Due to the world's heavy reliance on non-renewable, geopolitically sensitive energy sources such as natural gas and coal, coupled with the violent fluctuations in crude oil prices, ensuring alternative clean and renewable energy supplies has never been such urgent [1-6]. The world's primary energy consumption increased by 2.9% in 2018, which is the highest annual growth rate since 2010, almost double of the average annual growth rate in the last 10 years. Our main energy sources still depend on limited and non-renewable fossil fuels, such as oil, natural gas and coal. Besides, the combustion of fossil fuels has led to a series of serious environmental pollution issues, such as air pollution, water pollution and even global warming. Therefore, it is imperative to develop renewable, clean and environmentally-friendly energy sources. Solar energy is the most tremendous renewable natural resource, which is abundant, cheap and clean [7-11]. H₂ is considered as an important energy in the future, which is mainly due to its cleanness, non-toxicity and high energy density compared to other natural resources. Photocatalytic hydrogen production via solar splitting is one of the most promising solutions to address the environmental problems and sustainable clean energy issues.

Since the pioneer work of Fujishima et al. [12] on photoelectrochemical water splitting to produce hydrogen (H₂) and oxygen (O₂) separately under light irradiation by using TiO₂ in 1972, semiconductor photocatalytic water splitting to produce H₂ as a potential technology has received extensive attention and rapid development [11, 13-18]. In the past two decades, the semiconductor photocatalysts have been greatly developed by a large number of scholars from various countries and regions. Many photocatalytic materials have been investigated, including metal oxides or sulfides, such as TiO₂ [19-28], CdS [29-38] and ZnCdS [15, 39-42], and the metal-free semiconductors such as rGO [43] and g-C₃N₄ [5, 44-54]. However, the photocatalytic activities of the most promising semiconductor photocatalysts are still not quite satisfied due to the easy recombination between charge carriers. In order to suppress the charge recombination, many strategies have been developed for effective separation of electron-hole pairs, including surface loading of cocatalysts [55-57], turning the morphologies [48, 58, 59], heterojunctions/homojunctions [60-62], molecular design [63-67] and heteroatom doping [5, 68]. Cocatalyst loading is an effective strategy to promote the photocatalytic activity of semiconductor photocatalysts, in which way the stability of the photocatalysts, the separation efficiency of the photogenerated charges, the number of exposed active sites and the light harvesting ability could be largely enhanced. Therefore, cocatalysts play an important role in improving the performance of photocatalytic H₂ evolution. Until now, a large number of strategies were developed to modify semiconductor photocatalysts by cocatalysts, such as Type Ⅰ heterojunction [69, 70], Type Ⅱ heterojunction [61], p-n heterojunction [62, 71, 72], Schottky-junction [73, 74], Z-scheme heterojunction [75-78] and S-scheme heterojunction [14], which have been widely designed to form build-in electric field for boosting the charge carrier separation, thus achieving improved photocatalytic hydrogen production.

In this review, we focus on the roles and mechanisms of cocatalysts developed to date for photocatalytic H₂ evolution, including (1) metals, alloys, metal phosphides, metal oxides, metal hydroxides, metal sulfides, carbon-based, etc., as single cocatalysts for H₂ evolution, (2) the synergetic effect of the dual cocatalysts loading for reduction reaction, (3) the effect of Z-scheme for water splitting, and (4) the improvement of H₂ production by loading MOFs as cocatalysts. We pay attention to the roles and mechanisms of various cocatalyst systems with important and representative research being discussed and summarized.

In general, the reaction mechanism of photocatalytic water splitting for H₂ production involves three major steps [79]: (1) absorption of light by semiconductor photocatalysts to generate electron (e^–)-hole (h⁺) pairs; (2) charge separation and migration to the surface of semiconductor photocatalysts; and (3) surface reactions for water reduction and oxidation, respectively. The efficiency of photocatalytic H₂ evolution is determined by the kinetics of the above three steps altogether (Eq. 1):

(1)

where η₁ is the efficiency of light absorption, η₂ is the efficiency of charge separation, and η₃ is the efficiency of surface catalytic reaction.

Generally, the semiconductor photocatalyst is composed of vacant conduction band (CB) and filled valence band (VB), which are separated from each other by E_g. The photocatalytic process begins with light absorption, when the photon energy is larger than or equal to the band gap (E_g), photon absorption in the photocatalyst causes electronic transition and generates e^–-h⁺ pairs, then the electrons in the VB are excited to the CB, while the holes are left in the VB. The electrons in the CB participate in the reduction reaction to produce H₂, and the oxidation reaction is driven by the holes in the VB. The minimum band edge of the CB has to be more negative than the redox potential of H⁺/H₂ (0 V vs. RHE), while the maximum band edge of the VB should be more positive than the chemical redox potential of O₂/H₂O (Fig. 1). However, if there are no suitable activity sites or reaction center on the surface of the semiconductor photocatalyst, the photogenerated electrons and holes could quickly recombine in the bulk of the semiconductor. According to the previous reports [9, 80, 81] that the recombination rate of photoinduced electrons and holes in the bulk of the catalysts is very fast (ps to ms). Therefore, the electrons and holes should be quickly captured by the cocatalysts for efficient separation and transfer of photogenerated charges and participate in the reduction reactions on the surface, which is critically important for photocatalytic water splitting.

In photocatalytic H₂ evolution system, the cocatalysts have four pivotal roles in enhancing the activity of the semiconductor photocatalysts (Fig. 2): (1) Cocatalysts can enhance the light harvesting. Some plasmon metals (such as Au and Cu) can enhance visible light absorption due to the localized surface plasmon resonance (LSPR) effect [82, 83]. (2) Cocatalysts can be effective in facilitating the separation of electron-hole pairs. When a cocatalyst is loaded on the surface of the photocatalysts, the cocatalyst can attract electrons, therefore the separation and transfer of the charge carriers can be enhanced [11, 84]. (3) Cocatalysts can provide active sites for the photocatalytic reaction. Most cocatalysts can not only serve as electron sinks, but also provide proton reduction sites [85, 86]. (4) Cocatalysts can improve the stability of the photocatalysts. Some metal sulfides can be oxidized easily. After loading cocatalysts, the self-decomposition of sulfides can be suppressed by extracting the photogenerated holes for oxygen evolution reactions [81, 87].

Deposition of metals on the surface of semiconductors is one of the effective ways to improve the photocatalytic performance of photocatalysts. The photogenerated electrons migrate to the metal surface under the action of electric field formed by the relatively low Fermi level of the metal. The schematic of charge transfer for metal/semiconductor under light irradiation due to the existence of the Schottky junction is illustrated in Fig. 3(a). When a metal is in close contact with a semiconductor, Schottky barrier would be formed and resulted in band bending in the semiconductor at the interface of the contact, and the electrons must have a higher energy to overcome the barrier [81, 88, 89]. The photogenerated electrons will transfer from the CB of the semiconductor to the metal through the Schottky barrier until their Fermi levels are equal, and the holes will hold in the VB of the semiconductor, thereby achieving effective separation of e^–-h⁺ pairs. The metal not only acts as a cocatalyst to separate photogenerated electrons-holes, but also provides an active site for the photocatalytic H₂ evolution. In this section, the application of transition metal phosphides materials as cocatalysts for photocatalytic H₂ production is introduced, and Table 2 and Table 3 summarize the application of metals and alloys as cocatalysts in the photocatalyst system for H₂ evolution.

Several noble-metals such as Pt [54, 90-96], Pd [97-99], Ag [100-103] and Au [24, 104-108] have been investigated in recent years. Pt, with the largest work function among many noble-metal, (Table 1), is the best cocatalyst for trapping electrons.

In 2016, Li et al. [90] reported that isolated single Pt atoms as a new form of cocatalysts deposited on g-C₃N₄ to enhance the photocatalytic H₂ evolution activity. Under visible-light irradiation, bare g-C₃N₄ gives a low activity with H₂ evolution rate of 6.5 μmol·h^–1. While with 0.16 wt.% Pt loading, Pt-CN shows a H₂ generation rate up to 318 μmol·h^–1. The photocatalytic performance of H₂ production can be improved by adding isolated single Pt atom to the intrinsic change of the g-C₃N₄ surface trap state induced by Pt atoms. Zhu et al. [54] concluded that the size of Pt co-catalyst can be controlled to form single atoms by inducing a small amount of Pt precursor, which plays a pivotal role in enhancing the photocatalytic performance of g-C₃N₄. As shown in Fig. 4, the average size of Pt on g-C₃N₄ is less than 0.2 nm after loading 0.1 wt.% Pt. With increasing the loading amount of Pt, the size of Pt clusters is increased. When the size of the Pt clusters is the smallest, the photocatalytic H₂ generation rate of Pt_0.1-CN is 473.82 μmol·mg^–1 under the visible-light irradiation. It is ascribed to the enhanced utilization efficiency of Pt atoms.

Liu et al. [96] synthetized Pt/g-C₃N₄ by changing the sequence of adding TEOA on photo-deposition of Pt. The sample named CNS-Pt-A and CNS-Pt-B that TEOA is added into the photocatalytic system after and before photo-deposition of Pt, respectively. Both CNS-Pt-A and CNS-Pt-B show better photocatalytic H₂ production activity as shown in Fig. 5, and the CNS-Pt-A shows a higher photocatalytic H₂ generation rate of 4210.8 μmol·h^–1·g^–1, which is attributed to the formation of coordination complex between TEOA and Pt⁴⁺, inhibiting the photo-reduction of Pt⁴⁺ to metallic Pt.

Metals are modified on the surface of semiconductors to form metal/semiconductor systems, which act as co-catalysts to provide active sites and trapping electrons to separate photoinduced electron-hole pairs. When the frequency of the incident photons matches well with the inherent frequency of the free electrons on the surface of metal nanoparticles, which will lead to a collective oscillation, and the metal nanoparticles will have a strong absorption of photon energy, this effect is commonly known as localized surface plasmon resonance (LSPR) as shown in Fig. 2(b) [89, 115]. As previously mentioned, the light absorption ability of photocatalyst is an important factor that affects the photocatalytic hydrogen production, the more light that is absorbed, the more photo-generated charges are produced to enhance the rate of photocatalytic H₂ evolution. Au [24, 105] and Cu [116, 117] are the most popular materials used for plasmonic cocatalysts to enhance light absorption.

Guo et al. [104] reported Au as cocatalyst on g-C₃N₄ nanosheets with Au nanospheres (NSs) and nanorods (NRs) of different plasmon resonance wavelengths. Fig. 6a shows the SEM and TEM images of the Au nanocrystals on g-C₃N₄ nanosheets with Au loading amount of 2.0 wt.%. With the size of the Au nanocrystals increasing, the plasmon peak of the Au/g-C₃N₄ sample moves toward the long wavelength due to the reduction of the effective refractive index of the surrounding environment. The optimal 18 nm-sized Au nanospheres/g-C₃N₄ nanostructure exhibits a superior activity, with the H₂ generation rate reaching 540 μmol·h^–1·g^–1 under visible light. The introduced plasmonic Au nanocrystals not only enhance the photocatalytic activity, but also endow the hybrid photocatalysts with an extended light absorption range. Meanwhile, the improved H₂ production rates can be attributed to the LSPR effect of Au.

Although noble metal cocatalysts own high photocatalytic activity, their practical applications are limited by some factors such as resource shortages and high prices. Non-noble metals are rich in the earth's crust and some of their properties are similar to noble metals in enhancing the photocatalytic performance. Furthermore, non-noble metals also possess high work function as noble metals (Table 2), therefore non-noble metals play an important role as co-catalysis to trap and transfer electrons in photocatalytic reaction. Recently, non-noble metals such as Zn [118], Ni [119-123] and Co [124] have been utilized to effectively driven photocatalytic water splitting. Among the non-noble metals, the work function of Ni is closest to Pt, even higher than Au, therefore Ni is the most promising substitute for noble metals as co-catalysts for photocatalytic H₂ evolution. In 2016, Dong's group [121] successfully synthetized Ni/g-C₃N₄ photocatalyst, in which the non-noble metal Ni as cocatalyst was loaded on the surface of g-C₃N₄ nanosheets for photocatalytic water splitting under sun-light irradiation. In the absence of cocatalyst, the pure g-C₃N₄ nanosheets showed a low activity of photocatalytic hydrogen evolution due to the strong recombination trend of electrons and holes. When the loading content of Ni was 7.40 wt.%, the rate of H₂ evolution was 4318 μmol·h^–1·g^–1 with 2.10% AQY at wavelength of 400 nm, and kept stable after 48 h of irradiation. It is worth noting that the content of Ni²⁺ in the photocatalytic system B is the same as that in photocatalytic system A, as shown in Fig. 6a. Comparing with the rate of H₂ evolution, the photocatalytic activity of system A was 2.45 times higher than that of system B. The reason for this phenomenon is that Ni²⁺ and TEOA (as sacrifice reagent) could generate a new compound [Ni(TEOA)₂]Cl₂, which acts as the main co-catalyst, not metal Ni in the case of photocatalytic system B. Therefore, non-noble metal Ni could be helpful for the transfer of photogenerated electrons and high efficiency of photocatalytic H₂ evolution.

Metallic copper is common in our lives because of its earth-abundance and have been used as co-catalyst in different photocatalytic reactions. Some research groups reported that Cu nanoparticles loading on some semiconductors formed Schottky contact. To date, Cheng and co-worker [28] developed a highly efficient photocatalysis system for water splitting by loading metal Cu on TiO₂ microrods. As shown in Fig. 7(a, b), the TiO₂-Cu-2.5 wt.% hybrid photocatalyst exhibited superior rate of hydrogen evolution with 6046 μmol·h^–1·g^–1, which is 23 times higher than that of pure TiO₂ microrods. The significant increase in photocatalytic performance is due to the Schottky junction formed between metal Cu and TiO₂ microrods as shown in Fig. 7(c, d). The work function of Cu is larger than that of TiO₂, the e^– would transfer from TiO₂ to Cu. The movement of electrons causes the Fermi levels of metallic Cu and TiO₂ to reach the same level. In the equilibrium state, a depletion layer is formed at the TiO₂-Cu interface, wherein the Cu metal is negatively charged, and TiO₂ is positively charged due to electrostatic induction near its surface. The electric field would be formed between Cu and TiO₂, the photoexcited electrons would be difficult to pass through the Schottky barrier to recombination with holes, so that the electrons and holes could be efficiently separated to improve the rate of H₂ evolution. Therefore, metal Cu which is low-cost and earth-abundant could replace the noble metals as co-catalyst for photocatalytic water splitting.

Similar to mono-metallic, metal alloys such as PtPd [127], PtCo [128], AuPd [129, 130], AgCu [131], NiCd [132], NiMo [133] and NiCo [134] have also been synthesized and reported in the field of photocatalytic H₂ evolution, which act as co-catalysts to trap electrons from the semiconductor and improve the charge separation and transfer efficiency [89]. Compared with single-metal, alloy co-catalysts commonly show higher activity in photocatalytic H₂ evolution, which is mainly due to the synergistic effect between the two metals [135]. Bimetallic cocatalysts such as AuCu [135] and AuPd [129, 130], which contain plasmonic metals also possess LSPR effect.

In alloy NPs, the activity could be influenced by the different atomic arrangements, as exhibited in Fig. 8(a). Ding and co-workers [135] reported that the photocatalytic H₂ evolution rate of AuCu alloy nanoparticles/CaIn₂S₄ composite is up to 452.8 μmol·h^–1, which is 2.2 times higher than the AuCu core-shell structure/CaIn₂S₄ composite. The catalytic mechanism for charge separation is outlined in Fig. 8(b). The photocatalytic performance enhancement of AuCu alloy nanoparticles is better than that of AuCu core-shell structure due to the shorter distance of photoelectrons transfer from CaIn₂S₄ to Au, thus to Cu. Moreover, the existence of Cu shell in AuCu core-shell nanoparticles may decrease the visible absorption of Au core because the incident light can be scattered by Cu shell. Therefore, the AuCu alloy nanoparticles as cocatalyst is better than that of AuCu core-shell structure.

Our group [127] reported that PtPd alloy nanoparticles as cocatalyst were deposited on g-C₃N₄ nanosheets to enhance the photocatalytic H₂ evolution activity under visible-light irradiation as shown in Fig. 9. The 0.2 wt.% Pt_0.5Pd_0.5/g-C₃N₄ sample presented the highest H₂ evolution activity (1600.8 μmol·g^{– 1}·h^{– 1}), which is 800 times higher than that of pure g-C₃N₄. When K₂HPO₄ was added to the photocatalytic system, the H₂ production rate of the Pt_0.5Pd_0.5/g-C₃N₄ composite sample reached 2885.0 μmol·g^–1·h^–1. The AQY of Pt_0.5Pd_0.5/g-C₃N₄ composite sample at 420 nm was determined to be 5.5%. Therefore, our experimental results demonstrated that PtPd alloy NPs were a promising cocatalyst for g-C₃N₄, which can also be regarded as a potential functional cocatalyst in others semiconductor photocatalysts for enhancement of the photocatalytic H₂ production activities.

The high cost of noble-metals limited their application in industrial scale water splitting for H₂ evolution. Hence, it is necessary to choose a suitable photocatalyst with natural abundance cocatalyst that constructing an efficient material for water splitting. Non-noble metal elements of Ni or Co show the similar characteristics with noble Pt. Our group [128] reported that PtCo bimetallic alloy as cocatalyst to modified on g-C₃N₄ nanosheets via an in-situ chemical deposition method, which replace part Pt with Co to reduce the high cost. The 1.0 wt.% PtCo/g-C₃N₄ showed the highest rate of photocatalytic H₂ evolution of 960 μmol·g^–1·h^–1, which is 2.9 times higher than that of pristine bulk PtCo/g-C₃N₄ under visible light irradiation. Similarly, Peng et al. [138] found that the PtNi bimetallic alloy decorated on g-C₃N₄ nanotubes could improve the photocatalytic activity, and the H₂ generation rate is 104.7 μmol·g^–1·h^–1, which is about 47.6 and 1.1 times higher than that of pure g-C₃N₄ nanotubes and Pt/g-C₃N₄, respectively. Hence, PtCo and PtNi bimetallic alloy could replace Pt as a promising cocatalyst for photocatalytic H₂ evolution with better activity and lower cost.

For further cost reductions, the choice of two metals in the alloy are all non-noble metals. In 2018, Zhang's group [134] reported that Ni_0.8Co_0.2 alloy as cocatalysts decorated on Zn_0.75Cd_0.25S with optimized H₂ evolution rate of 211.4 mmol·h^{– 1}·g^–1 and AQY of 13.3% at 365 nm. Besides, the NiCd bimetallic nanoparticle can also be used as cocatalyst to enhance the rate of H₂ evolution with dropping on CdS nanorods which reported by Liu and co-workers [132]. The NiCd/CdS NRs sample exhibited a palpable improved in photocatalytic H₂ production performance, the optimization rate of H₂ evolution was determined to be 11.57 mmol·h^–1·g^–1, which is about 64.8, 17.2 and 2.3 times that of pristine CdS NRs, Cd/CdS and Ni/CdS, respectively.

Among the metal-based cocatalysts, metal phosphides are currently broadly studied in photocatalytic fields because of its good stability and high activity. Among the so many developed metal phosphides, metal phosphides composed by the elements of transition metals such as Co, Ni and Fe are getting more research attention. Firstly, phosphorus is abundant in the earth, which is easy to obtain and in low cost. Secondly, compared with other metal-based materials, the transition metal phosphides are easier to synthesize. Thirdly, the P-ligand in the phosphide (namely " the ensemble effect" ) has a moderate binding strength to hydrogen, which is important for the formation of H₂ molecules and the decrease of H-poisoning effects [142, 143]. Typical single transition metal phosphides including CoP [16, 41], Co₂P [144], Ni₂P [42], Ni₁₂P₅ [145] Cu₃P [146], FeP [147], etc.; multiple transition metal phosphides are mainly NiCoP [148]. In this section, the application of transition metal phosphide materials as cocatalysts for photocatalytic H₂ production is introduced, and Table 4 summarizes the application of non-metals as cocatalysts in photocatalyst system for H₂ evolution.

Han et al. [16] have synthesized CoP/g-C₃N₄ photocatalysts by in-situ chemical deposition for enhanced photocatalytic hydrogen production performance. The maximum rate of hydrogen production is 9.2 μmol·h^–1. In recent years, our group always committed to explore the transition metal phosphides as cocatalysts for H₂ evolution. Noble-metal-free photocatalyst Co₂P-CdS was synthesized via a facile in situ hydrothermal method [144]. The Co₂P-CdS sample with a loading ratio of 1.2 mol% Co₂P gave the highest H₂ production rate of 0.303 mmol·h^–1, after further introduction of K₂HPO₄ as sacrificial agent, the H₂ production rate reached 0.356 mmol·h^–1, and the AQE of the Co₂P-decorated CdS sample is about 13.88% at 420 nm. The application of Co₂P as co-catalyst has significantly improved the photocatalytic H₂ evolution performance of CdS. When CdS and Co₂P form intimate contact, the electrons transfer from Co₂P to CdS, which leads to a downward bending of the energy bands of CdS, the separation and transfer of photo charges can be boosted as a result of the band bending formed at the interface, which leads to a dramatically enhanced performance of photocatalytic H₂ production (Fig. 10).

Sun et al. [149] tested the performance of three different crystalline phases of Ni₂P, Ni₁₂P₅, and Ni₃P as cocatalysts to hybridize with g-C₃N₄ for photocatalytic H₂ generation in 2017. As shown in Fig. 11, all three phases of Ni_xP_y work as eximious cocatalysts for enhancement of visible light H₂ generation with g-C₃N₄. The highest H₂ generation activities by achieved with 5 wt.% Ni₂P/g-C₃N₄ was 0.162 mmol·h^–1·g^–1, which was approximately 34 times of the amount obtained on pure g-C₃N₄, and the rate was close to 5 wt.% Pt/g-C₃N₄ (0.168 mmol·h^–1·g^–1). In addition, the photocatalyst exhibited favorable stability after 9 hours illumination under visible light irradiation.

Dong's group [150] reported that noble-metal-free iron phosphide (Fe_xP) cocatalyst decorated graphitic carbon nitride (g-C₃N₄) as photocatalysts by two-step hydrothermal and phosphidation method for the highly efficient and stable H₂ evolution from water splitting under visible light irradiated. The g-C₃N₄/Fe_xP-0.08 showed the optimal rate of photocatalytic H₂ generation, the rate reached up to 166.4 μmol·h^–1·g^–1, which was about 277 and 139 times that over pristine g-C₃N₄ and g-C₃N₄/FeOOH precursor, respectively. The Fe_xP as cocatalyst significantly accelerates the separation and transfer of photogenerated charges which lead to the highly efficient photocatalytic H₂ evolution activity of g-C₃N₄.

In 2017, Qin et al. [148] synthetized NiCoP core/shell NPs as cocatalyst by a solid-state phosphorization reaction. NiCoP@NiCo-Pi/g-C₃N₄ exhibited much higher hydrogen production rate than Ni₂P@NiCo–Pi/g-C₃N₄ and CoP@NiCo-Pi/g-C₃N₄, which demonstrated better performance of NiCoP cocatalysts for photocatalytic H₂ evolution than the single transition metal phosphides.

Several metal oxides such as TiO₂ [158-161], CuO [22, 162-165], Cu₂O [49, 166, 167] and NiO [168] are well known as cocatalysts for photocatalytic H₂ production. Table 5 summarizes the application of metal oxides and hydroxides as cocatalyst in the photocatalyst system for H₂ evolution. Recently, Tan et al. [160] reported the nanostructured g-C₃N₄/TiO₂ was synthesized by controlling the growth of nanoscale g-C₃N₄ on TiO₂ NPs via one step vapor deposition method. The CN/TiO₂-24 composite shows the greatest H₂ evolution rate of 513 μmol·g^{– 1}·h^–1. Under visible light irradiation, the visible-light-excited electrons in the CB of g-C₃N₄ can transfer to the CB of TiO₂, the photoinduced charge could be efficiently separated, and therefore improve the photocatalytic H₂ evolution of the CN/TiO₂-24 composite.

For transition metal oxides, CuO and Cu₂O have also been proven to be beneficial for photocatalytic H₂ production. Wang and co-worker [167] reported that Cu₂O loading on TiO₂ displayed 14.48 times larger photocatalytic H₂ evolution activity than pure P25. Yu's group [162] found CuO can act as an effective cocatalyst to improve the photocatalytic H₂ production activity of TiO₂. When the loading content of CuO in the photocatalyst was increased to 1.3 wt.%, the rate of H₂ evolution reached 2061 μmol·h^–1·g^–1 with an AQY of 13.4% at 420 nm. The quantum size effect of CuO influences the electron transfer and its valence band edges, which enhance the photocatalytic activity. As shown in Fig. 12(a, b), the bulk CuO has a less negative potential than H⁺/H₂ potential. As the quantum size of CuO increases, the band gap decreases and the CB and VB edges are also shifted to the positions which permit transfer of photogenerated electrons, facilitating the interfacial electron transfer from CuO to H⁺ in the solution for photocatalytic H₂ evolution.

In 2015, Wu et al. [170] synthesized Ag₂O/g-C₃N₄ via facial hydrothermal reaction, the composite photocatalyst shows efficient properties for hydrogen evolution from water splitting under visible light irradiation, in which the rate of hydrogen evolution is 33.04 μmol·h^–1. For the mechanism, the Ag₂O as a cocatalyst plays an important role in charges transferred of photocatalytic progress. The reason relies on the relative position of CB and VB between g-C₃N₄ and Ag₂O, in which the CB of Ag₂O is lower than the redox potential of H⁺/H₂, so that the photogenerated electrons of g-C₃N₄ cannot be transferred to the CB of Ag₂O, which could be used to photocatalytic hydrogen evolution reaction. While the photogenerated holes in the VB of g-C₃N₄ can be transferred to the VB of Ag₂O, and the holes are consumed by TEOA. Thereby achieving the purpose of charge orientation separation.

Inspired by the research of NiO modified TiO₂, Yu's group reported that Ni(OH)₂ as cocatalysts decorated on TiO₂ [19], g-C₃N₄ [46] and CdS [171] for enhanced photocatalytic H₂ production activity which was synthetized by a simple precipitation reaction. Ni(OH)₂ is an outstanding and universal material as cocatalysts for photocatalytic water splitting under visible-light irradiation. The rate of H₂ evolution even is higher than Pt as cocatalyst deposited on CdS.

Our group [169] reported that plate-like Co(OH)₂ was incorporated onto the surface of TiO₂ nanosheets as an effective cocatalyst to increase the rate of H₂ production for water splitting. The sample of 0.5 mol% Co(OH)₂/TiO₂ shows a maximum hydrogen rate with 746.93 μmol·g^–1·h^–1, which is 20 times higher than pure TiO₂ nanosheets, along with a good stability for photocatalytic H₂ evolution upon continuous irradiation of 14 h. As shown in Fig. 12(c, d), the reason for photocatalytic performance improvement is the plate-like Co(OH)₂ act as electron collector to facilitate the migration of electron from TiO₂ to the plate-like Co(OH)₂, and reduce partial Co 2p to Co atoms, the as-formed Co⁰ species can work as a co-catalyst, therefore improve the separation efficiency and lifetime of the photoexcited electrons and holes.

The layered double hydroxide (LDH) can be represented by the general formula [M_1-x²⁺M³⁺_x(OH)₂](A^n‒)_x/n·mH₂O, where M²⁺ and M³⁺ are divalent and trivalent metal cations respectively and A^n‒ is the interlayer anion compensating for the positive charge of the brucite-like layers [172-176]. LDHs materials have been studied in photocatalysis due to their low cost, high photocatalytic activity and controllable electronic structure. For photocatalytic water splitting H₂ evolution, Zhang et al. [177] synthesized CdSe/ZnCr-LDH nanohybrids via an electrostatically derived self-assemble process, which exhibited a high photocatalytic hydrogen generation rate of 2196 μmol·h^–1·g^–1 with an AQY of 51.3% for H₂ generation under visible-light irradiation. The ZnCr-LDH as cocatalyst was also applied with g-C₃N₄, which is effective in enhancing the photocatalytic hydrogen generation activity, the H₂ production rate reached 155.7 μmol·h^–1·g^–1. ZnCr-LDH acts as an important role in improving the activity of photocatalytic hydrogen production by water splitting. As shown in Fig. 13, the electrons transfer from the CB of g-C₃N₄ to the lower-lying CB of ZnCr-LDH, the holes in the VB of ZnCr-LDH move to the upper-lying VB of g-C₃N₄, leading to an efficient separation of electrons and holes [178].

To date, many metal sulfides, such as MoS₂ [30, 184-188], NiS [45, 51, 184, 189], Cu_xS [190-196], Co₃S₄ [197], FeS [198], Ag₂S [199], Bi₂S₃ [200], WS₂ [201], Co₃Mo₇S [15] and NiMoS [202, 203], have been reported in photocatalytic H₂ evolution reaction. MoS₂ is an efficient cocatalyst on variety of semiconductor photocatalysts. Meng's group [184] load MoS₂ on ZnIn₂S₄ via hydrothermal method. Loading MoS₂ on ZnIn₂S₄ as a cocatalyst improved the rate of H₂ evolution by up to 28 times. In 2016, Li and co-worker [55] reported on the synthesis of novel bilayer junctions by assembling metallic MoS₂ monolayers selectively and chemically on oxygen-deficient (Bi₁₂O₁₇) end-faces of Bi₁₂O₁₇Cl₂ monolayers via oxygen-vacancy (OV) chemistry. Under visible light irradiation, the photocatalytic H₂ evolution rate reached 33 mmol·h^–1·g^–1 with an AQY of 36% at 420 nm. Furthermore, the band structures of Bi₁₂O₁₇Cl₂ and MoS₂ were shown in Fig. 14, Bi₁₂O₁₇Cl₂ in the photoexcited Bi₁₂O₁₇Cl₂-MoS₂ generated electrons and transferred from Bi₁₂O₁₇Cl₂ to MoS₂ in photocatalytic H₂ evolution process. In addition, in Bi₁₂O₁₇Cl₂, internal electric field can drive electrons to (Bi₁₂O₁₇) end-faces and the holes to (Cl₂) end-faces, achieving the atomic-level steering of charge separation.

Very recently, Xu et al. [188] found that MoS₂ could be facilely modulated through phosphorus doping, which can act as a cocatalyst and embedded onto the CdS nanorod for not only facilitating the separation of the charge carriers, but also promoting the separation efficiency of the charge carriers. The 3 wt.% CdS/P-MoS₂ sample exhibited a higher photocatalytic hydrogen evolution rate of 8.86 mmol·h^–1·g^–1 with the AQY of 32.4% at 420 nm.

In most reports, the NiS cocatalysts are generally decorated on semiconductors to synthetize hybrid photocatalysts. The NiS/g-C₃N₄ heterojunction was formed by one-step calcination method [189], hydrothermal method [45] and ion-exchange procedure [51]. All of the photocatalysts exhibited good photocatalytic hydrogen production performance in TEOA solution under visible light illumination, which is 250 times and 56 times higher than pure g-C₃N₄, respectively. This remarkable enhancement in the photocatalytic activity resulted from NiS cocatalyst, which acted as an important role in the NiS/g-C₃N₄ photocatalyst by boosting the electron separation and transfer from the CB g-C₃N₄ to the surface NiS, the reactions are presented in equations (2) and (3):

(2)

(3)

Furthermore, NiS was also decorated on TiO₂ [21], CdS [204, 205] and CdZnS [206] to enhance the catalytic activity.

The CuS/ZnS photocatalyst was prepared via a simple hydrothermal and cation exchange reaction between ZnS[en]_0.5 nanosheets and Cu(NO₃)₂ by Zhang in 2011 [191]. The optimal CuS loading content on ZnS is 2 mol% and the corresponding rate of H₂ production is 4147 μmol·h^–1·g^–1 with an AQY of 20% at 420 nm. As shown in Fig. 15(a), the photoinduced interfacial charge transfer (IFCT) from the ZnS VB to CuS causes partial reduction of CuS to Cu₂S, the holes in the VB of ZnS could be consumed by the sacrificial agents. The IFCT will retard the recombination of photoinduced electrons and holes due to the space separation and lead to H₂ evolution. The CuS/Cu₂S acts as an electron sink and cocatalyst to boost the separation and transfer of electrons from ZnS VB to CuS/Cu₂S, where H⁺ is reduced to H₂. The steps of electron transfer in this catalytic H₂ evolution process under visible light irradiation are listed is as the following equations:

(4)

(5)

(6)

In 2017, Zhang and co-workers [190] reported that a self-optimizing bifunctional core-shell CdS/Cu₂S heterojunction was synthesized via solvothermal method for photocatalytic hydrogen evolution under visible light irradiation in which the Cu₂S acts as cocatalyst. Compared with pure CdS, the rate of photocatalytic H₂ production has improved almost 25 times, the maximum rate of hydrogen production reached 14.4 mmol·h^–1·g^–1 with an AQY of 19.5% at 420 nm. Fig. 15(b, c) illustrated the mechanism for photocatalytic H₂ production by CdS/Cu₂S. Under visible light irradiation, e^–-h⁺ pairs were generated in CdS core, partial photo-electrons participate in H₂ evolution, while the others reduce Cu₂S into Cu⁰. The electrons transfer from the CB of CdS to Cu₂S/Cu⁰, the Cu₂S acts as active sites for photocatalytic H₂ production, while the light-reduced Cu⁰ accelerates the separation process of photo-generated electrons and holes by forming heterojunction.

In recent years, many studies have reported that the bimetallic sulfide play an important role in enhancing the photocatalytic H₂ evolution. Yu's group [207] found that the CoMoS co-catalyst could significantly improve the H₂ production activity of CdS photocatalyst, the rate of photocatalytic H₂ evolution is 178.5 μmol·h^–1, which is higher than that of pure CdS, MoS₂/CdS and Co/CdS by 26.3, 3.1 and 7.5 times, respectively. Furthermore, the CoMoS co-catalyst acts as a universal cocatalyst to improve the photocatalytic H₂ evolution activity of other semiconductors, such as TiO₂ and g-C₃N₄, their corresponding rate of is shown in Fig. 16. The exceptional visible light photocatalytic H₂ evolution performance is due to the presence of CoMoS, which acts as an efficient cocatalyst to prompt the transportation of the electrons for H₂ production, the detail process is described as the following 3 steps: (1) under visible light irritation, the CoMoS first work as an electron capturer to prompt the transportation of the electrons from the CB of CdS; (2) then play as an electron transfer mediator to efficiently transport the electrons; (3) finally serves as the activity site for H₂ production to boost the rate of photo-reduction of water.

Similar to MoS₂, Cu_xS, NiS and CoMoS_x, there are also many diverse metal sulfide cocatalysts which can decorated on semiconductors to effectively increase photocatalytic H₂-evolution activity, such as Ag₂S [199], Bi₂S₃ [200], NiMoS [202, 203], Co₃Mo₇S [15], Zn_0.23Cd_0.77S [208] and ZnIn₂S₄ [209], the application of metal sulfide co-catalysts for photocatalytic H₂ evolution is summarized in Table 6.

Common cocatalysts are mostly metal or their compounds loaded on semiconductor photocatalysts. However, metal-free carbon nanomaterials, which have many unique properties to be applications in the photocatalysts field, such as large work function, excellent electron receptors, large surface area and novel morphologies. In addition, they are very cheap, stable, abundant in earth and non-toxic. Many studies have shown that carbon-based materials can be used as an effective cocatalyst or an excellent conductor cocatalyst to enhance the activity of catalytic hydrogen production. Carbon-based materials including carbon quantum dots (CQDs) [27, 211-213], graphene [116, 214, 215] and carbon nanotube (CNT) [50, 216-221], and carbon-based material can act as cocatalysts in photocatalytic hydrogen production process, the related researches were summarized and showed in Table 7.

The CQDs were extensively studied due to some superior performances include the long-term stability, tunable size and band gap and size dependent photoluminescence. In 2018, Wang and co-workers [213] prepared tubular CQD-implanted g-C₃N₄ heterostructures (CCTs) via thermal polymerization of freeze-dried urea and CQDs precursor. The photocatalyst exhibited efficient photocatalytic activity with a high H₂ evolution rate of 3538.3 μmol·h^–1·g^–1 under visible-light irradiation and AQY was 10.94% at 420 nm. The integration of CQDs into CCTs optimizes the morphology and band structure, increases visible light absorption and reduces the energy barrier, thus improving the photocatalytic efficiency for H₂ evolution. Tang's group [211] reported the CQDs decorated g-C₃N₄ nanosheets via hydrothermal method. The photocatalytic H₂ evolution ability was tremendously improved by loading 10 wt.% CQDs as a cocatalyst. This study showed that CQDs/g-C₃N₄ displayed a maximum photocatalytic H₂ evolution rate of 219.5 μmol·h^{– 1}·g^{– 1} (UV-Vis) and 50.5 μmol·h^–1·g^–1 (Vis). In addition, the 10 wt.% CQDs/g-C₃N₄ has an AQY of 0.024% at 700 nm and 0.034% under 808 nm laser, the wavelength dependence of EQE clearly shows the extended active range toward the NIR region derived from the higher content CQD loading. The broadband absorption and high photocatalytic activity originate from the interaction of spectral and electronic spectrum, coupling these interactions helps to promote the wide range absorption and photoinduced charge separation Very recently, Sui et al. [27] found that the CQDs modified TiO₂ with a majority of (001) facet via a facile method. The CQDs/TiO₂-001 exhibited a superior photocatalytic H₂ evolution ability. CQDs loading on the surface of TiO₂-001 plays an important role of electron reservoir to trap photogenerated electron, which not only promotes efficient separation of electron-hole pairs, but also prevents the recombination of the photogenerated electron-hole pairs.

Graphene (GO), a two-dimensional macromolecular sheet of carbon atoms with a honeycomb structure, which has a wide applications in photocatalysis due to the outstanding performance with high thermal conductivity, excellent mobility of charge carriers and extremely high specific surface area [222-224]. Yu's group [214] reported the fabrication of graphene/g-C₃N₄ by combined impregnation-chemical reduction strategy. The optimal graphene content was 1.0 wt.% and the corresponding photocatalytic H₂ evolution rate was 451 μmol·h^-1·g^-1. The graphene acted as an acceptor of the photogenerated electrons, which could enhance the photocatalytic H₂ evolution under visible light irradiation. Zhang et al. [116] reported an in-site process for preparation of Cu NPs on rGO nanosheet. The Cu/rGO composite exhibited superior photocatalytic H₂ evolution activity which possesses a hydrogen evolution rate of 59 mmol·h^–1·g^–1 in lactic acid solution under visible light irradiation. Significantly, the Cu/rGO composite still have hydrogen production rate of 2.76 and 1.96 mmol·h^–1·g^–1 with the wavelength of 800 nm and 900 nm, respectively. The rGO co-catalyst acts as an electron mediator for enhancing the photoinduced carrier separation efficiency and promote the restrain of e^–-h⁺ recombination. To date, Zhang et al. [215] have prepared Cu₂O-RGO heterojunction with oxygen vacancy through hydrothermal method. Compared with pure Cu₂O, the rate of 2 wt.% Cu₂O-RGO heterojunction photocatalytic H₂ evolution was 64.3 μmol·h^–1·g^–1, which was 5 times higher than that of pure Cu₂O (12.8 μmol·h^–1·g^–1). The RGO as co-catalyst loading on Cu₂O to form heterojunction could enhance a broader absorbance under visible light and promote a faster electron transfer, which lead to the improved photocatalytic H₂ production activity.

Recently, Reddy and co-workers [216] prepared CNT/TiO₂ has enhanced photocatalytic H₂ evolution. After loading CNT on TiO₂, the photocatalytic H₂ evolution rate could be increased from 1.3 to 8.8 mmol·h^–1·g^–1. Yu et al. [219] investigation revealed that the CNT could serve as a highly efficient co-catalyst for Cd_0.1Zn_0.9S to promote the photocatalytic H₂ production under visible-light irradiation. The optimum rate of H₂ evolution is 78.2 μmol·h^–1 with an AQY of 7.9% at 420 nm, exceeding that of pure Cd_0.1Zn_0.9S by more than 3.3 times. Chen's group [50] found that introduce 0.2 wt.% CNT into g-C₃N₄ could enhance its photocatalytic activity for H₂ evolution under visible-light irradiation, the highest H₂ evolution rate is 39.4 μmol·h^–1, which is 2.4 times higher than the value obtained with pure g-C₃N₄. In 2012, Lou's group [218] reported that CNT@Ni₃S₂ nanostructures for photocatalytic H₂ evolution which composed of Ni₃S₂ nanosheets grown on carbon nanotube (CNT) backbone via multi-step transformation route, the synthesis procedure of CNT@Ni₃S₂ nanostructures is shown in Fig. 17. Under visible light irritation, the rate of photocatalytic H₂ evolution is 5.32 mmol·h^–1·g^–1 with an AQY of 11.1% at 420 nm, which is superior to pure Ni₃S₂. The high photocatalytic hydrogen production activity is due to the sensitization effect of ErY for effective visible light absorption, which would then be transformed into excited stated. The photoexcited electrons in ErY are quickly transferred to the Ni₃S₂ nanosheets due to the presence of the CNT backbone. Then the protons partake in the reduction reaction to product H₂ on Ni₃S₂ (Fig. 17). In this photocatalytic system, CNT not only acts as co-catalyst to collect and transfer electrons, but also provides the active sites for H₂ evolution.

Efficient charge separation is an effective way to improve the photocatalytic hydrogen evolution performance. The hydrogen production efficiency of water splitting over single photocatalyst is limited by the rapid recombination of photoexcited charge carriers. Cocatalyst loading can effectively accelerate the surface reaction to promote the water splitting process [225]. The electron-hole pairs are separated under the light irradiation, and the photoexcited electrons are the main active species that are used to reduce the proton to H₂. Various reductive cocatalysts have been broadly studied to transport electrons to water protons. Noble metals such as Pt and Pd with higher work function and lower impedance are usually primary candidates to act as reductive cocatalysts to transfer the electrons to protons. The higher work function equally to the lower Fermi level results in the separated electrons inclines to the noble metals, while the lower impedance can insure a faster electron mobility, which can prompt the photoinduced electrons transfer to the water proton. However, noble metals are suffering from high cost and low abundance in the earth. The emergence of various non-noble species has played an important role in the development of reductive cocatalyst. The nucleus on photocatalytic H₂ production are concentrated on the half reduction reaction of water splitting. The holes are unnecessary factor to charge separation. Oxidative cocatalyst, such as some metal oxides, CoO_x, IrO₂ etc., has a tendency to capture holes, while reductive cocatalysts acting as electrons trapping agents to pull photoinduced electrons away from the photogenerated charges [29, 226-228]. The applications of dual-cocatalysts photocatalyst systems for H₂ evolution are summarized in Table 8.

In 2016, Yu's group [229] reported that synergistic effect of dual amorphous cocatalysts, in which the Ti(Ⅳ)-hole cocatalyst and Ni(Ⅱ)-electron cocatalyst were loaded on CdS to improve the photocatalytic H₂ evolution activity under visible light irradiation. The maximum photocatalytic H₂ evolution rate is 3437.95 μmol·h^–1·g^–1, which is 2.1 times higher than that of pure CdS. The reason for the high performance is that when both of Ti(Ⅳ) and Ni(Ⅱ) cocatalysts are simultaneously modified on the CdS surface, the Ni(Ⅱ) cocatalyst can first act as an effective trapping site to accept the electrons, and then act as the reduction active site to boost the H₂ production reactions on the surface, while the Ti(Ⅳ) cocatalyst act as hole cocatalyst to quickly transfer the holes and boot the oxidation reaction (Fig. 18).

Wang et al. [230] reported the co-loading of graphene quantum dots (GQDs) and PdS dual cocatalysts on ZnCdS surface by hydrothermal and in-situ chemical deposition method for efficient photocatalytic H₂ evolution. The GQDs/ZnCdS/PdS hybrids exhibited the highest H₂ evolution rate of 517 μmol·h^–1 with an AQY of 22.4% at 420 nm, which were 15, 7 and 1.7 times higher than that of pure ZnCdS, GQDs/ZnCdS and ZnCdS/PdS, respectively.

Li's group [32] synthesized ternary structured photocatalyst which is composed by co-loading dual earth-abundant cocatalysts of metallic carbon black(CB) and NiS₂ on CdS via solvothermal and impregnation method. The superior photocatalytic H₂ evolution activity of CdS/CB/NiS₂ is 166.7 μmol·h^–1. As illustrated in Fig. 19(a), under visible light irradiation, the electrons are excited from the VB to the CB of CdS, then the electrons move to the carbon black due to its intimate interface contact and matched work functions, in which the carbon black acts as an electron mediator to transfer the collected electrons to the H₂ evolution active sites on NiS₂ owing to its excellent electrical conductivity and carrier mobility properties, leading to the effective separation and transfer of the electron-hole pairs. The effective separation of electron-hole pairs is the key factor in improving the photocatalytic hydrogen production activity.

The charge separation can be further promoted by loading reductive and oxidative cocatalysts [232]. The rational design of spatially-separated dual cocatalysts on photocatalysts has aroused an increasing attention. In most cases, the cocatalysts were randomly distributed on the surface of semiconductors, resulting in a random flow of the photogenerated electrons and holes and severe charges recombination. Li's group [233] reported an Pt-PdS/CdS photocatalyst in 2009. Visible-light-driven hydrogen production with the rational simultaneously design of oxidative and reductive cocatalysts could further enhance the photocatalytic activity. Wang et al. [231] have demonstrated the efficiency of Pt/TiO₂/CdS/Co₃O₄ in hierarchical structure, Pt and Co₃O₄ species were spatially decorated on the inner and outer surface of the TiO₂/CdS heterojunction shell. Its hydrogen production rate is 2000 μmol·h^–1·g^–1 under visible light irradiation, which is over 540 times higher than that of TiO₂/CdS double-shelled hollow spheres under the same conditions. The outstanding H₂ evolution rate can be attributed to the efficient separation of photogenerated carriers, transportation and interface reaction. The ultrathin shell thickness could contribute to the fast electron transportation and the synergistic effect of the separately loaded dual cocatalysts (Fig. 19(b)).

Sun et al. [234] synthesized a novel hollow porous CdS photocatalyst with Pd and PdS nanoparticles spatially distributed on the inner and outer surfaces by using microporous zeolites as a hard template. The Pd@CdS/PdS sample exhibits superior visible-light-driven photocatalytic H₂ evolution rate up to 144.8 mmol·h^–1·g^–1. This is the highest values among all of the reported CdS-based catalysts. This synthetic approach may be used to fabricate other highly efficient catalysts with spatially separated cocatalysts. Furthermore, the remarkably enhanced H₂ evolution rate can be result from the spatially separated reaction sites in the Pd@CdS/PdS sample, in which the metallic Pd pulls the photoexcited electrons away from CdS while PdS pushes the holes for more thorough oxidation of the sacrificial agent.

In 2016, Zhang and co-worker [235] reported that CoO_x/TiO₂/Pt photocatalysts produced by template-assisted atomic layer deposition for photocatalytic hydrogen evolution with CoO_x and Pt dual cocatalysts. Fig. 20 showed the reaction process of CoO_x/TiO₂/Pt photocatalysts with porous hollow TiO₂ nanotubes as a semiconductor, which the Pt and CoO_x deposited on the inner and outer surfaces of the nanotubes, respectively. The highest rate of 0.019 wt.% CoO_x/TiO₂/0.046 wt.% Pt photocatalysts is 275.9 μmol·h^–1, which is almost 5 times higher than that of porous hollow TiO₂ nanotubes. The spatially separated dual Pt and CoO_x cocatalysts not only facilitate electron-hole separation but also provide the activity sites for photocatalytic H₂ evolution on the surface.

An artificial Z-scheme photocatalyst system was inspired by the natural photocatalytic process, which was first reported by Bard in 1979 [252]. The schematic illustration of the redox-mediator Z-scheme system is shown in Fig. 21(a), there is no direct contact between the two semiconductors in the Z-scheme system, and the charge carrier transfer is achieved by ion redox shuttles in solution. Both the two semiconductors are photoexcited and generate e^–/h⁺ pairs in their CB and VB under light irradiation, respectively. And then the photogenerated e^– in the CB of PS Ⅱ is transferred to the VB of PS Ⅰ via D/A pairs. Finally, h⁺ in the VB of PS Ⅱ participate in oxidation reaction and e^– in the CB of PS Ⅰ participate in reduction reaction on the surface of photocatalysts, respectively. Commonly used D/A pairs are IO₃^–/I^– [253], Fe³⁺/Fe²⁺ [254, 255], I₃^–/I^– [256], NO₃^–/NO₂^– [257], [Co(phen)₃]^3+/2+ and [Co(bpy)₃]²⁺ [258].

In 2016, Ma and co-workers [256] reported a Z-scheme of Sm₂Ti₂S₂O₅ and co-loading with Pt to generate H₂, in which triiodide/iodide (I₃^-/I^-) redox couple were used as a shuttle electron mediator. Kudo's group [258] found a Z-scheme system was constructed using a Ru/BaTa₂O₆:Ir, La H₂ evolving photocatalyst combined with a [Co(bpy)₃]²⁺ as electron mediator. The redox potential of [Co(bpy)₃]²⁺ was reported as 0.32 V, and the value was not affected by pH. Due to the doping of Ir, the wide band gap material has a wider visible light absorption range. Yang et al. [259] revealed a carbon nanodots(CNDs)/WO₃ Z-scheme to improve the photocatalytic performance with IO₃^–/I^– as shuttle redox mediator. As shown in Fig. 22, photogenerated e^–/h⁺ pairs tend to recombine, and only a little of them participates in the photocatalytic reactions, which results in low photocatalytic activity. For the Z-scheme system, the photogenerated electrons transferred from the CB of WO₃ to the VB of CNDs through IO₃^–/I^– as shuttle redox mediator. Therefor the e^– reduce H⁺ to H₂ in the CB of CNDs. Because of the existence of the IO₃^–/I^–, the e^–/h⁺ pairs could be effectively separated without direct contact between the two semiconductors, thereby improving the photocatalytic activity of water splitting.

However, the redox-mediator Z-scheme photocatalytic system has some drawbacks. Namely, the redox mediator-induced back reactions are thermodynamically favorable, and can easily occur because the photogenerated electrons and holes with strong redox power are consumed by the shuttle redox ion pairs. In addition, the light-shielding effect, feasibility confined in solution systems and solution pH sensitivity could be limited its application in the field of photocatalytic water splitting. Thus, all-solid-state Z-scheme photocatalytic system attracted some scientific attentions.

Different from the photocatalytic mechanism of the redox-mediator Z-scheme, in the solid-state Z-scheme photocatalytic system, the two semiconductors are directly contacted via conductor as electron mediator (such as Au [260, 261], RGO [31, 262] and MWCNT [263]) to form Ohmic contacts with low contacting resistance. As shown in Fig. 21(b), both the photogenerated electrons in the CB of PS Ⅱ and photogenerated holes in the VB of PS Ⅰ transfer to the electron mediator and recombine with each other, leaving electrons in the CB of PS Ⅰ and holes in the VB of PS Ⅱ, which are mostly reserved for the targeting reduction and oxidation reactions, respectively. For the solid-state Z-scheme photocatalytic system, the distance of electrons transfer is reduced due to the presence of electron mediator, therefore reducing the recombine of the photogenerated e^–/h⁺ pairs. Without the D/A pairs, the solid-state Z-scheme photocatalytic system can work not only in the liquid phase, but also in the gas phase. However, when the electron mediator is noble metal, the development of the solid-state Z-scheme photocatalytic systems are limited by the expensive cost and shielding effect [264]. Table 9 shows the applications of all-solid-state Z-scheme photocatalytic system in hydrogen evolution applications.

In 2016, Ma's group [265] reported sandwich-structured Z-scheme g-C₃N₄-Au-CdZnS nanoparticles system for visible light photocatalytic hydrogen evolution with 24.6 mmol·h^–1·g^–1 hydrogen yield. Attribute to the Z-scheme charge-carrier transfer mechanism, Au is used as the electron transfer mediator between g-C₃N₄ and CdZnS, which improves the photocatalytic performance of the composite. Au acts as an electron transition mediator that induces the recombination of photogenerated electrons in CdZnS to the photogenerated holes in g-C₃N₄. Except noble metals, metal sulfide and oxides with significant conductivity have been employed as electron-mediator between the photocatalysts in Z-scheme system. Wang and his co-worker [266] reported a novel composite photocatalyst by sandwiching metallic 1T-phase MoS₂ nanosheets between MoO₃ and g-C₃N₄ for the first time, in which the metallic 1T-phase MoS₂ nanosheets act as electron mediator between MoO₃ and g-C₃N₄ to construct an all-solid-state Z-scheme photocatalytic system, and the photocatalytic H₂ evolution rate reaches to 513.0 μmol·h^–1·g^–1 under visible light irradiation. In addition, non-metal with significant conductivity such as reduced graphene oxide (RGO) have also been used as electron mediator to enhance the photocatalytic hydrogen evolution. Xue et al. [262] constructed ternary Cd_0.5Zn_0.5S/RGO/g-C₃N₄ all-solid-state Z-scheme via two-step method and obtained an H₂ evolution rate of 39.24 mmol·h^–1·g^–1 with an AQE of 37.88% at 420 nm. The enhanced photocatalytic hydrogen evolution performance can be ascribed to the synergistic effect between Cd_0.5Zn_0.5S nanospheres and g-C₃N₄ nanosheets in the existence of RGO, which plays a decisive role in the ternary system by acting as a mediator for charge carriers between the two semiconductors.

Although the electron mediators play an important role in transferring electrons between PS Ⅰ and PS Ⅱ, there are several adverse effects, such as metal-based electron mediators might affect light absorption, and Au could cause SPR effect, which would have great influence on the activity of the photocatalyst. Hence, mediator-free Z-scheme photocatalyst was constructed in 2013, Yu et al. [47] first proposed the concept of a direct Z-scheme photocatalyst. Although the band structures of direct Z-scheme photocatalyst systems are similar to the Type-Ⅱ heterostructures, the electron transfer path between the two semiconductors is completely different. As schematically illustrated in Fig. 21(c), the direct Z-scheme photocatalyst system was constituted by two closely contacted semiconductors, the electrons directly transfer across the interface for the CB of PS Ⅰ to the VB of PS Ⅱ without any mediator, and then participate in the redox reaction occurred on the surface of photocatalysts. As a result, the direct Z-scheme photocatalyst system has a strong redox ability for driving the photocatalytic reactions and spatially separated reductive and oxidative active sites. The application of direct Z-scheme photocatalyst system for H₂ production is summarized in Table 10. Yu's group [274] reported a novel TiO₂/NiS core-shell direct Z-scheme nanostructure hybrid for photocatalytic H₂ production with the rate of 655 μmol·h^–1·g^–1. As shown in Fig. 23, when the two semiconductors form an ohmic contact, a built-in electric field is formed at the interface between TiO₂ and NiS, the electrons in the CB of TiO₂ will recombine with the holes in the VB of NiS. The photogenerated electrons in the CB of NiS would effectively reduce H₂O to produce H₂. Meanwhile, the holes in the VB of TiO₂ oxidize the sacrificial agent of methyl. The TiO₂/NiS direct Z-scheme could effective separate electron-hole pairs to promote photocatalytic hydrogen production.

To date, Liu's group [275] conducted studies on the Bi₂WO₆/CdS Z-scheme photocatalyst system and obtained a photocatalytic H₂ evolution rate of 1223.364 μmol·h^–1 with AQY of 1.4% at 420 nm. The H₂ production rate was significantly higher than the one produced by using pure Bi₂WO₆ or CdS photocatalysts (Fig. 24). The excellent photocatalytic performance is due to the intrinsic properties of the direct Z-scheme Bi₂WO₆/CdS photocatalyst system, which replaces the electron mediator with an ohmic contact to shorten the distance of electron transfer, and thus improves the separation of electrons and holes. Pt acts as dual cocatalyst to improve the H₂ evolution activity in the photocatalyst system, because of the enhanced charge separation of photogenerated charge carriers and the efficient photocatalytic H₂ production. Pt is also applicable to other direct Z-scheme systems and acts as a cocatalyst to promote the activity of photocatalytic H₂ [276-279].

Metal organic frameworks (MOFs) are well known as porous materials consisting of metal nodes and organic ligands with unique catalytic natures [290, 291]. MOFs have been often used as electrocatalysts and photocatalysts because of their high design flexibility, large specific surface area and tunable porous channels, which are available for exposing numerous active sites to drive the redox reactions and shorten the transfer distance of photoexcited carriers to the pore surface [292]. In addition, MOF-derived materials have been found to serve as precursors for porous semiconductor materials, which includes metal phosphide, metal oxides, carbon materials and metal sulfides via facile thermal treatment, phosphorization and sulfidation process [293-296]. The as-prepared MOF-based materials can not only maintain the initial structural features of MOF materials that ensures the mono-dispersion of metals centers, but also improve the conductivity property of the materials. Thus, tremendous efforts have been made to explore MOF-derived materials in photocatalytic water splitting.

MOF-based materials as cocatalysts in the hydrogen evolution half-reaction system can serve various metal ions, and facilitate the doping of C and N to the MOF-derived semiconductors, and more important they can provide high surface areas to expose numerous active sites.

UiO-66 as a representative stable MOF owns intrinsic photocatalytic H₂ evolution activity. Jiang's group [297] has synthesized the composite of CdS-decorated UiO-66 that not only has taken the advantage of both CdS and MOF, but also avoid the drawbacks of each component. CdS/UiO-66 (Fig. 25) exhibits high H₂ production activity for water splitting of H₂ evolution under visible light irradiation, the key contribution of the composite is that CdS has the capability to adsorb visible light and UiO-66 provides a large surface area to disperse the CdS on the MOF surface, so that providing more adsorption sites and photocatalytic reaction centers. ZnIn₂S₄@NH₂-MIL-125(Ti) nanocomposites were obtained by a simple solvothermal method and ZnIn₂S₄ nanosheets were also highly dispersed on the surface of NH₂-MIL-125 (Ti) MOF [298]. The composite displayed higher photocatalytic performance towards H₂ production and process stability during the photocatalytic reactions under visible light irradiation. In addition, the g-C₃N₄/MOF (Fig. 26) heterojunctions that were synthesized by combined MOFs (UiO-66) also displayed outstanding photocatalytic performance toward H₂ production, as well as CO₂ reduction and organic pollutants degradation [299]. Interestingly, crystalline and porous covalent organic frameworks (COFs) and metal-organic frameworks (MOFs) materials are amazing combinations in the field of photocatalytic H₂ evolution. Zhang et al. [300] have successfully synthesized MOF/COF (Fig. 27) hybrid materials with high surface area, porous framework and high crystallinity by integrating COF (TpPa-1-COF) with stable MOF (NH₂-UiO-66). Especially, the resulting NH₂-UiO-66/TpPa-1-COF (4:6) exhibits excellent photocatalytic H₂ production rate of 23.41 mmol·g^–1·h^–1. This design will provide a new path to construct efficient photocatalysts for hydrogen evolution.

MOF-derived metal sulfides can also be well used as cocatalysts for photocatalysts. Kumar et al. [301] have incorporated MOF-derived hollow structure of Co₄S₃ with CdS particles (Fig. 28) via a facile wet chemical method. The optimized Co₄S₃/CdS composites exhibited superior photocatalytic performance and the H₂ evolution rate is 12, 360 µmol·h^–1·g^–1. The high catalytic performance can be ascribed to the abundant active sites and onion slice-type hollow structure in MOF-derived Co₄S₃, which were beneficial for effective separation and transfer of photo-induced charge carriers and boosting the surface shuttling properties.

Metal phosphide has been considered to be the most interesting co-catalysts with unique charging nature (positive charged in metal and negative charged in phosphorous [302, 303] for water splitting, especially for HER. Xu et al. [304] have constructed Fe₂P/Ni₂P-modified g-C₃N₄ photocatalysts, in which the C, N co-doped Fe₂P/Ni₂P cocatalyst was derived from NH₂-MIL-101(Fe)/Ni(OH)₂. The g-C₃N₄ nanosheets were deposited on the surface of Ni(OH)₂ anchored NH₂-MIL-101(Fe) MOF matrix and then were phosphidized. The photocatalyst showed excellent H₂ evolution rate of 13.81 mmol·h^–1·g^–1, which was contributed to the hollow polyhedrons structure that facilitated the efficient charges separation and transfer between the interface of g-C₃N₄, CN/Fe₂P and Ni₂P. Kumar et al. [305] also utilized nickel phosphide (Ni₂P) as cocatalysts to modify CdS, which is derived from metal-organic framework (Ni-MOF) (Fig. 29). The optimized composite material exhibited better photocatalytic activity than Pt loaded CdS.

Metal oxide nanostructures are considered to be the most important photocatalysts due to its low cost, low toxicity and high stability. Lin's group [20] have developed a system of Fe₂O₃@TiO₂ (Fig. 30) with MIL-101(Fe) as template and Fe precursor and then calcined in air to decompose the core. This material shows interesting photophysical performance due to its ability to split water under visible light irradiation. In addition, the MOF templating synthesizprocedure can be well modulated by modified varying the MOF and coating materials. Yao's group [306] also incorporated small amounts of uniform GO in MOF templated iron oxide (α-Fe₂O₃ and γ-Fe₂O₃), the heterojunction resulted in highly promoted photo-induced charge separation and transportation for synergistic effect between the reduced graphene and iron oxide. As a consequence, the as-prepared maghemite-carbonaceous composite showed an ideal photocatalytic H₂ evolution rate of 318 µmol·h^–1·g^–1.

Porous Pt-doped heterojunction, Pt-ZnO-Co₃O₄, Pt-ZnS-CoS' and Pt-Zn₃P₂-CoP (Fig. 31) can be obtained via facile oxidation, sulfurization and phosphorization treatment of ZnCo-zeolitic-imidazolate-framework (ZnCo-ZIF) and then doped Pt nanoparticles. The composite exhibits excellent photocatalytic performance towards H₂ production by water splitting, which is attributed to the porous structures of the MOF materials that can significantly enhance the light adsorption and afford abundant active sites. In addition, suitable band matching in heterojunctions facilitates the efficient photo-induced charge separation and transportation [307].

Apart from the abovementioned works, Table 11 summarizes other recently published MOF-based photocatalysts for photocatalytic HER, including pristine MOF-based materials, CdS/MIL-101(Cr) [308], 1TMoS₂/MIL-125-NH₂ [309], MOF-derived materials, MoS₂@TiO₂ [310], Cu₂O@C₃N [311], etc.

In the past decade, semiconductor-based photocatalytic hydrogen production has grown tremendously to be more efficient in converting solar energy into chemical energy. A high performance photocatalytic H₂ evolution reaction generally requires high charge separation and effectually charge transportation, which are necessary for loading cocatalysts. In this review, we summarized the research progress on several cocatalysts for photocatalytic hydrogen production in the past decade. It has been proved that integrating these cocatalysts with semiconductors to construct specific structures can significantly improve the photocatalytic activity for H₂ evolution. For comparison, TableS 2–11 summarize the latest developments in photocatalytic H₂ evolution over various kinds of cocatalysts including single, dual, Z-scheme and MOFs.

The cocatalysts have been extensively investigated in photocatalytic H₂ evolution to rapidly trap electrons from the bulk of the semiconductors, which results in efficienct separation of charge carries and suppressed recombination of e^–-h⁺ pairs. Additionally, the cocatalysts could supply active sites for water reduction and reduce the surface activation energy to catalyze the surface chemical reactions. Further, the cocatalysts usually possess good electrical conductivity, which could boost efficienct transfer of charge carriers.

At present, the cocatalysts used in photocatalytic hydrogen evolution systems has evolved from noble-metals to transition metals and alloys and from single systems to complex systems to enhance the production of H₂ evolution. In the above transitions, researchers have achieved a great progress in the field of photocatalytic hydrogen production via water spilling. However, the matching relationship between cocatalysts and photocatalysts is still needed to be further studied. We believe that with the continuous development of new cocatalysts and the further development of the related theories, the efficiency of solar photocatalytic hydrogen production will continue to increase, and the photocatalytic hydrogen production system will also achieve a greater development.