Recent Progress on Photoelectrochemical Water Splitting of Graphitic Carbon Nitride (g−CN) Electrodes

Graphitic carbon nitride (g−CN), a promising visible-light-responsive semiconductor material, is regarded as a fascinating photocatalyst and heterogeneous catalyst for various reactions due to its non-toxicity, high thermal durability and chemical durability, and “earth-abundant” nature. However, practical applications of g−CN in photoelectrochemical (PEC) and photoelectronic devices are still in the early stages of development due to the difficulties in fabricating high-quality g−CN layers on substrates, wide band gaps, high charge-recombination rates, and low electronic conductivity. Various fabrication and modification strategies of g−CN-based films have been reported. This review summarizes the latest progress related to the growth and modification of high-quality g−CN-based films. Furthermore, (1) the classification of synthetic pathways for the preparation of g−CN films, (2) functionalization of g−CN films at an atomic level (elemental doping) and molecular level (copolymerization), (3) modification of g−CN films with a co-catalyst, and (4) composite films fabricating, will be discussed in detail. Last but not least, this review will conclude with a summary and some invigorating viewpoints on the key challenges and future developments.


Introduction
Energy shortage and environmental pollution issues are receiving increasing attention due to the rapid industrialization and excessive consumption of resources by mankind [1,2]. At present, energy supply is mainly dependent on fossil fuels, such as coal, petroleum, and natural gas [3]. The utilization of these fossil fuels is inevitably accompanied by the release of various harmful gases, including carbon dioxide, nitrogen oxides and sulfur oxides etc. [4]. Therefore, developing environmentally friendly and renewable energies to replace conventional fossil fuels is considered a perfect solution to remove these issues, and hydrogen energy is one of the most promising candidates. However, over 90% of hydrogen gas is produced from steam reforming of methane in industry, which limits the development of hydrogen energy because methane is non-renewable energy [5]. In 1972, Fujishima and Honda discovered photoelectrochemical (PEC) H 2 production from water splitting via using a TiO 2 cell under ultraviolet (UV) light radiation [6]. Photoelectrochemical (PEC) water splitting by solar energy is generally considered a promising strategy to generate hydrogen gas because both water and solar energy are inexhaustible on the earth.
Carbon nitride-based polymers have attracted widespread attention thanks to their outstanding electronic properties, which have been exploited in numerous applications such as photocatalysis [7][8][9], CO 2 reduction [10], fuel cells [11], bioimaging [12], and so on. Modified graphitic carbon nitride (g−CN-based films), the focus of this review, can be

Water-Splitting Reaction Mechanisms
Gaining an elementary understanding of photoelectrochemical, charge-separation, and transfer mechanisms is the crucial step in designing photoelectrodes. Figure 2 show structural diagrams of the general photoelectrocatalytic processes for PEC water splitting. There are three primary components of a typical PEC device: (1) a working electrode consists of semiconductor material, a p-type semiconductor as a photocathode (Figure 2a), and an n-type semiconductor as a photoanode (Figure 2b), (2) a counter electrode, and (3) an electrolyte system. It will facilitate the transport of electrons and holes and ensure the separation of hydrogen and oxygen produced on different electrode surfaces. Before contact with the electrolyte, the energy level in the semiconductor is flat. Once in contact with the electrolyte, the different chemical potentials will drive the charge in the semiconductor Up to now, several insightful reviews have already focused on the fabrication and modification of the bulk or powder g−CN and their applications in dealing with the energy crisis and environmental problems [3,22]. However, only a handful of reviews can be traced that focused on the synthesis and applications of g−CN films for PEC water splitting. This scenario thus makes it very urgent to offer a critical and relatively comprehensive review of the state-of-the-art advances of g−CN films for PEC water splitting. In the present review, we initially focus on the latest advances in the fabrication, versatile properties, Nanomaterials 2022, 12, 2374 3 of 58 rational modification, and applications of g−CN-based films for PEC water splitting. Firstly, this review will conclude the different types of g−CN photoelectrodes according to their preparation pathways and then the rational modification methods to promote the PEC performance. Last but not least, the major challenges and some invigorating perspectives for future research on g−CN-based film materials are also concluded. We hope that this review will not only offer valuable insight into new g−CN-based film materials but also therefore promote the further developments of g−CN-based films in green and sustainable energy production and storage.

Water-Splitting Reaction Mechanisms
Gaining an elementary understanding of photoelectrochemical, charge-separation, and transfer mechanisms is the crucial step in designing photoelectrodes. Figure 2 show structural diagrams of the general photoelectrocatalytic processes for PEC water splitting. There are three primary components of a typical PEC device: (1) a working electrode consists of semiconductor material, a p-type semiconductor as a photocathode (Figure 2a), and an n-type semiconductor as a photoanode (Figure 2b), (2) a counter electrode, and (3) an electrolyte system. It will facilitate the transport of electrons and holes and ensure the separation of hydrogen and oxygen produced on different electrode surfaces. Before contact with the electrolyte, the energy level in the semiconductor is flat. Once in contact with the electrolyte, the different chemical potentials will drive the charge in the semiconductor to the electrolyte until the formation of the built-in electric field. Then, a depletion layer (WD) and a Helmholtz layer (WHL) present at the semiconductor interface and the electrolyte interface, respectively, to form a semiconductor-liquid junction (SCLJ). This charge transfer also gives rise to band bending at the semiconductor interface. For the n-type semiconductors, electrons flow from the semiconductor conduction band to the electrolyte, and the semiconductor surface forms a positively charged depletion layer, resulting in the upward band bending. In contrast, holes flow from the semiconductor to the electrolyte for the p-type semiconductors, which leads to the downward band bending.  In this section, the transport of excitons is discussed. Under a steady illumination, the semiconductor photoelectrode absorbs the photons whose energy is higher than the band gap energy. Then, the electrons become excited from the valence band (VB) to the conduction band (CB), leaving empty states that constitute positive holes in VB. The built-in electric field at the semiconductor/electrolyte interface and the band bending inside the semiconductor separate the electron-hole pairs and consequently generate the photovoltage (Vph) and photocurrent. For the semiconductor photocatalysts with a flat band potential above the H + /H2 energy level, the separation and transport of excitons (electron-hole pairs) do not require external bias to produce H2 and O2 (self-powered mode). However, the transport of excitons usually suffers from an over-potential across the semiconductor, potential drops at contacts, the resistance of the electrodes, and so on. Therefore, the pho- In this section, the transport of excitons is discussed. Under a steady illumination, the semiconductor photoelectrode absorbs the photons whose energy is higher than the band gap energy. Then, the electrons become excited from the valence band (VB) to the conduction band (CB), leaving empty states that constitute positive holes in VB. The built-in electric field at the semiconductor/electrolyte interface and the band bending inside the semiconductor separate the electron-hole pairs and consequently generate the photovoltage (V ph ) and photocurrent. For the semiconductor photocatalysts with a flat band potential above the H + /H 2 energy level, the separation and transport of excitons The band-gap of the laboratory fabricated g−CN is identified to be about 2.7 eV, corresponding to the absorption from the highest occupied molecular orbitals (HOMOs) to the lowest unoccupied molecular orbitals (LUMOs). The HOMO and LUMO levels originated from nitrogen pz and carbon pz orbitals, respectively [37]. This band-gap is large enough to surmount the thermodynamic barrier of the water-splitting, which requires 1.23 eV theoretically. Furthermore, the HOMO/LUMO levels of g−CN straddle the energy levels of water reduction and oxidation, which manifests that g−CN can thermodynamically function as a candidate semiconductor material for PEC [38,39].
Until now, thermal pyrolysis of nitrogen-rich precursors such as cyanamide, dicyanamide, urea, melamine, thiourea, guanidine hydrochloride, trithiocyanuric acid, etc. has been employed to synthesize g−CN. This type of nitrogen-rich precursor has a slight effect on the band-gap structure of g−CN [40]. The flat band potential generated by different precursors was measured using the Mott-Schottky plot, which is −0.6 V vs. RHE for urea [35], −0.8 V for cyanamide and dicyandiamide [41,42], and −0.7 V for melamine [43,44]. The valence band X-ray photoelectron spectroscopy plot, which calculates the energy level difference between the valence band and the Fermi level [45], produced a value of 1.56 V for urea, thiourea and melamine [35,40], and 1.74 V for dicyandiamide [46]. Ragupathi et al. [47] reported a band-gap of 2.89 eV, 2.78 eV, and 2.35 eV for thiourea, urea, and urea with glycine precursor to synthesize g−CN, respectively.

g−CN Thin-Film Preparation Methods
Numerous synthesis strategies have been developed to prepare g−CN films on substrates, which are classified as thermal vapor condensation (TVC), solvothermal route, The band-gap of the laboratory fabricated g−CN is identified to be about 2.7 eV, corresponding to the absorption from the highest occupied molecular orbitals (HOMOs) to the lowest unoccupied molecular orbitals (LUMOs). The HOMO and LUMO levels originated from nitrogen p z and carbon p z orbitals, respectively [37]. This band-gap is large enough to surmount the thermodynamic barrier of the water-splitting, which requires 1.23 eV theoretically. Furthermore, the HOMO/LUMO levels of g−CN straddle the energy levels of water reduction and oxidation, which manifests that g−CN can thermodynamically function as a candidate semiconductor material for PEC [38,39].
Until now, thermal pyrolysis of nitrogen-rich precursors such as cyanamide, dicyanamide, urea, melamine, thiourea, guanidine hydrochloride, trithiocyanuric acid, etc. has been employed to synthesize g−CN. This type of nitrogen-rich precursor has a slight effect on the band-gap structure of g−CN [40]. The flat band potential generated by different precursors was measured using the Mott-Schottky plot, which is −0.6 V vs. RHE for urea [35], −0.8 V for cyanamide and dicyandiamide [41,42], and −0.7 V for melamine [43,44]. The valence band X-ray photoelectron spectroscopy plot, which calculates the energy level difference between the valence band and the Fermi level [45], produced a value of 1.56 V for urea, thiourea and melamine [35,40], and 1.74 V for dicyandiamide [46]. Ragupathi et al. [47] reported a band-gap of 2.89 eV, 2.78 eV, and 2.35 eV for thiourea, urea, and urea with glycine precursor to synthesize g−CN, respectively.

g−CN Thin-Film Preparation Methods
Numerous synthesis strategies have been developed to prepare g−CN films on substrates, which are classified as thermal vapor condensation (TVC), solvothermal route, liquid-mediated growth, seed growth, electrophoretic deposition, disperse coating, and other deposition methods. Detailed descriptions of the various types of synthesis methods for the construction of g−CN films are provided below. The main advantages and disadvantages of the different synthetic methods are summarized in Table 1. Table 1. The main advantages and disadvantages comparison of different methods.

Thermal Vapor Condensation (TVC)
The TVC setups for the deposition of g−CN thin film are shown in Figure 4d,h. Owing to the different deposition sequences, TVC can be categorized into two forms: the one-step process and the two-step process. In a one-step TVC procedure (Figure 4d), the substrate is placed on top of a reactor. Taking nitrogen-rich precursor for an example, the precursor vapor generates if the temperature is higher than its sublimation point (above 300 • C). Then, the vapor inside the container will deposit on the substrate surface when the temperature reaches the condensation temperature (500-600 • C) of the precursor, resulting in the nucleation and growth of g−CN thin film [48]. While for the two-step TVC procedure, g−CN is condensed and deposited on the inner wall of the container upon the first vapor deposition. Subsequently, the substrate is placed on top of a reactor. The g−CN attached to the container is vaporized again and grown on the g−CN thin film on the substrate surface [49]. To date, g−CN thin film has mainly been prepared by TVC with different precursors, such as melamine [50], dicyanamide [49], thiourea [51], urea [52], and cyanamide [53], and the bulky g−CN is always utilized. Bian et al. discovered that melamine, compared with dicyandiamide, thiourea, and urea, shows the highest photocurrent density for the PEC system in a one-step TVC procedure [48]. The g−CN film fabricated with dicyanamide shows a photocurrent density of up to 63 µA cm −2 at the bias of 1.23 V versus RHE without a sacrificial reagent in a two-step TVC procedure (Figure 4f-g) [49]. density for the PEC system in a one-step TVC procedure [48]. The g−CN film fabricated with dicyanamide shows a photocurrent density of up to 63 μA cm −2 at the bias of 1.23 V versus RHE without a sacrificial reagent in a two-step TVC procedure (Figure 4f-g) [49].  The dependency of the properties of g−CN films on the influence factors, including deposition temperature, altering the substrate, precursor amount, and sacrificial reagent, will be compared and discussed. The vapor deposition method has been widely explored for depositing uniform and dense g−CN films on various substrates, such as indiumdoped tin oxide (ITO) glass, fluorine-doped tin oxide (FTO) glass, silicon slice, quartz glass, and glass. Bian et al. discovered that the morphology of the g−CN film is substratedependent, grown on ITO glass and silica are similar to that of on FTO glass; they are . TVC via melamine, illustrating AFM surface morphologies grown on (a) ITO glass, (b) silica and (c) glass, (d) TVC growth mechanism, and (e) transient photocurrent density of the thin film tested with and without adding Na 2 S at 1.55 V vs. RHE. Reprinted with permission from ref. [48] Copyright 2015 Elsevier. Two-step TVD procedure utilizing dicyanamide: (f) SEM images of the g−CN electrodes (the inset is the digital image), (g) the corresponding transient photocurrent density in 0.1 M Na 2 SO 4 at 1.23 V vs. RHE, and (h) two-step TVC growth mechanism. Reprinted with permission from ref. [49] Copyright 2017 Elsevier.
The dependency of the properties of g−CN films on the influence factors, including deposition temperature, altering the substrate, precursor amount, and sacrificial reagent, will be compared and discussed. The vapor deposition method has been widely explored for depositing uniform and dense g−CN films on various substrates, such as indium-doped tin oxide (ITO) glass, fluorine-doped tin oxide (FTO) glass, silicon slice, quartz glass, and glass. Bian et al. discovered that the morphology of the g−CN film is substrate-dependent, grown on ITO glass and silica are similar to that of on FTO glass; they are continuous films with many nanoparticles on the surface. However, inhomogeneous particles and nanorods formed on glass (Figure 4a-c) [48]. The sacrificial reagent is able to not only facilitate hole consumption, which can decrease the recombination of electron-hole pairs but also is significantly affected by the holes in the energy levels where they cannot react with OH − . The g−CN film prepared with melamine shows a photocurrent density up to 30 µA cm −2 versus RHE without sacrificial reagent under one sun illumination at 1.55 V. When sacrificial reagent was added to electrolytes, the highest photocurrent density reached up to 120 µA cm −2 (Figure 4e) [48]. Both the thickness and the optical light absorption of the films increase with the increasing mass precursor. The increased film thickness reduces the transfer resistance of electron-hole pairs between film and electrolyte interfaces, but a high mass precursor makes the films partly peel off [49]. Based on the experiment analysis, an increased deposition temperature is beneficial to enhance the optical light absorption and increase the thickness of the films. This indicates that the light absorption alongside the films shifts towards red-shift as the deposition temperature increases, which is very consistent with the absorption alongside the change of the bulk g−CN [54,55]. However, the high process temperature (above 650 • C) makes substrates largely deform [48], decreasing the uniformity and the PEC performance.
The luminescence mechanism of the g−CN thin film was reported by Urakami et al. [56], who successfully realized high-quality g−CN films on different types of substrates, such as SiO 2 , HfO 2 , and c-plane sapphire through TVC at 600 • C. A crystalline g−CN thin film can be deposited on both non-crystalline and crystalline surfaces by employing an amorphous-like buffer layer. The bonding states and stoichiometric chemical composition (N/C ratio was 1.36-1.39) are close to those of ideal g−CN. From the dependence of the photoluminescence (PL) intensity and peak shift on the excitationpower-density, electron injection from the sp 3 C-N σ to the sp 2 C-N π conduction band was strengthened under the above excitation condition (E ex > E g σ) (Figure 5a,b). Nanomaterials 2022, 12, 2374 9 of 60 thickness reduces the transfer resistance of electron-hole pairs between film and electrolyte interfaces, but a high mass precursor makes the films partly peel off [49]. Based on the experiment analysis, an increased deposition temperature is beneficial to enhance the optical light absorption and increase the thickness of the films. This indicates that the light absorption alongside the films shifts towards red-shift as the deposition temperature increases, which is very consistent with the absorption alongside the change of the bulk g−CN [54,55]. However, the high process temperature (above 650 °C) makes substrates largely deform [48], decreasing the uniformity and the PEC performance. The luminescence mechanism of the g−CN thin film was reported by Urakami et al. [56], who successfully realized high-quality g−CN films on different types of substrates, such as SiO2, HfO2, and c-plane sapphire through TVC at 600 °C. A crystalline g−CN thin film can be deposited on both non-crystalline and crystalline surfaces by employing an amorphous-like buffer layer. The bonding states and stoichiometric chemical composition (N/C ratio was 1.36-1.39) are close to those of ideal g−CN. From the dependence of the photoluminescence (PL) intensity and peak shift on the excitation-power-density, electron injection from the sp 3 C-N σ to the sp 2 C-N π conduction band was strengthened under the above excitation condition (Eex > Eg σ) (Figure 5a,b).

Figure 5.
Excitation-power dependence of (a) the integrated PL intensities and (b) PL peak shift for σ*-LP and π*-LP transition paths of the thin film. Reprinted with permission from ref. [56] Copyright 2019 Japan Society of Applied Physics.
Based on the above analysis, TVC is considered a very simple, convenient, and effective method to prepare g−CN films. In addition to morphology, the thickness and structure of the g−CN film can easily be controlled. However, the number of precursors that can be utilized is restricted. Other frequently reported precursors fail to form g−CN films on solid substrates [48]. As is well known, NH2 groups play an essential role in nucleation. However, a lack of one NH2 group in each precursor molecule is proven to fail to realize high-quality g−CN thin film [53]. Based on the above analysis, TVC is considered a very simple, convenient, and effective method to prepare g−CN films. In addition to morphology, the thickness and structure of the g−CN film can easily be controlled. However, the number of precursors that can be utilized is restricted. Other frequently reported precursors fail to form g−CN films on solid substrates [48]. As is well known, NH 2 groups play an essential role in nucleation. However, a lack of one NH 2 group in each precursor molecule is proven to fail to realize high-quality g−CN thin film [53].

Solvothermal Route
The solvothermal route is a procedure to arrange the covalent bond between the molecular precursors and the substrate surface [57,58]. It includes two essential steps ( Figure 6a): Firstly, the substrate is placed in an organic solvent that is filled with nitrogenrich precursor solutions. Subsequently, the g−CN films were annealed for the purpose of further polymerization and enhancing the contact between film and substrate. Solvothermal approach utilizing melamine and cyanuric chloride: SEM images of (b-d) g−CN sample before annealing and (e-g) g−CN sample after annealing and (h) UV-vis absorption spectra (the inset shows the Tauc plots). Reprinted with permission from ref. [57] Copyright 2016 Royal Society of Chemistry. Solvothermal approach using cyanuric chloride and cyanuric acid: SEM images of the top view (i) and cross-section (j) of annealed g−CN thin film (the inset is the digital image). Reprinted with permission from ref. [58] Copyright 2017 Royal Society of Chemistry. (k) Liquid mediated growth mechanism, (l-o) SEM images of the structured PCN films with different molar ratios (the mixture of melamine and trithiocyanuric acid) after condensation at 500 °C for 4 h, and (p) the growth mechanism of the PCN films by combination of sulfur and non-sulfur mixtures and mechanism for water splitting. Reprinted with permission from ref. [59] Copyright 2014 American Chemical Society.
Using the solvothermal method with post-annealing, g−CN films can be grown on different substrates, such as FTO [57,58], glass [58], and TiO2/FTO [58]. Xie et al. fabricated g−CN films on FTO via simply dissolving cyanuric chloride and melamine in acetonitrile, as described in Figure 6a. The solutions were maintained at 180 °C for 24 h, and the g−CN film followed the Stranski-Krastanov growth mode. A transition from the two-dimensional layer-by-layer mode (shorter than 12 h) to three-dimensional micro-spheres stacking mode was observed owing to the changes in the surface energy of substrates and the Figure 6. (a) Schematic illustration of the solvothermal synthesis route. Solvothermal approach utilizing melamine and cyanuric chloride: SEM images of (b-d) g−CN sample before annealing and (e-g) g−CN sample after annealing and (h) UV-vis absorption spectra (the inset shows the Tauc plots). Reprinted with permission from ref. [57] Copyright 2016 Royal Society of Chemistry. Solvothermal approach using cyanuric chloride and cyanuric acid: SEM images of the top view (i) and crosssection (j) of annealed g−CN thin film (the inset is the digital image). Reprinted with permission from ref. [58] Copyright 2017 Royal Society of Chemistry. (k) Liquid mediated growth mechanism, (l-o) SEM images of the structured PCN films with different molar ratios (the mixture of melamine and trithiocyanuric acid) after condensation at 500 • C for 4 h, and (p) the growth mechanism of the PCN films by combination of sulfur and non-sulfur mixtures and mechanism for water splitting. Reprinted with permission from ref. [59] Copyright 2014 American Chemical Society.
Using the solvothermal method with post-annealing, g−CN films can be grown on different substrates, such as FTO [57,58], glass [58], and TiO 2 /FTO [58]. Xie et al. fabricated g−CN films on FTO via simply dissolving cyanuric chloride and melamine in acetonitrile, as described in Figure 6a. The solutions were maintained at 180 • C for 24 h, and the g−CN film followed the Stranski-Krastanov growth mode. A transition from the twodimensional layer-by-layer mode (shorter than 12 h) to three-dimensional micro-spheres stacking mode was observed owing to the changes in the surface energy of substrates and the properties of solutions. However, this process can hardly form a complete tri-s-triazine structure g−CN film because of insufficient kinetic energy. An annealing process, annealed in N 2 at 520 • C for 1 h, for example, plays a key role in forming a tri-s-triazine structure. The interfaces of microspheres and nanoparticles become vague after a post-annealing treatment, manifesting the thermal cross-linking and condensation sequentially (Figure 6b-g). The obtained g−CN films showed higher activity in PEC performance of this technique, with photocurrent densities of 3.5 µA cm −2 at 1.23 V vs. NHE produced, compared with traditional paste-prepared g−CN. The improvement was mainly attributed to the following two factors. Firstly, the presence of covalent linkage between g−CN and the substrate generates synergistic effects. Secondly, a red shift from approximately 460 nm to 620 nm is accompanied by the diminished band gap of 2.0 eV, narrower than the 2.7 eV of the conventional g−CN (Figure 6h) [57].
Recently, a significant improvement was achieved by Gu et al. with a facile solvothermal method. Using cyanuric chloride and cyanuric acid as precursors, highly crystalline g−CN thin films were obtained on different substrates with post-annealing at 450 • C. In consequence, the photocurrent densities of 10 µA cm −2 at 1.23 V vs. RHE are produced. The resulting g−CN film exhibits enlarged conjugating structures of tri-s-triazines and s-triazines, allowing uniform and continuous interfacial contact (Figure 6i,j). Moreover, when the film was grown on the surface of anatase TiO 2 , the photocurrent densities were further enhanced owing to vectorial charge transfer induced by the layer-by-layer configuration [58]. This work manifests the vital effects of morphology and structure on the properties of carbon nitride films for PEC water splitting.
In addition, an effective tactic for controlling the thickness and density of g−CN films can be achieved by changing the growth time, precursor concentration, and post-heating temperature. The improvement of PEC performance was attributed to the presence of covalent linkage due to introducing either cyanuric acid or cyanuric chloride, which enhances the charge carrier transportation. Furthermore, both studies reported a diminished band gap due to improved π-electron delocalization of the conjugated structure, widening the light absorption range [57,58]. However, the obtained g−CN films for PEC water splitting were not as good as we expected, seemingly aroused by the rich surface defects of carbon nitride films. Developing new solvents and combining multiple precursors to tailor g−CN is particularly important.

Liquid-Mediated Growth
To obtain uniform, transparent, and reproducible g−CN solid thin films, the precursor in the liquid reaction system should be granted priority because liquid-mediated growth provides more intimate contact with the substrate than its common solid-phase counterpart.
The process of liquid-mediated growth is shown in Figure 6k [59]. Firstly, the substrate is fixed to the bottom of a crucible that is filled with nitrogen-rich solid precursors. Upon heating, this solid precursor melts into the liquid phase, obtaining a chummy contact with the surface of the substrate. Finally, post-annealing condensates the liquid phase and eventually turns it into a solid film. The obtained film is continuous, stable, and handin-glove adhered to the surface of various substrates, including FTO, ITO, glass, metal foil, and TiO 2 , providing great potential for the fabrication of high-performance devices. The supramolecular pre-assembly strategy has attracted extensive attention because its precursors show an identical chemical structure to g−CN and effectuate liquid-mediated growth [60]. Shalom et al. [61] utilized this strategy to form ordered structures of g−CN via hydrogen-bonded cyanuric acid-melamine supramolecular complex. The solid complex locals between two desired substrates were post-annealed to form homogeneous and comparatively organized g−CN rods. Besides its hydrogen bonds, the complex appears to possess free hydroxyl and amine groups which can be grown on varying substrates (glass, FTO, TiO 2 , ZnO, etc.) with diverse morphologies of g−CN. The electrode demonstrated current densities could reach up to 0.8 mA cm −2 at 0.6 V (vs. RHE) in neutral media.
In another similar work by Xu and co-workers, [59] the obtained g−CN film exhibited continuous porous network morphology on various substrates since no template was utilized during the fabrication. The mechanical robustness of the g−CN films fabricated utilizing a liquid-mediated pathway was very high, obtaining thin films that were not peeled off via sonication in water for 1 h. Moreover, g−CN films fabricated through supramolecular aggregates improve the optical and electronic properties because of their enhanced structure intactness. Inspired by these facts, Xu and co-workers [62] presented a liquid-mediated approach to preparing continuous phenyl-modified carbon nitride films via combining molten sulfur and 2,4-diamino-6-phenyl-1,3,5-triazine-cyanuric acid complexes. The photocurrent density of the prepared phenyl-substituted g−CN photoelectrode was up to 60 µA cm −2 , which was 20-fold higher than that of non-sulfur-processed. The high-temperature molten sulfur integrates phenyl groups into the thin film effectively to enhance the conductivity and charge transfer, which is beneficial to improving the PEC efficiency. The intermediate state plays a crucial role in the monomer's preorganization for the liquid-phase thin-film growth, owing to its influence on the supramolecular interactions [63]. Aside from film quality and intermediate state, the insufficient binding between the functional films and conductive substrate is also considered a major issue that hinders Polymeric carbon nitride (PCN) films for solar water splitting. To overcome this challenge, Fang et al. proposed a combination precursor of sulfur and non-sulfur complexes for the preparation of polymeric carbon nitride films [64]. It is found that sulfur exists on the PCN/FTO glass interface, which not only initializes the formation of the PCN films but also assists the charge migration between films and the substrate (Figure 6l-p). A preeminent photocurrent density of 100 µA cm −2 was achieved at 1.23 V (vs. RHE) under AM 1.5 illumination without sacrificial reagents. Inspired by the aforementioned method, the Ni chloride salt molecules were also introduced into the cyanuric acid and 2,4-diamino-6-phenyl-1,3,5triazine supramolecular aggregates. The band gap and catalytic properties of the obtained g−CN films could be regulated via altering the Ni amount, and the PEC properties were prominently improved. The Ni atoms within the carbon nitride layers significantly extend the light absorption range and enhance the charge transfer properties [65].

Seed Growth
Seed growth has been developed to fabricate g−CN films, and it includes three main steps: seeding the substrate, soaking the seeded layer into a hot supersaturated precursor solution, and calcination. The seeds are protected by the supersaturated precursor solution to prevent dissolution, which acts as nucleation sites for the crystallization of precursor from the solution. Upon cooling, precursor crystals spontaneously grew on the seeded substrate surface to form a well-covered film. Upon calcination of the as-prepared film, a g−CN layer was synthesized on the substrate. Shalom et al. [66] grew highly ordered carbon-nitrogen films via the seeded crystallization of carbon-nitrogen monomers, leading to a photocurrent density of 116 µA cm −2 at 1.23 V versus RHE (Figure 7a). A similar approach was used for the fabrication of g−CN-based films on conductive carbon paper [67]. Recently, Shalom et al. [68] developed a new approach to growing thick g−CN layer films via a spraycoated seeding layer composed of CN monomers, leading to an impressive photocurrent density of 300 µA cm −2 at 1.23 V versus RHE in the presence of a hole scavenger (Figure 7b), implying the high scalability potential of this method for the flexible material and device applications.

Electrochemical Deposition
To date, there are two representative routers of electrochemical deposition, including (1) electrodeposition (ED) in the involvement of ionic precipitation [35,69,70] and (2) electrophoretic deposition (EPD) based on the adsorption of charged granules onto externally biased electrodes [45,71]. In the last few years, there has been growing research interest in the fabrication of g−CN-based photoelectrodes by utilizing electrochemical deposition methods. Xu et al. [71] presented a mild pathway to construct a uniform g−CN film through the EPD approach under an anodic bias of 200 V, which can be successfully grown on different substrates such as FTO and carbon paper as well as nickel foam with complex 3D geometries. Toluene was used as a solvent owing to its large electrochemical window. The films deposited on carbon paper proved a higher photocurrent response of 12 µA cm −2 compared with 1 µA cm −2 on FTO under zero bias. The photocurrent was further improved significantly to 65 µA cm −2 by doping with barbituric acid. Therefore, tailored construction of these modified carbon nitride layers led to efficient contact with electrolytes due to the high surface area of 3D carbon paper and enhanced light-harvesting and conductivity via heteroatom doping ( Figure 7c

Disperse Coating
In spite of the fact that g−CN powder has been extensively and deeply researched, the preparation knowledge of g−CN film is still in its infancy. Recently, disperse coating strategy from colloid sources has also been reported extensively owing to its extraordi-

Disperse Coating
In spite of the fact that g−CN powder has been extensively and deeply researched, the preparation knowledge of g−CN film is still in its infancy. Recently, disperse coating strategy from colloid sources has also been reported extensively owing to its extraordinary features, including facile and cost-effectiveness. It has been applied to the deposition of carbon nitride thin film on various substrates via conventional filming technologies, such as spin coating [73], silk-screen printing [74], drop-casting [75], dip casting [76], doctor blade coating (DBC) [77], and Sol-gel process [42]. The uniformity and performance of the g−CN films mainly depend on the influenced various parameters of g−CN colloids in terms of chemical bonding and size of particles [42].
In order to fabricate the uniform g−CN thin films, the principal requirement is to obtain a stable dispersion in appropriate solvents such as alcohol, water, and methanol. Several researchers gained g−CN dispersions in water [78,79] or the combination of alcohol and water [80,81]. Nevertheless, the existence of water has always been restricted to obtain uniform thin films due to its high boiling point and surface tension. Furthermore, due to the poor solubility of g−CN powder in most solvents, the fabricated g−CN films usually present poor physical contact with the substrate as well as boundary defects, which results in a weak photocurrent response of less than 1 µA cm −2 [41,82,83].
As we know, exfoliation of two-dimensional structures shows abundant surface active sites and effectively captures the visible light compared to bulk g−CN [79,84,85]. From the practical application viewpoint, the alteration of bulk g−CN by two-dimensional g−CN nanosheets with a unilamellar structure via adopting a physical modification process (ultrasound-assisted or hydrolysis), chemical exfoliation with different solvents (acids, salts, bases, etc.), and physicochemical methods (heating and hot water) are widely used. Zhang et al. [77] demonstrated that the protonation of g−CN using strong acid (HNO 3 ) results in a sol, which allows the fabrication of films through facile dip coating. The melon units stack together through hydrogen bonds between the nitrogen atom and NH/NH 2 group, which in principle can be used as proton acceptors. To the best of our knowledge, a stable g−CN thin film with high photocurrent density not only benefits from the postannealing process but, more importantly, should preserve its unique physicochemical features after robust HNO 3 etching. As a result, there is great potential for applying the solgel technique for carbon-nitrogen PEC application, which shows enhanced PEC response compared to powder-based thin films.
Despite significant advances in the exfoliation of g−CN powders, the finding of ultrathin or monolayer g−CN nanosheets through strong acid treatment or chemical exfoliation undergo slight defects due to the fact that the preparation process is not environmentally friendly. Therefore, scientists suggested an urgent demand to obtain a stable and environmentally friendly dispersion for the preparation of g−CN films, which are solvents. Safaei et al. [86] demonstrated that methanol is preferable for layered g−CN due to the short drying time and significantly low surface tension. Inspired by this idea, a very promising work was reported by Mohamed and co-workers [87], who showed a simple synthetic method for the fabrication of a thinner layer by layer from bulk g−CN by employing a slow direct low thermal treatment (40 • C, 24 h) with ultrasonication, which was obtained as the highly stable g−CN suspension in a methanol solution. The exfoliation of nanosheets from bulk g−CN can form a uniform layer of g−CN thin films via using a spin-coating method followed by an appropriate post-annealing temperature at 350 • C. The obtained g−CN thin film presents a more negative conduction and valence band potentials as well as a much rougher surface (Sa roughness: 14.58 nm), which are generally helpful to PEC performance. Moreover, the photocurrent responses of the prepared g−CN films reached up to 10.21 µA cm −2 at 1.23 vs. (Ag/AgCl), which was two times higher than that of bulk g−CN.
The post-annealing temperatures also play an important role in the morphology and the chemical bonding of the g−CN thin film that ultimately influences the PEC efficiency. Zheng and co-workers [88] investigated that post-annealing transforms the hollow carbon nitride nanospheres into the triazine unit. Mohamed et al. [52] found that post-annealing improves the chemical bond structure and morphology of g−CN fabricated by the spin-coating method. In addition, there is a direct relationship between post-annealing temperature and band gap. The band gap reduces from 2.79 eV to 2.71 eV by increasing the post-annealing temperature from 150°C to 350 • C, while it increases from 2.71 eV to 2.85 eV at higher temperatures (350 • C to 500 • C), the energy band gap of the g−CN thin film was enlarged. This phenomenon is ascribed to the quantum confinement effect and the shift of the CB and VB in opposite directions [88].
Whereas using the supramolecular assembly affords thin dense layers, applying a paste made up of supramolecular precursors in dispersant (such as ethylene glycol) enabling hydrogen bonds leads to porous layers of controllable thickness. It is noteworthy that the carbon-nitrogen monomers can connect with triazine derivatives (e.g., barbituric acid and cyanuric acid) to form supramolecular paste via hydrogen bond interactions [22]. Changing the supramolecular precursors can control the elemental ingredient as well. Therefore, this method enables the application of the knowledge to progress from powder-based g−CN synthesis to g−CN-based electrodes. Based on the supramolecular pre-assembly approach, an ameliorative work based on g−CN films was reported by Shalom and co-workers, who presented a general and simple pathway for the construction of highly porous and large-scale g−CN films through forming a supramolecular paste, which can be successfully grown on different substrates using the doctor blade technique [89]. g−CN has a 2D layered structure connected by van der Waals forces. When it is stripped into multilayer or singlelayer nanosheets, unique chemical and physical properties are produced to further improve PEC efficiency, including (1) larger specific surface area to increase the density of the reactive center, (2) reduced thickness to reduce the carrier migration distance and effectively inhibit carrier recombination, and (3) unique two-dimensional flexible planar structure to enhance compatibility with heterojunction construction, cocatalyst modification, vacancy, and other modification strategies.

Other Deposition Methods
Some other ex situ methods have also been developed to fabricate g−CN electrodes successfully, including anodic aluminum oxide (AAO) membrane deposition [90], direct growth [61], vacuum filtration deposition [91,92], electrospinning (ESP) [77], vacuum magnetic filtered arc ion plating (VMFAIP) [93], ionothermal (molten salt) deposition [94], radio frequency magnetron (RFM)−based sputtering [95], and so on. A microcontactprinting process was reported by Liu et al., who fabricated g−CN nanostructures by injecting cyanamide into an anodized aluminum oxide (AAO) membrane sandwiched between two substrates and subsequently calcinating at 550 • C for 4 h (Figure 8a). The AAO is not a versatile template but a guide for gas-phase precursors to reach and react with the sandwiching substrates. The film thicknesses can be simply modulated by varying the concentration of cyanamide. The photocurrent density can reach about 30 µA cm −2 at 1.23 V vs. RHE (Figure 8a-e) [90].
Electrospinning as a practicably proven method is also used to fabricate the PEC photoanode, in which the ITO substrate is coated with prepared polyvinylpyrrolidone (PVP)/g−CN sol and annealed at 500 • C. The thickness of the films can be modulated by varying the electrospinning time and actually controlling the PEC performance. Furthermore, the photoelectrode demonstrates a stable photocurrent performance (6.64 µA cm −2 ) for 4000 s with no degradation (Figure 8f-h) [77]. Wang et al. developed a vacuum magnetic filtered arc ion plating (VMFAIP) approach, in which the carbon plasma emitted from a graphite cathode was deflected by a magnetic field and combined with N 2 to generate C-N compound on the FTO substrate. The resulting g−CN nanorod (NR) arrays eliminated the intralayer hydrogen bonds owing to the utilization of carbon and nitrogen sources alone. The change of the graphite target current allowed different lengths of the g−CN nanorod arrays from 550 nm to 1.1 µm. The gas-based approach via the VMFAIP system enabled the g−CN to form heterostructures easily with, including but not limited to, TiO 2 nanorod arrays (Figure 8i-k) [93]. Ladva et al. [94] synthesized g−CN thin films with a bandgap of approximately 1.73 eV by an ionothermal (molten salt) approach using dicyandiamide, KBr, and LiBr mixture precursors. Lou et al. [96] demonstrated a g−CN layer on the substrate by embedding g−CN in a matrix conductive polymer binder (PEDOT-PSS [96][97][98], Nafion [57,99,100], and others) [101]. Compared with Nafion, the g−CN film blended by PEDOT-PSS showed superior PEC activities because the PEDOT-PSS binder is a wonderful conducting polymer and can act as a hole-transport layer. Recently, Tejasvi et al. [95] synthesized stoichiometric g−CN thin films by radio frequency magnetron sputtering using a g−CN pellet as a target. Thin-film samples can be deposited on the borosilicate glass and TiO 2 nanotube array in Ar and N 2 plasma media. The PEC studies summarized that the g−CN/TiO 2 formed under Ar plasma (~290 µA/cm 2 at 0.9 V) obtained a higher current density as compared to that under N 2 plasma (~160 µA/cm 2 at 0.9 V). gration distance and effectively inhibit carrier recombination, and (3) unique two-dimen-sional flexible planar structure to enhance compatibility with heterojunction construction, cocatalyst modification, vacancy, and other modification strategies.

Other Deposition Methods
Some other ex situ methods have also been developed to fabricate g−CN electrodes successfully, including anodic aluminum oxide (AAO) membrane deposition [90], direct growth [61], vacuum filtration deposition [91,92], electrospinning (ESP) [77], vacuum magnetic filtered arc ion plating (VMFAIP) [93], ionothermal (molten salt) deposition [94], radio frequency magnetron (RFM)−based sputtering [95], and so on. A microcontactprinting process was reported by Liu et al., who fabricated g−CN nanostructures by injecting cyanamide into an anodized aluminum oxide (AAO) membrane sandwiched between two substrates and subsequently calcinating at 550 °C for 4 h (Figure 8a). The AAO is not a versatile template but a guide for gas-phase precursors to reach and react with the sandwiching substrates. The film thicknesses can be simply modulated by varying the concentration of cyanamide. The photocurrent density can reach about 30 μA cm −2 at 1.23 V vs. RHE (Figure 8a-e) [90]. Although various techniques have been employed to prepare g−CN electrodes, g−CN films prepared by ex situ methods mostly show poor photocurrent density (several µA cm −2 ), which can be ascribed to the uneven coverage and the weak adhesion of g−CN on the substrates [83]. In contrast, the g−CN films in situ synthesized on various conductive substrates by a bottom-up route generally exhibit good photocurrent density (10-120 µA cm −2 ) [66]. Shalom et al. [66] grew highly ordered g−CN films via the seeded crystallization of melamine monomers, leading to a photocurrent density of 116 µA cm −2 at 1.23 V versus RHE under AM 1.5G in 0.1 M KOH aqueous solution. This is the best value reported for pure g−CN electrodes without a sacrificial agent. Enormous space still exists for achieving a theoretical photocurrent density of 14.5 mA cm −2 and a maximum solar-to−hydrogen (STH) efficiency of 17.9% under AM 1.5G (100 mW cm −2 ) [24][25][26]. The various types of synthesis methods for the construction of g−CN films with the corresponding photocurrent comparison are summarized in Table 2. In addition, in order to produce g−CN with a uniform texture, casting g−CN embedded in a conductive polymer matrix tactics is highly worth considering. The method has two-fold advantages: it improves the charge transfer in the film and enhances the casting homogeneity. Despite this, enormous space still exists for devising efficient tactics to fabricate homogenous electrodes.

Heteroatoms Doping
When pure g−CN films are employed as photocatalysts, their practical applications are still hindered by limited theoretical photocurrents. The p-p* transitions in the conjugated aromatic system of g−CN result in a band gap of 2.7 eV, suppressing the absorption above 460 nm [75]. Doping of additional atoms and impurities into the g−CN matrix is considered an effective strategy to ameliorate structural defects and distinctly tune its physicochemical properties, including extending visible-light absorption scope and regulating redox potentials to further promote the PEC properties. There are two main types of elemental doping that will be explored in this section, namely nonmetallic elements doping and metallic elements doping.
Qi and co-workers [102] prepared Ag-doped g−CN films (Ag/g−CN) via a liquidbased reaction process employing AgNO 3 as the Ag dopant. The AgNO 3 strongly affects the PEC efficiency of the Ag/g−CN films. The optimized photocurrent density was as high as 6.40 µA mm −2 for the Ag/g−CN films (molar ratio of 1: 10), which was approximately 6.8−fold higher than that of a pristine g−CN film electrode. This is due to the efficient charge transfer and improved light-harvesting property of Ag-doped g−CN film. Apart from Ag, Karimi-Nazarabad and co-workers [103] developed a new type of Pd-doped g−CN photoanode via an electrophoretic deposition (EPD) strategy. A Pd-doping amount of 2 at% into g−CN exhibits a transient photocurrent density of 79.2 µA cm −2 , which is notably higher than those of Ag/g−CN and pristine g−CN films. Additionally, the photoelectrocatalytic oxygen evolution performance is improved due to the facilitated transfer efficiency of photo-generated charge carriers, which can be ascribed to the deeper band bending and the remarkable trapping capability for electrons. Recently, doping engineering with Thorium Nitrate (ThO 2 ) during g−CN polymerization altered the chemical bonding, with the improved PEC activity (Figure 9a) [104]. The Sr-doped g−CN showed enhanced PEC activity [105].
In addition to the metallic atom-doping mentioned above, the incorporation of nonmetallic atoms such as B, C, and S into the g−CN framework was also studied. Compared with metal doping, the superiority of nonmetal doping is distinct due to its capability of reserving the nonmetallic properties of g−CN.
Huang and co-workers [106] presented a facile TVC method to prepare B-doped g−CN films using boric acid as the B source. The XPS analysis clearly confirmed that the B atoms were successfully incorporated into the g−CN framework and replaced the C composition partially (Figure 9b). The B doping effectively reduced the band-gap by raising the valence band and prolonging the charge carrier lifetime by altering spatially orbital distributions. Optical band-gap energies decreased from 2.77 to 2.60 eV with the increasing boric acid content in the precursor (0 to 0.8 g), extending the photon absorption edge (Figure 9c). The photocurrent density of the optimized B-doped g−CN films was as high as 55 µ Acm −2 at 1.23 V vs. RHE, which was four-fold higher than that of pristine g−CN films. The type of species supplied from the precursors and the pyrolysis temperature has a slight effect on the chemical compositions of compound semiconductor alloys [107][108][109]. Therefore, to control additional atoms composition and band-gap energy, the influence of pyrolysis conditions on additional atoms incorporation into the C-N network should be expounded. Inspired by these facts, Urakami and co-workers [110] reported a simple thermal chemical vapor deposition (CVD) route to prepare the B-incorporated g−CN film by employing ammonia borane (H 3 N-BH 3 ) as the B source. It was also noted that the pyrolysis temperature had an intimate connection with the compositional ratios, and the preferable temperature for B doping ranged from 618 • C to 650 • C ( Figure 9d). Importantly, a shift of the PL peak from 2.7 eV to 3.60 eV was manifested by B-incorporation with a composition of 8.0%. The PL peak energies for σ*−LP and π*−LP shift routes were in accordance with the quadratic function, manifesting that the band-gap bowing appeared as in orthodox compound semiconductor alloys (Figure 9e,f). The green, blue, and pink colors represent C, N, and B atoms, respectively, and (c) the UV-vis absorption spectra (the inset shows the Tauc plots). Reprinted with permission from ref. [106] Copyright 2018 Elsevier. (d) Substrate temperature dependence of compositions, (e) typical RT-PL spectra for B-doped g−CN films with various compositions, the black line is the a peak energy corresponds to unintentionally doped g−CN films (π*-LP), the blue line is the a peak energy corresponds to unintentionally doped g−CN films (σ*-LP), and the pink line is the a peak energy corresponds to B-incorporated g−CN with a composition of 8.0% (σ*-LP) and (f) room temperature integrated PL intensity and PL peak energy depended on B composition. Reprinted with permission from ref. [110] Copyright 2020 Wiley-VCH.
In addition, a novel two-step surface boron decoration of g−CN is currently gaining attention as this approach shows active influences on the structural and optical properties [111]. The strong van der Waals force attractions between sp 2 carbon atoms make g−CN poorly dispersible in an aqueous solution, and the strong hydrophilicity is key to ensuring the full development of g−CN-based materials in an aqueous medium. A structure of hydrophilic and high surface area assembled by two-dimensional B-doped g−CN has been successfully prepared to effectively control the band gap and charge separation ( Figure  10a-c). The electron density difference (EDD) map and work function results show that B-doped g−CN has a higher distribution of electron density (Figure 10d-g), and a larger work function makes a higher degree of band bending, leading to the rapid transfer of electrons, and thus improving the PEC capability (Figure 10h,i) [112]. (d) Substrate temperature dependence of compositions, (e) typical RT-PL spectra for B-doped g−CN films with various compositions, the black line is the a peak energy corresponds to unintentionally doped g−CN films (π*-LP), the blue line is the a peak energy corresponds to unintentionally doped g−CN films (σ*-LP), and the pink line is the a peak energy corresponds to B-incorporated g−CN with a composition of 8.0% (σ*-LP) and (f) room temperature integrated PL intensity and PL peak energy depended on B composition. Reprinted with permission from ref. [110] Copyright 2020 Wiley-VCH.
In addition, a novel two-step surface boron decoration of g−CN is currently gaining attention as this approach shows active influences on the structural and optical properties [111]. The strong van der Waals force attractions between sp 2 carbon atoms make g−CN poorly dispersible in an aqueous solution, and the strong hydrophilicity is key to ensuring the full development of g−CN-based materials in an aqueous medium. A structure of hydrophilic and high surface area assembled by two-dimensional B-doped g−CN has been successfully prepared to effectively control the band gap and charge separation (Figure 10a-c). The electron density difference (EDD) map and work function results show that B-doped g−CN has a higher distribution of electron density (Figure 10d-g), and a larger work function makes a higher degree of band bending, leading to the rapid transfer of electrons, and thus improving the PEC capability (Figure 10h,i) [112]. Similarly to B, other non-metallic atoms such as C, P, and S were also examined. Bian and co-workers [53] fabricated a series of C-doped g−CN films by altering the weight percentage of 2, 6−Diaminopyridine and melamine precursors. The embedded gaseous sp 2 C groups influenced the energy band structure (absorption edge) and improved the diffusion length, resulting in the enhancement of PEC activity. The photocurrent density of Cdoped g−CN film (5 wt %) was 0.1 mA cm −2 at 1.23 V versus RHE, which was over fourfold higher than that of pure g−CN. Dang et al. [113] prepared P-doped g−CN by utilizing melamine as the g−CN precursor and triphenylphosphine oxide as the P source through a facile solid thermal melting method (Figure 11a). The optimum efficiency was achieved as 127.4 μA cm −2 , which was five-fold higher than that of pure g−CN. Similar methods were also prepared for the p-doped g−CN with enhanced PEC activity [114].
Other non-metallic atom-doped g−CN photoelectrodes have also attracted extensive attention in PEC applications. Ye et al., 2016 [115], grew a uniform and reproducible g−CN thin film on indium-tin oxide (ITO) substrates by incorporating quantitative thiourea into melamine through a CVD method with a slow heating rate. The S-doped g−CN films displayed a photocurrent of approximately 30 μA cm −2 , which was mainly ascribed to the Sdoping and particularly the modification of surface morphology, crystalline, and band structure by the polymerization of the precursors.
In addition, Fang et al. [116] modified the g−CN by the incorporation of gradient sulfur doping along the films through molten mediated polymerization of melamine and ammonium thiocyanate. As we know, the band edges have an important role in the thermodynamics of the redox reactions. An ideal design of the photoanodic film and the Similarly to B, other non-metallic atoms such as C, P, and S were also examined. Bian and co-workers [53] fabricated a series of C-doped g−CN films by altering the weight percentage of 2, 6−Diaminopyridine and melamine precursors. The embedded gaseous sp 2 C groups influenced the energy band structure (absorption edge) and improved the diffusion length, resulting in the enhancement of PEC activity. The photocurrent density of C-doped g−CN film (5 wt %) was 0.1 mA cm −2 at 1.23 V versus RHE, which was over four-fold higher than that of pure g−CN. Dang et al. [113] prepared P-doped g−CN by utilizing melamine as the g−CN precursor and triphenylphosphine oxide as the P source through a facile solid thermal melting method (Figure 11a). The optimum efficiency was achieved as 127.4 µA cm −2 , which was five-fold higher than that of pure g−CN. Similar methods were also prepared for the p-doped g−CN with enhanced PEC activity [114].
Other non-metallic atom-doped g−CN photoelectrodes have also attracted extensive attention in PEC applications. Ye et al., 2016 [115], grew a uniform and reproducible g−CN thin film on indium-tin oxide (ITO) substrates by incorporating quantitative thiourea into melamine through a CVD method with a slow heating rate. The S-doped g−CN films displayed a photocurrent of approximately 30 µA cm −2 , which was mainly ascribed to the S-doping and particularly the modification of surface morphology, crystalline, and band structure by the polymerization of the precursors.
In addition, Fang et al. [116] modified the g−CN by the incorporation of gradient sulfur doping along the films through molten mediated polymerization of melamine and ammonium thiocyanate. As we know, the band edges have an important role in the thermodynamics of the redox reactions. An ideal design of the photoanodic film and the corresponding charge separation and transfer after the visible absorption are shown in Figure 11b. The valence band maximum of the film should gradually increase from the bottom (close to the substrate) to the top (away from the substrate) versus vacuum potential (vs. Vac), while the conduction band minimum of the film should progressively decrease from the top to bottom (vs. Vac). Therefore, the photo-excited electron will transfer to the counter electrode for hydrogen evolution on the other side. Furthermore, the band-gap of the film should increase from bottom to top. Under illumination, the top of the photoanode will absorb the solar light with a low wavelength. Subsequently, the incident illumination with a higher wavelength will be captured by the bottom active film. As expected, the S-doped g−CN narrowed the band gap from 2.8 to 2.55 eV with a forceful absorption tail in the visible light region. In the depth study, charge separation and migration are enhanced via virtue of the inch-by−inch varied band structures, which makes an agreement with the theoretical example as shown in Figure 11c. As a result, the optimal photocurrent density is ca. 110 µA/cm 2 at 1.23 V vs. reversible hydrogen electrode. This manifestation acts as a platform to form the gradual variation of the band-gap employing metal-free films to enhance the performance of PEC for other extended applications [117][118][119][120].
Lately, g−CN film via co-doping of nonmetal and metal atoms has gained wide attention because this strategy shows a positive impact on the structural and optical properties. Jing et al. [121] reported the preparation of K and I co-doped g−CN with the use of dicyandiamide, KI, and I 2 as the starting materials. It was found that the K&I co-doping transforms the from amphoteric to n-type semiconductor property. The photocurrent density of K&I co-doped g−CN is approximately 50.3 µA cm −2 at 0.5 V versus Ag/AgCl. Luan et al. [122] reported a unique multi-layer doped non-metal g−CN photoanode using a one-pot TVC. In this structure, a P-doped top layer, B-doped middle layer, and pure g−CN bottom layer were designed to enhance the photocurrent density by nearly nine times that of the pure g−CN. This approach effectively forms dual junctions composed of ternary layers to optimize the charge carrier behavior. The calculated electrostatic potential along the c-axis clearly exhibits that the work function gradually decreases from pure g−CN to B-doped g−CN and P-doped g−CN layers, which is conducive to transfer carrier from bulk to surface (Figure 11d,e).
Based on the discussion above, doping additional atoms and impurities into the g−CN matrix was mainly focused on solid or liquid state reactions. A gas-phase reaction using phosphide as a P source was also achieved by Qin et al. [123]. Clearly, a gas reagent can diffuse into the interlayer and produce a more homogeneous modification. Moreover, a simple phosphide gas treatment of TiO 2 nanorod/g−CN not only narrows the band gap into a visible-light range from 2.70 to 2.13 eV but also forms a metal (Cu and Fe) phosphide water oxidation cocatalyst via one-step phosphide annealing process. As a result, the metal phosphide-modified sample yields an outstanding visible light (>420 nm) photocurrent response of approximately 0.3 mA cm −2 at 1.23 V versus RHE (Figure 11f,g). This strategy will provide new insight into other correlative and highly applicable 2D semiconducting materials.
In fact, heteroatom doping is deemed to introduce impurities into g−CN films, with the doping atoms either substituting lattice atoms or dwelling in the in-planar cave of g−CN. In addition, the incorporation of dopant orbitals into the g−CN molecular orbitals enables us to modulate the light absorption and redox band potentials. The above-discussed investigations on the novel doping approach can be further applied to other modified g−CN-based films, which hold immense potential for other applications such as organic pollutants degradation and CO 2 reduction. However, excessive doping of heteroatom is detrimental to PEC efficiency because the appearance of more defects reorganizes the light-generated electron/hole pairs due to the doping asymmetry [119,124,125].

Copolymerization of g−CN
The π−conjugated structure of g−CN is another factor that influences the PEC activity owing to the sp 2 −hybridization of carbon and nitrogen elements. The copolymerization approach allows the integration of structure-matching aromatic groups or organic additives into the g−CN framework to adjust its intrinsic properties [126][127][128]. It was found that molecular doping can modulate the conventional π systems, band structure, electronic properties, optical absorption, and, importantly, PEC performance [129]. Bian et al. [53] copolymerized melamine and 2,6−diaminopyridine (26D) to synthesize g−CN film by the TVC method. The XPS demonstrates that the 26D generates gaseous sp 2 C groups under high temperatures and increases the C/N ratio of g−CN film to promote electron transport. The resulting film shows strong light response properties as well as an extended light absorption region, resulting in the improved photocurrent density from approximately 0.024 to 0.1 mA cm −2 at 1.23 V versus RHE (nearly four times).
In another work, Jia et al. [130] prepared a compact 2D carbon nitride polymer film (CNP) on FTO through an evaporation polymerization strategy, utilizing melamine and cyanuric chloride as raw materials. The photocurrent density of the CNP film is 230 μA cm −2 at 1.23 V (vs. RHE) under visible light irradiation without sacrificial reagents and cocatalysts, which is 7.5, 176.9 and 67.6 times higher than that of the CN-Me (melamine as a precursor), CN-CC (cyanuric chloride as a precursor), and bulk g−CN powders,

Copolymerization of g−CN
The π−conjugated structure of g−CN is another factor that influences the PEC activity owing to the sp 2 −hybridization of carbon and nitrogen elements. The copolymerization approach allows the integration of structure-matching aromatic groups or organic additives into the g−CN framework to adjust its intrinsic properties [126][127][128]. It was found that molecular doping can modulate the conventional π systems, band structure, electronic properties, optical absorption, and, importantly, PEC performance [129]. Bian et al. [53] copolymerized melamine and 2,6−diaminopyridine (26D) to synthesize g−CN film by the TVC method. The XPS demonstrates that the 26D generates gaseous sp 2 C groups under high temperatures and increases the C/N ratio of g−CN film to promote electron transport. The resulting film shows strong light response properties as well as an extended light absorption region, resulting in the improved photocurrent density from approximately 0.024 to 0.1 mA cm −2 at 1.23 V versus RHE (nearly four times).
In another work, Jia et al. [130] prepared a compact 2D carbon nitride polymer film (CNP) on FTO through an evaporation polymerization strategy, utilizing melamine and cyanuric chloride as raw materials. The photocurrent density of the CNP film is 230 µA cm −2 at 1.23 V (vs. RHE) under visible light irradiation without sacrificial reagents and cocatalysts, which is 7.5, 176.9 and 67.6 times higher than that of the CN-Me (melamine as a precursor), CN-CC (cyanuric chloride as a precursor), and bulk g−CN powders, respectively. The improved PEC performance is ascribed to the enhanced charge transport, increased number of terminal amino groups, and the efficient visible light harvesting caused by the intimate contact (Figure 12a,b). According to the report by Xiong et al. [131], a synergistic protocol is employed to combine the formaldehyde modification and sulfur doping methods to tune (100) and (001)  respectively. The improved PEC performance is ascribed to the enhanced charge transport, increased number of terminal amino groups, and the efficient visible light harvesting caused by the intimate contact (Figure 12a,b). According to the report by Xiong et al. [131], a synergistic protocol is employed to combine the formaldehyde modification and sulfur doping methods to tune (100) and (001)   In view of the advantages of chemical structure modification and copolymerization, Kumar et al. [132] synthesized C3N5 polymer by applying melamine-derived melem (2,5,8−triamino-s−heptazine) and inorganic monomer (hydrazine hydrate) to generate intermediate melem hydrazine (MH) through hydrothermal reaction, following annealing at 450 °C for 2 h. In the C3N5 framework, two s-heptazine units are bridged together by azo linkage (−N≡N−), which is a completely different bonding mode compared to C3N4, where three heptazine units are connected with tertiary nitrogen. Owing to overlap between the π orbitals of azo-bridged units and the π system of heptazine motif, the bandgap of C3N5 was markedly reduced to 1.76 eV, which enables the optical light absorption to expand up to 670 nm. The photocurrent response for the C3N5 and g−C3N4 blended TiO2 electrode was proved to be 152 and 100 μA cm −2 at 1.23 V vs. NHE (Figure 13a-e). Recently, a high dispersible carbon nitride material functionalized with polycyclic aromatic hydrocarbons (PAH-CNs) was first introduced by Karjule et al. [133]. New starting monomers with substituted triazines were fabricated through a microwave (MW) irradiation strategy using dicyandiamide and aryl-nitrile as raw materials, which forms PAH-CNs by thermal condensation at 350 °C. Furthermore, changing the PAH group within the g−CN scaffold enables the modulation of their optical and electronic properties. The photocurrent density is improved from approximately 2 to 27 µ A cm −2 (nearly 13 times) owing to the more extended conjugation to promote electron transport and expand light-harvesting In view of the advantages of chemical structure modification and copolymerization, Kumar et al. [132] synthesized C 3 N 5 polymer by applying melamine-derived melem (2,5,8−triamino-s−heptazine) and inorganic monomer (hydrazine hydrate) to generate intermediate melem hydrazine (MH) through hydrothermal reaction, following annealing at 450 • C for 2 h. In the C 3 N 5 framework, two s-heptazine units are bridged together by azo linkage (−N≡N−), which is a completely different bonding mode compared to C 3 N 4, where three heptazine units are connected with tertiary nitrogen. Owing to overlap between the π orbitals of azo-bridged units and the π system of heptazine motif, the band-gap of C 3 N 5 was markedly reduced to 1.76 eV, which enables the optical light absorption to expand up to 670 nm. The photocurrent response for the C 3 N 5 and g−C 3 N 4 blended TiO 2 electrode was proved to be 152 and 100 µA cm −2 at 1.23 V vs. NHE (Figure 13a-e). Recently, a high dispersible carbon nitride material functionalized with polycyclic aromatic hydrocarbons (PAH-CNs) was first introduced by Karjule et al. [133]. New starting monomers with substituted triazines were fabricated through a microwave (MW) irradiation strategy using dicyandiamide and aryl-nitrile as raw materials, which forms PAH-CNs by thermal condensation at 350 • C. Furthermore, changing the PAH group within the g−CN scaffold enables the modulation of their optical and electronic properties. The photocurrent density is improved from approximately 2 to 27 µA cm −2 (nearly 13 times) owing to the more extended conjugation to promote electron transport and expand light-harvesting range (Figure 14a-c). This breakthrough work can permit researchers to fabricate well-dispersed g−CN materials with tunable electronic and improved light-harvesting properties for PEC or other applications.

2022, 12, 2374
24 of 60 range (Figure 14a-c). This breakthrough work can permit researchers to fabricate welldispersed g−CN materials with tunable electronic and improved light-harvesting properties for PEC or other applications.   To sum up, doping is used to modulate the π conjugated system including light absorption, electron band structure-properties, and PEC activity by introducing heteroatoms, aromatic groups, or organic additives. It can be divided into atomic doping and molecular doping according to the doping levels. In general, atomic doping involves the insertion of non-metallic atoms into the g−CN lattice, partially replacing C or N atoms. On the other hand, the metal atoms can be inserted into the triangular gap cavity of the g−CN lattice. Compared with metal doping, non-metallic doping has obvious advantages because it can retain the non-metallic properties of g−CN. Co-doping and gradient doping have also attracted extensive attention. However, excessive heteroatomic doping will lead to doping asymmetry and a large number of defects, which will lead to the recombination of photo-generated carriers, which will hinder the improvement of PEC efficiency. Copolymerization is a molecular doping method in which the addition of structure-matching organic reagents (such as organic comonomers containing cyanide and benzene rings) in the copolymerization of nitrogen-rich precursors is an effective method to adjust their band structure at the molecular level of g−CN.

Cocatalysts
The modification of semiconductors with a cocatalyst is also a fascinating endeavor to enhance PEC performance. Cocatalysts mainly play two important roles, namely, acting as a reactive site and facilitating the separation of photo-induced charge. The typical reduction cocatalysts employed in the g−CN photoelectrodes are noble metals such as Pt, Rh, Ru, and Au. Nevertheless, it is not practical for commercial g−CN-based PEC applications owing to their high cost and rarity. In terms of economic and practical production, much research has been conducted on the use of transition metals (such as Ni [13] and Cu [134]) or alloying Pt with less expensive and abundant metals as co-catalysts. For example, nickel-decorated sulfur-doped graphitic carbon nitride with bismuth oxybromide (Ni/S-g−CN/BiOBr) shows a photocurrent density of 177.2 uA/cm 2 at 1.23 V vs. RHE, which is nearly three-fold higher than that of S-g−CN/BiOBr [13]. In addition, bimetallic co-catalysts have a great advantage in promoting the PEC reaction in comparison with monometallic analogues. The excited electrons from semiconductors can be injected into bimetallic alloys more easily. The bimetallic Pt-Ni incorporated g−CN nanotubes exhibit an excellent and stable photocatalytic performance, which is higher than that of Pt/g−CN nanotubes [135]. This result shows that bimetallic is expected to replace noble metal with better performance and lower cost. Furthermore, hydroxides [136] and oxyhydroxides [137,138] are effective cocatalysts for the g−CN modification to improve the redox reactions activity [139,140]. The photocurrent density could be availably improved from 89 µA cm −2 to 0.12 mA cm −2 after loading FeOOH on the g−CN film surface [51]. In another work, Fan et al. [141] modified g−CN film with NiCo-layered double hydroxide (NiCo-LDH) co-catalysts via cathodic electrochemical deposition. The resulting photoelectrode after incorporation of the NiCo-LDH decreases the charge migration resistance at the interfaces, which helps to promote the reaction kinetics and charge separation. The optimized g−CN/NiCo-LDH photoelectrode shows a photocurrent up to 11.8 µA cm −2 at 0.6 V vs. SCE, which was 2.8−fold higher than that of pristine g−CN (Figure 15a,b).
Because of the superior electrical conductivity of sp 2 −bonded carbon atom, carbonbased materials, e.g., carbon dots, carbon nanotubes (CNT), and graphene oxide, etc., are promising substrates and excellent reductive cocatalyst to integrate with g−CN, inhibiting the recombination of charge carriers [142][143][144]. Graphene, a honeycomb nanostructured 2D macromolecular structure of carbon atoms, has received tremendous attention because of its excellent optical and electrical features, large surface area, and good chemical stability. There are growing numbers of research on the fabrication of graphene/g−CN hybrid photoelectrodes for PEC applications by considering 2D/2D face-to−face interaction. In one of the pioneer studies [145], a hybrid material with the carbon nitride layer on the reduced graphene oxide layer (CN-rGO) was synthesized by a doctor-blade technique using a supramolecular precursor followed by thermal calcination. The incorporated graphene behaves as the electron-conducting channel to improve both photo-generated charge separation and electron diffusion length. With increasing graphene contents, the hybrid material exhibited a higher photocurrent response of 72 µA cm −2 at 1.23 V versus RHE (electrode thickness up to 36 µm) and subsequently decreased for a thicker photoelectrode. Photocurrent measurements with holes scavenger (TEOA) display a photocurrent response of 660 µA cm −2 at 1.23 V versus RHE and the external quantum yield of 60% at 400 nm (Figure 15c,d).   Recently, graphene derivatives (e.g., chemically doped graphene) have been integrated with g−CN to form a heterojunction photoelectrode. In particular, various types of dopants have triggered keen interest in the preparation of N-doped graphene as it can ameliorate the optical, electronic, and chemical reactivity of graphene by making use of the lone-pair electrons of nitrogen atoms and the π−conjugated electrons of 2D graphene. Hou and coworkers [142] employed a facile hydrothermal method with urea to integrate N-deficient porous g−CN nanosheets (DPCN) and Nickel-iron−layered double hydroxide (NiFe-LDH) into 3D N-doped graphene (NRGO) framework. A record-high photocurrent density of 162.3 µA cm −2 at 1.4 V and remarkable photostability (>10 h) are realized (Figure 15e,f). The enhancement in the efficiency of this novel aerogel photoanode is based on: (i) the optimized energy band structure and improved light absorption derived from the 2D DPCN with deficient structures, (ii) 3D NRGO-based aerogel provides a hierarchical porous structure to improve the charge transport property, and (iii) a close-contacting NiFe-LDH contributes to the effective kinetic transportation of photo-generated electron-hole pairs via introducing the spatially separated interfaces.
In addition, 1D CNT is another carbon allotrope with extraordinary electronic properties. Hitherto, there have been several investigations to integrate g−CN with CNT as the composite films for promising PEC applications. Miao et al. [91] employed a vacuum filtration self-assembled process and subsequent annealing treatment to synthesize g−CN−CNTs thin film by incorporating g−CN nanosheets supernatant with CNTs aqueous solution. The photocurrent density could be availably improved from approximately 0.5 to 1.8 µA/cm 2 (nearly 3.7 times) via the recombination of CNTs to promote electron transport. The enhanced PEC performance was accredited to the excellent electrons transport capability by CNTs, fundamentally resolves the poor electronic conductivity of the conglomerated g−CN nanosheets structure and allows valid electrons separation and collection (Figure 15g,h).
Among the surface modification methods to improve the PEC performance of g−CN, the supported co-catalyst is an effective one to improve the separation efficiency of photogenerated carriers. The co-catalyst can not only promote the activity of the substrate semiconductor but also inhibit the photo-corrosion that may occur during the PEC process. However, non-noble metal-based co-catalysts still encounter low activity and low stability. In addition, the energy level structure, electronic structure and interface structure also require reasonable construction for the g−CN/co-catalyst system. There are many factors affecting the PEC performance of the g−CN/cocatalyst system, including load, particle size, interface binding force, etc. Only after reasonable consideration and design can the photogenerated carriers be efficiently generated from g−CN and across the contact interface to the cocatalyst surface in the right direction. In addition, a photocurrent density comparison of the functionalization of g−CN-based photoelectrodes by heteroatoms doping and textural design is shown in Table 3.

Composite Films of g−CN as Highly Valid Photoelectrodes for PEC Application
As aforementioned, similar to the conventional inorganic semiconductor photocatalysts, the pristine g−CN films usually suffer from several intrinsic shortcomings, such as insufficient visible-light absorption, the rapid recombination rate of charge carriers, and low electrical conductivity [105,146,147]. Therefore, various strategies are employed to restrain the recombination of charge carriers in g−CN films to enhance the PEC performance. In addition to the heteroatoms doping and textural design, as discussed in Section 5, the construction of g−CN-based heterojunction composite films for PEC application has elicited a great deal of attention, which will be across-the−board reviewed in the following subsections.
Typically, a g−CN-based heterojunction photoelectrode is engineered by combining two different semiconductors with appropriate valence band (VB) and conduction band (CB) potentials, which results in a new electronic structure after thermal equilibration. In other words, band bending forms at the heterojunction interface on account of potential differences between two sides of composite films [148]. This transformation induces a built-in electric field to spatially separate and transfer the photo-generated electrons and holes. Subjected to the separation mechanism of charge carriers and VB and CB potentials of semiconductors, the g−CN-based composites can be categorized into four classifications, i.e., Type I, Type II, Type III, and Z-scheme ( Figure 16) [3]. For the Type I heterojunction (Figure 16a), both the VB and CB potentials of semiconductor B with a smaller band gap are within the larger band-gap (semiconductor A), creating a straddling band alignment. The band edge positions of Type I heterojunction lead to the transfer of charge carriers in a single direction. Since all charges are accumulated in a semiconductor B, this will result in the commonly impaired photoredox efficiency. For the Type II heterojunction, the band edge potentials are staggered between both semiconductors (Figure 16b). As a result, the photo-generated electrons migrate to a less negative CB potential of semiconductor A, while the holes migrate to a less positive VB potential of semiconductor B. This significantly im-proves the spatial separation and migration of carriers to restrain the charge recombination [149,150]. However, the enhancive separation efficiency can be obtained based on the Type II heterojunction. The main drawback was caused by the chemical potential between both semiconductors, leading to a decrease in oxidation and reduction capacity. Recently, a new Z-scheme heterojunction was reported to surmount this shortfall. According to the electron mediator, the g−CN-based Z-scheme mechanism can be categorized into two typical forms: the semiconductor-semiconductor Z-scheme heterojunction (Figure 16d) and semiconductor-conductor−semiconductor Z-scheme heterojunction (Figure 16e). The solid electron mediator includes reduced graphene oxide [151,152] and noble-metal nanoparticles such as Au [153,154], Ag [155][156][157] and Cd [158], etc. For the former [159][160][161], the electron in the CB of semiconductor B directly transfers to the VB of semiconductor A, leading to the valid separation of the electrons in the CB of semiconductor A and the holes in semiconductor B. While for the latter, a conductor acts as an electron mediator and can facilitate the electron migration from semiconductor B to semiconductor A [162]. Early studies also demonstrated that interface engineering by introducing an electron mediator is beneficial to improving the photocatalytic performance of the Z-scheme heterojunction compared to the direct Z-scheme one [163,164]. In the type III heterojunction (Figure 16c), the VB and CB of semiconductors A and B do not overlap with each other [165].
Based on the above discussions, Type II and Z-scheme pathways that effectively separate the electron-hole pairs are ideal in the preparation of the g−CN-based film. In view of these two types of heterojunctions, selecting an appropriate semiconductor material to couple with g−CN is of utmost significance not only to enhance the optical light absorption but also to facilitate the charge migration and separation to suppress the recombination rate. In the following section, we will critically exemplify recently published studies on these two types of g−CN-based composite films for PEC water splitting. Nanomaterials 2022, 12, 2374 29 of 60 reduction capacity. Recently, a new Z-scheme heterojunction was reported to surmount this shortfall. According to the electron mediator, the g−CN-based Z-scheme mechanism can be categorized into two typical forms: the semiconductor-semiconductor Z-scheme heterojunction ( Figure 16d) and semiconductor-conductor−semiconductor Z-scheme heterojunction (Figure 16e). The solid electron mediator includes reduced graphene oxide [151,152] and noble-metal nanoparticles such as Au [153,154], Ag [155][156][157] and Cd [158], etc. For the former [159][160][161], the electron in the CB of semiconductor B directly transfers to the VB of semiconductor A, leading to the valid separation of the electrons in the CB of semiconductor A and the holes in semiconductor B. While for the latter, a conductor acts as an electron mediator and can facilitate the electron migration from semiconductor B to semiconductor A [162]. Early studies also demonstrated that interface engineering by introducing an electron mediator is beneficial to improving the photocatalytic performance of the Z-scheme heterojunction compared to the direct Z-scheme one [163,164]. In the type III heterojunction (Figure 16c), the VB and CB of semiconductors A and B do not overlap with each other [165]. Based on the above discussions, Type II and Z-scheme pathways that effectively separate the electron-hole pairs are ideal in the preparation of the g−CN-based film. In view of these two types of heterojunctions, selecting an appropriate semiconductor material to couple with g−CN is of utmost significance not only to enhance the optical light absorption but also to facilitate the charge migration and separation to suppress the recombination rate. In the following section, we will critically exemplify recently published studies on these two types of g−CN-based composite films for PEC water splitting.

Type II Heterojunctions
The flexibility, as well as their 2D flat morphology, favors g−CN films as an accessible platform for the construction of heterojunction with various components. Based on the

Type II Heterojunctions
The flexibility, as well as their 2D flat morphology, favors g−CN films as an accessible platform for the construction of heterojunction with various components. Based on the reported literature, a series of categories of g−CN-based heterojunctions have been successfully fabricated: (i) metal oxides/g−CN heterojunction, (ii) metal sulfides/g−CN heterojunction, and (iii) complex compound/g−CN heterojunction. The detailed studies for g−CN-based heterojunctions, efficiency, and applications will be discussed.

Metal Oxides/g−CN Hybrid Heterostructures Films
Until now, various metal oxide semiconductors have been coupled with g−CN to form type II heterojunction for PEC applications. Among those metal oxides, TiO 2 is one of the most suitable candidates, and the triumphant formation of TiO 2 /g−CN heterojunctions has been reported to improve the PEC activity.
Albero et al. [166] analyzed the electron transfer kinetics between g−CN and mesoporous TiO 2 using laser transient absorption spectroscopy (L-TAS) (path 2, Figure 17a), suggesting the effective separation of the electron. However, the ineffective diffusion of the hole to the electrolyte (path 3, Figure 17a) was found to be the restricting process during the PEC performance. Enlightened by this conception, combining TiO 2 with g−CN is the most feasible strategy to achieve a large-scale g−CN-based heterojunction photoelectrode with high PEC efficiency. Sima et al. [167] fabricated a mesoporous TiO 2 /g−CN hybrid nanostructure film, in which exfoliated g−CN were immobilized on mesoporous TiO 2 via a spray coating approach. In this architecture, the g−CN exposes its edges to the visible light and dispersion liquid that facilitates fast electrons and holes transport. The prepared TiO 2 /g−CN hybrid film possessed a high photocurrent density of 0.32 mA cm −2 , which was higher than those of pristine g−CN and pure TiO 2 films. Inspired by the aforementioned approach, Xu and co-workers [168] coated g−CN on the mesoporous TiO 2 through an in situ vapor-transport growth approach by calcining a supramolecular mixture of cyanuric acid-melamine (CM) precursor. As expected, its photocurrent reached 1.4 mA cm −2 , which was a 30−fold enhancement compared to pristine TiO 2 (0.045 mA cm −2 ). They also studied the impact of barbituric acid molecules on the mixture precursor on the performance of g−CN/TiO 2 . A relatively high carbon doping content augments the defect levels and causes higher recombination of charge carriers, resulting in a decreased photocurrent to 0.7 mA cm −2 . Until now, various metal oxide semiconductors have been coupled with g−CN to form type II heterojunction for PEC applications. Among those metal oxides, TiO2 is one of the most suitable candidates, and the triumphant formation of TiO2/g−CN heterojunctions has been reported to improve the PEC activity.
Albero et al. [166] analyzed the electron transfer kinetics between g−CN and mesoporous TiO2 using laser transient absorption spectroscopy (L-TAS) (path 2, Figure 17a), suggesting the effective separation of the electron. However, the ineffective diffusion of the hole to the electrolyte (path 3, Figure 17a) was found to be the restricting process during the PEC performance. Enlightened by this conception, combining TiO2 with g−CN is the most feasible strategy to achieve a large-scale g−CN-based heterojunction photoelectrode with high PEC efficiency. Sima et al. [167] fabricated a mesoporous TiO2/g−CN hybrid nanostructure film, in which exfoliated g−CN were immobilized on mesoporous TiO2 via a spray coating approach. In this architecture, the g−CN exposes its edges to the visible light and dispersion liquid that facilitates fast electrons and holes transport. The prepared TiO2/g−CN hybrid film possessed a high photocurrent density of 0.32 mA cm −2 , which was higher than those of pristine g−CN and pure TiO2 films. Inspired by the aforementioned approach, Xu and co-workers [168] coated g−CN on the mesoporous TiO2 through an in situ vapor-transport growth approach by calcining a supramolecular mixture of cyanuric acid-melamine (CM) precursor. As expected, its photocurrent reached 1.4 mA cm −2 , which was a 30−fold enhancement compared to pristine TiO2 (0.045 mA cm −2 ). They also studied the impact of barbituric acid molecules on the mixture precursor on the performance of g−CN/TiO2. A relatively high carbon doping content augments the defect levels and causes higher recombination of charge carriers, resulting in a decreased photocurrent to 0.7 mA cm −2 . The improved PEC efficiency was in virtue of the prominent interface-bound energy transition of g−CN/TiO 2 . In addition, the direct Ti-O-C bond creates supererogatory molecular orbitals at the interface, leading to a strong red-shift of the optical visible-light absorption (Figure 17b-e). The contribution of TiO 2 /g−CN photoelectrodes is obviously affected by their shape, size, and nanostructures of TiO 2 . Among various nanostructures, one-dimensional (1−D) TiO 2 nanostructures such as nanowires and nanotube arrays have drawn comprehensive interest owing to their high specific surface area and pore volume [169][170][171][172]. Sun et al. [173] reported the synthesis of g−CN/TiO 2 nanotube array heterojunction by the vapor-pyrolysis of anodized Ti sheets over melamine. The obtained g−CN/TiO 2 nanotube heterojunction exhibited strong optical absorption and reduced impedance compared to the pure TiO 2 nanotube. In addition, it also demonstrated enhanced PEC activity and stability. In another work, Liu and co-workers [174] reported g−CN/TiO 2 nanotube array (TNTAs) heterojunction with a g−CN shell and TNTAs core through TVC employing melamine and cyanuric acid supramolecular complex as the precursors. A noticeably enhanced photocurrent density of 0.86 mA cm −2 was realized, which was 4.5−fold and 2−fold higher than those of g−CN/amorphous TiO 2 heterojunction and pure TiO 2 nanotubes. It is worth emphasizing that the crystalline TiO 2 nanotubes offer a more distinct photo-response due to their unique one-dimensional nanostructure in comparison with the amorphous TiO 2 film.
In addition, Bhat et al. [175] reported the combination of g−CN and TiO 2 nanorods using a dip-coating process with cyanuric acid-melamine-thiourea (CMT) complexes as starting materials. Using CMT complexes for the fabrication of TiO 2 /g−CN heterojunction is more advantageous than the individual materials. The photocurrent density can reach up to 2.74 mA cm −2 at 1.23 V (vs. RHE) (Figure 18a-c). In another work, Su and co-workers [176] synthesized modified vertically aligned TiO 2 NTs with P-doped g−CN via anodization and wet-dipping followed by post-annealing. 1−butyl−3−methylimidazolium hexafluorophosphate (BmimPF 6 ) was added as a cheap phosphorus source. P-CN/TiO 2 NTs exhibited the photocurrent response (1.98 mA cm −2 at 0 V vs. Ag/AgCl), which was higher than that of pristine TiO 2 NTs. This promotion was because of the synergic mechanism between the introduction of P heteroatoms and the improvement of π−electron delocalization by 2−aminobenzonitrile (ABN). A previous study revealed that hydrogenation can improve the visible-light absorption range, structure matching, and stability [177]. Inspired by this, Zhou et al. [178] reported a g−CN/S-doped hydrogenated TiO 2 NTAs (g−CN/H-S−TiO 2 NTAs) photoelectrode using a dip-coating approach. The results exhibited a higher transient photocurrent response of 1.64 mA cm −2 with good stability compared to TiO 2 NTAs (0.29 mA cm −2 ). Hydrogenation reduces the energy gap of TiO 2 due to the introduction of Ti 3+ and oxygen vacancies. S-doping further forms a new impurity level that upshifts the TiO 2 VB edge. The introduction of g−CN can form type II heterojunction with H-S−TiO 2 NTAs, realizing an effective separation of charge and extended visible-light absorption (Figure 18d,e). Recently, the structure of high surface area assembled by chemical tailoring g−CN has been successfully prepared (Figure 18f). Additionally, combining with N,S co-doped TiO 2 significantly improves the PEC activity [179]. Improvement of the efficiency of TiO 2 −based photoanodes using nickel-modified polymeric carbon nitride has also been explored [180]. Quantum dot (0D) semiconductors help to widen band structures and enhance redox capability and open-circuit voltage compared with their bulk counterparts (3D) [181,182]. Recently, g−CN quantum dots (CNQDs) have been attracting a lot of attention because of the narrow size distribution, as well as high fluorescence [183]. Su et al. [184] fabricated TiO2 nanotube arrays (NTAs) with CNQDs via a delicate organic molecular linkage. The process included immersing TiO2NT in CNQDs aqueous solution with 1 v% 3−mercaptopropionic acid (MPA) at 80 °C for 12 h. The resulting products were rinsed and dried in air. The MPA-modified TiO2NT/CNQDs achieved a photocurrent response of 1.34 mA cm −2 at 0.3 V vs. Ag/AgCl, which was mainly ascribed to the effective heterojunction structure and particularly the strong affinity of TiO2 NTAs.
In addition to the frequently reported TiO2/g−CN photoelectrode, ZnO is another pivotal metal oxide with a wide band gap (3.2eV), which can be valid coupled with g−CN [185][186][187]. In one study by Fang et al. [188], g−CN was coated on yttrium-doped zinc oxide Nanorod (Y:ZnO NR) Arrays via in situ thermal vapor polymerization. The prepared Y:ZnO/g−CN heterostructure possessed a photocurrent density of 0.4 mA cm −2 , over 50−fold that of pure g−CN films achieved by powder. The result reveals the photocurrent retention capability can be preserved at 95% for 160 min. The charge collector layer in type II alignment remarkably reduces the traveling distance of the carrier and minimizes the defect-trapping effect, thereby suppressing the recombination effect (Figure 19a-d). In another similar work, Park et al. [189] loaded the g−CN layer on vertically aligned ZnO nanorods through thermal vapor condensation. In comparison with the individual bare ZnO nanorods or g−CN, the effective charge separation and transportation emerge from the formation of Type II heterostructures. This photoanode demonstrated a photocurrent density of 0.12 mA cm −2 and excellent stability up to 5 h. Quantum dot (0D) semiconductors help to widen band structures and enhance redox capability and open-circuit voltage compared with their bulk counterparts (3D) [181,182]. Recently, g−CN quantum dots (CNQDs) have been attracting a lot of attention because of the narrow size distribution, as well as high fluorescence [183]. Su et al. [184] fabricated TiO 2 nanotube arrays (NTAs) with CNQDs via a delicate organic molecular linkage. The process included immersing TiO 2 NT in CNQDs aqueous solution with 1 v% 3−mercaptopropionic acid (MPA) at 80 • C for 12 h. The resulting products were rinsed and dried in air. The MPA-modified TiO 2 NT/CNQDs achieved a photocurrent response of 1.34 mA cm −2 at 0.3 V vs. Ag/AgCl, which was mainly ascribed to the effective heterojunction structure and particularly the strong affinity of TiO 2 NTAs.
In addition to the frequently reported TiO 2 /g−CN photoelectrode, ZnO is another pivotal metal oxide with a wide band gap (3.2eV), which can be valid coupled with g−CN [185][186][187]. In one study by Fang et al. [188], g−CN was coated on yttrium-doped zinc oxide Nanorod (Y:ZnO NR) Arrays via in situ thermal vapor polymerization. The prepared Y:ZnO/g−CN heterostructure possessed a photocurrent density of 0.4 mA cm −2 , over 50−fold that of pure g−CN films achieved by powder. The result reveals the photocurrent retention capability can be preserved at 95% for 160 min. The charge collector layer in type II alignment remarkably reduces the traveling distance of the carrier and minimizes the defect-trapping effect, thereby suppressing the recombination effect (Figure 19a-d). In another similar work, Park et al. [189] loaded the g−CN layer on vertically aligned ZnO nanorods through thermal vapor condensation. In comparison with the individual bare ZnO nanorods or g−CN, the effective charge separation and transportation emerge from the formation of Type II heterostructures. This photoanode demonstrated a photocurrent density of 0.12 mA cm −2 and excellent stability up to 5 h. However, the continuous g−−CN layer obtained via in situ thermal vapor polymerization blocks the porous structure. In addition, the as-designed heterostructures also impeded the electrolyte from penetrating into the structure, which may inhibit water oxidation under visible light [168]. Therefore, enhancement of the PEC performance can only be acquired by developing modified synthetic tactics that allows uniform coverage while preserving the porous structure. Hajduk et al. [190] loaded the g−CN layer on the ZnO Nanowires by forming a seed layer of cyanuric acid-melamine (CM) supramolecular complex, followed by their conversion to carbon nitride at high temperature. The successful retaining of the porous structure in the ZnO/g-CxNy (1.0) sample was verified by SEM image analysis (Figure 19e-h). The resulting ZnO/g-CxNy (1.0) photoelectrode possesses a photocurrent density of approximately 0.25 mA cm −2 , more than 3.5−fold that of the pure ZnO. Mahala et al. [191] loaded the g−CN QDs on the ZnO 2D nanosheets through a dipcoating method. In comparison with pure ZnO, the ZnO/g−CN QDs heterojunction shifts the optical absorption to higher wavelength ranges with stronger absorbance in the visible light and improves the charge-carrier separation efficiency. The best-obtained ZnO/g−CN photoanode exhibits approximately 2.29 times as high photocurrent density (0.952 mA cm −2 ) compared to bare ZnO at 0.5994 V versus RHE (Figure 19i,j).
Other than the recombination with UV-active TiO2 and ZnO semiconductor, WO3 has However, the continuous g−CN layer obtained via in situ thermal vapor polymerization blocks the porous structure. In addition, the as-designed heterostructures also impeded the electrolyte from penetrating into the structure, which may inhibit water oxidation under visible light [168]. Therefore, enhancement of the PEC performance can only be acquired by developing modified synthetic tactics that allows uniform coverage while preserving the porous structure. Hajduk et al. [190] loaded the g−CN layer on the ZnO Nanowires by forming a seed layer of cyanuric acid-melamine (CM) supramolecular complex, followed by their conversion to carbon nitride at high temperature. The successful retaining of the porous structure in the ZnO/g-C x N y (1.0) sample was verified by SEM image analysis (Figure 19e-h). The resulting ZnO/g-C x N y (1.0) photoelectrode possesses a photocurrent density of approximately 0.25 mA cm −2 , more than 3.5−fold that of the pure ZnO. Mahala et al. [191] loaded the g−CN QDs on the ZnO 2D nanosheets through a dip-coating method. In comparison with pure ZnO, the ZnO/g−CN QDs heterojunction shifts the optical absorption to higher wavelength ranges with stronger absorbance in the visible light and improves the charge-carrier separation efficiency. The best-obtained ZnO/g−CN photoanode exhibits approximately 2.29 times as high photocurrent density (0.952 mA cm −2 ) compared to bare ZnO at 0.5994 V versus RHE (Figure 19i,j).
Other than the recombination with UV-active TiO 2 and ZnO semiconductor, WO 3 has received prodigious attention as a promising candidate owing to its relatively lower band gap (2.5-2.8 eV), which lies in the visible-light range, thus increasing the potential for wider utilize in the PEC applications [192][193][194].
Li et al. [195] deposited 2D/2D WO 3 /g−CN nanosheet arrays (NSAs) on an FTO substrate through a hydrothermal growth and post-annealing process. It drastically suppressed the recombination of carriers and prolonged the lifetime of electrons manifested by EIS analyses for the WO 3 /g−CN heterojunction. The photocurrent density reached 0.73 mA cm −2 at the bias of 1.23 V vs. RHE (Figure 20a-c). Zhan et al. [193] synthesized a g−CN/WO 3 heterostructure plate array film by a combination of hydrothermal and dipping-annealing strategies. When the content of g−CN was 2.13 at%, the obtained g−CN/WO 3 heterojunction films exhibited the highest photocurrent density of 2.10 mA cm −2 at +2.0 V (vs. RHE), which was three-fold higher than that of pure WO 3 under illumination. Li et al. [196] uniformly coated g−CN nanosheet on WO 3 nanorod surface through the dip-coating method, which can achieve efficient visible-light−governed PEC water splitting. The photocurrent response of the g−CN/WO 3 photoanode achieved 1.92 mA/cm 2 at 1.23 V vs. RHE. The heterojunction was approximately stable, but distinct degradation (only decreases to 1.85 mA cm −2 ) was observed after 400 min of illumination. The stability performance promotion was realized via dipping WO 3 nanorod into g−CN nanosheet dispersion for various impregnation times without intermediary drying steps. The enhanced PEC performance is attributed to the synergic effect between g−CN and WO 3 originating from their well-matched overlapping band structures, which can availably boost the charge separation and the improvement of electrons' lifetime.
Apart from that, combining g−CN with Fe 2 O 3 has also recently been studied for PEC application. Liu et al. [197] fabricated g−CN/Fe 2 O 3 photocatalysts via electrodeposition followed by chemical vapor deposition. By altering the filling degrees of the melamine during the annealing process, the morphology of the g−CN/Fe 2 O 3 changes from regular nanosheets toward porous cross-linked nanostructures. The photocurrent density of the g−CN/Fe 2 O 3 is almost 70−fold higher than that of individual Fe 2 O 3 (Figure 20d-f). The enhancement of PEC efficiency was assigned to the morphology change and the synergy effect of the heterojunction. In previous work, the Ti-doped Fe 2 O 3 showed an ameliorative conductivity, thus achieving an enhancement in the photocatalytic performance in contrast to bulk Fe 2 O 3 [198]. Inspired by these facts, Liu et al. [199] established synthetic tactics to construct g−CN−modified Ti-doped Fe 2 O 3 nanosheets with different percentages, in which g−CN dispersion in water was immobilized on Ti-doped Fe 2 O 3 nanoarrays via a dip-coating approach. This strategy yielded tremendous enhancement in photocurrent density, reaching approximately 2.55 mA cm −2 at 1.23 V vs. RHE (Figure 20g-i). The results were prospectively higher than those for TVC deposition of undoped Fe 2 O 3 .
Apart from that, other endeavors have been dedicated to combining promising candidates with g−CN to construct various heterojunction photoelectrodes. These include SnO 2 /g−CN [73], NiO/g−CN [200], Sn-doped In 2 O 3 /g−CN [201], Cu 2 O/g−CN [202], MoO 3 /g−CN [203], and all of them showed enhanced PEC activity compared to pure g−CN. The construction of metal oxide-based thin films containing 2D structures is also an effective way to improve PEC activity [204,205]. Apart from that, other endeavors have been dedicated to combining promising candidates with g−CN to construct various heterojunction photoelectrodes. These include SnO2/g−CN [73], NiO/g−CN [200], Sn-doped In2O3/g−CN [201], Cu2O/g−CN [202], MoO3/g−CN [203], and all of them showed enhanced PEC activity compared to pure g−CN. The construction of metal oxide-based thin films containing 2D structures is also an effective way to improve PEC activity [204,205].

Metal Sulfides/g−CN Hybrid Heterostructures Films
Along with the wide band-gap metal oxides, the exploitation of g−CN coupled with a narrower band-gap metal sulfide has become a hot research area [206]. CdS, as one of the most indispensable candidates, has attracted comprehensive attention owing to its narrow band gap (~2.4 eV) and great visible-light response (520 nm). In addition, the highly matched energy levels with g−CN are beneficial for the formation of Type II heterostructures, assisting the charge migration and separation under irradiation [207]. In an early study, Li et al. [208] hybridized CdS nanorods (NRs) (5 nm diameters and 20-50 nm lengths) with g−CN nanosheets (NSs) by a facile hydrothermal route and then the dip casting approach. The Cd atoms are embedded steadily in the polymer CN backbone that offers numerous binding sites for CdS nanorods to form intimate contact, resulting in the promoted PEC efficiency. The enhancement was mainly because of increased interfacial contact and a sequential network offered by the long axis of CdS NRs, resulting in effective charge dissociation and transport on both components. On the other hand, CdS seriously suffers from photocorrosion that the sulphide ion (S 2− ) tends to be self-oxidized under strong radiation, and their corresponding photo-generated electron-hole pairs tend to be recombined [209]. In order to improve its stability, protecting CdS from photocorrosion

Metal Sulfides/g−CN Hybrid Heterostructures Films
Along with the wide band-gap metal oxides, the exploitation of g−CN coupled with a narrower band-gap metal sulfide has become a hot research area [206]. CdS, as one of the most indispensable candidates, has attracted comprehensive attention owing to its narrow band gap (~2.4 eV) and great visible-light response (520 nm). In addition, the highly matched energy levels with g−CN are beneficial for the formation of Type II heterostructures, assisting the charge migration and separation under irradiation [207]. In an early study, Li et al. [208] hybridized CdS nanorods (NRs) (5 nm diameters and 20-50 nm lengths) with g−CN nanosheets (NSs) by a facile hydrothermal route and then the dip casting approach. The Cd atoms are embedded steadily in the polymer CN backbone that offers numerous binding sites for CdS nanorods to form intimate contact, resulting in the promoted PEC efficiency. The enhancement was mainly because of increased interfacial contact and a sequential network offered by the long axis of CdS NRs, resulting in effective charge dissociation and transport on both components. On the other hand, CdS seriously suffers from photocorrosion that the sulphide ion (S 2− ) tends to be self-oxidized under strong radiation, and their corresponding photo-generated electron-hole pairs tend to be recombined [209]. In order to improve its stability, protecting CdS from photocorrosion via coating g−CN to form core-shell nanostructures has proven to be an ideal solution. Li et al. [210] synthesized novel nanoarrays composed of CdS/g−CN core-shell nanorods (CSNRs) through a hydrothermal treatment and heating process. The optimal photocurrent density of the obtained CdS/g−CN CSNRs reached up to 1.16 mA cm −2 , which was approximately 2.5 times that of pure CdS NRs. After 3600 s continuous illumination, there was no obvious loss observed in the photocurrent density (Figure 21a-d).
via coating g−CN to form core-shell nanostructures has proven to be an ideal solution. Li et al. [210] synthesized novel nanoarrays composed of CdS/g−CN core-shell nanorods (CSNRs) through a hydrothermal treatment and heating process. The optimal photocurrent density of the obtained CdS/g−CN CSNRs reached up to 1.16 mA cm −2 , which was approximately 2.5 times that of pure CdS NRs. After 3600 s continuous illumination, there was no obvious loss observed in the photocurrent density (Figure 21a-d). In another work, Wang and co-workers [211] fabricated an interlocking g−CN/CdS heterojunction film on FTO by electrophoretic deposition and chemical bath deposition strategies. The optimal g−CN/CdS−2h photoanode revealed a maximum photocurrent density of 5.4 mA cm −2 at 0.0 V vs. Ag/AgCl under visible light illumination. From their research, what was of even more interest was that the interlocking g−CN/CdS photoanode indicated high stabilities for the 9 h long time consecutive PEC H2 production test and even stabilized in ultrasonic treatment.
Apart from the CdS modification, 2D−layered transition metal sulphide (MoS2) has recently become an ideal co-catalyst on account of its high mobility, low cost, and narrow band gap (1.2-1.9 eV) [212,213]. Planar growth of MoS2 above the g−CN surface is realized due to the well-matched lattice and the similar layered structure [214,215]. The MoS2/Sdoped g−CN heterojunction film is fabricated through CVD deposition of S-doped g−CN film on ITO glass employing thiourea and melamine as the raw materials and subsequent growth of a MoS2 layer by hydrothermal treatment. The obtained film showed evidently In another work, Wang and co-workers [211] fabricated an interlocking g−CN/CdS heterojunction film on FTO by electrophoretic deposition and chemical bath deposition strategies. The optimal g−CN/CdS−2h photoanode revealed a maximum photocurrent density of 5.4 mA cm −2 at 0.0 V vs. Ag/AgCl under visible light illumination. From their research, what was of even more interest was that the interlocking g−CN/CdS photoanode indicated high stabilities for the 9 h long time consecutive PEC H 2 production test and even stabilized in ultrasonic treatment.
Apart from the CdS modification, 2D−layered transition metal sulphide (MoS 2 ) has recently become an ideal co-catalyst on account of its high mobility, low cost, and narrow band gap (1.2-1.9 eV) [212,213]. Planar growth of MoS 2 above the g−CN surface is realized due to the well-matched lattice and the similar layered structure [214,215]. The MoS 2 /S-doped g−CN heterojunction film is fabricated through CVD deposition of S-doped g−CN film on ITO glass employing thiourea and melamine as the raw materials and subsequent growth of a MoS 2 layer by hydrothermal treatment. The obtained film showed evidently improved PEC property with the anodic photocurrent density of 1.2 × 10 −4 A/cm 2 at an applied potential of +0.5 V (vs. Ag/AgCl), which is approximately twice the S-doped g−CN film [216]. In addition to metal sulfide, metal phosphide semiconductors with inexpensive and earth-abundant elements have also been used to prepare g−CN-based photoelectrode [217]. Recently, Li and co-workers [206] reported the merging of gradient concentration of cobalt disulfide (CoS 2 ) on PCN films by one-pot synthesis, resulting in remarkably promoted PEC performances (Figure 21e,f).

Complex Compound/g−CN Hybrid Heterostructures Films
As we know, a complex compound consists of at least three elements. Generally, the incorporation of a complex compound into g−CN is deemed to be a promising technique for promoting photocatalytic activity, especially for the PEC, because of the efficient carrier separation and transfer [218][219][220]. For example, BiVO 4 is an n-type semiconductor with a narrow band gap (2.4 eV), nontoxicity, low cost, and being more active in visible light [221][222][223], which is an available candidate for the development of g−CN-based heterojunction film.
Wang et al. [224] prepared nanostructured g−CN/BiVO 4 composite films, employing facile electrospinning with a post-annealing technique. The exfoliated g−CN nanosheets prevented the agglomeration of BiVO 4 , forming heterojunctions with BiVO 4 as well as profiting the formation of nanostructures. The obtained BiVO 4 /g−CN composite film displayed a photocurrent density of 0.44 mA cm −2 at 0.56 V, which was higher than that of the pristine g−CN and BiVO 4 photoanodes. Inspired by this, Feng et al. [225] synthesized ultrathin g−CN nanosheets (g−CN−NS) and coupled them with nanoporous BiVO 4 photoanodes via a facile ethylene glycol dispersion and impregnation. The photocurrent density of the ultrathin BiVO 4 /g−CN−NS photoanode was up to 3.12 mA/cm 2 at 1.23 V (vs. RHE), which was approximately 7−fold and 4−fold higher than those of BiVO 4 and BiVO 4 /bulk g−CN, respectively. The improved PEC performance was ascribed to the fact that the g−CN−NS serves as an efficient cocatalyst due to its enlarged band-gap (3.03 eV). Moreover, the intimate contact of ultrathin g−CN−NS and BiVO 4 in core shell structure facilitated the migration and separation of the charge and restrained the interfacial charge recombination. This work accentuates the importance of geometry engineering of heterojunction film, which can be outspread to other ternary oxide semiconductor such as LaTiO 3 .
Nevertheless, less emphasis is placed on comprehending the interface interaction and charge migration of g−CN/BiVO 4 heterojunction. Jiang et al. [226] obtained g−CN/BiVO 4 heterojunction photoelectrode through modified sol-gel tactics via varying the weight ratio of g−CN. The obtained g−CN/BiVO 4 (30%) sample exhibited improved photocurrent density compared to pure BiVO 4 . Such augment can be ascribed to the effective charge separation and fast interfacial charge transfer of the type-II hetero-interface. The corresponding electronic structures and charge transfer properties were also discussed based on systematic experimental studies and density functional theory (DFT) calculations (Figure 22a-e).
By combining the elemental metal doping and hybridization with other semiconductors, Zeng et al. [227] coated the Mo-doped nanoporous BiVO 4 surface uniformly with ultrathin g−CN through a facile electrostatic self-assembly method with improved the PEC capability. By mixing cation precursors (Bi and Fe) in different stoichiometric proportions, Yang et al. [228] performed the mixed-phase Bi x Fe 1-x VO 4 heterostructures through a onepot hydrothermal synthesis. In addition, the impregnated Bi x Fe 1-x VO 4 on the sulfur-doped g−CN (SCN) enhances the charge separation on account of the formation of heterojunction and attains a wide range of solar absorption due to the decrease of band gap (Figure 22f-h).
A typical ABO 3 perovskite oxide semiconductor such as LaTiO 3 (band gap of 3.51 eV) is also well matched with g−CN to form g−CN/LaTiO 3 heterostructure with improves the PEC activity [229]. Xing et al. [230] prepared neodymium (Nd)−doped g−CN−bismuth oxyiodide (BiOI) (Nd-doped g−CN/BiOI) through a two-step thermal poly-condensation and hydrothermal method (Figure 23a). The incorporation of Nd into the g−CN lattice decreases the band gap and also absorbs visible light, culminating in higher PEC properties (15.50 mA/cm 2 ). Furthermore, the enhancement could be because of the formation of heterojunction (Figure 23b-d).
oxyiodide (BiOI) (Nd-doped g−CN/BiOI) through a two-step thermal poly-condensation and hydrothermal method (Figure 23a). The incorporation of Nd into the g−CN lattice decreases the band gap and also absorbs visible light, culminating in higher PEC properties (15.50 mA/cm 2 ). Furthermore, the enhancement could be because of the formation of heterojunction (Figure 23b-d).  Many other complex compounds incorporate with g−CN materials have also been investigated recently, for example, g−CN/NaNbO3−NF [231], g−CN/CuInS2 [232], CdMoO4/g−CN [233], PbTiO3/g−CN [234], and all of them showed enhanced PEC activity compared to pure g−CN. Although the above materials can be used to fabricate com- Many other complex compounds incorporate with g−CN materials have also been investigated recently, for example, g−CN/NaNbO 3 −NF [231], g−CN/CuInS 2 [232], CdMoO 4 /g−CN [233], PbTiO 3 /g−CN [234], and all of them showed enhanced PEC activity compared to pure g−CN. Although the above materials can be used to fabricate compounds with g−CN, there still are some disadvantages: TiO 2 can not absorb the visible light effectively due to the large band gap. Therefore, most researchers use a modified TiO 2 to composite with g−CN, accelerating charge separation and charge transport as well as enhancing the visible light absorption of the composite material. As a sulfide photo-catalyst, the band gap of CdS is well matched with g−CN, but the photo-corrosivity and toxicity of CdS still need to be improved. Finally, the morphology of g−CN also has a great influence on PEC performance. Compared with other morphologies of g−CN, 0D semiconductor is generally more conducive to broaden the band structure, enhance the redox capacity and open circuit voltage.

G−CN-based Z-Scheme Heterojunction Composite Films
Although enhancive separation efficiency can be obtained based on the Type II alignment, the main drawback is that the redox ability of photo-generated charges is minimized upon the charge migration process [3,159]. Therefore, enhancing the reduction/oxidation ability is a challenge to be accomplished. Recently, the electrons migration mechanism in the Z-scheme heterojunction has been demonstrated to surmount this shortfall. By maintaining the reductive electrons in the higher CB edge and oxidative holes in the lower VB edge, it can not only enhance the charge separation efficiency but also realize a remarkably high redox ability of photo-generated charges [235][236][237][238]. The energy band structure of the Z-scheme heterojunction is similar to that of the traditional II-scheme heterojunction system. At present, the electron-hole transfer mechanism in the Z-scheme heterojunction is still controversial, and usually needs to be differentiated and verified by combining experimental and theoretical simulation. Free radical types were detected by electron paramagnetic resonance (EPR) and fluorescence spectra (PL), and the mechanism of photo-generated carrier transfer was predicted by free radical capture detection. Pan's group [239] established synthetic tactics for the construction of g−CN/TiO 2 heterojunction, in which g−CN was deposited on the surface of Corn-shaped TiO 2 nanofilms via an in-site microwave-heating technique. It can be revealed that the g−CN/TiO 2 heterojunction obviously has a higher photocurrent density compared with pure TiO 2 . The improved properties were primarily attributed to the Z-scheme heterojunction interfaces governed by the full-matched energy level between g−CN and TiO 2 . In addition, the as-designed Corn-shaped TiO 2 nanofilms also provided a special nanostructure that may benefit from light scattering and deposition of g−CN.
Vacancy defects are reported to play a critical role in the modification of the electronic structure. Inspired by this, Xiao et al. [240] fabricated a 0D/1D g−CN nanoparticles/TiO 2 nanotube arrays hybrid nanostructure containing an interfacial oxygen vacancy layer via anodic oxidation, NaBH 4 reduction, and CVD methods. The prepared 0D/1D g−CN/OV−TiO 2 possessed an optimal photocurrent density (0.72 mA/cm 2 ), which was 8−fold higher than that of g−CN/TiO 2 material (without interfacial oxygen vacancy layer). The obviously enhanced activity was accredited to the construction of Z-scheme heterostructure with an interfacial oxygen vacancy layer between TiO 2 Nanotube arrays and g−CN nanoparticles. Furthermore, the presence of an interfacial oxygen vacancy layer provides sufficient charge carrier transfer to heterostructure.
In another similar work by the same research group [241], the proton-functionalized g−CN (p-CN) nanosheet was deposited on the TiO 2 nanotube array containing oxygen vacancies (OV−TiO 2 ). Additionally, the protonation pretreatment was undertaken by ultrasonic bulk g−CN in the presence of hydrochloric acid, realizing larger contact areas and a close-knit contact interface between two semiconductors. The optimal photocurrent density of the obtained p-CN/OV−TiO 2 Z-scheme heterostructure reached 96 µA cm −2 , which was approximately 1.4 times that of CN/OV−TiO 2 samples (Figure 24a-e). Safaei et al. [35] reported Z-scheme g−CN/BiVO 4 photoelectrode prepared through elec-trodeposition of BiVO 4 and then spin-coating on g−CN thin films surface. The optimal photocurrent of the obtained g−CN/BiVO 4 reached 0.42 mA cm −2 at 1.23 V vs. RHE. For comparison, a TiO 2 /BiVO 4 photoelectrode was also devised. The acquired photocurrent density of g−CN/BiVO 4 is higher than that of TiO 2 /BiVO 4 , demonstrating the superiority of g−CN to TiO 2 . The enhancement was mainly because of the formation of the Vander Waals type interaction, resulting in effective charge separation at the g−CN/BiVO 4 interface (Figure 24f-h). Furthermore, the presence of hydrogen bonds in g−CN induced the production of oxygen vacancy sites, culminating in its higher PEC properties. In another similar study, Murugan et al. [242] designed a novel graphitic carbon nitride-bismuth vanadate (g−CN−BiVO4) heterojunction with different wt. % of g−CN by a simple hydrothermal approach. It can be seen from Figure 25a that the CB offset (1.51 eV) is higher than the gap between the VB of g−CN and the CB of BiVO4 (1.11 eV). This structure provides a strong driving force for the photo-generated electrons to transfer from the CB of BiVO4 to the VB of g−CN easily, leaving the photo-generated electrons and holes in the CB of g−CN and the VB of BiVO4, respectively. Then, the heterojunction restricts the electron-hole pair recombination, which was testified by EIS and PL spectra ( Figure  25b,c). This manifests the construction of mediator-free direct Z-scheme heterojunction between g−CN and BiVO4. The photoanode with 10 wt. % of g−CN resulted in a photocurrent of 2.02 μA cm −2 at 1.23 V vs. RHE, which was higher than that of the pure BiVO4 and g−CN, respectively. In another similar study, Murugan et al. [242] designed a novel graphitic carbon nitridebismuth vanadate (g−CN−BiVO 4 ) heterojunction with different wt. % of g−CN by a simple hydrothermal approach. It can be seen from Figure 25a that the CB offset (1.51 eV) is higher than the gap between the VB of g−CN and the CB of BiVO 4 (1.11 eV). This structure provides a strong driving force for the photo-generated electrons to transfer from the CB of BiVO 4 to the VB of g−CN easily, leaving the photo-generated electrons and holes in the CB of g−CN and the VB of BiVO 4 , respectively. Then, the heterojunction restricts the electron-hole pair recombination, which was testified by EIS and PL spectra (Figure 25b,c). This manifests the construction of mediator-free direct Z-scheme heterojunction between g−CN and BiVO 4 . The photoanode with 10 wt. % of g−CN resulted in a photocurrent of 2.02 µA cm −2 at 1.23 V vs. RHE, which was higher than that of the pure BiVO 4 and g−CN, respectively. Tailored exposure of semiconductor materials to ionizing radiation (e.g., X-rays, α, β, γ, and neutron) can adjust the bond structure and turn the optical, electrical, and structural properties so that a higher portion of the solar spectra is captured [243]. Compared with other ionizing radiations, γ irradiation has relatively higher penetration power, no contamination and radioactive material, as well as the alteration to the chemical bonds, which is potential for solar applications [244]. A few studies have explored the positive role of γ radiation on materials. Mohamed et al. [245] reported the effect of γ irradiation on the PEC solar water splitting for the g−CN and g−CN/BiVO4 Z-scheme heterojunction. It has been discovered that the γ radiation alters the structural and bonding structure of g−CN, which ultimately decreases the band-gap energy from 2.82 to 2.76 eV (Figure 26a). The γ irradiated g−CN/BiVO4 heterojunction exhibited a high photocurrent (1.38 mA cm −2 ) at 1.23 V versus Ag/AgCl, which was higher than that of the γ irradiated g−CN or nonradiated heterojunction. Fast electron migration took place on the g−CN/BiVO4 heterojunction interface, which resulted from the effects of γ radiation.
NiTiO3, a typical ABO3 perovskite oxide, presents a narrow band gap (Eg = 2.2 eV), high stability in oxidizing environments, and a large absorption coefficient [246][247][248]. Moreover, NiTiO3 possesses appropriate band edges (ECB = 0.21 eV, EVB = 2.43 eV), which match well with g−CN (ECB = −1.2 eV, EVB = 1.5 eV) to form a direct Z-scheme heterojunction system [31,249,250]. Huang et al. [251] synthesized NiTiO3/g−CN under ultrasonication and subsequent calcination treatment at 400 °C for 2 h. Benefiting from a unique nanostructure and g−CN-based direct Z-scheme charge transfer mechanism, such Ni-TiO3/ g−CN composite films exhibit enhanced PEC performance (0.25 mA cm −2 ) compared with the pure NiTiO3 and g−CN. This study provides meaningful guidance for the research of inorganic perovskite materials for water splitting, and it is essential to further inquiries into this type of photoelectrodes.
Early studies also demonstrated that the solid-state−mediators (Ag, Au, and Cu) incorporated into the interface of two components in the Z-scheme structure act as electron mediators to improve electron migration efficiency. For example, Xiao and co-workers [252] fabricated the ternary photoanode via loading g−CN on the surface of ZnO nanowire (ZnONWs) to form a junction to promote the electron tunneling effect. Whereafter, Tailored exposure of semiconductor materials to ionizing radiation (e.g., X-rays, α, β, γ, and neutron) can adjust the bond structure and turn the optical, electrical, and structural properties so that a higher portion of the solar spectra is captured [243]. Compared with other ionizing radiations, γ irradiation has relatively higher penetration power, no contamination and radioactive material, as well as the alteration to the chemical bonds, which is potential for solar applications [244]. A few studies have explored the positive role of γ radiation on materials. Mohamed et al. [245] reported the effect of γ irradiation on the PEC solar water splitting for the g−CN and g−CN/BiVO 4 Z-scheme heterojunction. It has been discovered that the γ radiation alters the structural and bonding structure of g−CN, which ultimately decreases the band-gap energy from 2.82 to 2.76 eV (Figure 26a). The γ irradiated g−CN/BiVO 4 heterojunction exhibited a high photocurrent (1.38 mA cm −2 ) at 1.23 V versus Ag/AgCl, which was higher than that of the γ irradiated g−CN or non-radiated heterojunction. Fast electron migration took place on the g−CN/BiVO 4 heterojunction interface, which resulted from the effects of γ radiation. NiTiO 3 , a typical ABO 3 perovskite oxide, presents a narrow band gap (Eg = 2.2 eV), high stability in oxidizing environments, and a large absorption coefficient [246][247][248]. Moreover, NiTiO 3 possesses appropriate band edges (E CB = 0.21 eV, E VB = 2.43 eV), which match well with g−CN (E CB = −1.2 eV, E VB = 1.5 eV) to form a direct Z-scheme heterojunction system [31,249,250]. Huang et al. [251] synthesized NiTiO 3 /g−CN under ultrasonication and subsequent calcination treatment at 400 • C for 2 h. Benefiting from a unique nanostructure and g−CN-based direct Z-scheme charge transfer mechanism, such NiTiO 3 /g−CN composite films exhibit enhanced PEC performance (0.25 mA cm −2 ) compared with the pure NiTiO 3 and g−CN. This study provides meaningful guidance for the research of inorganic perovskite materials for water splitting, and it is essential to further inquiries into this type of photoelectrodes.
Early studies also demonstrated that the solid-state−mediators (Ag, Au, and Cu) incorporated into the interface of two components in the Z-scheme structure act as electron mediators to improve electron migration efficiency. For example, Xiao and co-workers [252] fabricated the ternary photoanode via loading g−CN on the surface of ZnO nanowire (ZnONWs) to form a junction to promote the electron tunneling effect. Whereafter, platinum nanoclusters are further incorporated at the g−CN/ZnONWs interface (g−CN/Pt/ZnO) or surface (Pt/g−CN/ZnO) by photo-depositing tactics. The photocurrent density of the Pt/g−CN/ZnO sample (120 µA cm −2 at 0.5 V vs. Ag/AgCl) was higher than that of g−CN/ZnO due to the cocatalyst promotion. In addition, the Pt/g−CN/ZnO photoanode showed a much lower photocurrent response than that of g−CN/Pt/ZnO, manifesting the charge carrier transportation and separation by the solid-state−mediators. Inspired by this, the 3D urchin-like ZnO/Au/g−CN Z-scheme heterostructure was prepared and applied as an excellent photocathode for PEC water splitting by Wen et al. [253]. Under simulated sunlight irradiation, both ZnO and g−CN were excited to generate electron-hole pairs. Due to the high work function (−5.30 eV) of Au nanoparticles (Au NPs), the photo-generated electrons on the CB of ZnO were quickly injected into that of Au NPs, followed by further migration into the VB of g−CN to quench the holes. The sandwiched Au NPs acted as electron transfer mediators to enhance vectorial electron migration in direct Z-scheme ZnO/g−CN heterojunction (Figure 26b-e). This work provides a brand-new insight into the application of the Z-scheme photosystem-conductor−photosystem (PS-C−PS) system for PEC water splitting, in which the high redox capacity and rapid electron transfer channel are retained. Constructing multi-heterojunction photocatalysts with multi-sources for carrier generation is one of the effective methods for enhancing the PEC activity and MO degradation. Zhao et al. [254] designed a novel TiO 2 /g−CN/Ag-AgBr (TCNAAB) multi-heterojunction by depositing Ag-AgBr nanoparticles onto 3D spherical TiO 2 /g−CN (Figure 26f-g). platinum nanoclusters are further incorporated at the g−CN/ZnONWs interface (g−CN/Pt/ZnO) or surface (Pt/g−CN/ZnO) by photo-depositing tactics. The photocurrent density of the Pt/g−CN/ZnO sample (120 μA cm −2 at 0.5 V vs. Ag/AgCl) was higher than that of g−CN/ZnO due to the cocatalyst promotion. In addition, the Pt/g−CN/ZnO photoanode showed a much lower photocurrent response than that of g−CN/Pt/ZnO, manifesting the charge carrier transportation and separation by the solid-state−mediators. Inspired by this, the 3D urchin-like ZnO/Au/g−CN Z-scheme heterostructure was prepared and applied as an excellent photocathode for PEC water splitting by Wen et al. [253]. Under simulated sunlight irradiation, both ZnO and g−CN were excited to generate electronhole pairs. Due to the high work function (−5.30 eV) of Au nanoparticles (Au NPs), the photo-generated electrons on the CB of ZnO were quickly injected into that of Au NPs, followed by further migration into the VB of g−CN to quench the holes. The sandwiched Au NPs acted as electron transfer mediators to enhance vectorial electron migration in direct Z-scheme ZnO/g−CN heterojunction (Figure 26b-e). This work provides a brandnew insight into the application of the Z-scheme photosystem-conductor−photosystem (PS-C−PS) system for PEC water splitting, in which the high redox capacity and rapid electron transfer channel are retained. Constructing multi-heterojunction photocatalysts with multi-sources for carrier generation is one of the effective methods for enhancing the PEC activity and MO degradation. Zhao et al. [254] designed a novel TiO2/g−CN/Ag-AgBr (TCNAAB) multi-heterojunction by depositing Ag-AgBr nanoparticles onto 3D spherical TiO2/g−CN (Figure 26f-g).  The key discrepancy between Z-scheme and type II heterojunction is the electronically regulated medium for photo-carriers transfer. Compared with the traditional type II heterojunction system, the Z-scheme heterojunction maintains a strong redox capacity, achieves a wide spectral response, and possesses high stability. The main advantages of Z-scheme heterojunction are presented as follows: Firstly, there is no intermediate, it is low cost, there are fewer components, and it is easy to prepare. Secondly, the contact between the two kinds of semiconductors forms a strong interaction, which can decrease the carrier transfer resistance and improve the PEC activity. Finally, in addition to noble metal, cheaper carbon-based materials can also be used as charge modulators. By optimizing the electronic medium, electronic interface resistance, carrier transfer efficiency, and PEC stability can be enhanced. In particular, graphene and its derivatives stand out in the Z-scheme heterojunction system due to their excellent electrical conductivity and electron transport in recent years. However, Z-scheme g−CN-based heterojunction photoelectrodes are still in their elementary stage of development, and they are beneficial for future experimentation.

Design and Construction of Multicomponent Heterojunctions
As mentioned previously, the coupling of g−CN with appropriate semiconductors enables an efficient amelioration in the separation and migration of charge carriers via a heterojunction-transfer mechanism. Therefore, ternary g−CN-based heterojunction films will possess more meaningful applications owing to their synergistic effects.
For example, Reddy et al. [255] fabricated a g−CN−decorated CdS-nanoparticle−doped Fe 3 O 4 nanocube (g−CN/CdS-Fe 3 O 4 ) hybrid material electrode driven by a two-stage hydrothermal method (Figure 27a). Due to the well-matched electronic potentials of Fe 3 O 4 nanocube, CdS NPs, and g−CN, the e− in the CB of g−CN were quickly migrated to that of CdS NPs, followed by further transfer into the CB of Fe 3 O 4 nanocube to increase the charge carriers lifetime. Next, the accumulated e− in the Fe 3 O 4 nanocube can directly transfer into the ITO-coated glass. Meanwhile, the holes on the VB of the Fe 3 O 4 nanocube occur as a result of water oxidation. Evidently, the highest PEC efficiency in terms of linear sweep voltammetry (LSV)and photocurrent response (~0.0235 mA/cm 2 ) was achieved using the g−CN/CdS−Fe 3 O 4 electrode as the small band-gap (2.17 eV) led to effective interfacial charge transfer, which was higher than those of individual single and binary catalysts under visible-light irradiation (Figure 27b-d). In another work, g−CN/NiFelayered double hydroxide (LDH) was modified by graphdiyne (GDY) to form a ternary hierarchical mesoporous g−CN/GDY/NiFe-LDH heterostructure [256]. Owing to the high charge mobility, unique 2D structure with sp and sp 2 −hybridized carbon atoms, and a band-gap of~2.1 eV, the GDY can not only extend the light absorption range but also form a build-in electric field to extract and transfer the holes from g−CN to NiFe-LDH. The optimum photocurrent density for the photocathode was as high as 178.66 µA cm −2 at 1.4 V (Figure 27e,f). Bi 2 O 2 CO 3 /NiFe-LDH/g−CN [257] and CuTi-LDH/g−CN [258] showed enhanced PEC activity compared to pure g−CN.
In addition, Zhang et al. [259] designed a 0−D Co 3 O 4 decorated TiO 2 @P-CN (coreshell) photoanode by combining the hydrothermal strategy and successive chemical bath deposition process to enhance PEC performance. The ternary photoanode led to the rapid separation of charge to suppress the recombination rate (Figure 28a,b). The core/shell structure of CdS/g−CN/ZnO heterojunction was successfully prepared to improve the PEC activity [260].
Bashiri et al. [261] employed a two-stage hydrothermal method followed by the chemical bath deposition technique to develop a novel tailored TiO 2 /SrTiO 3 /g−CN ternary system. The photoelectrode manifests improved visible-light absorption and increased spatial separation of charge carriers. Gopalakrishnan et al. [262] employed a spin-coating strategy and subsequent annealing treatment to synthesize silicon nanowire-based hybrid heterostructures by incorporating perovskite SrTiO 3 nanoparticle (NPs) with g−CN nanosheets (NSs) as the interfacial layer. The hybrid heterojunction as photocathode accomplished the higher photocurrent density of 28 mA cm −2 at 1.23 V (vs. RHE) with ABPE 5.4% at 0.8 V versus RHE. Moreover, EIS measurements suggest that the hybrid photocathode holds a reduced interfacial charge transfer resistance with a prolonged lifetime of photon-excited electrons due to the well-aligned photocathode/electrolyte interface (Figure 28c-e). In addition, Zhang et al. [259] designed a 0−D Co3O4 decorated TiO2@P-CN (coreshell) photoanode by combining the hydrothermal strategy and successive chemical bath deposition process to enhance PEC performance. The ternary photoanode led to the rapid separation of charge to suppress the recombination rate (Figure 28a,b). The core/shell structure of CdS/g−CN/ZnO heterojunction was successfully prepared to improve the PEC activity [260].
Bashiri et al. [261] employed a two-stage hydrothermal method followed by the chemical bath deposition technique to develop a novel tailored TiO2/SrTiO3/g−CN ternary system. The photoelectrode manifests improved visible-light absorption and increased spatial separation of charge carriers. Gopalakrishnan et al. [262] employed a spin-coating strategy and subsequent annealing treatment to synthesize silicon nanowire-based hybrid heterostructures by incorporating perovskite SrTiO3 nanoparticle (NPs) with g−CN nanosheets (NSs) as the interfacial layer. The hybrid heterojunction as photocathode accomplished the higher photocurrent density of 28 mA cm −2 at 1.23 V (vs. RHE) with ABPE 5.4% at 0.8 V versus RHE. Moreover, EIS measurements suggest that the hybrid photocathode holds a reduced interfacial charge transfer resistance with a prolonged lifetime of photon-excited electrons due to the well-aligned photocathode/electrolyte interface (Figure 28c-e). In another work, Xu et al. [263] employed a layer-by−layer assembly method to synthesize Cu 2 O/g−CN/WS 2 triple-layer photocathode. The g−CN not only protected the inner layer but also enhanced the light-to−electricity conversion efficiency because it can form a p-n junction with Cu 2 O, while exfoliated WS 2 was used as an out-layer catalyst for the further amelioration of the photocathode. The optimum photocurrent density under visible-light irradiation for the photocathode was as high as −9.5 mA cm −2 at −0.55 V versus RHE (pH 6.0). Taking advantage of CoFeO x as a cocatalyst, Bhandary and coworkers [264] employed a drop-casting method to integrate hematite dendrite/carbon nitride composite (HD-CN) on an ITO glass substrate. Subsequently, CoFeO x was used to modify the HD-CN through a one-step electrodeposition process. The photocurrent density could be availably improved from approximately 27 to 600 µA/cm 2 (nearly 22 times) via the recombination of the CoFeO x layer to promote electron transport. A plausible illustration for such a remarkable increment in PEC activity can be given in terms of the construction of a p-n junction between the HD-CN and CoFeO x layer. The as-prepared heterojunction can help facilitate photo-generated charge separation and inhibit the electron/hole recombination. In another work, Xu et al. [263] employed a layer-by−layer assembly method to synthesize Cu2O/g−CN/WS2 triple-layer photocathode. The g−CN not only protected the inner layer but also enhanced the light-to−electricity conversion efficiency because it can form a p-n junction with Cu2O, while exfoliated WS2 was used as an out-layer catalyst for the further amelioration of the photocathode. The optimum photocurrent density under visible-light irradiation for the photocathode was as high as −9.5 mA cm −2 at −0.55 V versus RHE (pH 6.0). Taking advantage of CoFeOx as a cocatalyst, Bhandary and coworkers [264] employed a drop-casting method to integrate hematite dendrite/carbon nitride composite (HD-CN) on an ITO glass substrate. Subsequently, CoFeOx was used to modify the HD-CN through a one-step electrodeposition process. The photocurrent density could be availably improved from approximately 27 to 600 μA/cm 2 (nearly 22 times) via the recombination of the CoFeOx layer to promote electron transport. A plausible illustration for such a remarkable increment in PEC activity can be given in terms of the construction of a p-n junction between the HD-CN and CoFeOx layer. The as-prepared heterojunction can help facilitate photo-generated charge separation and inhibit the electron/hole recombination.
In order to achieve efficient PEC performance, a variety of modifications have been taken to improve g−CN-based materials, for example: adjusting morphologies, metal/non-metal element doping, copolymerization, loading cocatalyst, heterojunction, In order to achieve efficient PEC performance, a variety of modifications have been taken to improve g−CN-based materials, for example: adjusting morphologies, metal/nonmetal element doping, copolymerization, loading cocatalyst, heterojunction, etc. Ternary g−CN heterojunction film will have a wider application prospect due to its synergistic effect. However, the reasonable construction of the ternary g−CN heterojunction system needs to be optimized from the aspects of energy level structure, electronic structure and interface structure. A performance comparison of g−CN-based heterojunction composite film photoelectrodes is shown in Table 4.Cocatalysts also play an important role in ternary heterojunction. The efficient cocatalysts should meet the following conditions: (1) extract electrons from photocatalysts and transfer them to protons (H + ); (2) excellent PEC activity for HER. To meet the first requirement, the Fermi level of the cocatalyst must be lower than the conduction band bottom of the photocatalyst, resulting in the smooth flow of photoexcited electrons and facilitating the separation of electron-hole pairs. For the second requirement, the Gibbs free energy of hydrogen adsorption on the cocatalyst should be close to zero so that proton adsorption and rapid release of hydrogen molecules with the lowest activation energy occur at neutral thermal conditions.

The Stability of the g−CN Photoelectrodes for PEC Applications
The stability of photoelectrode, namely, the PEC activity does not decrease significantly during repeated recycles, is a key factor for practical PEC applications. Time-dependent current (I-t) curves under fixed bias are commonly used to evaluate stability. For semiconductor photocatalyst with poor stability, the photocurrent decreases with the increasing illumination time, and the photocatalytic properties disappear gradually (the so-called photo-corrosion). By doping with the elemental atoms, the stability of the semiconductor photocatalyst can be improved. Bian et al. [34] grew g−CN films on FTO substrates through TVC using melamine and 2,6−diaminopyridine as precursors. The successful incorporation of sp 2 carbon groups into the g−CN film enhances the stability of photoelectrodes. The photocurrent density of the modified g−CN film is approximately 46% of that at the initial stage after 1 h and maintains 42% in the second cycle. Huang et al. [106] successfully introduced boron-doped g−CN films using boric acid as the B source. The boron-doped g−CN film retains 90% initial photocurrent under 10 h illumination while only 7% for the pristine one. This is mainly due to the efficient extraction of photo-generated holes from the surface region, reducing the chemical attack on the C-N bond. The cocatalyst is also a fascinating strategy to enhance the stability of the semiconductor photocatalyst and improve the PEC performance. Cocatalysts act as a protection layer to catalyze the surface reactions by minimizing the activation energies and facilitating the separation of photo-induced charge. For example, the nickel incorporated sulfur-doped g−CN with bismuth oxybromide (Ni/S-g−CN/BiOBr) photoelectrode exhibits a photocurrent density of 1 ± 0.2 mA cm −2 under illumination for 6000 s [13]. To achieve a long-term stability, the modified technologies should be further studied, including protecting g−CN from photo−/chemical-corrosion by protective layers and using both co-catalyst and protective layers. The reduced graphene oxide (rGO) can act as an electron transfer channel for reducing the recombination of the electron-holes and a protective layer for protecting g−CN from photo−/chemical-corrosion. Therefore, the rGO can be introduced into g−CN-based photoelectrodes in order to achieve good stability.

Conclusions, Perspective, and Outlook
The low price, high physicochemical stability, and attractive electronic band structure of g−CN materials, together with the recent developments in their controlled growth on various substrates, contribute to expectations of an infusive future for g−CN materials in PEC applications. The overview of the topics involved in this review is illustrated in Figure 29. In contrast to the many groundbreaking results on g−CN powders and their demonstrated functionality in various applications to date, the research reports on g−CN materials in photoelectrochemical applications are still in their infancy. To this end, to achieve a significant advancement in this topic and to obtain efficient photoelectric devices requires resolving the following points:

Conclusions, Perspective, and Outlook
The low price, high physicochemical stability, and attractive electronic band structure of g−CN materials, together with the recent developments in their controlled growth on various substrates, contribute to expectations of an infusive future for g−CN materials in PEC applications. The overview of the topics involved in this review is illustrated in Figure 29. In contrast to the many groundbreaking results on g−CN powders and their demonstrated functionality in various applications to date, the research reports on g−CN materials in photoelectrochemical applications are still in their infancy. To this end, to achieve a significant advancement in this topic and to obtain efficient photoelectric devices requires resolving the following points: Figure 29. Schematic illustration of the topics involved in this review.

Improvement in the deposition methods:
Currently, it is still difficult to achieve a final film that is both homogenous and with good contact with the conductive substrate while concurrently controlling morphology, electronic structure, as well as optical and PEC properties. The deposition methods should be further modified in the aspect of the starting materials design, interfaces, and reactor parameters optimization for achieving the g−CN layers in a more controllable manner. Morphology regulation is the most direct, simple and effective method to improve the electronic structure, surface characteristics, and photoelectrical properties of g−CN. The three-dimensional (3D) sphere or cage structure of g−CN has a larger specific surface area and sufficient reactive sites but has not been explored in the field of PEC water splitting.

Improvement in the deposition methods:
Currently, it is still difficult to achieve a final film that is both homogenous and with good contact with the conductive substrate while concurrently controlling morphology, electronic structure, as well as optical and PEC properties. The deposition methods should be further modified in the aspect of the starting materials design, interfaces, and reactor parameters optimization for achieving the g−CN layers in a more controllable manner. Morphology regulation is the most direct, simple and effective method to improve the electronic structure, surface characteristics, and photoelectrical properties of g−CN. The three-dimensional (3D) sphere or cage structure of g−CN has a larger specific surface area and sufficient reactive sites but has not been explored in the field of PEC water splitting. Compared with other morphologies (bulk, flake and thin film), 0D semiconductor is generally more helpful in broadening the band structure, enhancing the redox ability and open-circuit voltage.

Tuning the band gap of g−CN:
Decreasing the band gap of g−CN would result in increased light absorption and photocurrent density. In general, it can be acquired by optimizing the C/N ratio or the spatial organization of the N atoms or by doping with heteroatoms and halogens. The electronic structure of g−CN materials can be modified by inducing vacancy defects or tailoring monomer design.
3. Improvement of photo-excited charge separation and suppression of the recombination of the carrier: The low electronic conductivity of the g−CN layer leads to a high photo-excited charge recombination rate and to a low diffusion length. The latter restricts the thickness of the active layer and hence limits light absorption. While increasing the thickness of the active harvesting layer results in enhanced light absorption, this also results in higher charge recombination rates. This trade-off is crucial in the PEC efficiency and is restricted by light absorption and charge extraction properties. Replacing some C=N with C=C bonds or doping with heteroatoms/metals could highly improve the conductivity of electrons. Plenty of theoretical and experimental research has proved that the formation of well-designed g−CN-based heterostructures can improve its charge separation and transfer within the electrode. Constructing multi-heterojunction photocatalysts with multi-sources for carrier generation is one of the effective methods for enhancing the PEC activity. Cocatalysts also play an important role in enhancing charge carrier transfer. Two-dimensional transition metal sulfide has a unique electronic structure and large work function, which is considered a potential cocatalyst to replace precious metals. It has the following advantages: (1) a large number of active sites located at both edge and substrate; (2) strong conductivity to promote the rapid transfer of photogenerated electrons; (3) the edge position has higher H + adsorption capacity and accelerates the adsorption and desorption of hydrogen.
4. Enhancing the stability of g−CN: In a photoanode, water oxidation on the semiconductor surface is still the bottleneck that needs to obtain more attention in future studies. The introduction at the g−CN/electrolyte interface of a co-catalyst that can both serve as the catalytic sites for oxygen evolution and act as an extract for holes from the g−CN was demonstrated to be a successful tactic. There are several factors that should be taken into account when incorporating a co-catalyst onto a g−CN layer. One problem is the nature of the contact, in which different types of co-catalysts change the transport of photo-induced charges across the g−CN/electrolyte interface. On the other hand, the electronic properties of the co-catalyst need to match the VB potentials of the CN for promoting hole transfer. To achieve a long-term stability, the modified technologies should be further studied, including protecting g−CN from photo− or chemical-corrosion by protective layers and using both a co-catalyst and a protective layer is possible.
To summarize, although there is a mass of obstacles hindering and taking advantage of carbon nitride materials in PEC devices, we believe that their excellent and unique physicochemical properties will enable us to conquer the current limitations. We believe that enhancements in the synthesis and deposition tactics, as well as a better understanding of their photoelectrochemical and charge separation and transfer processes by the study of structure-activity relationships, will result in their exploitation in many optoelectronic devices.