Latest progress in g-C3N4 based heterojunctions for hydrogen production via photocatalytic water splitting: a mini review

Graphitic carbon nitride based heterojunction photocatalysts have gained increasing attention in producing the clean energy source of hydrogen. Coupling carbon nitride (g-C3N4) with other semiconductor materials or metals as co-catalysts is considered as an effective strategy to overcome the drawbacks of g-C3N4 such as the quick recombination of photogenerated charges. In this review, the recent research advancements in the construction of g-C3N4-based heterojunctions as well as their different charge separation/transfer mechanisms will be systematically discussed, making special emphasis on the design and fabrication of type-II, Z-scheme, S-scheme and Schottky heterojunctions and their application towards H2 generation from water splitting. Finally, a summary and some crucial issues, which should be further resolved for developing advanced g-C3N4-based heterojunction photocatalysts, are presented.


Introduction
The increased demand for more energy as well as the environmental pollution issues due to fossil fuels consumption are among the major problems faced by humanity in the 21st century. Thus, it is necessary to explore new energy resources that are clean and eco-friendly to replace traditional fossil fuels [1]. At present, photocatalytic hydrogen (H 2 ) production via water splitting is an effective remedy to overcome the energy and environmental issues [2,3]. Since the Fujishima-Honda report in 1972 on photoelectrochemical H 2 evolution [4], photocatalytic water splitting based on photoactive material has attracted considerable interdisciplinary attention. Although a plethora of potential semiconductor photocatalysts such as TiO 2 , WO 3 and CdS have been extensively employed [5][6][7][8], there is still no available photocatalytic material that meets the specific requirements such as cost-effective and high efficiency for the commercial applications of photocatalytic H 2 generation via water splitting process [9,10]. Therefore, the development of low-cost and highly effective semiconducting materials is necessary for practical applications.
Over the past few years, graphitic carbon nitride (g-C 3 N 4 ), a 2D-structured metal-free semiconductor, has attracted increasing attention as one of the most promising photocatalytic systems for H 2 production [11][12][13]. The structure of g-C 3 N 4 consists of a long pair of N-atom-based triazine (C 6 N 7 ) and tri-s-triazine (C 3 N 3 ) tectonic units and both the N-atoms in these two rings are sp 2 hybridized (figures 1(a) and (b)). It can be synthesized by direct condensation of low-cost N-rich feedstocks, including melamine [14,15], urea [16,17], thiourea [18] and cyanamide [19,20]. In recent years, various methods have been effectively used in the preparation of g-C 3 N 4 such as thermal polycondensation, chemical or physical vapor deposition, solvothermal synthesis and microwave heating. Among them, microwave heating technique with such advantages as time saving, low energy consumption and high economic feasibility is more effective and reliable. Remarkably, the microwave-synthesized g-C 3 N 4 revealed higher thermal stability, larger surface area and stronger emission intensity than traditional preparation methods. By contrast with the widely used  [34], (c) Charge transfer mechanism over pure g-C3N4 (adapted from Patnaik et al) [9], (d) Band structures of some typical semiconductor materials (adapted from Li et al) [35].
photocatalyst titanium dioxide, g-C 3 N 4 possess a mild band gap of 2.7 eV [21], rendering it a fascinating visible-light driven photocatalyst. Under irradiation, g-C 3 N 4 is directly excited by absorbing photons with energy ⩾e.g. (band gap = 2.7 eV), resulting in the creation of photogenerated electron-hole owing to the transfer of electron from the valence band (VB) to the conduction band (CB) of g-C 3 N 4 . The photoinduced electrons and holes act as reactive species that are highly oxidizing and reducing. They may recombine or migrate to the active sites on the surface, thereby splitting the water into oxygen and hydrogen (figure 1(c)). Despite its good photoresponse to visible light as well as its high thermal and chemical stabilities, pristine g-C 3 N 4 still does not meet the requirements of a wider range of applications because of its low surface area, fast recombination and low migration ability of its photoexcited charges [22,23]. Hence, significant efforts have been made to improve the photocatalytic efficiency of pristine carbon nitride such as doping with other elements [24][25][26], introducing defects [27,28] and combining with other materials to form heterojunction interfaces [29][30][31][32][33].
To achieve overall water splitting, the band structure of a semiconducting material should meet specific requirements. The VB maximum level must be higher than the oxidation potential of water (E O2/H2O , 1.23 V vs. NHE) and the minimum position of the CB must be more negative than the reduction potential of hydrogen (E H2/H2O , 0 V vs. NHE) [36]. Hence, the semiconductor should have a bandgap higher than 1.23 eV to meet the minimum dynamic requirement. Nevertheless, in practice, water-splitting reaction requires higher energy (1.8 eV) due to overpotentials for hydrogen and oxygen evolution reactions, which greatly limits the utilization of a broad solar light spectrum. Therefore, it is hard to achieve a reasonable water splitting efficiency with a single semiconductor. Up to now, there are only a few examples of single materials that could achieve the two half-reactions (O 2 and H 2 evolution) ( figure 1(d)). Owing to its suitable band gap as well as CB and VB levels, pristine g-C 3 N 4 could achieve the full water splitting under visible light. Nevertheless, its photocatalytic water splitting efficiency was still unsatisfactory. Actually, the formation of heterojunctions by coupling g-C 3 N 4 with other semiconductors or metals as co-catalysts is considered as an effective strategy for improving its photocatalytic properties since the heterojunction formation with suitable band alignment could realize better visible light-absorption and promote the separation/transfer of photoexcited charge carriers.
In recent years, several excellent review papers on g-C 3 N 4 -based materials dealing with the design of pure g-C 3 N 4 and its promising photocatalytic applications have been published [34,[37][38][39], whereas, there are few review papers that report the application of g-C 3 N 4 based heterojunctions (CNBH) and their role in improving the photocatalytic hydrogen generation process. Therefore, this review will focus on the latest development in the application of CNBH for hydrogen production via photocatalytic water splitting. In particular, the construction and the advantages and disadvantages of different CNHB, including g-C 3 N 4 /semiconductor and g-C 3 N 4 /metal systems are emphasized.

Different types of g-C 3 N 4 -based heterojunctions
The construction of composites/heterojunctions is considered as the most viable method to improve the charge separation/transfer process and thereby increasing the overall photocatalysis efficiency. Due to its appealing physico-chemical properties, many metal nanoparticles and semiconductors could easily be combined with g-C 3 N 4 semiconductor. Coupling g-C 3 N 4 and other semiconducting materials with different favorable properties induces the creation of band bending at the interface of the heterojunction with an inner electric field so that the transfer and the separation of free charges can be significantly improved [40][41][42]. In general, CNBH systems could be divided into four main categories, e.g. type II heterojunction, Z-scheme heterojunction, S-scheme heterojunction and Schottky junction (figure 2). The following section will summarize and discuss the latest advances in CNBH systems for photocatalytic hydrogen production.

Type II heterojunctions
For the most CNBH systems, type II heterojunction is the most common type. As displayed in figure 2, type II heterojunction is commonly formed by the combination of two semiconductors (SC 1 and SC 2) possessing wide and small band gaps; among them one must has a more negative CB level. Upon light irradiation, when the CB and VB levels of SC 1 are greater than the CB and VB positions of SC 2, photogenerated holes (h + ) migrate from the VB of SC 2 to the VB of SC 1, meanwhile photexcited electrons (e -) migrates from the CB of SC 1 to the CB of SC 2. In this way, the transfer of photoinduced eand h + in opposite directions results in an efficient charge separation, which greatly enhance the photocatalytic efficiency.
Recently, many efforts have been invested for preparing C 3 N 4 -based type II heterojunctions for photocatalytic H 2 production. In principle, the band alignment and the close interface contact between the different semiconductors are the crucial factors for the fabrication of an efficient type II heterojunction photocatalysts [43]. Shen et al [44] synthesized Black TiO 2 nanobelts/g-C 3 N 4 (b-TiO 2 /g-C 3 N 4 ) type II heterojunction by hydrothermal-calcination method. The formation of laminated heterojunctions was clearly demonstrated by SEM and TEM results, which show that g-C 3 N 4 surfaces are coated of TiO 2 nanobelt (figures 3(a)-(c)). The intimate contact between b-TiO 2 and g-C 3 N 4 promotes the charge separation in the b-TiO 2 /g-C 3 N 4 photocatalyst (figure 3(d)), resulting in superior H 2 generation activity compared to single semiconductors g-C 3 N 4 and TiO 2 . Lin et al [45] investigated the electronic and optical properties of g-C 3 N 4 /TiO 2 type II heterojunction photocatalyst using density-functional theory (DFT) calculations to understand the mechanism behind its enhanced H 2 production activity. The calculation results of TDOS and PDOS (figures 4(a)-(c)) indicated that the close contact between the two components leads to band gap narrowing, which improves the visible light-harvesting of g-C 3 N 4 /TiO 2 . The results show that the VBM of the g-C 3 N 4 /TiO 2 system is dominated by C 2p and N 2p states instead of O 2p states for the TiO 2 , which reduces its bang gap, while the CBM of g-C 3 N 4 /TiO 2 system is occupied by Ti 3d states (figure 4(d)). Interestingly, the CBM and VBM of TiO 2 are lower than those of carbon nitride, making the type II heterojunction feasible in g-C 3 N 4 /TiO 2 heterojunction. The improved photoactivity of the g-C 3 N 4 /TiO 2 system is due to the suitable type II band alignment structure, which promotes the eand h + separation efficiency.
In another work, Li et al [46] have prepared Cu 2 O/g-C 3 N 4 heterostructure via combined solvothermal and chemisorption process. The Cu 2 O/g-C 3 N 4 heterostructure showed higher photocurrent intensity compared to that of Cu 2 O and pure g-C 3 N 4 (figure 5(a)), indicating the high separation and transport efficiency of photoexcited species through the interfacial connection of g-C 3 N 4 and Cu 2 O. As displayed in figure 5(b), due to the large contact area and the intimate contact between the two semiconductors, the excited eon Cu 2 O semiconductor could migrate to the CB of g-C 3 N 4 , which promotes the charge separation, resulting in superior photoactivity for H 2 production. Other semiconductors such as MoS 2 [47,48], In 2 O 3 [49], TiO 2 [50][51][52][53], ZnS [54], Bi 2 MoO 6 [55], SiC [56], CeO 2 [57], CoO [58] and CdS [59,60] have been employed to combine with g-C 3 N 4 for the fabrication of advanced type II heterojunction materials with high photocatalytic efficiency (table 1).
Although all g-C 3 N 4 -based type II heterojunction photocatalytic materials exhibit enhanced photoactivity in H 2 production, the type II heterojunctions weakens the redox ability of eand h + because of the migration of photoinduced species to the lower-lying CB and VB [13,77]. To surmount this shortfall of type II heterojunction, a novel type of heterojunction, named as Z-scheme heterojunction, is proposed [78]. In this new charge-transfer mechanism, the photoinduced eon the lower-lying CB of one SC could recombine with the photoinduced h + on the lower-lying VB of the other SC [79][80][81] (figure 2). In this way, eremains at the more negative CB potential and the h + remains at the more positive VB potential, which maintains the high redox powers of the free charge carriers. Therefore, the Z-scheme heterojunction systems can realize an effective charge transfer and separation with high redox abilities of photoinduced of e --h + pairs.

Z-scheme heterojunctions
Inspired by the natural photosynthesis system, constructing CNBH systems has emerged as a promising option to enhance the photoactivity of g-C 3 N 4 in H 2 production. Figure 6 shows the typical forms of Z-scheme mechanisms with or without electron mediators. Notably, the electron mediator plays a crucial role in promoting the separation of the carriers as it affords sites for coupling photoinduced charges with weak redox ability. The original Z-scheme heterojunction system was applied in liquid phase using shuttle redox mediators including Fe 2+ /Fe 3+ , I − /IO 3− and [Co(bpy) 3 ] 3+/2+ [78,[82][83][84]. Nevertheless, the soluble redox shuttles-mediated Z-scheme heterojunction photocatalysts are only applicable in liquid phase; hence, the development of Z-schemes with solid electron mediators is very desirable (figure 6(a)). Recently, a new mediator-free Z-scheme heterojunction system, named as direct Z-scheme, was proposed [85,86]. Contrary to the mediated Z-scheme heterojunction system where a conductor is employed to afford sites for charge recombination and relay, in the direct Z-scheme system, the photoexcited charges on the lower-lying CB of SC 2 and VB of SC 1 can directly recombine at the interface between the two semiconductors [87][88][89].
Recently, many researchers have reported the use of various electron mediators, including metals and carbonaceous materials as efficient electron mediators to construct g-C 3 N 4 -based Z-scheme heterojunction systems with high H 2 production efficiency. Li et al [61] synthesized Z-scheme Ag/g-C 3 N 4 -Ag-Ag 3 PO 4  heterostructure. In the SEM and TEM images (figures 7(a)-(c)), the Ag/g-C 3 N 4 -Ag-Ag 3 PO 4 heterojunction can be found, where Ag NPs are lied between Ag 3 PO 4 and g-C 3 N 4 and also located on other side of carbon nitride. An optimal Ag/g-C 3 N 4 -Ag-Ag 3 PO 4 heterostructure exhibited the highest hydrogen production  (221.64 µmol) over three times that the rate of pristine g-C 3 N 4 . Interestingly, Ag NPs that lie between the two semiconductors promote the migration of efrom the CB of Ag 3 PO 4 to the VB of g-C 3 N 4 , meanwhile, Ag NPs on the other side of carbon nitride act as co-catalyst and thereby enhancing the photocatalytic efficiency ( figure 7(d)). Moreover, Au is also considered as an efficient electron mediator [62,63]. In our previous work [63], a Cd x Zn 1−x S/Au/g-C 3 N 4 system was developed for photocatalytic H 2 production, which accomplished a significantly enhanced H 2 generation activity. PL spectra and time-resolved photoluminescence results (figures 8(a) and (b)) verified the efficient spatial separation of photogenerated eand h + in Cd x Zn 1−x S/Au/g-C 3 N 4 photocatalysts mainly originated from the Z scheme charge flow ( figure 8(c)). Shi et al found that nanocarbon is an effective electron mediator in the Z-scheme g-C 3 N 4 /nanocarbon/ZnIn 2 S 4 heterojunction [64]. Namely, nanocarbon could enhance the conductive capability, enhancing the charge separation/transfer process and thus improving the H 2 production activity. In another work, Jo and co-workers reported the preparation of Z-scheme CdS/RGO/g-C 3 N 4 photocatalyst using RGO as electron mediator [65]. Owing to its good conductivity, RGO could accelerate the electron migration at the CdS/g-C 3 N 4 interface, which resulted in high H 2 production activity.
Although their important role in promoting the spatial charge separation in Z-scheme heterojunctions, solid electron mediators have some drawbacks such as backward reactions as well as photocorrosion. In particular, it has been well established that noble metals suffer from shielding effects, which limit the photoactivity of the mediated Z-scheme systems. In regard of this, the creation of direct Z-scheme systems (figure 6(b)) is put forward, which is able to resolve the shortcomings of the mediated Z-scheme systems. The typical CNBH systems, including the mediated and direct Z-scheme systems are summarized in table 1.
It is noteworthy that the close contact between SC 1 and SC 2 in the direct Z-scheme heterojunction system plays a crucial role because the interface has the same role as the electron mediators. Hence, the effective combination of SC 1 and SC 2 is essential for preparing efficient direct Z scheme heterojunction system. Most recently, Ye et al [66] fabricated a direct Z-scheme CoTiO 3 /g-C 3 N 4 photocatalyst, which showed high photocatalytic hydrogen production activity of 858 µmol h −1 g −1 due to the close contact between CoTiO 3 and carbon nitride. In another work, Yu and co-workers [67] reported the construction of direct Z-scheme g-C 3 N 4 /WO 3 , which showed enhanced hydrogen production rate of 3120 µmol h −1 g −1 due to the formation of internal electric fields in the heterojunction system and the covalent W-O-N-(C) 2 , which accelerate the transfer of photoexcited carriers. Generally, it is necessary to investigate the charge transfer pathway for any heterojunction system because of the competition between Z-scheme and the type II heterojunction mechanisms. To understand the charge transfer mechanism, DFT calculations of charge carrier effective mass were conducted as depicted in figure 9. The results showed that the charge carriers in g-C 3 N 4 could only move through the 2D planes, whereas electrons in WO 3 tend to move from bulk to interface and then migrate to carbon nitride, which support the direct Z-scheme mechanism (figure 9(e)).
She et al [68] have prepared α-Fe 2 O 3 /g-C 3 N 4 nanocomposite with 2D/2D geometry using a small amount of α-Fe 2 O 3 as a catalyst to promote the formation of ultrathin 2D g-C 3 N 4 nanosheets as revealed by HRTEM and AFM measurement (figures 10(a) and (b)). The α-Fe 2 O 3 nanosheets successfully formed a Z-scheme heterojunction with 2D g-C 3 N 4 nanosheets owing to their matched energy levels. Interestingly, this Z-scheme junction photocatalyst showed unprecedented hydrogen generation activity of 31 400 µmol g −1 h −1 owing to the direct contact between α-Fe 2 O 3 and g-C 3 N 4 nanosheets, which effectively boost the emigration from the CB of α-Fe 2 O 3 to the VB of g-C 3 N 4 nanosheets ( figure 10(c)). More importantly, to verify the direct Z-scheme mechanism, ESR spin-trapping experiments were performed. If the α-Fe 2 O 3 /g-C 3 N 4 system follows the double charge transfer mechanism, ein the CB of g-C 3 N 4 will move to the CB of α-Fe 2 O 3 , while the h + will move to the VB of g-C 3 N 4 . Taking into consideration the bands edge potentials of g-C 3   pathway. Several other semiconducting materials have been coupled with carbon nitride to fabricate efficient direct Z-scheme heterojunction systems for H 2 evolution reaction, including Nb 2 O 5 [69], W 18 O 49 [70], Bi 2 O 2 CO 3 [71], 2D WO 3 [72], SnFe 2 O 4 [73], N-ZnO [74], SnS 2 [75], 0D-ZnO [76].

S-scheme heterojunctions
As aforementioned in the previous sections, traditional type-II heterojunctions could extremely facilitate the separation of charge carriers due to staggered band alignment. However the redox ability of the photogenerated electrons and holes is weakened, which is unfavorable for photocatalytic reactions. Although, Z-scheme heterojunctions are capable of spatially separating the photoinduced charge carriers and preserving the electrons and holes with high redox ability, they still have some drawbacks. In Z-scheme heterojunction system, the conductor promotes the migration of e-from the photocatalyst with higher CB potential to the photocatalyst with lower VB potential, however, e-can also migrate through the conductor to the surface if the conductor is not situated between the two photocatalysts. Thus, the charge transfer pathway in Z-scheme heterojunction system might be different from what it is claims to be. In order to overcome the shortcomings of Z-scheme mechanism, a new step-scheme (S-scheme) heterojunction concept was proposed by Yu et al [90], which illustrates the charge transfer route clearly and vividly.
Typically, S-scheme heterojunction system comprises two n-type semiconductor photocatalysts with staggered band structure. As shown in figure 11, SC 1 and SC 2 represent oxidation photocatalyst with more positive VB position and reduction photocatalyst with more negative CB position, respectively. When SC 1 and SC 2 come into contact, e-in SC 2 with higher Fermi level could transfer to SC 1 across their interface, thereby forming an interfacial internal electric field and its direction is from SC 2 to SC 1. In addition, the alignment of the Fermi levels of SC 1 and SC 2 to the same level after contact leads to an upward and downward shift in the Fermi levels of SC 1 and SC 2, respectively [91,92]. Under irradiation, the internal electric field and the band bending offer a driving forces to accelerate the recombination of the useless photoinduced e-of SC 2 and h + of SC 1, meanwhile useful e-and h + with higher redox ability are remained  in the CB of SC 2 and in the VB of SC 1, respectively, resulting in a spatial separation of photoinduced charge carriers [93][94][95][96]. In S-scheme heterojunction, the charge transfer route is like a 'step' type from macroscopic viewpoint. Such heterojunction system can greatly promote charge separation and acquire strong photoredox ability.
As a typical n-type semiconductor with unique structural properties, g-C 3 N 4 can be manipulated to construct efficient g-C 3 N 4 -based S-scheme heterojunction systems for photocatalytic H 2 production. For instance, Li et al [97] reported a novel CdS/g-C 3 N 4 S-scheme heterojunction photocatalyst with 2D-2D surface contact prepared by in situ hydrothermal growth of CdS nanosheets on g-C 3 N 4 nanosheets. It was found that the work function of CdS (5.04 eV) is larger than that of g-C 3 N 4 (4.34 eV), which induces the formation of internal electric field and band-edge bending at the interface of the two photocatalysts. This internal electric field promotes the charge carriers separation and diffusion as illustrated in figures 12(a) and (b). Moreover, the formed electric field directing from g-C 3 N 4 to CdS thermodynamically facilitates the migration of e-from the CB of CdS to VB of g-C 3 N 4 and prevents h + in the VB of CdS to diffuse into the VB of g-C 3 N 4 and e-in the CB of g-C 3 N 4 to diffuse into the CB of CdS, giving rise to an S-scheme system ( figure  12(c)). As a result, the CdS/g-C 3 N 4 photocatalysts showed a superior photocatalytic H 2 evolution rate, which is about 3060 times higher than that of pristine g-C 3 N 4 . In addition, Yu et al [90] reported the construction of 2D-2D WO 3 /g-C 3 N 4 S-scheme heterojunction photocatalyst via an electrostatic self-assembly process, which showed an excellent photocatalytic performance towards H 2 generation. Theoretical calculations and band diagram analysis along with in situ EPR results confirm a S-scheme charge transfer pathway for WO 3 /g-C 3 N 4 heterojunction system. Owing to this unique charge transfer route, the photogenerated charge One-step hydrothermal 373 [99] (2020) g-C3N4/Zn0.2Cd0.8S-DETA in-situ growth method 6690 [100] (2020) carriers with strong redox abilities are spatially separated and reserved, contributing to the excellent photocatalytic performance.
Recently, Ma et al [98] reported the construction of a novel S-scheme MCN/UCN isotope heterojunction by using melamine and urea as g-C 3 N 4 precursors. It was found that H 2 generation rate of optimum MCN/UCN is almost 5.5 and 1.8-fold higher than those of MCN and UCN, respectively, with adding Pt NPs as cocatalyst. The mechanism study shows that the improved photocatalytic activity mainly results from the effective charges separation driven by the internal electric field that formed at the interface of MCN and UCN. Under illumination, the useless efrom the CB of UCN recombine with the photogenerated holes from the VB of MCN due to the formed electric field at the interface, while the reserved ein the CB of MCN further diffuse to Pt NPs for conducting water reduction reaction ( figure 13(a)). Additionally, the TEM results (figures 13(b) and (c)) confirm that most Pt NPs are located at MCN surface, indicating that MCN surface is the electron-rich region, which further confirms the S-scheme mechanism. Various g-C 3 N 4 -based S-scheme heterojunction photocatalysts for H 2 generation reported in recent years are listed in table 2.
The examples mentioned above showed that S-scheme heterojunction systems possess strong redox abilities with thermodynamic advantages in photocatalytic water splitting for H 2 generation, however, as is known, photocatalytic reactions are not only affected by thermodynamics, but also by dynamics. Thus, the thermodynamic advantages of S-scheme heterojunction systems are not sufficient. Highly efficient g-C 3 N 4 -based S-scheme heterojunction systems may be achieved by employing some strategies such as doping, which is indispensable to adjust and widen the Fermi level difference between the two semiconductors constructing S-scheme heterojunction. Moreover, cocatalyst loading is an effective approach to promote the charge separation and decrease the kinetic barrier of the photocatalytic reaction.

Metal/g-C 3 N 4 schottky junction systems
Hybridization of g-C 3 N 4 with metal nanoparticles is also an attractive strategy for improving its photocatalytic properties [101][102][103][104]. In principle, when a metal comes into close contact with a semiconductor, a potential difference called Schottky barrier and a depletion layer will be created at the interface between the metal and the semiconductor due to their different work functions and Fermi levels. The Schottky barrier, which can serve as an effective electron skin, prohibits the charge carriers' recombination and prolongs the lifespan of the electrons [105,106]. For photocatalytic hydrogen evolution reaction, platinum Pt is the most used co-catalyst [107][108][109][110][111][112]. Liu and co-workers [110] have synthesized Pt/g-C 3 N 4 by photodeposition of Pt onto g-C 3 N 4 surface, which displayed a H 2 generation activity of 4210.7 µmol g −1 h −1 . In the Pt/g-C 3 N 4 system, the eexcited from the CB of g-C 3 N 4 will move to the surface of Pt, leading to efficient charge separation and enhanced photoactivity for H 2 generation.
It has been found that Pt NPs size can significantly influences the hydrogen evolution rate of Pt/g-C 3 N 4 photocatalyst. In our previous work [111], we studied the photocatalytic activity of Pt/g-C 3 N 4 in hydrogen production with different sizes of Pt NPs from single atoms to nano-clusters. The obtained results indicated that single-atom Pt exhibits superior photoactivity compared to nano-clusters Pt. Particularly, single-atom Pt can provide the most abundant active sites and enable maximum accessibility of Pt atoms leading to the highest reaction efficiency. Moreover, it has been proved that the shape of Pt nanoparticle greatly influences the hydrogen generation activity of Pt/g-C 3 N 4 system. Cao et al [112] reported the construction of Pt/g-C 3 N 4 with Pt of different shapes by a colloidal adsorption-deposition method for photocatalytic H 2 production (figure 14(a)). As illustrated in figures 14(b)-(d), TEM and HRTEM images confirm the successful deposition of 3 types of Pt NPs (nanocubes, nanooctahedrals and nanospheres) onto g-C 3 N 4 surface. The Pt NPs have similar sizes of about 10 nm, however they exposed different facets. Interestingly, the Pt/g-C 3 N 4 hybrids with spherical Pt nanoparticle shape exhibited the highest photocatalytic hydrogen generation activity owing to its favorable exposed facets.
Because of the high costs of Pt noble metal, considerable efforts have been made to reduce its amount of usage with maintaining its excellent activity for photocatalytic H 2 production via replacing a part of Pt with transition metals [113][114][115][116]. In light of this, Han et al [116] synthesized Pt-Co/g-C 3 N 4 by in-situ chemical deposition method, which showed hydrogen generation activity of 960 µmol h −1 g −1 better than that of Pt/g-C 3 N 4 . The increased hydrogen production efficiency is attributed to the synergistic effect between Pt and Co, which modifies the Fermi level position of Pt-Co alloy co-catalyst in comparison to Pt, resulting in improved electron trapping ability of the Pt-Co co-catalyst and thus enhancing the photocatalytic efficiency. In order to prepare cost-effective CNBH photocatalysts for hydrogen generation, loading carbon nitride with non-noble-metals, which are low-cost and abundant elements, has also been explored [117][118][119][120]. Recently, Han et al [120] found that loading g-C 3 N 4 with Ni-Mo alloy resulted in improved hydrogen production activity of 1785 µmol h −1 g −1 , which is close to Pt/g-C 3 N 4 photoactivity. This work demonstrates the successful Pt-free approach for photocatalytic H 2 production.

Conclusions and prospects
In this mini review, the recent advancements in g-C 3 N 4 based heterojunctions for H 2 production via photocatalytic water splitting have been reviewed. Over the past two decades, various types of CNBH systems have been explored as photocatalysts in photocatalytic hydrogen generation. Up to now, the construction of CNBH could effectively improve the efficiency of photocatalytic systems. However, despite the considerable progress achieved in this field, the existing apparent quantum efficiency and water splitting efficiency (STH) by g-C 3 N 4 based heterojunctions are still far from satisfactory. As such, coupling g-C 3 N 4 with low bandgap materials such as black phosphorus and WS 2 , etc could achieve better usage of solar energy and attain a high quantum efficiency.
With type-II and Z-scheme heterojunction systems, there is still much vagueness in the charge transfer mechanism and it maybe different from what it is claimed to be. Therefore, it is necessary to clarify and understand the interfacial charges transfer pathway. Besides direct experimental support by ultrafast spectroscopic studies, in this regard, DFT computation should also be well developed and applied to further understand and clarify the charges transfer pathway and even the mechanism of photocatalytic reaction over g-C 3 N 4 -based heterojunctions photocatalysts.
In order to overcome the shortcomings of type II and Z-scheme heterojunction, S-scheme heterojunction concept was proposed, which explains and illustrates the charge transfer process clearly and vividly. S-scheme system enables high charges redox ability and efficient charge separation, resulting in boosted photocatalytic performance. Despite considerable progress has been achieved in C 3 N 4 -based heterojunction photocatalysts, S-scheme heterojunction system is still at the primary stage of study. In the hopeful future, highly efficient g-C 3 N 4 -based S-scheme heterojunction systems can be methodically realized by employing some strategies such as doping and co-catalyst loading. Furthermore, to make g-C 3 N 4 a cost effective photocatalyst the application of non-noble metal as co-catalyst should be further investigated.