Promotion effect of metal phosphides towards electrocatalytic and photocatalytic water splitting

Hydrogen evolution from water splitting over semiconductors has been considered one of the most promising ways to address energy shortages and environmental pollution. Searching for low-cost, highly efficient, and durable catalysts is the key to improve the hydrogen production rate. Expensive noble metals, such as Pt and Au, are generally loaded onto semiconductors to promote photocatalytic activity. Metal phosphides are promising candidates to replace noble metals in hydrogen generation via electrocatalytic or photocatalytic water splitting due to their low hydrogen-producing overpotential, tunable electronic structure, high electrical conductivity, and low price. In this review article, the characteristics and synthetic methods of metal phosphides are briefly introduced, and the development of metal phosphides for electrocatalytic or photocatalytic water splitting is presented. Finally, the chal-lenges and future directions of metal phosphides are discussed.


| INTRODUCTION
Energy shortages and environmental pollution are scientific problems that urgently need to be solved in today's society. As the global energy structure is transforming to clean and low carbonization, hydrogen energy, as a secondary energy source with abundant reserves, flexibility, efficiency, and zero emission, is an important breakthrough to address the energy crisis and environmental pollution and promote energy transformation and upgrading. The International Hydrogen Council forecasts that hydrogen (H 2 ) will account for 18% of global energy consumption by 2050. Solar energy with abundant reserves is clean energy that will not cause environmental pollution in the process of its utilization and can be developed locally without transportation problems. Hence, photocatalytic water splitting to produce H 2 under the irradiation of solar energy is an effective way to solve energy shortages and environmental pollution.
The key to this process is to develop a photocatalyst with high activity and stability to achieve efficient and continuous H 2 production.
Since the first report of photoelectrochemical water splitting for H 2 evolution with TiO 2 photoanodes, 1 photocatalytic decomposition of water with semiconductors to produce H 2 has attracted increasing attention. Currently, different kinds of catalysts, such as TiO 2 , 2 ZnO, 3 BiVO 4 ,4 CdS, 5 ZnIn 2 S 4 , 6 Mn x Cd 1−x S, 7 and g-C 3 N 4 , 8 have been applied in photocatalytic water splitting. However, when these photocatalysts are used alone for the H 2 production reaction, the electrons and holes are more easily recombined, inducing inferior photocatalytic performance. Hence, a series of modification methods, such as cocatalyst loading, semiconductor coupling, elemental doping, and defect engineering, are introduced to accelerate the reaction kinetics. Among them, cocatalyst loading is usually used to accelerate the separation and transfer of charge carriers. Metal phosphides are promising candidates to replace noble metals in photocatalytic H 2 production by water splitting due to their low hydrogenproducing overpotential, tunable electronic structure, high electrical conductivity, and low price. For example, Ni 2 P, 9 CoP, 10 Cu 3 P, 11 MoP, 12 and FeP 13 have been demonstrated to be promising cocatalysts in photocatalysis.
Metal phosphides were first reported in the 18th century. After a long time, they gradually became popular with expanding application scope, such as hydrodesulfurization (HDS), hydrodenitrogenation (HDN), and hydroprocessing (HPC). 14 Since Liu et al 15 predicted the outstanding behavior of Ni 2 P (001) towards the hydrogen evolution reaction (HER) by density functional theory (DFT) calculations, metal phosphides were widely used in the HER and oxygen evolution reaction (OER) with high reactivity and stability. The types of metal phosphides and their applications are presented in Figure 1. Currently, an increasing number of metal phosphides are employed in electrochemical water splitting and photocatalytic water splitting.
In this minireview, an overall summary of metal phosphides as promising bifunctional catalysts and highly active cocatalysts for overall electrochemical water splitting and photocatalytic water splitting is presented. Different kinds of metal phosphides acting as electrocatalysts and cocatalysts with excellent electrocatalytic and photocatalytic activities, respectively, are summarized in Tables 2 and 3. Finally, the existing problems, challenges, and future directions of metal phosphides are discussed. It is expected that this review could provide guidance for researchers to design and construct low-cost and high-efficiency metal phosphide systems.

| CHARACTERISTICS OF METAL PHOSPHIDES
Metal phosphides are usually composed of metal and phosphorus. The outer valence electron configuration of phosphorus is 3s 2 3p 3 , and there are five valence electrons and five vacant 3D orbitals in the third shell. Its intrinsic activity determines that phosphorus can form compounds with most metals. Metal phosphides have shown superior catalytic activity, low H 2 evolution overpotential, and tunable composition and structure, which makes them popular among researchers. 16 On the one hand, the electronegative P atoms in metal phosphides can capture the positively charged H* in the electrolyte to produce H 2 , leading to the enhancement of the catalytic activity. On the other hand, by properly adjusting the atomic ratio of metal and P, metal phosphides may have excellent electrical conductivity. 17 Metal phosphides, labeled M x P y (for instance, Ni 3 P, Ni 12 P 5 , Ni 2 P, Ni 5 P 4 , NiP, NiP 2 , and NiP 3 ), with different molar ratios of M and P are divided into metal-rich (x/y ≥ 1) and phosphorus-rich (x/y < 1) metal phosphides. The former is ascribed to covalent compounds with metallic properties, which exhibit excellent thermal and electrical conductivity and high mechanical strength; the latter belongs to semiconductors, whose thermal stability and chemical stability are worse than those of the former. 14,18 According to the Hagg rule, the crystal configuration depends largely on the atomic radius and electronegativity. To achieve geometric stability, the ratio of the atomic radius of the nonmetal to that of the metal is usually between 0.41 and 0.59. The atomic radius of phosphorus (0.100 nm) is larger than that of carbon (0.070 nm) or nitrogen (0.065 nm). The ratio of the atomic radius of phosphorus to that of the metal is not within this scope; hence, it is difficult for phosphorus to form stable structures with metal atoms. Metal phosphides with more coordination unsaturated surface atoms present a higher catalytic activity than carbides and nitrides. 19 Metal phosphides with different crystalline phase structures present various F I G U R E 1 Types and applications (photocatalysis, electrocatalysis, batteries, HDS, HDN, and HPC) of metal phosphides electron conductivities, thermal and chemical stabilities, and catalytic reactivities. 20 M and P in M x P y possess a partial positive charge (δ+) and a partial negative charge (δ−), respectively. 21,22 P atoms in metal phosphides with more electronegativity can draw electrons from metal atoms, which can trap positively charged protons. 23 The crystalline structures of representative metal phosphides are shown in Figure 2. The various crystal structures make metal phosphides show different catalytic activities. Monodispersed nickel phosphide nanocrystals (NCs) were prepared through a phase-controlled synthesis strategy by changing the molar ratio of P and Ni precursors. Hollow-structure Ni 12 P 5 NCs with an average particle size of 17.55 ± 2.25 nm were obtained at a P:Ni precursor molar ratio of 0.65. 24 When the molar ratio of the P:Ni precursors increased to 1.1, Ni 12 P 5 and Ni 2 P coexisted in the products. Upon further increasing the P:Ni precursor ratio to 2.18, Ni 2 P presented a hollow structure with an average particle size of 9.19 ± 1.16 nm. Moreover, when the molar ratio of P:Ni precursors was increased to 8.75, Ni 5 P 4 solid spheres with an average particle size of 600 nm were obtained. Ni 5 P 4 displayed much better catalytic activity than Ni 12 P 5 and Ni 2 P due to the larger positive charge of Ni and stronger ensemble effect of P in Ni 5 P 4 .

| PREPARATION OF METAL PHOSPHIDES
The synthesis of metal phosphides is more complex than that of metal sulfides, metallic oxides, and metal hydroxides. Some methods, such as in-situ phosphating transition metal oxides, hydroxides, and metal-organic frameworks (MOFs) with the PH 3 generated from the decomposition of sodium hypophosphite (NaH 2 PO 2 ), 25,26 the ion exchange strategy 27 and the solvent thermal process with red phosphorus 28 and white phosphorus, 29 have been developed for the fabrication of metal phosphides. High temperature, the release of toxic gases, or the use of toxic materials was involved in this process, making it difficult to achieve large-scale preparation and application. Hence, it is necessary to develop a simple and environmentally friendly strategy for the preparation of metal phosphides.
Generally, the phosphorus source used in the hightemperature calcination preparation process is NaH 2 PO 2 , which decomposes to release PH 3 when heated over 250 C. Then, PH 3 can further react directly with a variety of metal oxides, hydroxides, and MOFs to form metal phosphides. 23 Zhao et al 21 fabricated g-C 3 N 4 /Ni 2 P photocatalyst with NaH 2 PO 2 ÁH 2 O as the phosphorus source and g-C 3 N 4 /Ni (OH) 2 as the precursor. Fe 2 P nanoparticles were synthesized by annealing a mixture of Fe 2 (SO 4 ) 3 and NaH 2 PO 2 ÁH 2 O at 300 C for 2 hours in argon. 30 Porous FeP nanosheets were also obtained by an anion exchange reaction of Fe 18 S 25 -TETAH (TETAH = protonated triethylenetetramine) nanosheets with P ions, 27 which provided guidelines for the transformation of solid inorganic-organic hybrid precursors into nanoporous products while maintaining the original morphology. Single MOF precursor-derived Ni 2 P catalysts were also constructed, in which triphenylphosphine organic ligands were inherently incorporated into the structure of F I G U R E 2 Crystalline structures of A, Ni 3 P; B, Ni 12 P 5 ; C, Ni 2 P; D, Ni 5 P 4 ; E, NiP; F, NiP 2 ; and G, NiP 3 the phosphine coordination materials. 31 Ni(NO 3 ) 2 Á6H 2 O and yellow phosphorus were employed to fabricate Ni 2 P. 32 In addition, Mi et al 29 selectively constructed Ni 2 P and Ni 12 P 5 via a solvent thermal process in the presence of sodium dodecylbenzene sulfonate with NiCl 2 and white phosphorus as the starting reactants. The formation process of Ni 2 P and Ni 12 P 5 was also proposed as follows: Ni 2 + + PH 3 ! Ni 12 P 5 =Ni 2 P + H + : However, the above preparation methods involve high temperature, the release of poisonous gases, such as PH 3 , or the use of toxic substances, such as triphenylphosphine, white phosphorus, and yellow phosphorus, which makes the synthesis process very dangerous. To solve the above problems, researchers replaced white phosphorus with red phosphorus to synthesize Ni 2 P, Ni 12 P 5 , and a mixture of them under mild conditions, 33,34 avoiding the use of toxic substances and high temperature.
In addition, Ni 2 P, Ni 12 P 5 , and Ni 2 P/Ni 12 P 5 biphase nanocomposites were successfully constructed by changing the initial molar ratio of phosphorus and NiCl 2 Á6H 2 O. 33 Compared with bare Ni 12 P 5 and Ni 2 P, Ni 2 P/Ni 12 P 5 biphase nanocomposites exhibited higher catalytic activity for the reduction of 4-nitrophenol. Nanosized nickel phosphides with controllable crystal structures and morphologies were fabricated by Deng et al 35  O, methanol, ethanol, glycol, glycerol, and polyglycol) and using nontoxic red phosphorus as the P 3− ion source, pure crystal phases of Ni 2 P and Ni 12 P 5 with various morphologies were fabricated, as presented in Figure 3.
In addition, MOFs, a kind of functional material exhibiting large specific surface areas and pore size distributions, multiple active sites, and clipping structures, were usually used as precursors to prepare metal phosphides. [36][37][38] MOF-derived nickel phosphide (Ni 2 P) was prepared via a simple, cost-effective procedure using inexpensive and abundant elements. 39 The synthetic process of Ni 2 P is displayed in Figure 4. Ni-BTC MOFs were first prepared via hydrothermal treatment, and certain amounts of Ni-BTC and NaH 2 PO 2 were mixed together, placed in a covered ceramic crucible, and heated to 275 C at a ramp rate of 1 C. In addition, MOF-derived nitrogen-doped FeP nanorods were fabricated through the combination of phosphorization and thermal decomposition. The P atoms in FeP could effectively attract electrons from Fe atoms, serving as the collector of protons. The incorporation of N into FeP could increase the bond strength between FeP and oxygen. 40 Kang et al 41 proposed a universal but facile and controllable sol-gel strategy ( Figure 5) to fabricate carbonsupported metal phosphides with different metal ions as metal sources. During the sol-gel process, citric acid chelated with metal ions to form cross-linked networks in which the inorganic species were homogeneously dispersed. Through the universal strategy, a series of metal phosphides, such as CoP@C, MoP@C, FeP@C, Cu 2 P@C, Ni 2 P@C, PtP 2 @C, NiFeP@C, NiCoP@C, FeNiP@C, CoNiP@C, and FeCoNiP@C, were successfully constructed. This work provided a general methodology for preparing carbon-supported metal phosphides.

| METAL PHOSPHIDES AS BIFUNCTIONAL CATALYST FOR HER AND OER
Electrocatalytic water decomposition is regarded as an efficient and green method because of its convenience and high purity and involves two half reactions: the OER and HER of water. Freshwater has usually been employed as the raw material for water electrolysis in experiments and is scarce in some parts of the Earth. However, the reserves of seawater are enormous, accounting for approximately 97.5% of total water resources. Therefore, it is vital to develop an economical and feasible method for the electrocatalytic decomposition of seawater to produce hydrogen. Generally, catalysts with excellent catalytic activity are noble metals, such as Pt, IrO 2 , and RuO 2 ; however, their high price and low content prevent their large-scale application. Therefore, researchers have replaced precious metals with nonprecious metals to minimize cost and achieve sustainability. 42 The electrocatalytic activities of metals in the periodic table for the HER (following Sabatier's F I G U R E 3 Product distribution of Ni x P y under different conditions. Reproduced with permission: Copyright 2013, The Royal Society of Chemistry 35 principle) and OER vary with the change in the atomic number. 43 The relationship of logarithmic of exchange current densities (log i 0 ) vs bonding adsorption strength (4E M-H ) was calculated by Trasatti ( Figure 6A). A more thorough computation of the "volcano" curve (dotted line) for H 2 evolution (−log i 0 ) is displayed in Figure 6B. [44][45][46] An increasing number of theoretical calculations have confirmed nickel hydroxides, nickel phosphides, nickel oxides, and nickel sulfides to be effective catalysts. Currently, various metal phosphides have been demonstrated to be effective electrocatalysts for the HER and OER. In this section, we mainly introduce the application of metal phosphides in electrocatalytic water splitting.
In an electrolytic cell, the two-electron HER and fourelectron OER proceed at the cathode and anode, respectively, and usually require at least a working potential of 1.23 V. [47][48][49] Stable and efficient catalysts are highly desired for the HER and OER to reduce the overpotential. Generally, catalysts that work for both the HER and OER are often diverse, and their optimal operating conditions, such as electrolyte and pH, are also different; therefore, it is difficult to achieve overall water splitting in a system with two kinds of catalysts under optimal conditions. On the one hand, bifunctional catalysts can be used for both the HER and OER; on the other hand, the use of bifunctional photocatalysts greatly reduces the cost of material preparation. Consequently, the design of bifunctional catalysts that are effective in an electrolyte for both the HER and OER is urgently needed. 50,51 For the decomposition of seawater, the seawater electrolysis rate is controlled by the sluggish OER accompanied by competitive side reactions, such as chloride electrooxidation reactions (CERs). Chloride is present in seawater at high concentrations ($0.5 M), and its oxidation products, such as Cl 2 and OCl − , are toxic and should be minimized to achieve maximum OER selectivity. The reactions are as follows: 52 The CER is more difficult than the OER to occur from a thermodynamic point of view; however, the CER has faster reaction kinetics than the OER. Hence, the OER and CER always proceed simultaneously. 53,54 Hence, different methods have been employed to promote the catalytic activity of the HER and OER while avoiding the occurrence of side reactions. For example, Co 3 O 4modified MnO 2 exhibited a smaller OER overpotential of only 450 mV, which is lower than that of hypochlorite formation (480 mV). Moreover, the iodine titration test also proved that no hypochlorite was produced during the OER process. Excellent catalytic activity was obtained in alkaline seawater splitting. The Tafel slope values and the current densities were ≈40 mV dec −1 and 13 mA cm −2 , respectively, which was superior to those of most Mn 2 O 3 -based catalysts. This work provided guidelines for selective seawater electrolysis. 42 Metal phosphides such as Ni 2 P, Fe 2 P, and CoP have been demonstrated to be highly efficient electrocatalysts for a single HER or OER. Exploring novel materials with both superior HER and OER performances remains a top priority. Currently, there has been increasing attention on using metal phosphides such as Ni 2 P-Fe 2 P, 17 NiCo-P, 55 and Co 2 P 56 as bifunctional catalysts towards overall electrochemical water splitting. However, when metal phosphides are used alone for water splitting, their activity is unsatisfactory. Researchers have devised various strategies, such as ion doping, noble metal loading, and microstructure regulation, to boost catalytic activity.
Based on this, S-incorporated Co 2 P was constructed via an economical and ecofriendly urea-phosphate-assisted strategy, in which doping with S with a higher electronegativity than Co and P led to a more positive charge on Co δ+ and a more negative charge on P δ− and S δ− , as supported by XPS results. 56 DFT calculations confirmed the reduced density of states (DOS) of metallic Co, as observed from the contraction of the conduction/valence band near the Fermi level in S:Co 2 P compared with that in pure Co 2 P, as shown in Figure 7, which signified the enhanced hydride acceptor (Co δ+ H δ− ) and proton acceptor (P/S δ− H δ+ ) properties after the doping of S. Hence, a stable current density of 100 mA cm −2 at 1.782 V was obtained with S:Co 2 P@NF as both the cathode and anode in alkaline electrolytes, which was superior to that obtained with the Pt/C and IrO 2 systems. S-doped Co 2 P displayed smaller Tafel slopes and higher current densities (62 mV dec −1 and 0.133 mA cm −2 for the HER and 71-82 mV dec −1 and 1-3 mA cm −2 for the OER) than primary Co 2 P (113 mV dec −1 and 0.116 mA cm −2 for the HER and 94 mV dec −1 and 1.3 mA cm −2 for the OER) for the HER and OER, respectively, illustrating that significantly improved carrier transfer dynamics were achieved after the incorporation of S. This research provided a general approach to fabricate phase-pure Co 2 P without the use of expensive and toxic phosphines. In addition, Chen et al 57 designed an ultralow Ru (1.08 wt%)-loaded transition metal phosphide supported by nickel foam (Ru-MnFeP/NF), which presented excellent electrocatalytic activity, requiring overpotentials of only 191 and 35 mV to obtain current densities of 20 mA cm −2 and 10 mA cm −2 for the OER and HER, respectively. The Tafel slope of Ru-MnFeP/NF for the HER (36 mV dec −1 ) and OER (69 mV dec −1 ) was the smallest among the Tafel slopes of all electrocatalysts, implying its outstanding HER and OER kinetics due to its intrinsic catalytic activity. The amount of H 2 and O 2 produced in 6 minutes was exactly in line with the theoretical value, implying that the Faradic efficiency was close to 100% and that the overall water splitting process was highly selective. DFT calculations indicated that Ru-loaded Fe 2 P or Mn 2 P possessed a lower *H adsorption capabilities, and the electronic structure between Ru and phosphides would change after the incorporation of Ru, inducing the improvement of HER activity due to the interaction of Ru and Fe 2 P/Mn 2 P.
It was reported that ternary bimetallic phosphides with the advantages of individual metal phosphides showed superior catalytic activity relative to the corresponding binary metal phosphides, which is beneficial for the configuration of electronic structures. 58 For example, nanoporous (Ni x Fe 1−x ) 4 P 5 with a tunable Ni/Fe ratio was constructed by Xu et al 59 via electrochemical dealloying. A current density of 10 mA cm −2 was obtained for the OER and HER over the np-(Ni 0.67 Fe 0.33 ) 4 P 5 catalyst at overpotentials of 245 mV and 120 mV in 1 M KOH, respectively, with corresponding Tafel slopes of 32.9 and 41.8 mV dec −1 for the OER and HER, respectively. The CV curves confirmed that the current density increased as the number of aging cycles increased, implying that the catalytic activity was significantly boosted. Ni oxidation, such as Ni x Fe 1−x OOH, was generated in the aging process, which indicated that Ni x Fe 1−x OOH would be the real active site for the HER. However, the aged sample displayed lower catalytic activity than the fresh one towards the HER, which suggested that the metal phosphides might be the active sites for the HER. In addition, the electrochemically active surface area demonstrated that amorphous (Ni x Fe 1−x ) 4 P 5 showed superior catalytic activity than crystalline (Ni x Fe 1−x ) 4 P 5 because of its abundant active sites and disordered atomic arrangement. The maximum shift in Ni 2p and P 2p caused the change in the electronic structure for both Ni and P atoms, which was beneficial for the improvement of catalytic activity.
Generally, catalytic reactions occur at the active sites on catalyst surfaces; the more active sites exposed, the faster the catalytic reaction rate. The number of exposed catalytic active sites is determined by the apparent morphology and microstructure of the catalyst. To increase the number of active sites exposed, researchers fabricated metal phosphides possessing structures including nanoparticles, 59 nanofibers, 60 nanoframes, 58 and nanocubes 61 with a large specific surface area and pore structure, which favors contact between the electrolyte and catalysts. For instance, Lv et al 61 fabricated doubly functionalized carbon carved hollow nanocubes, C-(Fe-Ni)P@PC/(Ni-Co)P@CC. The overpotentials required to attain a current density of 10 mA cm −2 in 1 M KOH were 142 and 251 mV, with Tafel slopes of 98 and 56 mV dec −1 for the HER and OER, respectively. The faradaic efficiency of C-(Fe-Ni)P@PC/(Ni-Co)P@CC was nearly 100%. The overpotential of C-(Ni-Co)P@CC was 308 mV at 10 mA cm −2 , which was much higher than that of C-(Fe-Ni)P@PC/(Ni-Co)P@CC (251 mV) and (Fe-Ni) P@PC/(Ni-Co)P@CC (280 mV), illustrating that the exterior (Fe-Ni)P was the main active substance for the OER. Meanwhile, the HER performance results proved that (Ni-Co)P was the dominant active material towards the HER. The synergistic effects between (Fe-Ni)P and (Ni-Co)P jointly promoted the improvement in HER and OER performance. In addition, the hollow structure of the catalyst could provide more electrocatalytic active sites, and the holes located at the eight vertices of the cube could reduce the transfer distance of electrolytes and bubbles. This novel bifunctional carved hollow nanocube structure provided guidelines for the design of F I G U R E 7 Total-DOS and projected DOS of A, Co 2 P and B, S: Co 2 P. C, The electron distribution of S: Co 2 P. D, The mechanism of photocatalytic water decomposition. Reproduced with permission: Copyright 2018, American Chemical Society 56 bifunctional electrocatalysts. Using metal phosphides as bifunctional catalysts to design electrolyzers has also achieved amazing results. Yu et al 62 constructed porous Ni 2 P and FeP supported on Ni foam with Fe(NO 3 ) 3 , Ni foam, and phosphorus as the Fe, Ni, and P sources, respectively. Ni 2 P and FeP supported on Ni foam required a very low cell voltage of 1.42 V to afford 10 mA cm −2 in alkaline water electrolyzers. A voltage of only 1.72 V was needed to achieve a commercially practical current density of 500 mA cm −2 and superior catalytic stability for more than 40 hours. The lowest Tafel slope of FeP/Ni 2 P for the HER was only 24.2 mV dec −1 , which is even lower than that of Ni 2 P (117.3 mV dec −1 ) and Pt (36.8 mV dec −1 ) due to FeP/Ni 2 P having a larger active surface area and smaller resistance. The number of active sites in FeP/Ni 2 P was approximately 2.5 times that in Ni 2 P, demonstrating that the incorporation of FeP on Ni 2 P contributed to the improvement in HER activity. FeP/Ni 2 P displayed outstanding catalytic activity and stability, paving the way for promising large-scale H 2 generation.
In addition, the bimetallic heterostructure Ni 2 P-Fe 2 P was fabricated by dipping commercial Ni foam in HCl (3 M) and Fe(NO 3 ) 3 Á9H 2 O (0.1 M) aqueous solutions. The bifunctional Ni 2 P-Fe 2 P heterostructure displayed excellent catalytic activity and stability towards overall water/ seawater splitting. The Ni 2 P-Fe 2 P electrocatalysts achieved current densities of 100 mA cm −2 in KOH (1 M) and KOH seawater (1 M) at voltages of 1.682 and 1.811 V, respectively, which were superior to those of the IrO 2 jjPt/C. The Tafel slopes of Ni 2 P-Fe 2 P/NF were 58 and 86 mV dec −1 for the OER and HER, respectively. In addition, the enhanced stability and corrosion resistance of Ni 2 P-Fe 2 P due to the alloying of Ni, Fe, and P atoms enable them to be highly efficient in seawater. Only H 2 and O 2 could be found by gas chromatography, the amount of H 2 and O 2 was consistent with the theoretical value, and no Cl 2 signal was observed, which demonstrated that the faradaic efficiency of Ni 2 P-Fe 2 P/NF in overall seawater splitting reached nearly 100%. In contrast to the hydrothermal and electrochemical deposition methods reported in other literature to prepare bimetallic compounds, the preparation of Ni 2 P-Fe 2 P could be carried out at room temperature, the loading of Fe was adjusted by the concentration of the Fe(NO 3 ) 3 Á9H 2 O solution, and the reaction solution could be reused, which was conducive to large-scale production. This study provided a new way for the large-scale synthesis of bifunctional catalysts for the efficient decomposition of water/ seawater to produce hydrogen. 17 For electrocatalytic water splitting, the wettability of water is very important for catalytic reactions; the larger the contact angle with water is, the more favorable the contact between the electrolyte and the electrode surface. The high wettability can improve contact between the electrode and the electrolyte, promote the migration of photogenerated carriers and reduce ohmic losses. 63,64 NiFePi and NiFePi/P possessed smaller contact angles (63 ± 3 , 69 ± 3 ) than NiFe (100 ± 3 ) and NiFeP (127 ± 3 ), implying that the affinity of the catalyst for water remarkably improved after phosphorylation. In addition, NiFePi/P presented the lowest charge transfer resistance (Rct) (0.57 Ω) and the lower series resistance (Rs), as displayed in Figure 8, which confirmed that phosphorylation could greatly improve the electrical conductivity and charge transfer by improving the surface wettability. For NiFePi/P, Rct was the lowest, but Rs was not. NiFePi/P presented the lowest overpotential of 230 mV at 10 mA cm −2 , with a Tafel slope of 57 mV dec −1 , which further demonstrated that wettability is more vital than electrical conductivity in contributing to the OER catalytic activity. The XPS results confirmed that a synergistic effect between metal phosphates and phosphates can be achieved by changing the amount of P-O-P and M-O-P in NiFePi/P, leading to a change in the electron environment around metal ions, an improvement in catalytic activity, and an increased number of Fe 3+ and Ni 3+/4+ active sites. 65 Surface phosphorylation of metal phosphides is helpful in improving the wettability of water on metal phosphides. It was reported that the distorted-tetrahedral cobalt geometry derived from the phosphate/pyrophosphate groups was beneficial to water adsorption. 66 The surface wettability of NiFe/NiFe:Pi (contact angle of 44 ± 3 ) remarkably improved relative to that of NiFe hydroxide (contact angle of 129 ± 5 ). Moreover, the electrochemical polarization also changed the surface wettability of the sample. NiFe/NiFe:Pi became superhydrophilic without a contact angle after long-term catalysis. For NiFe, the contact angle changed from $129 to $45 under electrochemical polarization conditions. The current density of NiFe/NiFe:Pi increased by two times at a potential of 1.65 V relative to that of the NiFe electrode, which confirmed the synergistic effect between NiFe and NiFe:Pi. The Tafel slope of NiFe/NiFe: Pi was 38 mV dec −1 , which is smaller than that of NiFe (48 mV dec −1 ) and Ni:Pi (50 mV dec −1 ). 64

| METAL PHOSPHIDES AS COCATALYSTS FOR PHOTOCATALYTIC WATER SPLITTING
The process of photocatalytic H 2 evolution from water splitting is as follows: (a) absorption of photons; (b) separation and transfer of photoinduced electrons and holes; and (c) oxidation and reduction reactions. 20 Currently, the major scientific issues of photocatalytic water decomposition are the low light absorption efficiency, severe photogenerated carrier recombination, and the slow surface reaction rate. Based on the main problems of photocatalytic water splitting, the catalyst was designed and modified in the following ways to improve the rate of H 2 production by researchers: (a) researchers introduced an impurity level between the forbidden bandwidth by metal-or nonmetal-doping or regulated the defects of materials to increase the response range of visible light, 67,68 (b) designed the microstructure and surface morphology of catalysts to speed up the transport of carriers or reduce their migration distance, 7,69 (c) loaded cocatalysts to accelerate surface chemical reactions and inhibit the fast recombination of charge carriers, 70 (d) coupled two different semiconductors to take advantage of their respective advantages, such as the construction of Z-scheme heterojunctions and the formation of p-n junctions, (e) doped foreign atoms in semiconductors to induce the formation of vacancies or defects ( Figure 9). Among them, loading cocatalysts on the surface of semiconductors has usually been employed to improve the surface reaction efficiency. Metal phosphides with excellent electrical conductivity, low H 2 production overpotentials, and high stability have been extensively studied by researchers.
Generally, metal phosphides have physical properties similar to those of metal carbides and nitrides, such as excellent thermal and electrical conductivity and high mechanical strength. Compared with carbon and nitrogen atoms, phosphorus atoms possess a larger atomic radius, and it is hard to reside in the interstitial spaces between metal atoms to form a simple mesenchymal crystal structure. 14 Hence, metal phosphides tend to form a more isotropic crystal structure, as displayed in Figure 2, which makes them possess more unsaturated surface atoms for coordination. Consequently, metal phosphides exhibit much better catalytic activity than metal carbides, nitrides, and sulfides. 50 Major scientific issues and modification strategies for photocatalytic water decomposition overpotential, tunable electronic structure, high electrical conductivity, and low price are promising candidates to replace noble metal cocatalysts in photocatalytic H 2 production by water splitting. The work function (W f ) refers to the minimum energy required to move an electron from the inside of a solid just to the surface, reflecting its ability to bind electrons. It affects the catalytic performance by affecting the injection of carriers. A high work function is an important index to evaluate the excellent catalytic activity of catalysts. The larger the work function is, the more favorable it is for the catalyst to capture electrons. 71,72 The barrier layer and anti-barrier layer would form at the interface between the metal and semiconductor due to the difference in the work functions of the metal and semiconductor, as presented in Table 1. When a metal is in contact with an n-type semiconductor, if the work function of the metal (W m ) is greater than the work function of the semiconductor (W s ), electrons enter the metal from the semiconductor, and the energy band at the surface of the semiconductor bends upward to form a surface barrier, which is a high-resistance area called a barrier layer. The built-in electric field is oriented from the semiconductor to the metal, which is not conducive to the transfer of electrons from semiconductors to metals. Otherwise, if W m is smaller than W s , electrons enter the semiconductor from the metal, and the energy band at the surface of the semiconductor bends downward to form an antibarrier layer, which is beneficial for electron migration from semiconductors to metals. In addition, for p-type semiconductors, the opposite is true, as displayed in Table 1. Hence, it is very important to choose metals and semiconductors with matching work functions.
For n-type semiconductors, choose a metal with small W f , and for p-type semiconductors, choose a metal with large W f , which can reduce the Schottky barrier height at the interface between the metal and the semiconductor, benefitting the injection of carriers. For example, the work function of Cu (111) was 4.99 eV, which perfectly matches that of WC (001) (5.11 eV) relative to Ag (111) (4.74 eV), Au (111) (5.31 eV), Al (111) (4.24 eV), Ni (111) (5.35 eV), and Fe (111) (4.74 eV). The absolute work function difference between Cu and WC was the smallest among these metals, implying the preservation of a high work function for the WC (001) surface. In addition, 4G H* was reduced to −0.36 eV from −0.72 eV, which favored the improvement in HER performance. 73 A Schottky junction will form at the interface when metal phosphides touch the semiconductor because of the difference in W f , which will induce the bending of the conduction band. 74 For example, when Fe 2 P, Co 2 P, and Fe 2 P-Co 2 P are coupled with g-C 3 N 4 , the electrons flow from metal phosphides to g-C 3 N 4 because of the difference in W f until their Fermi levels are equal, causing the band to bend downward for g-C 3 N 4 , as displayed in Figure 10. 75

| Nickel phosphide
Nickel phosphides are the metal phosphide cocatalysts most commonly combined with semiconductors in photocatalytic water splitting due to their low H 2 evolution overpotential and unique structural features. 76 DFT calculations have shown that hollow Ni sites in Ni 2 P made Ni 2 P possess more active sites for anchoring reactants. 15 Noble-metal-free Ni 2 P cocatalyst-decorated g-C 3 N 4 composites were fabricated through a two-step hydrothermal and phosphidation method by Zhao et al. 21 Under the optimal conditions, the H 2 evolution rate reached 5.67 μmol h −1 , which is approximately 1418 times greater than that of pure g-C 3 N 4 and even higher than that of the noble metal Pt-loaded g-C 3 N 4 . The Ni and P in the Ni 2 P cocatalyst acted as the hydride-acceptor and proton-acceptor centers, respectively, inducing an increase in the rate of H 2 production. Similarly, Indra et al 77 prepared Ni 2 P-modified mesoporous graphitic carbon nitride (sg-CN) by a sol-gel method, which showed a much higher catalytic activity than the physical mixture of Ni 2 P and sg-CN or metallic nickel on sg-CN under similar conditions. In this integrated system, the loading of Ni 2 P not only decreased the charge recombination rate but also accelerated the surface chemical reaction. Zeng et al 78 developed a novel solution-phase method to construct highly monodisperse zero-dimensional (0D) nickel phosphide (Ni 2 P) nanoparticles and then anchored them on the surface of two-dimensional (2D) porous g-C 3 N 4 nanosheets via a solution-phase self-assembly process. As presented in Figure 11, monodisperse Ni 2 P nanoparticles with good crystallization and small particle sizes were produced by the phosphorization of intermediate monodisperse Ni. Ni 2 P nanoparticles were tightly anchored on the surface of porous g-C 3 N 4 nanosheets. g-C 3 N 4 /3.5% Ni 2 P presented the highest H 2 evolution rate of 474.7 μmol h −1 g −1 because Ni 2 P with excellent electrical conductivity was an effective cocatalyst for accelerating the separation of holes and electrons.  In view of the above problems, Li et al 25 constructed 2D/2D Ni 2 P/ZnIn 2 S 4 nanohybrids to increase their contact area and reduce the carrier diffusion distance, which confirmed that Ni 2 P was an effective cocatalyst to improve the photocatalytic performance of H 2 generation for sulfides. In addition, Ni 2 P was also demonstrated to be an effective cocatalyst for Zn 0.5 Cd 0.5 S, 79 CdS, 80 g-C 3 N 4 , 77 ZnO, 81 and red phosphorus 34 in the photocatalytic field. To solve the photocorrosion problem of sulfide and enhance its stability, Ni 2 P@CdS composites with core-shell structures were constructed by Zhen et al 82 via a two-step solvent thermal process, as depicted in Figure 12. In this system, artificial gills were applied to remove the O 2 generated during the process of photocatalysis, and the occurrence of CdS photocorrosion was inhibited to a certain extent. Meanwhile, the reverse reaction of H 2 and O 2 was prevented. ICP results confirmed that the Ni 2 P shell played a vital role in the protection of CdS from photocorrosion.
It has been reported that the microstructure of a material has a great effect on catalytic activity. Quantum dots (QDs) with smaller diameters possessed unique physical and chemical properties due to the quantum confinement effect, which could reduce the transfer distance of carriers. For example, Ni 2 P QDs-modified ultrathin g-C 3 N 4 nanosheets were fabricated by Lu et al 83 . The maximum H 2 production rate was 1503 μmol h −1 g −1 at the optimal Ni 2 P loading amount, which was much better than that of noble metal Ptloaded g-C 3 N 4 (560 μmol h −1 g −1 ). As displayed in the TEM images of Figure 13, ultrasmall Ni 2 P QDs were tightly anchored on the surface of ultrathin g-C 3 N 4 nanosheets, and the contact between them was close. According to the characterization results, the valence band value of g-C 3 N 4 shifted up after the modification of Ni 2 P QDs, enhancing the photocatalytic reduction ability. The light absorption capacity of Ni 2 P/g-C 3 N 4 was significantly enhanced relative to that of pure g-C 3 N 4 , which was beneficial for improving the H 2 evolution rate. In addition, the XPS results confirmed the formation of Ni N chemical bonds between Ni 2 P QDs and g-C 3 N 4 nanosheets, which was favorable for the timely migration and separation of photonic carriers. Hence, the photocatalytic and photoelectric activity of g-C 3 N 4 significantly improved after the incorporation of Ni 2 P QDs. This work showed that metal phosphides are a promising and lowcost cocatalyst for the replacement of noble metals.
In addition, QDs have been regarded as one of the most promising structures due to their small particle size. On the one hand, a smaller particle size is advantageous to the timely transfer and separation of photocarriers, inhibiting the recombination of electrons and holes; on the other hand, QDs possess a large surface area, providing more catalytically active sites. 83,84 Ni 2 P QDs have also been employed as cocatalysts in combination with other semiconductors, such as TiO 2 and red phosphorus. A 0D/2D Ni 2 P/TiO 2 architecture was fabricated by Luo et al 85 via a facile solvothermal method, and the H 2 generation rate of TiO 2 significantly improved after the introduction of Ni 2 P. Ni 2 P dispersed on ultrathin TiO 2 nanosheets acted as the electron acceptor, reducing H + into H 2 , and the intimate interface between Ni 2 P and TiO 2 was advantageous to the timely transfer and separation of photocarriers, accelerating the reaction rate. Liang et al 34 employed Ni 2+ as a chemical scissor to construct a 0D/2D Ni 2 P/red P photocatalyst using a facile and mild in situ hydrothermal method. Ni 2 P in the form of QDs was loaded on the surface of red P nanosheets. During the hydrothermal process, Ni 2+ not only served as a nickel source to fabricate Ni 2 P but also functioned as a chemical scissor during the formation of red P nanosheets.
In addition to Ni 2 P, Ni 12 P 5 is usually employed as a cocatalyst and loaded on the surface of semiconductors for water splitting. Ni 12 P 5 /g-C 3 N 4 composites were prepared by Wen et al. 86 At the optimal Ni 12 P 5 loading amount, the rate of H 2 generation was 126.61 μmol h −1 g −1 , which was up to 269.4 times higher than that of g-C 3 N 4 . Ni 12 P 5 nanoparticles on the surface of g-C 3 N 4 trapped electrons from the conduction of g-C 3 N 4 , acted as the active sites to promote the separation of charge carriers and accelerated the H 2 evolution kinetics due to its low H 2 evolution overpotential. 87 In addition, colloidal Ni 12 P 5 nanoparticles were successfully embedded into porous g-C 3 N 4 nanosheets through a facile solution-phase approach. The H 2 evolution rate reached 535.7 mmol h −1 g −1 at the optimal content of Ni 12 P 5 . As displayed in Figure 14, Ni 12 P 5 nanoparticles evenly inserted into the porous g-C 3 N 4 nanosheets, which acted as excellent catalytic activity sites for H 2 generation, inducing the improvement in H 2 evolution. 88 This study was beneficial to the development of effective and robust g-C 3 N 4 -based photocatalysts. Different crystal structures also lead to different H 2 production rates. Sun et al 22 fabricated three different crystalline phases of Ni 2 P-, Ni 12 P 5 -, and Ni 3 P-hybridized g-C 3 N 4 composites and systematically investigated the effect of the phase structure of Ni x P y on H 2 production activity. It was demonstrated that the H 2 generation rate of all types Ni x P y /g-C 3 N 4 composites obviously improved after the incorporation of Ni x P y , implying that Ni x P y was an effective cocatalyst for g-C 3 N 4 . In particular, Ni 2 P/g-C 3 N 4 hybrids showed the highest rate of H 2 evolution among them. Since the P atoms in Ni x P y played an essential role in trapping the positively charged proton and delivered H 2 , Ni 2 P with more P atoms could provide more Ni P bonds, leading to an increase in the photocatalytic activity, as displayed in Figure 15. In addition, amorphous Ni x P-modified CdS nanorods were constructed via a facile photoreduction method using NiCl 2 and NaH 2 PO 2 as the nickel and phosphorus sources, respectively, in which Ni x P acted as a low-cost cocatalyst uniformly distributed onto the CdS nanorods. The presence of Ni x P in Ni x P/CdS improved the light absorption performance, boosted the photocurrent intensity, lowered the photoelectric impedance, reduced the fluorescence intensity, and prolonged the fluorescence lifetime of CdS, effectively promoting the migration and separation of charge carriers. The abovementioned research confirmed that Ni x P y was an inexpensive and efficient cocatalyst in the photocatalytic decomposition of water. 89 Unlike the above binary photocatalyst system, ternary photocatalysts can achieve the efficient and timely separation of electrons and holes in space, which can significantly increase the rate of H 2 production. In addition, ternary photocatalysts possess an enhanced light absorption capacity and a faster carrier transfer rate than binary photocatalysts. Thus, researchers designed ternary photocatalysts to enrich electrons and holes on different catalysts, inhibiting their recombination. NiO/Ni 2 P/g-C 3 N 4 (NiO/Ni 2 P/CN) ternary photocatalyst was constructed by Shi et al 90 via a one-step in situ phosphating strategy. As displayed in Figure 16, Ni 2 P acted as a cocatalyst to receive electrons from the conduction band of CN, reducing H + into H 2 because of its low H 2 evolution overpotential. Meanwhile, NiO accepted holes from the valence band of CN, inducing the timely transfer and separation of photocarriers in space. The maximum H 2 evolution rate (5.04 μmol h −1 ) was 126 times higher than F I G U R E 1 3 TEM images of A, g-C 3 N 4 and B-E, g-C 3 N 4 /Ni 2 P. Size distribution of Ni 2 P in g-C 3 N 4 /Ni 2 P (inset of B). F-I, Elemental mapping of g-C 3 N 4 /Ni 2 P. Reproduced with permission: Copyright 2018, Elsevier 83 that of CN, which was ascribed to the intimate contact among NiO, Ni 2 P, and CN and the timely transfer and separation of photocarriers. This work might provide guidelines for the construction of highly efficient ternary photocatalysts. Jiang et al 91 synthesized Ni 2 P/Ni(PO 3 ) 2 /g-C 3 N 4 3D heterojunction photocatalysts by the hydrothermal and calcination method, in which Ni 2 P/Ni(PO 3 ) 2 nanoparticles were evenly dispersed on the surface of g-F I G U R E 1 4 TEM images of A, g-C 3 N 4 ; B, Ni 12 P 5 ; and C-F, 5Ni 12 P 5 -g-C 3 N 4 . E, HAADF image and G-K, elemental mapping images of the C, N, Ni, and P of 5Ni 12 P 5 -g-C 3 N 4 . Reproduced with permission: Copyright 2017, The Royal Society of Chemistry 88 C 3 N 4 nanosheets. When Ni 2 P and Ni(PO 3 ) 2 existed together, the H 2 evolution rate of g-C 3 N 4 reached the maximum, demonstrating the existence of a synergetic effect between them. In addition, an N-TiO 2 /g-C 3 N 4 heterostructure modified with a Ni x P cocatalyst was fabricated by Wu et al 92 via ammonia etching and photoreduction deposition method. N-doped TiO 2 nanosheets formed and stood upright on the surface of g-C 3 N 4 nanosheets with a 3D structure. The loading of the Ni x P cocatalyst and the formation of intimate nanoheterojunctions between TiO 2 nanosheets and g-C 3 N 4 nanosheets promoted charge transfer. This work provided some perspectives for the design and construction of the microstructure of catalysts.
To inhibit the photocorrosion of sulfides, Qin et al 93 developed a Ni 2 P-Cd 0.9 Zn 0.1 S/g-C 3 N 4 core-shell heterostructure with Cd 0.9 Zn 0.1 S crystalline core and g-C 3 N 4 shell layer, achieving timely effective carrier separation in space, as displayed in Figure 17. The electrons and holes transferred to Cd 0.9 Zn 0.1 S and g-C 3 N 4 , respectively, due to the formation of a type II heterojunction at their interface. Meanwhile, Ni 2 P as a cocatalyst received electrons from g-C 3 N 4 , reducing H + into H 2 . The g-C 3 N 4 layer covering the surface of Cd 0.9 Zn 0.1 S collected holes from Cd 0.9 Zn 0.1 S, which could effectively suppress the photocorrosion of Cd 0.9 Zn 0.1 S, inducing a significant improvement in H 2 production activity and stability. The maximum H 2 production rate was up to $2100 μmol h −1 mg −1 , with an apparent quantum yield of 73.2% at 420 nm and an excellent photocatalytic H 2 evolution stability of 90 hours. In addition, Yu et al 94 fabricated Ni 2 P/Cd 0.5 Zn 0.5 S/Co 3 O 4 (Ni 2 P/CZS/ Co 3 O 4 ) composites that showed better photocatalytic activity than bare CZS. The electrons transferred from the CB of Co 3 O 4 to the CB of CZS, the holes shifted from the valance band (VB) of CZS to the VB of Co 3 O 4 , and Ni 2 P as a cocatalyst received the electrons from the CB of CZS and reduced H + into H 2 , achieving efficient space charge separation in time. This study provided a deep understanding of the fabrication of photocatalysts with unique microstructures to achieve timely space charge separation.
2D black phosphorus has a large surface area and excellent electrical conductivity, which makes it very attractive in the catalytic field. For example, 2D-black phosphorus (BP)-supported Ni 2 P (2D-Ni 2 P@BP) was integrated with 2D mesoporous graphitic carbon nitride (CN). 95 In the integrated system, BP acted not only as a catalyst but also as a source of phosphorus for the synthesis of Ni 2 P, inducing close contact between Ni 2 P and BP. The obtained 2D-Ni 2 P@BP/CN displayed superb photocatalytic activity, which was attributed to the synergetic effect of fast carrier separation, the fast transfer of electrons from CN to Ni 2 P@BP and rapid H + reduction into H 2 on Ni 2 P. Meanwhile, 2D Ni 2 P@BP combined F I G U R E 1 5 Photocatalytic H 2 production mechanism of Ni x P y /g-C 3  with CN led to a greater interface contact area, significantly reducing the carrier transfer distance. Yan et al 96 designed ternary Ni 2 P/rGO/g-C 3 N 4 nanotubes with a maximum H 2 generation rate of 2921.9 mmol h −1 g −1 , which was approximately 35, 16, and 9 times higher than that of g-C 3 N 4 , rGO/g-C 3 N 4 , and Ni 2 P/g-C 3 N 4 , respectively. In this integrated system, rGO acted as an electron transfer medium, and Ni 2 P worked as a reaction site. This work confirmed that Ni 2 P was an effective cocatalyst in photocatalytic H 2 evolution and that the fabrication of ternary photocatalysts was beneficial to the improvement of H 2 generation.
MOFs have attracted much attention due to their unique physical and chemical properties. Ni-based MOFs were employed as the source of nickel to fabricate Ni 2 P/ Ni nanoparticles encapsulated in carbon/g-C 3 N 4 composites via in situ pyrolysis and phosphidation. 97 The maximum H 2 evolution rate was up to 18.04 mmol h −1 g −1 with eosin Y (EY)-sensitization, which was 13 times higher than that of pristine g-C 3 N 4 . In this system, carbon acting as an electron transport bridge accepted the electrons from excited EY and g-C 3 N 4 and then transferred them to Ni 2 P and Ni. The rapid separation of carriers, intimate contact interface, and matching bandgap array among g-C 3 N 4 , Ni, and Ni 2 P, as well the accelerated proton reduction reaction promoted by Ni 2 P/Ni NPs, were the main reasons for the improvement in catalytic activity. The CdS/MoS 2 @Ni 2 P ternary photocatalyst was prepared via a hydrothermal and MOF template strategy. The optimal CdS/MoS 2 @Ni 2 P displayed an excellent H 2 production rate, which was 69.29 times that of pure CdS. In this system, MoS 2 and Ni 2 P acted as electron acceptors and cocatalysts, respectively, and the electrons captured by MoS 2 transferred to Ni 2 P and then reacted with H + to generate H 2 . 98 In addition, a g-C 3 N 4 /C@Ni 3 S 4 /Ni 2 P hybrid photocatalyst was fabricated via the in situ sulfuration and phosphidation of g-C 3 N 4 /Ni-MOF, and the highest photocatalytic H 2 evolution rate was 14.49 mmol h −1 g −1 with EY as the sensitizer, which was ascribed to the staggered band alignment among each component and the synergistic effect between Ni 2 P and Ni 3 S 4 . 99 The above research supplied some guidance for the construction of MOF-derived cocatalysts.

| Cobalt phosphide
It has been reported that the photocatalytic reaction mechanism of Co-based phosphides is similar to what hydrogenases naturally do. 100 DFT simulations confirmed that the rate-limiting step in the process of H 2 generation was the H 2 desorption energy from the cobalt phosphide (CoP) surface, which needed an energy input of 0.63 eV, as shown in Figure 18. This value is lower than that of the noble metal Pt (0.74-0.86 eV), which indicated that noble metal-free CoP is a promising candidate for H 2 evolution. 101 Furthermore, DFT calculations also confirmed that the electrical conductivity of CoP was mainly derived from Co atoms. 102 Luo et al 10 constructed CoP/g-C 3 N 4 composites via the phosphorization of the Co 3 O 4 QDs/g-C 3 N 4 precursor. The highest H 2 generation rate was 1.074 mmol h −1 g −1 , which is 283-fold higher than that of pure g-C 3 N 4 . The characterization results showed that the formation of a new electron transfer route created by the interaction of Co with N atoms promoted the fast transfer and separation of carriers, that the lower energy barrier of CoP than Pt for H 2 generation made CoP a robust and low-cost H 2 production cocatalyst, and that the homogeneously dispersed CoP nanoparticles offered many more reaction sites for H 2 evolution, which together promoted the H 2 production rate. This work offered an effective method to fabricate highly dispersed cocatalysts in the matrix of host catalysts.
In addition, CoP combined with a CdS photocatalyst was fabricated by Xu et al 103 and applied for the oxidation of aryl alcohols to form aromatic aldehydes, which was accompanied by H 2 evolution in water. The species driving the catalytic reaction were hydroxyl free radicals derived from the decomposition of water. Meanwhile, the isotope labeling experiment confirmed that water worked as a hydrogen source. Qiu et al 104  The rapid recombination of electrons and holes was the main factor hindering the improvement in the H 2 production rate; reducing the transfer distance of carriers and accelerating the migration rate of carriers were efficient methods for promoting photocatalytic performance. For example, a 2D/2D CoP/g-C 3 N 4 interface was designed to shorten the transmission distance of the photogenerated charges. Theoretical calculations confirmed that the Schottky effect existed between CoP and g-C 3 N 4 . The unique 2D/2D CoP/g-C 3 N 4 heterostructure enlarged the Schottky effect and shortened the transmission distance of the photogenerated charges, inducing the promotion of catalytic activity. 105 The construction of a 0D/2D CoP/black phosphorus (CoP/BP) heterostructure was also an effective strategy for increasing the intimate interface contact area between CoP and BP, which effectively reduced the carrier transfer distance. 106 Xiang et al 107 fabricated 2D-2D CoP/ZIS nanohybrids via an electrostatic self-assembly method. The unique 2D structure of CoP and ZIS made the large intimate contact interface between them, effectively reducing the diffusion distance of carriers. In addition, CoP QDs were also employed as cocatalysts loaded on the surface of g-C 3 N 4 via a simple pyrolysis-based method and electroless plating. 100 This work provided a simple and feasible method of synthesizing metal phosphide QDs.
The intimate contact between each component was crucially important for the timely transfer of carriers. Based on this, Co 2 P-modified CdS was fabricated by Li et al 108 via an in situ hydrothermal strategy. In the integrated system, the Co 2 P nanoparticles were homogeneously dispersed on the surface of CdS with unique intimate contact. The characterization results confirmed that great bonding occurred between the interface of Co 2 P and CdS during the in situ hydrothermal process, which was beneficial to the improvement in photocatalytic H 2 evolution. Dong et al 109 loaded Co x P onto CdS nanorods by a photochemical strategy with Co salt and NaH 2 PO 2 as the Co and P sources, respectively. The whole process was completed within an hour, which was safe, rapid, and energy savings. The optimized H 2 production rate of Co x P/CdS was approximately 500 mmol h −1 g −1 , which was attributed to the fact that the Co x P cocatalyst could effectively prevent the recombination of electrons and holes. This study supplied a fast and feasible strategy for fabricating metal phosphides. In addition, photochemical synthesis strategies have also been applied for the synthesis of amorphous Co-P alloymodified ZnIn 2 S 4 , and the test results showed that ZnIn 2 S 4 presented improved photocatalytic activity after modification with an amorphous Co-P alloy, which was attributed to the elevated photocurrent density, reduced impedance, weakened fluorescence intensity and extended carrier life. 110 This work demonstrated that an amorphous Co-P alloy was a robust and efficient cocatalyst in the photocatalytic field.
MOFs are often employed as precursors for the synthesis of metal phosphides due to their excellent physical and chemical properties. For example, CoP/g-C 3 N 4 composites were constructed by Sun et al 102 via a two-step calcination procedure. The preparation process is shown in Figure 19. ZIF-67 and g-C 3 N 4 were prepared via simple precipitation and thermal condensation processes, respectively, in which the precursor mixed with NaH 2 PO 2 was heated to 300 C. The introduction of CoP could obviously boost the photocatalytic activity of g-C 3 N 4 due to the enhanced light absorption ability, accelerated carrier migration, and separation rates. Zhang et al 111 constructed CoP/CeVO 4 nanohybrids through a simple onestep chemical precipitation method, in which CeVO 4 particles were firmly attached to the surface of CoP particles derived from ZIF-9 to form a "small point" to "large point" heterojunction. Mechanistic studies showed that the Schottky barrier led to the bending of the bands of CoP and CeVO 4 and the formation of an internal electric field created by the heterojunction between CoP and CeVO 4 . Hence, the synergistic effect between CoP and CeVO 4 promoted the improvement in photocatalytic activity. This study presented some guidelines for modulating the electronic structure and carrier behavior of transition metal phosphide-based photocatalysts.
In practice, the recombination of electrons and holes was still the main factor hindering the improvement in photocatalytic H 2 production activity. It is vital to increase the H 2 evolution rate to improve the timely transfer and separation of photogenerated carriers in space. Based on this, Lin et al 112 developed dual cocatalyst-modified photocatalysts by selectively loading Au nanoparticles and CoP nanosheets onto the inside and outside surfaces of a three-dimensionally ordered macroporous (3DOM) g-C 3 N 4 framework. In the CoP/3DOM g-C 3 N 4 /Au sample, the spatially separated dual cocatalyst could provide two paths for electron migration in opposite directions to dramatically boost the rate of electron migration. As shown in Figure 20, one was electron movement from g-C 3 N 4 to CoP nanosheets on the outside surface of g-C 3 N 4 , and the other was electron transfer from g-C 3 N 4 to Au nanoparticles on the inside surface of g-C 3 N 4 . Meanwhile, the holes were consumed by scavengers, which could achieve fast carrier separation at different reactive sites, resulting in an obvious improvement in photocatalytic H 2 evolution.
A p-n-n tandem heterostructure was constructed by combining n-type CdS with n-type WS 2 nanosheets and p-type CoP nanoparticles, which acted as a bifunctional photocatalyst for simultaneously degrading pollutants and generating H 2 . 113 Reddy et al 114 designed multicomponent hierarchical dandelion-flower-like CdS/RGO-MoS 2 @CoP composites with a maximum H 2 generation rate of 83 907 μmol h −1 g −1 . In this composite, the covered RGO nanosheets acted as good electron collectors and transporters. At the same time, the CoP and MoS 2 nanostructures worked as electron acceptors and cocatalysts, respectively, achieving the effective separation of charge carriers.

| Iron phosphide
Iron phosphide (FeP) as a cocatalyst has become increasingly popular in photocatalytic reactions since iron is the most abundant transition metal. In addition, the vacant 3d orbital and/or 3p lone pair electrons of P atoms regulate the surface charge state of Fe atoms, facilitating electron transfer to the surface of FeP nanoparticles. 115 Zhao et al 116 designed ultrasmall 0D FeP nanodots anchored on layered 2D g-C 3 N 4 nanosheets by the low-temperature phosphidation method, and the resulting composites had well-defined nanostructures and an intimate 0D/2D interface. The FeP nanocrystals in the composites played a cocatalyst role, accelerating the transfer and separation of charge carriers. The presence of FeP improved the visible light response capability and accelerated the charge carrier transfer dynamics of g-C 3 N 4 .
The particle size and dispersibility of cocatalysts are also one of the main factors affecting the catalytic activity; the smaller the particle size is, the higher the dispersity, the more active sites exposed, and the better the catalytic activity. Sub-5 nm ultrafine FeP nanodotmodified porous graphitic carbon nitride (g-C 3 N 4 ) F I G U R E 1 9 Schematic diagram of the fabrication of CoP/g-C 3 N 4 . Reproduced with permission: Copyright 2018, The Royal Society of Chemistry 102 F I G U R E 2 0 Photocatalytic H 2 production process over CoP/3DOM g-C 3 N 4 /Au. Reproduced with permission: Copyright 2018, Elsevier 112 composites were fabricated by Zeng et al 117 via the gasphase phosphorization of a Fe 3 O 4 /g-C 3 N 4 precursor in an argon atmosphere. The introduction of sub-5 nm FeP nanodots with uniform sizes and a high dispersity boosted the transfer rate of electrons and acted as reactive sites for reducing H + into H 2 . In addition, the DFT calculations showed that the FeP/g-C 3 N 4 hybrids exhibited a moderate adsorption-desorption capacity (ΔG H* = −0.09 eV), as displayed in Figure 21, which was similar to that of the noble metal Pt, suggesting superior HER kinetics.
Sun et al 118 fabricated FeP nanoparticle-coupled CdS nanosheets via an in situ phosphorization method. In the synthesized heterojunction photocatalysts, 0D FeP NPs, were tightly attached to the surfaces of 2D CdS nanosheets, achieving efficient photocatalytic H 2 evolution. The characterization results demonstrated that the fabrication of 0D/2D heterojunctions with close contact between FeP and CdS could boost the transfer and separation of carriers, promoting the improvement in photocatalytic performance. FeP nanoparticle-modified CdS nanocrystal photocatalysts were applied for photocatalytic H 2 generation and achieved a maximum H 2 evolution of 202 000 μmol h −1 g −1 , which is 3-fold higher than that of Pt/CdS. In addition, the system presented excellent stability for more than 100 hours. 119 In addition to binary FeP composites, Qi et al 120 designed a g-C 3 N 4 /Fe 2 O 3 @FeP hybrid material by annealing and phosphidation of g-C 3 N 4 /Fe MOFs. During the preparation process, as shown in Figure 22, heterogeneous structure construction and cocatalyst loading were realized with compact contact between each component. The optimized g-C 3 N 4 /Fe 2 O 3 @FeP sample presented a maximum H 2 evolution rate of 12.03 mmol h −1 g −1 in the presence of EY, which was ascribed to the following reasons: (a) improved light absorption capacity; (b) efficient separation and transfer of electrons and holes due to the formation of a type-II heterojunction between Fe 2 O 3 and g-C 3 N 4 ; and (c) increased numbers of H 2 production sites provided by FeP.

| Copper phosphide
Copper phosphide (Cu 3 P), a p-type semiconductor with a bandgap of 1.3-1.4 eV, has been applied to lithium-ion batteries, HER and photocatalytic H 2 evolution by water splitting due to its abundant reserves. In particular, Cu 3 P functioned as a p-type semiconductor coupled with an n-type semiconductor, forming p-n heterojunction photocatalysts. For example, Yue et al 121 developed an earth-abundant Cu 3 Pmodified TiO 2 "P-N" heterojunction by mechanical mixing and calcining processes. An optimized H 2 generation performance was obtained over Cu 3 P/TiO 2 (7940 μmol h −1 g −1 ), which was attributed to the fast separation and transfer of charge carriers due to the presence of a built-in electric field created by the p-n heterojunction. In addition, Shen et al 122 constructed Cu 3 P nanoparticle-decorated g-C 3 N 4 nanosheets. The study found that Cu 3 P functioned as a cocatalyst at a low content (1.5 wt%) and served as a p-type semiconductor when its content reached 20 wt%, as shown in Figure 23. Comparatively speaking, the role of the p-type semiconductor for Cu 3 P in increasing the photocatalytic H 2 production rate in Cu 3 P/g-C 3 N 4 composites was more obvious than that of the cocatalyst for Cu 3 P. A maximum H 2 evolution rate of 159.41 μmol h −1 g −1 was achieved over Cu 3 P/ g-C 3 N 4 , which was 1014-fold higher than that of g-C 3 N 4 .
Hua et al 123 fabricated a p-type Cu 3 P/n-type g-C 3 N 4 heterojunction via phosphorization of a CuCl(OH) 3 /g-C 3 N 4 precursor. In this system, Cu 3 P performed the following two functions: accelerating the separation of carriers and lowering the hydrogen evolution overpotential. In addition, control experiments confirmed that the electrons accumulated on Cu 3 P nanoparticles and that holes were enriched on g-C 3 N 4 nanosheets, which demonstrated that the carrier transfer path was Z scheme. Rauf et al 124 designed Bi 2 WO 6 -Cu 3 P Z-scheme composites through a simple ball-milling complexation strategy, producing a mediator-and cocatalyst-free photocatalyst system. The study demonstrated the significant role of interfacial solid-solid contact and the well-matched energy level positions between Bi 2 WO 6 and Cu 3 P for solar-water splitting.

| Other metal phosphides
In addition to the metal phosphides mentioned above, other metal phosphides have been developed for F I G U R E 2 1 The free energy of the HER at the equilibrium potential. Reproduced with permission: Copyright 2019, American Chemical Society 117 photocatalytic H 2 evolution. Nanosized MoP was fabricated via a phosphorylation process under an ambient air atmosphere. The optimal MoP/g-C 3 N 4 photocatalyst possessed an H 2 evolution rate of 3868 μmol h −1 g −1 . In this system, the introduction of MoP into g-C 3 N 4 could broaden the absorption range of visible light and build a conducive highway (Mo (δ + )-N (δ − ) bond) for electron transfer from g-C 3 N 4 to MoP, inducing an improvement in the H 2 production rate. 125 In addition, Chen et al 126 anchored ruthenium on the CN framework and then further phosphorized it, acquiring the target product Rh-P/ g-C 3 N 4 . Structural analysis showed that Rh-P was atomically dispersed on the framework of CN, and theoretical calculations confirmed that the single Rh-P site on g-C 3 N 4 provided convergence centers for photogenerated electrons and reduced the H 2 evolution potential.
In addition, Zhang et al 127 synthesized tungsten phosphide (WP) nanoparticles through the traditional temperature programming reduction method, and then, they were loaded on the surface of CdS. After the incorporation of WP at a content of 4.0 wt%, the H 2 generation rate of CdS reached a maximum of 155.2 μmol h −1 , which was 11.67 times higher than that of CdS. Jin et al 128 designed EY-sensitized EY-UiO-66(Zr)/WP photocatalysts via the ultrasound-assisted impregnation method. In the integrated system, 3D UiO-66 with a large specific surface area, porosity, and adsorption capacity was favorable for the loading of WP nanoparticles and the adsorption of a large amount of EY molecules. EY dye molecules acting as photosensitizers extended the absorption range of visible light. WP nanoparticles acting

| Bimetallic phosphides
It was reported that the high catalytic activity of metal phosphides is mainly attributed to the unbalanced surface charge distribution derived from partially positively charged Ni (Ni δ+ ) and partially negatively charged P (P δ− ) due to the electronegativity difference between Ni and P. 129 Bimetallic phosphides possess a more unbalanced surface charge distribution than single metal phosphides. 130 For example, bimetallic phosphide NiCoPmodified g-C 3 N 4 exhibited an obviously improved H 2 evolution rate compared with Ni 2 P/g-C 3 N 4 and Co 2 P/g-C 3 N 4 , which was mainly attributed to the synergistic effect of the Schottky barrier and the lower overpotential relative to that of the Ni 2 P or Co 2 P counterparts. 131 Ternary metal phosphide Ni x Co 1−x P-decorated Zn 0.5 Cd 0.5 S nanorod photocatalysts were constructed by Li et al 132 through an in situ phosphating method, in which Ni x Co 1 −x P nanoparticles were homogeneously dispersed on the surface of Zn 0.5 Cd 0.5 S nanorods. DFT calculations confirmed that the proposed photocatalytic mechanism agreed well with the experimental results. The formation of a coordination bond between Zn 0.5 Cd 0.5 S and Ni 0.1 Co 0.9 P at their contact interface played a vital role in the separation and transfer of charges during the photocatalytic process.
In addition, bifunctional NiCoP-modified g-C 3 N 4 photocatalysts were successfully fabricated by Qin et al. 133 In the integrated system, a crystalline NiCoP core was surrounded by an amorphous nickel cobalt phosphate (NiCo-Pi) shell. The mechanistic study suggested that the NiCoP core acted as a reductive reaction site and that the NiCo-Pi shell behaved as an oxidative reaction site for separate photocatalytic H 2 and O 2 generation, as presented in Figure 24. It was demonstrated that the NiCoP@NiCo-Pi core/shell cocatalyst accelerated the F I G U R E 2 5 A, P 2p XPS spectra of CoP/Zn 0.5 Cd 0.5 S. B, The mechanism of photocatalytic H 2 production over Co-Pi/CdS. Reproduced with permission: Copyright 2017, Elsevier. 136  transfer and separation of carriers and promoted the surface reaction. Shen et al 134 constructed carbon black (CB)-and Co 1.4 Ni 0.6 P-comodified graphitic CN via sonochemical loading and high-temperature phosphatizing. In the integrated system, the synergetic effect between the Schottky heterojunctions and Co 1.4 Ni 0.6 P could improve the separation efficiency of photogenerated electron-hole pairs, inducing the improvement in activity and stability. Fe-Ni-P nanotubes derived from Fe-Ni-MIL-88 nanorods presented superior photocatalytic H 2 and O 2 evolution activity in the presence of different dyes. 135 Moreover, the catalytic activities of such dye-sensitized systems could be regulated by changing the molar ratio of Fe and Ni in Fe-Ni-P nanotubes.

| DISCUSSION
Generally, metal phosphides reported in most literature are covered by metal phosphates since metal phosphides are easily oxidized when exposed to air. CoP cocatalystdecorated Zn 0.5 Cd 0.5 S exhibits a 20-fold higher H 2 production rate than the primary Zn 0.5 Cd 0.5 S catalyst. Although the phosphide in topic is in the form of CoP, the XPS results demonstrated that the peak area of oxidized P species was significantly larger than that of metal phosphides due to the superficial oxidation of CoP, as shown in Figure 25A. 136 The peak of oxidized P at 133.8 eV was in accordance with that of phosphate (133.7 eV) in the Co-Pi cocatalysts, further confirming the presence of Co-Pi on the surface of Zn 0.5 Cd 0.5 S. 137 A similar phenomenon was observed in other studies, 105 which suggested that most superficial metal phosphides were oxidized into metal phosphate. However, there is little explanation of the action of phosphates in the article, so it is confusing which part plays a major role in the improvement in H 2 production activity. However, Di et al 137 considered that Co-Pi deposited on the surface of CdS acted as an oxidation cocatalyst to capture the holes derived from CdS, inducing the partial oxidation of Co 2+ to Co 3+ due to the more positive valence band potential of CdS than the oxidation potential of Co 2+ /Co 3+ , and then, Co 3+ was reduced into Co 2+ by lactic acid with a perfect cycle ( Figure 25B). Meanwhile, the electrons at the CB of CdS participated in the reduction reaction to reduce H + into H 2 . The function of metal phosphates covering the surface of metal phosphides was indistinct, and the true source of the improvement in catalytic activity was not clear. Similar reports have been provided by other researchers. 138 Hence, it is urgent to explore the main reasons for the improvement in photocatalytic activity.
2D-2D CdS-CoP photocatalysts were successfully fabricated. In the integrated system, CoP acting as a reducing cocatalyst collected electrons from CdS and reduced H + into H 2 . 139 In addition, core-shell amorphous cobalt phosphide (CoP x ) with integrated cadmium sulfide nanorods was synthesized by a simple solvothermal method and exhibited exceptional photocatalytic H 2 production under visible light. This work showed that amorphous CoP x served as an important substitute for precious Pt in photocatalytic H 2 production. Nevertheless, the XPS results confirmed that phosphate species were not the active sites and could be dissolved during the process of photocatalytic H 2 evolution. 140 However, Pan et al 141 considered that CoP and Pt nanoparticles on the surface of g-C 3 N 4 nanosheets served as proton reduction sites and O 2 evolution sites, respectively, which is contradictory to the above discussion and other reports. 136 No conclusive conclusions have been reached regarding the role of phosphates in photocatalysis. Hence, it is absolutely imperative to explore the role of metal phosphides in photocatalysis and electrocatalysis.

| CONCLUSION AND OUTLOOK
In summary, metal phosphides can lower the H 2 evolution overpotential and accelerate the transfer and separation of charge carriers in the photocatalytic process; hence, they served as cocatalysts and electrocatalysts for improving the photocatalytic H 2 generation performance and electrocatalytic H 2 evolution reaction, respectively. This review summarizes the promotion of metal phosphides in electrocatalytic and photocatalytic H 2 evolution. As presented in Tables 2 and 3, different kinds of transition metal phosphides have been explored for electrocatalytic and photocatalytic H 2 production by the decomposition of water. Although some progress has been made, the actual rate of H 2 production is still too low for industrial use. The ultimate goal of these efforts is to seek stable, efficient, and low-cost catalysts for commercial electrocatalytic and photocatalytic H 2 evolution from water splitting. At present, metal phosphides applied to electrocatalytic and photocatalytic H 2 evolution by water splitting have attracted wide attention and have good development prospects in dealing with energy shortages and environmental pollution due to their low cost, abundance, and low H 2 evolution potential. However, they still fall far short for the largescale application required for industrialization. A stable and efficient catalyst is one of the key factors necessary to realize the conversion of solar energy into hydrogen energy. [158][159][160] In recent years, numerous metal phosphides have been employed as effective catalysts displaying stable and highly efficient H 2 evolution activity, and some of them are even superior to traditional noble metals, such as Pt and Pd. However, metal phosphides also have the following problems.
1. Metal phosphides supported on the surface of photocatalysts are more easily oxidized when exposed to air.
Most of the metal phosphides reported in the literature were covered by some metal phosphates due to the oxidation of partial metal phosphides. The role of metallic phosphates during the photocatalytic process is unclear and unexplored. 2. It has been reported that the morphology and microstructure of catalysts have a great influence on the catalytic activity. Few studies have focused on the effect of morphology and the microstructure of phosphidedecorated photocatalysts on H 2 production activity. 3. There is a lack of research on the root cause of the improvement in the H 2 production rate of phosphidedecorated photocatalysts. The mechanism of improving the photocatalytic H 2 production performance of phosphate-based photocatalysts is still unclear. The function of metal phosphides and metal phosphates is not definite.
Based on the problems mentioned above, metal phosphide photocatalysts should be modified to further improve their solar-to-hydrogen conversion efficiency. In future studies, the design, preparation, and modification of metal phosphides can be carried out based on the following aspects: 1. New methods and synthetic strategies should be explored to fabricate phosphide-modified photocatalytic systems to prevent the oxidation of the metal phosphide surface in the air. In addition, the role of metal phosphates in the photocatalytic process should also be explored. 2. Nanostructure design and the optimization of metal phosphides should be achieved to improve their activity and stability in photocatalytic H 2 production. The effect of the apparent morphology and microstructure of the catalyst on the rate of H 2 production should be systemically studied. 3. In situ analytical techniques, such as XPS, scanning probe microscopy, Raman spectroscopy, and the corresponding theoretical calculations of metal phosphides, should be implemented to thoroughly understand the mechanism of the enhancement in photocatalytic activity. 4. Due to their distinctive properties, metal phosphides exhibit superior photocatalytic activity. More theoretical calculations should be conducted for metal phosphide materials to further understand the inherent mechanism to boost photocatalytic water splitting.
Science Foundation of China (Nos. U1663228 and 51972165), and the Priority Academic Program Development of Jiangsu Higher Education Institutions.