**1. Introduction**

The conversion of solar-emitted electromagnetic waves to useful forms of energy is a very promising research area in the field of renewable energy production. Although roughly 32 × 10<sup>24</sup> J of solar energy reaches the Earth's surface per year, only 0.001% of the incoming solar energy is used for human needs [1]. The conversion of solar light to useful forms of energy is still challenging at the scientific and engineering level in terms of energy production for the needs of human beings. Even though there are many technologies for renewable energy [2], including solar cells, solar collectors and solar fuel reactors (water-splitting), the major challenges we face are to improve efficiency and stability in the conversion of solar energy to other energy forms. Currently, one of the popular research technologies to tackle solar energy conversion is trying to convert photons into chemical energy [3] by using artificial photoelectrochemical (PEC) processes.

Metal-oxide nanomaterials have been thoroughly studied for the conversion of solar energy to hydrogen molecules due to their chemical and physical stability, optical and electronic properties, easy fabrication and low cost. They have shown good properties for use in photoelectrochemical devices such as TiO2 [4–6], α-Fe2O3 [7–9], BiVO4 [10–12], ZnO [13–15] and WO3 [16–18]. On the above properties of semiconductor materials, both suitable bandgap positions to generate hydrogen and oxygen gases and the ability to absorb a reasonable portion of the solar light spectrum are critical for water-splitting. However, a single metal-oxide photocatalyst cannot simultaneously satisfy all the requirements for solar-to-hydrogen-driven systems since it encounters many problems (including fast recombination of charge carriers, photo corrosion, instability in aggressive electrolytes, short lifetime of charge carriers, improper bandgap or di ffusion length of photogenerated electrons and holes). As a result, most of the metal oxide semiconductors are not suitable to split water at the visible light irradiation, which occupies 54% of whole solar spectrum since they either do not have proper bandgaps or only absorb the UV light region. However, the problems stated above have been successfully addressed by introducing heterojunctions, composite nanomaterials, coupling wide band and narrow band materials, doping, surface–interface engineering, dye sensitization, etc.

Among the metal oxides, WO3 is a promising semiconductor for PEC water-splitting with favorable properties. (These properties include: suitable bandgap (~2.6 eV), good chemical stability under strong solar exposure, oxygen-evolution capability, long minority carrier di ffusion length (~500 nm–6 μm [19,20]), absorption of visible light (~12%) and low cost.) The conduction band energy position of WO3 is 0.25 eV, which is not suitable for reorientation of bonds of hydrogen atoms from the aqueous phase to the gaseous (0 V vs. NHE). On the other hand, the valence band, located at 2.7 eV, is more positive than the oxidation potential of oxygen (1.23 V vs. NHE) and is suitable for oxygen evolution. Although the WO3 photocatalyst su ffers from some limitations such as sluggish charge transfer [21], boosting charge separation can be achieved by modifying WO3 photoanode with numerous materials including Ag nanoparticles [22] and Au plasmonic particles [23]. Many papers have reported on using WO3 photoanodes for O2 evolution study [24,25]. During the study of hydrogen evolution from aqueous phase, various photoelectrochemical systems and configurations integrated with WO3 and its composites have been developed.

Among the published materials in this prospect, numerous amounts of work can be distinguished: Ji et al. reported a triple layer heterojunction BiVO4/WO3/SnO2 material with a perovskite solar cell [26], Liu. et al. prepared a WO3 photoanode with a tandem cell [27] and Lee used dye-sensitized solar cells to produce hydrogen with bare WO3 photoanodes [28]. Zhang fabricated the WO3@a–Fe2O3/FeOOH photoanode, which exhibits a 120 mV negative shift in onset potential and yields a photocurrent density of 1.12 mA/cm<sup>2</sup> at 1.23 V vs. reversible hydrogen electrode (RHE) [29]. Moreover, some systems use free-particle WO3 heterostructures based on photochemical cell reactions. Despite the fact that WO3 cannot generate hydrogen, there are some publications where scientists show high photocatalytic activity for CdS-WO3 [30] and non-stoichiometric WO3−<sup>x</sup>/CdS heterostructures for e fficient hydrogen generation [31].

Thus, we have conducted a literature survey on the WO3-based photocatalytic system and found a dramatic increase of publications recently (Figure 1). This indicates that the WO3 is a very important material for designing e fficient photocatalytic systems. The analysis of scientific articles, reviews and conference materials found in the authoritative database revealed few review papers in the use of tungsten trioxide photocatalyst for water-splitting. As shown in Figure 1, the trend of the published papers in the WO3 photocatalytic research is increasing exponentially. Therefore, in our opinion, it is essential to present a review article to our scientific community with recent research progress of WO3 in the photocatalytic water-splitting. Although there are some review papers that included the WO3 and their water-splitting applications, from the best of our knowledge, few papers have been specifically focused on sole WO3/nanocomposites and their recent photocatalytic application.

**Figure 1.** Statistics analysis from Web of Science indicates increase of recent publications in WO3 photocatalytic areas.

The primary focus of our review article is to deliver the recent progress of tungsten-based photocatalytic systems that have been developed. More specifically, it discusses the morphology, crystal, doping, surface–interface engineering effect of WO3 on the heterostructured photocatalytic system, and all of the results in different conditions including electrolyte, power, applied bias, morphology, and synthesis approaches were tabulated for the researchers to check. Therefore, in this review, we try to give comprehensive information on WO3 including the physical chemistry property, crystal structure, and nanomorphology along with their composites including binary and ternary structures used in the particulate, PEC, Z-scheme and tandem configuration for effective water-splitting applications.

#### **2. Basic Principles of the Water-Splitting Reaction**

#### *2.1. Thermodynamics of Water-Splitting*

In the reaction of water-splitting, solar energy can be directly converted into chemical energy form, hydrogen gas [32–34]. The hydrogen acts as a green energy carrier since it possesses high energy density. When used in a fuel cell, water is the only byproduct.

As early as 1923, J. B. S. Haldane, a British scientist, proposed a concept of photocatalytic hydrogen production. Seeing that there is no naturally produced pure hydrogen on Earth, its resource is highly abundant throughout the universe. Like fossil fuels, water or biomass can be utilized to produce hydrogen or other chemical fuels. Hydrogen gas can be further used in hydrogen fuel cell-powered vehicles, which are much more environmentally friendly than the commonly used nonrenewable fuel options. Increasing the efficiency of water-splitting devices for hydrogen fuel production has a potential to decrease its dependence on using fossil fuels and importation.

There are some other approaches for hydrogen production, however, the most environmentally sustainable, "green" method is photocatalytic or photoelectrochemical water-splitting. The PEC water-splitting works similarly to a solar cell. The main difference is that it converts solar energy to a chemical bond instead of converting directly to electric power, which is beneficial to store energy for later use. PEC consists of three main components: an anode, a cathode and an electrolyte (aqueous media). At the anode, water is oxidized to generate oxygen via the oxygen evolution reaction (OER), whereas at the cathode hydrogen ions are reduced into hydrogen gas via a hydrogen evolution reaction (HER). Based on the configuration of the PEC cell, either the cathode or anode, or both, can be photoactive semiconductors which absorb light. Furthermore, water can also split via connecting a p–n junction solar cell in parallel with a photoelectrochemical cell. This process not only avoids the complicated manufacturing process, but it also reduces the system's cost [35]. Although extensive research has been conducted using many semiconductor configurations, there is still so much that needs to be done to reach the targeted efficiency and stability goals. For a particle-based photo catalytical

system, the ideal solar to hydrogen (STH) efficiency should be 10% [36,37]. This efficiency brings cheaper H2 production.

In 1972, Japanese scientists, Fujishima and Honda first studied TiO2 as a photonic material and proved that water can be decomposed under UV-light exposure [38]. Since then, scientists have been studying a variety of light-sensitive material, including all inorganic and organic dyes [39–41].

Decomposing water into H2 and O2 is an endothermic reaction thermodynamically (+237.2 kJ/mol). This means that additional energy is required to perform the decomposition reaction (E-1):

$$
\Delta \mathbf{G}^0 = -\mathbf{n} \mathbf{F} \Delta \mathbf{E}^0 = +237.2 \text{ kJ/mol } \mathbf{H}\_2 \tag{1}
$$

where:

F—Faraday's constant (F = 96,485 C/mol),

n—Number of transferred electrons (n = 2)

ΔE0—standard potential of the electrochemical cell (ΔE<sup>0</sup> = 1.229 V).

The amount of Gibbs free energy required to split a molecule of water into hydrogen and oxygen is ΔG = 237.2 kJ/mol, which is corresponded to ΔE<sup>0</sup> ≈ 1.23 eV per electron, transforming the Nernst equation under standard conditions. This means a minimum energy of 1.23 eV per electron should be supplied by the photocatalyst. This process can be written in the following two half-reactions (E-2; E-3; E-4):

$$\text{Water oxidation: }\text{H}\_2\text{O} + 2\text{h}^+ \rightarrow \frac{1}{2}\text{ O}\_2 + 2\text{H}^+\text{ (HER)}\tag{2}$$

$$\text{Water reduction: } 2\text{H}^+ + 2\text{e}^- \rightarrow \text{H}\_2\text{ (ORE)}\tag{3}$$

$$\text{H}\_2\text{O} \rightarrow \frac{1}{2}\text{ O}\_2 + \text{H}\_2\text{ }\Delta\text{G} = +237.2\text{ kJ/mol}\tag{4}$$

The bandgap (Eg) is the main parameter that defines the light-harvesting ability of an absorber. Photons alone with energies higher than the bandgap can excite electrons in the valence band to the conduction band. The excess energy or the difference in the energy of the absorbed photon and the band gap energy (*E*–*Eg*), is lost as phonons. The absorption coefficient of the semiconducting materials is another parameter which shows how efficiently a photocatalyst can harness the solar spectrum. One crucial point that needs to be taken into consideration as quantifying the optimal minimum band gap value is the intrinsic loss (Eloss), associated with the solar energy conversion process. These losses are connected with the fundamental loss caused by thermodynamics because of non-ideality (kinetic losses) in the conversion process [35,38]. The former loss results from the second law of thermodynamics. In fact, the following equation shows how bandgap energy (Eg) corresponds to the change in internal energy, which is related to the change in Gibbs energy (E-5):

$$
\Delta \mathbf{G} = \Delta \mathbf{U} + \mathbf{P} \Delta \mathbf{V} - \mathbf{T} \Delta \mathbf{S} \tag{5}
$$

where U, P, V, T and S indicate the internal energy, pressure, volume, temperature and entropy, respectively. When the semiconductor absorbs photons, increasing excited states can be created in addition to ground states, increasing the entropy of the ensemble. The change in entropy or ΔSmix, occurs because there are existing excited states along with the ground states. A volume change (ΔVmix) is also caused by the mixture of excited and ground states. However, this is not true for the ideal chemical system (ΔVmix = 0). Thus, the band gap energy should be greater than the available work under ideal conditions (Gibbs energy change per electron), at least Eloss = TΔSmix with a minimum of 0.3–0.5 eV. In reality, Eloss reaches higher values (roughly 0.8 eV) as a result of kinetic losses and due to non-ideality (overpotential at the anode and cathode, reduction in resistance at the electrolyte, electron–hole pair recombination). Therefore, in order to maximize the chemical conversion efficiency, materials commonly used as photoelectrodes in PEC cells require a band gap of 2.0 to 2.25 eV [35,38].

When UV and/or visible sunlight shine onto a semiconductor photocatalyst, the semiconductor absorbs photons and excites electrons from the semiconductor's valence band to its conduction

band, leaving a hole in the valence band, i.e., electron–hole pairs (Figure 2a). This is the so-called, "photo-excited" semiconductor phase. The bandgap is the difference between the maximum valence band energy and the minimum conduction band energy. Ideally, semiconductors have a bandwidth greater than 1.23 V as well as a more negative conduction band relative to the water reduction potential and a more positive valence band relative to the water oxidation potential.

**Figure 2.** (**a**) Schematic illustration for WO3-particle-based photocatalyst system; (**b**) principle of photoelectrochemical water-splitting.

A typical one-step PEC configuration for water decomposition consists of either a photoanode or a photocathode. N-type tungsten trioxide is mostly represented as a photoanode and the basic principles of such cell can be depicted in Figure 2b. Process of PEC water decomposition is initiated via accepting light photons by photoactive materials. Then, this step is accompanied by generating excitons (electron–hole pairs) inside semiconductors. Photogenerated holes on the surface of WO3 can oxidize water while electrons flow to the Pt electrode to produce hydrogen (Figure 2b). Due to the improper positioning of the conduction and valence bands with respect to the potentials of water reduction and oxidation, external bias voltage is used to separate excitons.

Another thermodynamic precondition is the position of the band edges. For the oxidation reaction to occur, holes move from the photoelectrode to the interface between the semiconductor and the solution freely. The top edge of the valence band must be more positive than the oxidation potential of O2/H2O as seen in Figure 2b. Likewise, the reduction reaction happens if the bottom edge of the conduction band is more negative than the reduction potential of H+/H2.

Figure 3 shows the band structure and bandgap values of some semiconductors [42]. While wide band gap (Eg > 3 eV) photocatalyst can harvest only UV light (a small portion of the solar spectrum, less than 4%), its band gap can be easily engineered to absorb the visible light range via metal and nonmetal doping. Furthermore, narrow bandgap materials (e.g., WO3, Fe2O3) are not able to drive the water reduction and oxidation reactions at the same time since their bandgap energy positions are not properly positioned to the water redox potentials. Therefore, they are commonly used to construct tandem cell structures for the water-splitting reaction.

**Figure 3.** Bandgap positions of typical semiconductors for water-splitting. Reproduced from [42], with permission from MDPI, 2016.

#### *2.2. Device Requirements and Calculation of Their E*ffi*ciency*

Various types of overall water-splitting techniques include: particulate systems [43], Z-schemes [44–46] and photoelectrochemical cells [46]. The photoelectrodes of PEC water decomposition such as a photoanode and a cathode electrode are made up of photocatalysts. However, fabrication of stable photoelectrodes under the influence of strong sunlight is still a challenging one in PECs. Right now, it is preferred to fabricate the film via direct growth on the photoelectrode, which provides a relatively stable photoactive films.

In order to design an efficient and affordable solar hydrogen production PEC system, the electrode requires low cost materials with the capability of efficient light harnessing and long term stability.

To date, large band gap semiconductors (UV-active specifically) and metal oxides have been extensively investigated for the photocatalytic water-splitting studies due to their robustness and suitable band gap energy positions. One challenge to using these materials is the limitation of solar light harnessing to a small portion of the solar spectrum.

The following formula helps us calculate the theoretical maximum photocurrent Jmax, which is proportional to the solar–hydrogen conversion efficiency (STH):

$$\mathbf{J}\_{\text{max}} = \mathbf{q} \int \Phi\_{\lambda} \left[ 1 - \exp(-\alpha\_{\lambda} \mathbf{d}) \right] d\lambda \tag{6}$$

where λ, q, d and αλ represent wavelength, electron charge, sample thickness and absorption coefficient under the photon flux of the AM 1.5 G solar spectrum, respectively. Considering the conversion, reflection and other losses, obtaining the goal of 10% STH conversion efficiency is very challenging.

The following efficiencies are usually reported for PEC cells. They are STH conversion efficiency, applied bias photon-current efficiency (ABPE), external quantum efficiency and internal quantum efficiency (IQE) or absorbed photons to current efficiency (APCE).

STH efficiency is commonly used to evaluate PEC device performance and is expressed in the following way:

$$\text{STH} = \text{[(H}\_2 \text{ production rate)} \times \text{(Gibbs free energy per H}\_2\text{)}] \text{([Incident energy])}\tag{7}$$

The E-8 formula can be applied to calculate ABPE:

$$\text{ABPE} = \text{[I}\_{\text{ph}} \times (1.23 - \text{V}\_{\text{b}})\text{]} \text{[P}\_{\text{total}}\tag{8}$$

where Jph is the photocurrent density as a bias Vb is applied and Ptotal is the total incident solar light power.

External quantum efficiency defines the photocurrent generation per incident photon flux under a certain irradiation wavelength. Solar-to-hydrogen conversion efficiency can be evaluated via applying the external quantum efficiency data over the total solar spectrum in a two-electrode system. However, applying external quantum efficiency data obtained in a three-electrode system under a bias to estimate solar-to-hydrogen conversion efficiency is not considered to be a valid method. However, it is still considered to be a useful approach for finding PEC cell material properties. The external quantum efficiency (EQE) is expressed by equation (E-9):

$$\text{EQE} = (\text{J}\_{\text{ph}} \times \text{hc}) / (\text{P}\_{\text{mcono}} \times \lambda) \tag{9}$$

where Jph is the photocurrent density, h is Planck's constant, c is the light speed, Pmono is the power of calibration and monochromatic illumination, and λ is the wavelength of monochromatic light.

#### **3. WO3 and Its Nanocomposites for Particle-Based Photocatalytic Systems**

#### *3.1. Half Reaction Systems*

Nowadays, the pursuit for highly efficient photocatalytic materials to produce hydrogen fuel under the exposure of light photons is still in the active stage. In a photoelectrochemical cell, it is required to create sufficient voltage between the anode and cathode to perform the water decomposition reaction. However, most of wide bandgap semiconductor materials are not able to respond to the visible part of the spectrum. Absorption of ultraviolet radiation alone is an undesirable property of photocatalyst operating in terrestrial conditions. One of the exciting ways to solve the above contradictions is the creation of photocatalytic systems consisting of a series of photocatalysts. That is why researchers try to use photochemical systems, where water can be decomposed using colloid particles without any external voltage. Many papers have reported [47] that hydrogen can be generated, even though the efficiency is very low. Evolution of oxygen is difficult because it requires process of four electrons and four H<sup>+</sup> transfers.

Under solar illumination, although photoexcited electrons and holes are produced, they simultaneously experience recombination and back reaction, which are competitive processes of photogeneration. Hence, most works focus only on half reactions where either H2 or O2 evolution is possible in the presence of sacrificial electron donor or acceptor.

CdS/WO3 photocatalysts produced a high hydrogen evolution rate of 369 μmol/gh with lactic acid as an electron donor [30]. Further modification of CdS/WO3 with Pd particles increased the hydrogen evolution rate to 2900 μmol/gh, 7.9-fold higher than for CdS/WO3.

Furthermore, the surface plasmon resonance (SPR) effect of non-stochiometric WO3−<sup>x</sup> was demonstrated [31] from CdS/WO3−<sup>x</sup> heterostructures photocatalysts via photoinduced electron injection for hydrogen evolution. The non-elemental metal plasmonic material WO3−<sup>x</sup> has intense SPR in the visible/NIR region (Figure 4b). Free electrons in the conduction band of WO3−<sup>x</sup> can be generated from oxygen vacancies that are results of chemical reduction during synthesis. Further excitation of electrons can happen by SPR and then they can become hot electrons for the hydrogen generation as shown in Figure 4a. Photo-excited electrons on CdS inject into conduction band of WO3−x, so that the SPR of the photocatalyst WO3−<sup>x</sup> is stable and some hot electrons participate in hydrogen evolution reactions (Figure 4c).

In addition, to choose the photoanode material for the half-reaction of water-splitting, attention should be paid to the selection of the electrolyte. For sulfide semiconductors and composites, a Na2S/Na2SO3 mixture is used as an absorbing hole agent. In type II heterojunctions, for example, WO3–NS/CdS–NR, with high conductivity, WO3 provides efficient charge collection and, therefore, reduces the rate of space charge recombination, which leads to the accumulation of holes in cadmium sulfide. An electrolyte based on Na2S/Na2SO3 provides fast hole collection, which allows the half-reaction to occur without degradation of the photoanode [48]. The effect of some electrolyte solutions on the oxidative half-reactions of WO3-based photoanodes was studied on [49]. James C. Hill and Kyoung-Shin Choi studied photo-oxidative processes in chloride solutions, acetate solutions, phosphate solutions, perchlorate solutions, sulfate solutions and solutions with K<sup>+</sup> and Li<sup>+</sup> cations. The electrodeposited porous WO3 layers were used as a photoanode. The results show that the presence of acetate and chloride ions suppressed the release of O2. In a phosphate solution, the release of O2 and the formation of peroxides was the main result of photooxidation. The oxidation of water in perchlorate electrolytes was accompanied by the release of O2 and the formation of peroxides. In this case, the photocurrent density in such a system was lower in comparison with phosphate electrolytes. The authors also showed that cations have a significant effect on the efficiency of conversion of the photocurrent to O2. For example, Li<sup>+</sup> ions adsorbed on the surface of WO3 serve as blockers of water oxidation centers, while K<sup>+</sup> ions increase oxygen evolution in perchlorate, sulfate and phosphate solutions.

**Figure 4.** (**a**) Graphic illustration of charge transfer for CdS/WO3−<sup>x</sup> composite; (**b**) DRS spectra of WO3−x, CdS and CdS/WO3−x−10; (**c**) hydrogen generation for CdS nanowires, WO3−<sup>x</sup> and CdS/WO3−x−<sup>10</sup> composites in 20 vol% lactic solution under illumination. Reproduced from [31], with permission from Elsevier, 2018.

The effect of tungsten trioxide layers on hydrogen reduction processes also demonstrates positive dynamics. When combining Cu2O with WO3, a semiconductor p–n junction is created and that generates the conditions for the rupture of photogenerated excitons. Thus, in the Cu2O/WO3 heterostructure, an enhancement of the half-reaction of reduction is demonstrated in comparison with the sole Cu2O photocatalyst.
