O₂ via water oxidation with AgNO₃ as a sacrificial reagent. Even though the efficiency achieved on Ta₃N₅ for photocatalytic water splitting is low in the presence of sacrificial reagents, it represents one of the quite few materials with a band gap of 2.1 eV together with the ability to split water without applied bias. The photocatalytic activity of Ta₃N₅ for water oxidation was improved by six times via modifying Ta₃N₅ with alkaline metal salts [101], indicating the potential of further improving the photocatalytic performance of Ta₃N₅ with the proper synthesis technique.

Apart from photocatalytic water splitting, Ta₃N₅ has been extensively investigated as a photoanode material for PEC water oxidation [102,103]. Abe et al. used a simple electrophoretic deposition method for the fabrication of a Ta₃N₅ photoelectrode [104]. In this approach, Ta₃N₅ powder was first prepared and then assembled on the FTO substrate with the electrophoretic deposition method. This as-deposited Ta₃N₅ photoelectrode showed negligible photocurrent. Then a post-necking treatment of the Ta₃N₅ photoelectrode in NH₃ was employed to ensure better contact between the Ta₃N₅ powder and the Ta₃N₅ powder with the FTO substrate. The as-prepared film demonstrated drastically enhanced photocurrent after this post-necking treatment. The loading of IrO₂ as a cocatalyst on Ta₃N₅ was found to further drastically enhance the photocurrent by more than three times using Na₂SO₄ as the electrolyte. However, the Ta₃N₅ photoelectrode degraded upon light irradiation due to self-photooxidation. The stability of the Ta₃N₅ photoelectrode was significantly improved by using a Co₃O₄ nanoparticle water oxidation cocatalyst and an alkaline electrolyte [105].

In order to further promote the PEC performance of Ta₃N₅, several strategies have been used to prepare the Ta₃N₅ photoanode [106,107,108,109,110,111,112,113,114,115,116]. Li et al. reported the preparation of vertically aligned Ta₃N₅ nanorod arrays via nitridation of Ta₂O₅ nanorod arrays prepared in situ on Ta foil using a through-mask anodization approach (Figure 7). The as-prepared Ta₃N₅ electrode exhibited a photocurrent density of 3.8 mA cm⁻² at an applied potential of 1.23 V vs. RHE under AM 1.5G simulated sunlight at 100 mW cm⁻² and a maximum IPCE of 41.3% at 440 nm. This current density is 3.2 times higher than that of a thin film Ta₃N₅ photoelectrode. Moreover, by coating the Co-Pi cocatalyst on Ta₃N₅ nanorod arrays, the Ta₃N₅ electrode demonstrated quite high stability during the PEC reactions [110]. Zhen et al. reported the simple template-free synthesis of a Ta₃N₅ nanorod array on Ta substrate by combining a vapor-phase hydrothermal process and subsequent nitriding. After modification with a Co (OH)_x cocatalyst, the Ta₃N₅ nanorod array films exhibited a stable photocurrent density of 2.8 mA cm⁻² at 1.23 V vs. RHE under AM 1.5G simulated sunlight. The IPCE at 480 nm is ca. 37.8% [109].

2.2.2. MTaO₂N (M = Sr, Ba)

A series of alkaline earth tantalum oxynitrides with a chemical formula of MTaO₂N (M = Ca, Sr, Ba) have been developed for photocatalytic water-splitting reactions. Among these materials, SrTaO₂N and BaTaO₂N have band gaps of 2.1 and 2.0 eV, respectively [117,118,119,120]. The conduction band and valence band compositions for both materials are the same as that of Ta₃N₅, while the participation of Sr and Ba will lead to different electronic structures [121,122,123,124]. Both materials can generate H₂ from an aqueous methanol solution under visible light. Under visible light irradiation, SrTaO₂N will undergo self-oxidative decomposition in the presence of I⁻ and unsuitable in the Z-scheme system, while Pt/BaTaO₂N can split pure water when used as an H₂-evolution photocatalyst in combination with a Pt/WO₃ photocatalyst in a Z-scheme system. Moreover, Pt-BaTaO₂N is also reported to be photoactive at wavelengths up to 660 nm, representing the first example of an overall water-splitting system where visible light at wavelengths longer than 600 nm is effectively absorbed for H₂ evolution [125,126].

2.2.3. LaTiO₂N

Ti-based oxynitride, LaTiO₂N, has a band gap of 2.1 eV [127,128]. The top of the valence band consists predominantly of N 2p orbitals with a small contribution from O 2p orbitals, while the bottom of the conduction band is composed of empty Ti 3d orbitals. The conduction and valence bands of LaTiO₂N are suitable for water reduction and oxidation reactions, respectively. When loading Pt as a cocatalyst, LaTiO₂N can produce H₂ from an aqueous methanol solution under visible light irradiation. Moreover, LaTiO₂N showed activity for photocatalytic O₂ production with AgNO₃ as the sacrificial reagent [129,130,131,132]. LaTiO₂N with both single-crystalline mesoporous and macroporous architectures was prepared and then modified with a CoO_x cocatalyst. The as-obtained CoO_x/LaTiO₂N photocatalyst demonstrated drastically enhanced efficiency for photocatalytic water oxidation reactions. A quantum efficiency of 27.1 ± 2.6% at 440 nm was achieved on the optimized CoO_x/LaTiO₂N photocatalyst [133].

LaTiO₂N has also been investigated as a photoanode material for PEC water splitting [134,135]. Domen’s group proposed a particle transfer (PT) approach for fabricating LaTiO₂N photoelectrodes [136,137]. LaTiO₂N powder was first coated onto a primary glass substrate. A thin contact layer and a thick contact layer were subsequently deposited onto the LaTiO₂N layer with radio-frequency (RF) magnetron sputtering. The above three layers were adhered to a second glass plate by epoxy resin, then peeled off the primary glass plate. Even though this approach requires complex procedures and facilities, it will ensure good electrical contact with the semiconductor particles and facilitate the charge transfer. Under AM 1.5G irradiation, a photocurrent of ca. 3 mA cm⁻² at 1.23 V_RHE and nearly 100% Faradic efficiency was obtained, much higher than that obtained with the previous method. The problem with this material is its relatively low stability during the PEC reactions, even though it was modified with IrO₂ as the oxygen evolution cocatalyst.

2.3. Sulfide and Oxysulfide-Based Semiconductor Materials

2.3.1. CuInS₂

Copper indium disulfide (CuInS₂) has a band gap of ca. 1.5 eV and a large absorption coefficient (ca. 10⁵ cm⁻¹) [138]. The conduction band of CuInS₂ is suitable for the water reduction reaction, while the valence band is unsuitable for the water oxidation reaction [139,140]. Xie et al. investigated the photocatalytic H₂ production on CuInS₂ powder prepared by a facile solvothermal method. The as-obtained CuInS₂ powder exhibited monodisperse hierarchical microarchitectures. Under visible light irradiation, the Pt/CuInS₂ powder can produce H₂ at 20 μmol h⁻¹ with Na₂S-Na₂SO₃ as sacrificial reagents [141]. Ikeda et al. investigated PEC water splitting on a modified CuInS₂ photocathode [142]. The CuInS₂ photocathode was synthesized from electrochemically stacked metallic Cu and In layers followed by sulfurization. A Cu/In ratio of 1.3–1.4 was used to obtain CuInS₂ film with high crystallinity and large grain sizes. The bare CuInS₂ photocathode showed negligible cathodic photocurrent under illumination. After loading the Pt cocatalyst, the CuInS₂ photocathode showed drastically enhanced photocurrent. The introduction of an n-type buffer layer, such as CdS and ZnS, on the CuInS₂ photocathode before the Pt loading will further enhance the performance of the CuInS₂ photoelectrode (Figure 8). IPCE as high as 20% at wavelengths ranging from 500 to 750 nm was obtained.

2.3.2. Ln₂Ti₂S₂O₅ (Ln = Pr, Nd, Sm, Gd, Tb, Dy, Ho, and Er)

Domen et al. systematically investigated a series of Ti, Ga, and In-based oxysulfide materials for photocatalytic water splitting due to their strong absorption in the visible region [143,144,145,146,147,148]. Sm₂Ti₂S₂O₅ is the first reported oxysulfide material capable of splitting water photocatalytically in the presence of sacrificial reagents [143]. It was prepared by heating a stoichiometric mixture of Sm₂S₃, Sm₂O₃, and TiO₂ in a sealed vacuum quartz tube at 1273 K [143]. The band gap of the resulting Sm₂Ti₂S₂O₅ was determined to be ca. 2.0 eV from the UV-Vis spectra, which is much smaller than that of Sm₂Ti₂O₇ (Figure 9). DFT calculations indicated that the contribution of the S 3p orbitals to the top part of the valence band of Sm₂Ti₂S₂O₅ led to a smaller band gap. Under visible light irradiation (λ > 420 nm), Sm₂Ti₂S₂O₅ can reduce water to produce H₂ from aqueous solutions containing sacrificial electron donors (Na₂S-Na₂SO₃ or methanol). More interestingly, when loaded with IrO₂ cocatalyst, Sm₂Ti₂S₂O₅ can also oxidize water to produce O₂ in the presence of a AgNO₃ sacrificial reagent. This property of oxysulfide is unique when compared with sulfide materials that are incapable of oxidizing water due to the instability problem. PEC measurement indicated that Sm₂Ti₂S₂O₅ was an n-type material and showed a quite small photocurrent for PEC water splitting. From these results, the conduction and valence band edge levels of Sm₂Ti₂S₂O₅ are determined to be suitable for water reduction and oxidation reactions. Moreover, the flat band potential of Sm₂Ti₂S₂O₅ was found to be almost independent of pH, which is similar to that of sulfide material but totally different from that of oxide material. The problem with the Sm₂Ti₂S₂O₅ material is its low efficiency for photocatalytic water splitting. In the subsequent work, the modification of Sm₂Ti₂S₂O₅ with metal ions such as Ag⁺ and Mg²⁺ was found to increase the photocatalytic activity of Sm₂Ti₂S₂O₅ by seven times [148,149]. Similar to Sm₂Ti₂S₂O₅, a series of Ti-based oxysulfide materials with a chemical formula of Ln₂Ti₂S₂O₅ (Ln = Pr, Nd, Sm, Gd, Tb, Dy, Ho, and Er) with band gaps of 1.9–2.1 eV were synthesized and demonstrated photocatalytic activity for the water reduction and oxidation reactions in the presence of sacrificial reagents, respectively [145].

2.4. Selenide-Based Semiconductor Materials

2.4.1. CuIn_1-xGa_xSe₂ (CIGS)

CIGS are a series of small band gap semiconductor materials. By tuning the ratios of In and Ga, the resulting materials will have band gaps ranging from 1.0 to 1.7 eV. These materials have several attractive properties for solar energy conversion applications, including good carrier transport properties and a direct band gap [29,150]. The conduction band edge level of CuIn_1-xGa_xSe₂ is high enough for the hydrogen evolution reaction, while the valence band edge level is too shallow for oxide water. Marsen et al. reported the preparation of semitransparent CuGaSe₂ film [151]. The 0.9 μm thick films were grown on an FTO substrate by vacuum coevaporation of elemental Cu, Ga, and Se and showed polycrystalline structures with grain sizes of tens of nanometers. Under outdoor one sun irradiation, the as-prepared CuGaSe₂ photocathode showed a photocurrent density of up to 13 mA cm⁻². Significant IPCE throughout the visible spectrum, peaking at 63% at 640 mm, was obtained. Moreover, under AM 1.5G light irradiation, the CuGaSe₂ photocathode showed a notably stable photocurrent, indicating the high stability nature of the film at cathodic potentials (Figure 10). Similar results were obtained in other studies [152]. Yokoyama et al. reported the preparation of CIGS films onto Mo-coated soda-lime glass using a molecular beam epitaxy apparatus [153]. They observed a similar phenomenon in the CuGaSe₂ film mentioned above. Under irradiation by monochromatic light (λ = 600 nm), the IPCE of CIGS, Pt/CIGS, and Pt/CdS/CIGS were 0.03, 19, and 59%, respectively, at −0.24 V vs. RHE. Although all the above studies claim that CIGS is a good candidate for practical water splitting, the high cost of In and Ga will be a big barrier toward this goal.

2.4.2. WSe₂

Tungsten diselenide (WSe₂) is a layered semiconductor material with a direct band gap of 1.4 eV and an indirect band gap of 1.2 eV. WSe₂ has been traditionally investigated as a photoanode material for PEC water splitting. It was reported that n-type WSe₂ had shown stability and attractive >17% PEC energy conversion efficiency in aqueous polyhalide-based redox systems [154]. Due to the excellent properties of WSe₂, Mckone et al. fabricated p-type WSe₂ photocathodes and investigated their performance for PEC water splitting [155]. In their work, crystalline p-type WSe₂ crystals were first grown using a chemical vapor transport method. The as-grown crystals were then cut into small pieces and used as the photocathode after proper treatment. To enhance the performance of the WSe₂, Ru and Pt cocatalysts were deposited onto WSe₂ using a photo-electrochemical method. PEC measurements showed that the bare p-type WSe₂ photocathode exhibited negligible photocurrent density, indicating the photocathodes produced negligible net conversion of light energy to H₂. However, after Ru or Pt cocatalyst loading, the p-type WSe₂ photocathode exhibited drastically enhanced photocurrent of up to 24 mA cm⁻² at 0 V vs. RHE under AM 1.5G illumination (Figure 11). Overall thermodynamic photoelectrode solar energy-conversion efficiency of >7% was achieved using the Pt-Ru composite catalysts on the p-type WSe₂ photocathode. Moreover, the p-type WSe₂ photocathode demonstrated high stability under cathodic conditions for at least 2 h in both acidic and alkaline electrolytes. The problem for the p-type WSe₂ photocathode is the short effective minority-carrier diffusion lengths of only ca. 1 μm, which results in low quantum efficiencies at photon energies below the direct band gap. It was suggested that higher energy conversion efficiencies could be achieved with enhanced doping control and further improvements in crystal purity [156,157,158].

2.5. Phosphide-Based Semiconductor Materials

InP

InP is an important material with a band gap of 1.3 eV. InP’s conduction band edge level is more negative than the hydrogen evolution potential, fulfilling the thermodynamic requirement for the water reduction reactions [159,160]. InP has been investigated as a photocathode semiconductor for PEC water splitting, and an energy conversion efficiency of ca. 12% has been obtained in the InP photocathode [161]. Lee et al. investigated the effect of nanostructuring on the PEC performance of InP photocathode for PEC water splitting [162]. Figure 12 shows the process flow for the fabrication of InP nanopillar arrays. InP nanopillars were first obtained by treating planar p-type InP (100) bulk wafer with a reactive ion etching and surface cleaning process. A TiO₂ passivation layer was then deposited on the InP nanopillars with atomic-layer deposition (ALD) to improve the stability of InP. Finally, a Ru film with a thickness of ca. 2 nm was sputtered on the surface to act as the H₂-evolution cocatalyst. The reflectance of InP nanopillar structures was found to be drastically reduced compared to the planar InP wafer due to the nanostructuring of the surface. Under AM 1.5G illumination, the InP photocathode with nanopillar structure shows a higher current density and more positive onset potential (ca. 37 mA cm⁻², 0.73 V vs. NHE) than that with the planar morphology (ca. 27 mA cm⁻², 0.5 V vs. NHE). This high current density achieved on the InP nanopillars approaches the theoretical limit, and high energy conversion efficiency of ca. 14% was obtained. Moreover, the stability of the nanostructured InP photocathode was much improved due to the low surface energy of the InP nanopillars. A stable cathodic photocurrent of ca. 37 mA cm⁻² can be achieved for at least four hours. Therefore, the nanostructuring of the planar InP photocathodes was found to drastically enhance its PEC performances due to the improved light absorbance and surface properties related to the desorption of the as-evolved H₂ gas.

2.6. Iodide-based Semiconductor Materials

2.6.1. BiOI

Bismuth oxyhalides, BiOX (X = Cl, Br, I), are a series of Bi-based semiconductor materials with varying band gaps related to the type of X. BiOI has the smallest band gap among these materials, with an indirect band gap typically between 1.7 and 1.9 eV. BiOI with different morphologies has been synthesized and extensively investigated [163,164,165]. Hahn et al. investigated the PEC properties of n-type BiOI photoanode for water splitting [166]. Their work employed a spray pyrolysis approach to preparing BiOI nanoplatelet film photoelectrodes on FTO substrates at different temperatures. The as-obtained BiOI films showed an n-type property, most likely due to the existence of anion vacancies. The BiOI film prepared at 533 K showed the optimum photocurrent of 0.9 mA cm⁻² at 0.4 V vs. Ag/AgCl and IPCE of more than 20% in 0.25 M NaI acetonitrile solution due to the oxidation of I⁻ to I₃⁻during the PEC reactions (Figure 13). The photons in the visible region (λ > 420 nm) contributed ca. 75% of the total photocurrent under AM 1.5G irradiation. However, the BiOI photoanode showed poor durability in aqueous solutions due to the oxidation of I⁻ in BiOI and the subsequent formation of an insulating bismuth hydroxide layer. It is anticipated that the modification of BiOI with a cocatalyst layer or the insulation of BiOI with a proper protection layer will lead to stable PEC water splitting, which has been observed in several semiconductor materials such as Cu₂O, WSe₂, and InP detailed in this review.

2.6.2. CH₃NH₃PbI₃

Recently, halide perovskite materials have emerged as promising candidates for high-performance optoelectronic devices [167]. The general formula of halide perovskite materials is ABX₃, where A is usually CH₃NH₃⁺ (MA), CH (NH₂)₂⁺ (FA), or Cs⁺, B is Pb²⁺ or Sn²⁺, X is a halide anion (e.g., I⁻, Br⁻, Cl⁻) [168]. The compositions of the halide perovskite materials determine their band gaps. Typically, the I-based perovskites have band gaps smaller than 2.1 eV, including MAPbI₃ (1.6 eV), FAPbI₃ (1.4 eV), and CsPbI₃ (1.8 eV), among others [169].

The halide perovskite materials were first discovered in 1978 [170]. However, the rapid development of these materials was encouraged by Miyasaka in 2009 [171]. Miyasaka et al. first employed MAPbI₃ in solar cells with a power conversion efficiency (PCE) of 3.8%. Then many efforts have been devoted to enhancing efficiency [172]. Up to now, the certified PCE has reached up to 25.7% [173]. The high PCEs of perovskite solar cells were ascribed to their excellent optoelectronic characteristics [174], including high light absorption coefficient, tunable band gap, and long charge carrier lifetime, which are also desirable properties for photocatalytic or PEC hydrogen evolution. However, due to the extreme susceptibility to water, halide perovskite materials are hindered from solar fuel production [175,176,177].

Nam et al. first used MAPbI₃ particles for photocatalytic hydrogen evolution in a saturated HI aqueous solution [178]. They found that MAPbI₃ powders can be stable in an aqueous solution when finely controlling the concentrations of H⁺ and I⁻ due to the formation of a dynamic equilibrium between the dissolved species and the precipitate. Therefore, a photocatalytic HI splitting reaction was realized by MAPbI₃ particles to generate hydrogen. After the modification of Pt as a cocatalyst, the performance could be enhanced, and a solar-to-fuel conversion efficiency of 0.81% was achieved.

Other researchers have widely employed this dynamic equilibrium strategy [177]. For example, Huang and co-workers demonstrated that reduced graphene oxide (rGO) grafted on MAPbI₃ could act as an electron acceptor to significantly improve hydrogen evolution activity [179]. The photogenerated electrons from MAPbI₃ can, in situ, photoreduce graphene oxide to rGO, and then a composite of rGO and MAPbI₃ (MAPbI₃/rGO) was formed. The MAPbI₃/rGO showed 67 times enhancement in photocatalytic activity compared to bare MAPbI₃, ascribed to efficient extraction of photogenerated electrons from MAPbI₃ by rGO. Furthermore, the MAPbI₃/rGO showed good stability for photocatalytic hydrogen evolution with a continuous reaction of 200 h. Our group also reported that Pt/TiO₂ nanoparticles could drastically promote charge separation and transportation of MAPbI₃ particles (Figure 14) [180]. TiO₂ pre-deposited with Pt was introduced to MAPbI₃ particles to establish a dynamically-existing heterojunction, which enhanced photocatalytic activity. The rate of hydrogen evolution on Pt/TiO₂-MAPbI₃ was 89-fold higher than that of Pt/MAPbI₃. The solar-to-fuel conversion efficiency of Pt/TiO₂-MAPbI₃ was 0.86%. In addition, considering that Pt was a noble metal, substantial earth-abundant materials, such as Ni₃C, MoS₂, and CoP have been reported as alternative cocatalysts [181,182,183,184].

Halide perovskite materials have also been investigated as photoelectrode materials for PEC water splitting. However, the inherent instability of halide perovskite materials in water hinders their application in PEC reactions. Therefore, many strategies have been developed to protect perovskite-based photoelectrodes. Zheng et al. first reported a Ni-protected MAPbI₃ photoanode [185]. A Ni layer deposited on the top of MAPbI₃ can improve the stability of PEC S²⁻ oxidation. The photoanode lost most of its activity after a 1000 s continuous reaction. Yang et al. used hydrophobic alkylammonium cations to functionalize MAPbI₃ photoanodes [186]. The stability of the photoanode was enhanced to 30 min. Other strategies, such as atomic layer deposited TiO₂ [187], carbon nanotube/polymer composite film encapsulation [188], and mesoporous carbon protection layer [189], were also employed to improve the stability of perovskite-based photoelectrodes. Up to now, one of the most effective strategies has been demonstrated by Reisner et al. [190]. They used a Field’s metal (an alloy of In, Bi, and Sn) to encapsulate the MAPbI₃ photocathode. The Field’s metal can well-protect MAPbI₃ to avoid the permeation of water while still efficiently transferring photogenerated electrons from the photocathode to the electrolyte. The photocathode showed a current density of −9.8 mA cm⁻² at 0 V vs. RHE with enhanced durability of 1 h under continuous reaction. Following Reisner’s strategy, many works have been reported [191,192,193]. Our group developed a Ti foil-encapsulated MAPbI₃ photocathode with a sandwich-like structure [194]. As a hydrogen evolution cocatalyst, Pt was pre-deposited on the Ti foil using the magnetron sputtering method. The sandwich-like photocathode displayed a current density of −18 mA cm⁻² at 0 V vs. RHE. Moreover, it showed excellent stability under 12 h of continuous reaction in water. This work paves the way for employing halide perovskite materials in solar fuel production.

2.7. Silicon and Silicide-Based Semiconductor Materials

2.7.1. Si

Si (Silicon) has a band gap of 1.12 eV and can utilize solar energy of up to 1000 nm. The conduction band edge of Si is high enough to reduce water to H₂, while an external power source is needed to drive overall water splitting. Si has been commercially used in the photovoltaic industry as an earth-abundant semiconductor material. Due to the great potential for the fabrication of efficient photoelectrode for PEC water splitting, Si has been extensively investigated as the building block material for photoelectrode [195,196].

One problem of a planar Si-based film is that ca. 25% of incident light is reflected away from the Si surface, which limits the collection of the incident photons. Lewis and Atwater et al. reported the preparation of an efficient p-type silicon (p-Si) wire array on p⁺-Si (111) substrates from SiCl₄ and BCl₃ precursors using a copper-catalyzed, vapor-liquid-solid-growth process. Energy conversion efficiencies of up to 3% and internal quantum yields of at least 0.7 were observed for these wire arrays in contact with an aqueous methyl vilolgen^2+/+ electrolyte during PEC water splitting [197]. They showed that compared with planar-sheet absorbers, Si wire arrays have beneficial optical features for photovoltaic applications, such as reasonable absorption of sunlight despite low areal packing fractions, enhanced near-infrared absorption, and effective optical concentration over a wide range of incidence angles [198]. These properties may be useful in improving the efficiency of the Si photoelectrode or reducing the material consumption for PEC water-splitting applications. In the following work, a radial n⁺ emitter shell was formed on the surface of B-doped p-Si microwires, followed by the deposition of 1.5-nm-thick Pt film [199]. The as-prepared electrode yielded energy conversion efficiency >5% under one sun solar simulation in 0.5 M K₂SO₄ (pH = 2) solution, which is much higher than that of the analogous p-Si wire-array electrodes (<0.2%). The higher photovoltage generated at the n⁺ p-Si/H₂O junction is supposed to be the main reason for its higher PEC performance. Oh et al. prepared a nanoporous black silicon photocathode with a simple single-step metal-assisted etching technique [200]. Compared with traditional planar silicon photoelectrode, the nanoporous photoelectrode has several advantages, such as broadband anti-reflection that enables efficient absorption of the incident light, higher effective surface reactions sites in contact with water, lower overpotential required to drive PEC water reduction, and less need for surfactant in the electrolyte to promote H₂ evolution. Similarly, the Si nanowire array photocathode was fabricated by metal-catalyzed electroless etching, which demonstrated superior PEC performance than the planar Si photoelectrode [201].

The problem for Si is its instability during long-term PEC water reactions, which is especially serious for n-type Si photoanodes under anodic conditions. Therefore, different strategies have been developed to improve the stability of Si electrodes [202]. The common way to improve the stability of Si electrodes is to deposit a charge tunneling layer or catalytic layer on the Si surface to avoid the direct contact of Si with an aqueous solution or promote a water catalysis reaction. Chen et al. reported the deposition of a protective 2 nm thick TiO₂ layer with the atomic layer deposition method on Si photoanode followed by coating with a 3 nm iridium catalyst layer. The as-prepared photoanode exhibited sustained water oxidation capability for many hours under harsh pH and oxidative conditions, while silicon photoanodes without the TiO₂ layer quickly failed [203]. Brian et al. sputtered a thin layer of Ti on the n⁺ p-Si photocathode, followed by the deposition of a molybdenum sulfide catalyst. The Ti layer can protect the photoelectrode from oxidation, and molybdenum sulfide can work as a substitute for Pt to promote water reduction performance [204]. Other strategies, such as the coating of the metal oxide protective layer, Ni-Mo, and CoPi catalyst, have demonstrated their capacity for enhancing the PEC performance of Si-based photoelectrode [205,206,207,208,209,210,211,212,213,214].

2.7.2. TiSi₂

Ritterskamp et al. reported the photocatalytic water splitting on commercial titanium disilicide (TiSi₂) derived semiconducting catalyst under solar radiation [215,216]. Different from the classical semiconductor materials, which exhibit small band gap spreads, the dark grey TiSi₂ can absorb light within a wide range of 360 to 800 nm. The Fermi level of TiSi₂ was measured to be located at −0.43 eV and −0.41 eV before and during the reaction at pH = 7, respectively. Therefore, the energy level is high enough to fulfill the thermodynamic requirement for the proton reduction reactions. Under solar light irradiation, two phases (A and B) of H₂ evolution from pure water were observed when TiSi₂ power was allowed to react under standard conditions. As shown in Figure 15, H₂ evolution due to the surface passivation and photocatalytic water splitting occurred in phase A, during which the catalytic centers for water oxidation and proton reduction were in situ generated. In phase B, H₂ production from the photocatalytic process contributed to more than 96% of the total H₂ produced. One interesting finding for the TiSi₂ catalyst is that O₂ can store in the TiSi₂ catalyst and can be liberated upon heating to 373 K. Therefore, TiSi₂ is a unique photocatalyst with the capability of self-separating H₂ and O₂ products from water splitting under solar irradiation.

Banerjee et al. investigated the PEC performance of a photoanode composed of TiSi₂ nanoparticles that were assembled onto a TiO₂ nanotube array [217]. The TiO₂ nanotube array was synthesized using a sonoelectrochemical anodization method. TiSi₂ nanoparticles were prepared by ball milling commercial TiSi₂ particles followed by ultrasonication processing. The TiSi₂ nanoparticles were then assembled onto the TiO₂ nanotube array using the multiple impregnation method. The resulting TiSi₂-TiO₂ heterostructural composite photoanode showed a photocurrent of 3.49 mA/cm² at 1.2 V vs. NHE under AM 1.5G irradiation conditions, while only photocurrents of 0.9 mA/cm² and 0.26 mA/cm² were achieved on TiO₂ nanotubes or TiSi₂ photoelectrodes alone at the same conditions. The enhanced solar light absorption of TiSi₂ and increased charge transport abilities of one-dimensional TiO₂ NTs are supposed to be responsible for the excellent photoactivity of this hybrid photoanode.

3. Conclusions and Prospects

The development of small band gap semiconductor materials with suitable energy levels is crucial for efficient solar H₂ production from PEC or photocatalytic water splitting. Significant efforts have been devoted to exploring and modifying visible-light-responsive semiconductor materials, and several important signs of progress have been witnessed. For example, various material systems have demonstrated the capability for overall photocatalytic water splitting under visible light, which has been supposed to be one of the “Holy Grails” of chemistry. Although this proof-of-concept evidence is encouraging, no photocatalytic water splitting system can achieve solar energy conversion efficiency of more than 10%, even though different strategies have been proposed to improve photocatalytic performance. This is partially due to the extremely stringent thermodynamic requirement on the materials themselves and the unavoidable severe recombination of photogenerated charges. On the other hand, several small band gap semiconductor material systems have shown appreciable energy conversion efficiency of more than 10% for PEC water splitting under applied bias. Even though these efficient PEC devices could be used in some special conditions, the expensive material supply and the high fabrication cost will limit their wide application. In order to achieve a PEC system that can be met with the economic feasibility for practical solar H₂ production, developing small band gap (≤2.1 eV), cost-effective, environmentally benign, and stable semiconductor materials should be the future research focus. It is worth noting that computational models and theoretical simulations can investigate the materials’ band structures and electronic states, which could help us rationally synthesize semiconductors. Moreover, high-throughput computational screening efficiently identifies potential candidates for photocatalytic and PEC hydrogen evolution. The combination of experimental and theoretical results will facilitate the development of semiconductor materials for solar water splitting.