PEM water electrolysis for hydrogen production: fundamentals, advances, and prospects

Hydrogen, as a clean energy carrier, is of great potential to be an alternative fuel in the future. Proton exchange membrane (PEM) water electrolysis is hailed as the most desired technology for high purity hydrogen production and self-consistent with volatility of renewable energies, has ignited much attention in the past decades based on the high current density, greater energy efficiency, small mass-volume characteristic, easy handling and maintenance. To date, substantial efforts have been devoted to the development of advanced electrocatalysts to improve electrolytic efficiency and reduce the cost of PEM electrolyser. In this review, we firstly compare the alkaline water electrolysis (AWE), solid oxide electrolysis (SOE), and PEM water electrolysis and highlight the advantages of PEM water electrolysis. Furthermore, we summarize the recent progress in PEM water electrolysis including hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) electrocatalysts in the acidic electrolyte. We also introduce other PEM cell components (including membrane electrode assembly, current collector, and bipolar plate). Finally, the current challenges and an outlook for the future development of PEM water electrolysis technology for application in future hydrogen production are provided.


Introduction
Carbon neutrality means having net-zero carbon dioxide emissions. Current issues on climate change and global warming have been an ongoing concern for the world, environmental issues will strongly affect us if we do not start taking them seriously [1][2][3]. In order to reduce carbon emissions and achieve carbon neutrality, it is an urgency to develop clean and sustainable fuels from renewable energy resources such as wind and solar power, geothermal, and biomass energy to replace traditional fossil fuels. In this respect, hydrogen fuel, with high gravimetric energy density (140 MJ kg −1 ) and only generates water as a byproduct without any environmental concern, has been hailed as the unrivaled clean energy carrier to reduce conventional fossil fuels consumption and relieve aggravating environmental problems [4,5]. Currently, hydrogen is mainly produced through steam reforming of fossil fuels under harsh conditions, which are restrained by the complex operation and emission of carbon dioxide. Electrochemical water splitting is an effective and clean method to produce high-purity hydrogen by using renewable energy, which has ignited new interests in the past decades. During the water electrolysis process, hydrogen evolution reaction (HER) occurs at the cathode and oxygen evolution reaction (OER) operates at the anode [6][7][8]. Theoretically, it only needs the voltage of 1.23 V to drive the water electrolysers to obtain hydrogen production. However, it usually needs a much larger voltage of 1.8-2.0 V in the practical electrolysis process, due to the overpotentials on both cathode and anode. Therefore, various kinds of electrocatalysts have been developed to reduce the overpotentials for both HER and OER [9][10][11].
Electrochemical water electrolysis can be classified into three main types based on the type of electrolytes and operating conditions: alkaline water electrolysis (AWE), solid oxide electrolysis (SOE), and proton exchange membrane (PEM) water electrolysis, as shown in Fig. 1. AWE operates at the high concentrations of alkaline solutions and employs diaphragm (such as asbestos, polyethersulfone (PES), microporous polymer membrane, glass reinforced polyphenylene sulfide (PPS) compounds et al.) to separate the produced gases from the cathode and anode, which suffers from low operating pressure, sluggish anion transport kinetics and limited current density. SOE operates at high pressure and high temperature of 500-850°C, but is plagued by unsatisfactory catalyst durability and membrane degradation. PEM electrolysis is considered a splendid method for high-purity hydrogen production in the future industrial application, due to its high current density, greater energy efficiency, smaller gas crossover, wider operating temperatures (20-80°C) [12], smaller mass-volume characteristic, more importantly, the specialty of adaptive to renewable energy volatility. In the past decades, significant progress has been made in PEM electrolysis technology. For example, a high cell electrolysis efficiency of 94.4% and a very low cell voltage of 1.567 V were obtained by Marshall [13] to reach the current density of 1 A cm −2 . A stable voltage for over 5000 h during 2 A cm −2 and a peak operating current density of 20 A cm −2 has been acquired [14]. Despite the many advantages and bright prospects of PEM electrolysers, the practical industrial application of this technology is plagued by the prohibitively exorbitant cost and extreme degradation of electrocatalysts under an acidic environment, especially for OER catalysts that simultaneously experience the acidic and strongly oxidative environments. To address these problems, great research efforts have been directed towards reducing the usage of noble metals and exploiting noble-metal-free catalysts, improving the intrinsic catalytic performance and the durability of catalysts in the acidic electrolyte. To date, the materials for acidic HER are mainly Pt-based material, transition metal dichalcogenides (TMDs), and transition metal phosphides (TMPs). The most popular acidic OER electrocatalysts are focused on precious Ir and Ru-based materials, perovskites, and nonprecious metal oxides. At present, the electrocatalytic performance and stability for both HER and OER have made significant progress, stimulating the development of PEM water electrolysis.
Here, we provide recent progress of the PEM water electrolysis for hydrogen production. In this review, we first compare the distinction between AEW, SOE, and PEM to emphasize the advantages of PEM water electrolysers in producing hydrogen. Then, we focus on the recent development of acidic HER electrocatalysts and OER electrocatalysts for PEM water electrolysis. Then, we introduce other key materials for PEM water electrolysis, such as membrane electrode assembly (MEA), current collector, and bipolar plate. Finally, we propose the current challenges and perspectives on the development of PEM water electrolysis technology for application in future hydrogen production.

Alkaline water electrolysis (AWE)
AWE is a well-established technology for hydrogen production and this phenomenon was first found by Troostwijk and Diemann in 1789 [12]. AWE usually operates at lower temperatures between 30-90°C and uses alkaline KOH/NaOH aqueous solution with a typical concentration of 20-30% as the electrolyte. In alkaline electrolysers, the diaphragm and porous nickel-based materials are usually used as asbestos and electrode. Alkaline electrolysis is a mature process, which only needs low investment and operation costs. However, the AWE usually exhibits a low operating current density of 200-400 mA cm −2 and an energy efficiency of approximately 62-82% [15]. Moreover, the pressure between the anode and cathode sides needs to keep balanced, avoiding the hydrogen or oxygen generated by the electrolysis penetrating the diaphragm to the other side and resulting in the explosion risk.

Solid oxide electrolysis (SOE)
SOE has received an enormous amount of attention because of its high efficiency and environmental friendliness, the needed electrical power could be provided by sustainable energy sources [16]. However, SOE works at high temperatures (500-850°C) and pressure, water is in the form of steam during the SOE process. The most commonly used electrolyte material is yttria-stabilized zirconia, the conventional high-temperature oxygen-ion conductor. However, the high temperatures result in the fast deterioration of catalytic performance, and achieving the long-term operation of SOE is a challenge. Meanwhile, the produced hydrogen can form a mixture with steam and needs an additional procedure to acquire high purity hydrogen.

Proton exchange membrane water electrolysis (PEM)
PEM electrolysis has been known for over sixty years and was developed by General Electric, where a solid sulfonated polystyrene membrane is used as the electrolyte. This technology is also named polymer electrolyte membrane electrolysis. The proton exchange membrane acts as both the gas separator and the electrolyte. In the electrolysis process, only deionized water is injected into the cell without any electrolytic additives. There are a variety of advantages of PEM electrolysis, such as high current density, greater energy efficiency, low gas permeability, wider operating temperatures (20-80°C) [12], easy handling and maintenance. However, the popular electrocatalysts for PEM electrolysis are precious metals (Pt, Ir, Ru), the exorbitant cost of these precious metal makes PEM electrolysis a relatively higher investment than alkaline electrolysis. In this regard, researchers have paid tremendous efforts for reducing the cost and improving the intrinsic activity of PEM electrocatalysts. As shown in Table 1, listing the typical characteristics of three water electrolysis technologies.

Electrocatalysts for hydrogen evolution reaction (HER)
The predicament for HER in PEM electrolysis is the development of high-activity and stability catalysts for the cathode. Presently, platinum (Pt) and Pt-based materials are recognized as the state-of-the-art electrocatalysts for the cathode side in PEM electrolysers [17,18], which exhibit decent electrochemical activity for the HER in acidic electrolytes. However, the large-scale  [19][20][21][22][23][24][25][26][27]. Lim and co-workers [28] prepared a Pt@N-containing carbon core-shell catalyst on carbon nanofibers (Pt@CS/CNF), which exhibited excellent catalytic activity and durability. They found that the graphitic N and pyridinic N on the carbon shell could not only offer additional catalytic active sites for HER, but also impede the active Pt core from dissolution and agglomeration, improving the stability (Fig. 2a-d). Zhong and co-workers [21] adopted an ultralow-temperature solution method to prepare a single atom Pt anchored on carbon nanotubes (SA-Pt-CNTs), which reduced Pt usage and improved the intrinsic HER activity in acidic media, a low overpotential of only 41 mV to achieve the current density of 10 mA cm −2 . The aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) demonstrates Pt atoms were atomically dispersed. Density functional theory (DFT) demonstrated the high electrocatalytic activity is attributed to the interaction between CNTs and Pt atoms, which leads to the downshift of the d-band center and facilitates the H desorption ( Fig. 2e-g). Xu et al. [22] rationally designed atomically dispersed Pt anchored on amorphous MoO x (Pt-SA/α-MoO x ), which unfolded a low overpotential of 19 mV at 10 mA cm −2 and excellent mass activity of 52.0 A mg Pt −1 at the overpotential of 50 mV in 0.5 M H 2 SO 4 . Additionally, it also exhibited a low Tafel slope of 123 mV dec −1 and long-time durability. DFT results revealed that the remarkable performance is attributed to the synergistic effect between unsaturated Pt atoms and amorphous MoO x , which are beneficial for atomic hydrogen adsorption. Xia and co-workers [29] reported a series of carbon-supported ultrasmall highentropy alloy nanoparticles (us-HEA NPs) through a general and ultrafast chemical co-reduction method. In 0.5 M H 2 SO 4 electrolyte, this us-HEA NPs exhibited an eminent mass activity of 28.3 A mg −1 at -0.05 V, which shows a promising prospect for HER in PEM electrolysis.
In recent years, various transition metal dichalcogenides have been developed as acidic HER electrocatalysts [30], such as MoS 2 [31][32][33][34][35], WS 2 [36], MoSe 2 [37][38][39], and WSe 2 [40], which have shown to be splendid cathodic materials for HER. Nørskov and co-workers [25] investigated the HER performance of MoS 2 and fabricated an MEA based on a Nafion proton exchange membrane, which showed a lower current density of 10 mA cm −2 at the overpotential of 175 mV. Kim et al. [41] reported a simple one-step electrodeposition method to prepare amorphous molybdenum sulfide (MoS x ) on carbon paper and used as the gas-diffusion layer for PEM water electrolysis, which displayed acceptable cell performance. Because sulfur atoms on the catalyst surface with different binding energy values, they could extremely affect the  [28]. e A scheme showing the synthesis of Pt-CNTs through ultralow-temperature solution reduction. f HAADF-STEM image of Pt-CNTs. g Calculated ∆G H* on different sites for HER [21] intrinsic activity of MoS x and facilitate the HER performance. The 1T phase of TMDs possesses high electrical conductivity and a larger number of active sites compared to the 2H phase [42][43][44], which could result in a higher charger transfer rate in the electrochemical process. Recently, Kang and co-workers [45] successfully synthesized Mo 1-x V x Se 2 alloy nanosheets through a colloidal reaction, which could lead to the phase transition from a semiconducting 2H phase MoSe 2 to the metallic 1T phase VSe 2 (Fig. 3). DFT calculations suggested the decent electrocatalytic activity attributed to the V and Se vacancies, which are beneficial for the phase transition from the 2H phase to the 1T phase and forming the intermediates in HER. Additionally, various cation-doping is also an effective strategy to tune the surface electronic structure of catalysts and optimize the hydrogen adsorption free energy, improving the electrocatalytic HER activity [46,47]. Morozan et al. [24] adopted microwave irradiation or heat treatment in a furnace under inert and reductive atmospheres to prepare the bimetallic FeMoS catalyst, which presented a low overpotential of 140 mV at the current density of 10 mA cm −2 . Combining with  [45] carbon nanotubes (CNTs), the composite catalyst could operate at 500 mA cm −2 using a voltage of less than 300 mV compared to using Pt as the cathode catalyst in PEM electrolyser. Zhao et al. [48] prepared an ultrastable Pd-doped MoS 2 material and supported on a vertical graphene network (VGN@Pd-MoS 2 ), which exhibited a stable operation over 1000 h at 10 mA cm −2 and provide a prospect to replace Pt for PEM electrolysis.
Apart from Pt and TMDs, researchers also devoted to developing various alternative electrocatalysts for the hydrogen evolution reaction in acidic electrolytes [26,27,[49][50][51]. Li and co-workers [26] proposed a universal strategy to prepare a series of Ni-activated transition metal carbides (TMCs) on Ni foam coated with graphene (Ni-GF) and further optimize the HER activity of these TMCs. After the introduction of Ni atoms, these Ni-GF/ TMCs exhibited much reduced overpotentials and intriguing durability toward HER, as displayed in Fig. 4ac. They found that the superior HER performance hinges on the unique structural characteristics and optimized electronic synergy of Ni-activated TMCs. The doping of Ni atoms on the surface of TMCs could effectively regulate the d-electron structure, which can not only significantly enhance the exposed active sites, but also improve the intrinsic catalytic activity (Fig. 4d, e). Additionally, the high specific surface areas and unique 3D skeleton structure could facilitate a fast electron transfer rate and efficient release of hydrogen bubbles. Sapountzi et al. [50] prepared highly crystalline FeP NPs with predominantly exposed [010] facets on a conductive carbon support and presented high activity in HER. They also assembled a full-cell PEM water electrolysis by using FeP as cathode and a commercial IrRuO x as anode, which could reach a current density of 0.2 A cm −2 at a voltage of 2.06 V and operate steadily for up to 36 h continuously. Ito and coworkers [27] further improved the acid stablility through graphene-covering approach and investigated the effect of graphene-covering layers on HER activities of Ni and Cu. They found that the graphene could optimize the HER activity and impede the corrosion, and the ability of protons to penetrate graphene manipulates the HER performance of these graphene-covered non-noble metals.

Electrocatalysts for oxygen evolution reaction (OER)
Because of the pivotal role of OER in water electrolysis, the energy community has put substantial efforts into exploiting admirable OER catalysts. Due to the acidic and strongly oxidative environments, developing high-active and robust electrocatalysts for OER in PEM electrolysers is an arduous challenge. To date, Ir-based and Ru-based catalysts are the most popular OER electrocatalysts in acid. Ru-based catalysts exhibit super OER activity but unsatisfactory stability compared with Ir-based catalysts. In this section, we systematically summarize the reports about Ir-based and Ru-based catalysts in the past decades ( Table 2).  [26] Ir metal catalysts Ir metal is a vital electrocatalyst for OER under an acidic environment and has gained explosive interest because of its low overpotentials and unique stability. The catalytic performance is largely credited to the morphology, particle size, crystal structure, and substrate materials. As a result, various Ir-metal materials have been explored to optimize the OER activity and reduce the amount of precious metal.
Decreasing particle size or preparing ultrathin structure is an effective method to maximize surface area and expose more active sites [80], beneficial for improving the catalytic activity [81,82]. Ultrafine monodisperse Ir NPs with narrow size distribution and high dispersion were successfully synthesized through a colloidal method [52]. Due to the ultrafine particle size and uniform dispersion, this catalyst exhibited startling catalytic performance for acidic OER and exceeded commercial IrO 2 catalyst. Carbon nanobowls supported ultrafine Ir nanocrystal (Ir@ HEDP/CNBs) catalyst was prepared by Chen and coworkers [53], Fig. 5a-c show the ultrafine size and homogeneous distribution of Ir nanocrystal, as well as the high surface area and conductivity of CNBs, rendering it low OER overpotential of 290 mV at 10 mA cm −2 and excellent stability. Single atom catalysts (SACs) can realize the maximum atom utilization efficiency and reduce the amount of precious metal [83,84]. In view of the severe activity decay of SACs in harsh acidic solutions, high-stable Ir SAC onto Fe NPs coated with nitrogen-doped carbon nanotubes (Ir-SA@Fe@NCNT) was prepared, which realized the ultralow Ir loading and superior OER performance with an overpotential of only 250 mV to reach 10 mA cm −2 in acidic solution. As shown in Fig. 5d-f, Qiao and co-workers [55] also reported a short-range ordered Ir SA integrated into Co oxide spinel structure and exhibited much higher acidic OER activity and excellent stability. A 3D Ir superstructure catalyst consisting of ultrathin Ir nanosheets to increase the accessible active  [79] sites and active surface area was reported by Huang and co-workers [56]. This 3D Ir catalyst exhibited an unexpectedly low overpotential of 240 mV and intriguing durability with ignorable voltage degradation ( Fig. 5g-i).
A robust and appropriate catalyst support is also important to improve activity and stability. Graphene, with decent electrical conductivity and chemical stability, high mechanical strength, and flexibility, has been used as a promising substrate material in diverse fields. Velasco-Vélez and co-workers [85] prepared electrochemically active Ir NPs on graphene and improved the acidic OER activity and stability. Shao and co-workers further adopted graphitic carbon nitride (g-C 3 N 4 ) and nitrogen-doped graphene (NG) as a substrate to anchor monodispersed Ir NPs (Ir/g-C 3 N 4 /NG). This g-C 3 N 4 /NG substrate could effectively prevent the aggregation of Ir NPs and improve the reaction kinetics. The Ir/g-C 3 N 4 / NG catalyst can reach the current density of 10 mA cm −2 at a low overpotential of 287 mV, much superior to the pure Ir. Montmorillonite (MMT, (Na,Ca) 0.33 (Al,Mg) 2 (S i 4 O 10 )(OH) 2 ·nH 2 O) is a natural mineral, which is easily available and cost-efficient. Boshnakova et al. [86] proved that MMT is a distinguished substrate to replace conventional carbon supports, the unique morphology and layered structure are conducive to enhancing the accessible active sites and improving the stability.
The preparation of catalysts with abundant porous structures or defects is an effective approach to expose more active surface area to the electrolyte environment and enhance the amount of the active sites. As shown in Fig. 6, Yamauchi and co-workers [58] reported an unprecedented type of 2D mesoporous Ir nanosheets through the formic acid assisted method, which served as both a reducing agent and a shape-directing agent. They found that the mesoporous structure was robust and this mesoporous Ir nanosheets also presented unrivaled electrocatalytic OER activity. This work opened a new dimension in the preparation of other porous 2D materials and further exploration of the paramount of porous structure on 2D materials for OER performance. Amorphous catalysts are often prepared due to their abundant active sites, inherent disordering, and an unsaturated coordination structure, which usually displayed more impressive catalytic performance than corresponding crystalline  [53]. d Synthetic scheme of Ir SA catalyst. e LSV curves and (f) mass activity of various catalysts [55]. g Schematic illustration of synthesis 3D Ir superstructure. h OER performance and (i) Tafel slopes of 3D Ir superstructure, Ir NPs, and Pt/C catalysts in HClO 4 solution [56] materials [78,87,88]. Wu et al. [59] prepared a series of amorphous metal catalysts through a general and facile method. They found that the amorphous Ir nanosheets depicted intriguing acidic OER activity than crystalline Ir nanosheets and benchmark IrO 2 catalyst, which only required an overpotential of 255 mV to reach the current density of 10 mA cm −2 and exhibited robust durability with ignorable activity alleviation even after 5000 cycles in 0.1 M HClO 4 solution.
Ru metal catalysts Ruthenium (Ru) has been evaluated to be one of the promising active materials for acidic OER by virtue of its appropriate adsorption strength toward oxygenated intermediate species. Additionally, Ru is abundant in natural resources and cheaper than Ir metal. Unfortunately, the reported Ru-based catalysts are unstable, which is likely due to the oxidation of lattice oxygen and generation of oxygen vacancy, leading to the over-oxidation of Ru-based materials [61,77]. To overcome this limit, great efforts have been dedicated to exploiting robust and efficient Ru-based electrocatalysts for OER in acid media [62,89,90].
Recently, Ru SACs have ignited much interest and exhibited extraordinary performance for acidic OER. It has demonstrated that the substrate and coordination environment have a significant effect on the electrocatalytic activity. As displayed in Fig. 7, a Ru SAC with the Ru 1 -N 4 configuration anchored on nitrogen-carbon support displayed an exceptionally intrinsic activity, the mass activity could reach 3571 A g metal −1 and a high turnover frequency of 3348 O 2 h −1 , as well as a low overpotential Fig. 6 a Formation mechanism of the mesoporous Ir nanosheets. b,c SEM and (d) TEM images of Ir nanosheets. e-g OER polarization curves, Ir mass activity and chronopotentiometry curves of catalysts [58] [60]. Moreover, it also revealed decent stability and could operate steadily for 30 h in 0.5 M H 2 SO 4 solution. Operando synchrotron radiation X-ray absorption fine structure (XAFS) spectroscopy and Fourier transform infrared (SR-FTIR) spectroscopy validated that adsorbed oxygen atom on Ru site to form O-Ru 1 -N 4 site under the operating condition. The oxygen adsorption could downshift the Ru 4d band and increase the average valence state of Ru, which is responsible for reducing the limiting reaction barrier. Wu and co-workers [61] constructed a series of PtCu x alloys that supported Ru SACs through sequential acid etching and electrochemical leaching. Ru 1 -Pt 3 Cu as the best catalyst in this work, only need a low overpotential of 220 mV to reach a current density of 10 mA cm −2 and displayed ultra-long lifetime than benchmark RuO 2 . From XPS and operando XAFS results, they demonstrated the valence state of Ru has no evident variation even after the stability test, preventing the over-oxidation of Ru. Further, they captured the OOH intermediate during the electrochemical process by using in-situ attenuated total reflection infrared (ATR-IR) spectroelectrochemical measurements, indicating the adsorbate evolution mechanism (AEM) instead of lattice oxygen evolution reaction (LOER) during the acidic OER reaction over Ru 1 -Pt 3 Cu. DFT calculations suggested that with the increase of Pt to Cu ratio, the compressive surface strain would lead to the upshift of the d-band center of Ru and enhancement of oxygen adsorption, resulting in the deteriorated performance.
Interface engineering is also an admirable approach to modulating the electronic environment, exposing abundant active sites, promoting electron transfer, and optimizing the electrocatalytic activity [91,92]. Based on this, Mu and co-workers [62] prepared a laminar Ru/ RuS 2 heterostructure catalyst through the synchronous reduction and sulfurization under the eutectic salt system for the first time. Theoretical calculation results suggested engineering of this heterostructure would endow it with higher electrical conductivity, charge redistribution over interfaces, and optimized adsorption strength of key intermediates, sequentially reducing the thermodynamic energy barriers.
The OER catalytic performance of Ru nanocrystals with different structures was investigated to explore the relationship between catalytic activity and crystal structures [63]. It is reported that the face-centered cubic (fcc)-Ru octahedral nanocrystals exhibit the best performance compared with fcc-Ru truncated cubes, fcc-Ru particles, and conventional hexagonal close-packed (hcp)-Ru particles. Fig. 7 a HAADF-STEM image of Ru SAC. b The Ru k-edge k 2 -weighted Fourier transform spectra for Ru foil, RuCl 3 , RuO 2 , and Ru SAC. c R-space curve-fitting of ex-situ Ru-N-C. d,e OER performance of Ru SAC, RuO 2 /C, and N-C catalysts. f The effect of oxygen adsorption on the electronic structure of Ru SAC [60] Metal alloy catalysts Although the OER performance of mono-metallic Ir or Ru has achieved some progress, the activity and stability are still far from practical application. Alloying is an outstanding approach to significantly alter the electronic structures and optimize the adsorption energy of reaction intermediates on the catalyst surface, which in turn extremely improves the OER activity. Simultaneously, the alloy strategy could noticeably reduce the usage of precious metals Ir and Ru.
A wide range of Ir-based alloy structures has been synthesized and applied in acidic OER, such as nanodendrite [93,94], thin film [95], nanowires [64], nanotube [96], and porous structure [65,97]. For example, a series of IrM (M = Ni, Co, Fe) nanoporous nanowires (IrM NPNWs) were prepared through a novel eutecticdirected self-templating method [64], the subsequent dealloying process introduced the abundant nanoporous structure. The IrM NPNWs after the introduction of Fe, Co, and Ni all have an eminent performance in 0.1 M HClO 4 , of which the Ni is the most promising element and IrNi NPNWs displays the finest catalytic performance and durability. DFT calculations confirmed that alloy formation can shift down the d-band center and alter the adsorption strength of O-related intermediates, elucidating the relationship between the d-band center and bond strength. Apart from promoting the intrinsic activity of catalysts, increasing the number of active sites is another typical strategy to optimize catalytic performance via morphology and structure control. As shown in Fig. 8a-c, a nanovoid incorporated Ir 3 Cu metallic aerogel was formed by a galvanic displacement reaction and gelation process [65]. The electrochemical active surface area (ECSA) derived from the double-layer capacitance of Ir 3 Cu is much larger than that of commercial IrO 2 and other counterparts, revealing the much more active sites of Ir 3 Cu toward OER. Additionally, the introduction of Cu weakens the binding energy of O-related species and promotes mass transfer. Downsizing the catalyst NPs diameter is also a general method to increase the specific surface area and active sites [66,67]. A scalable microwave-assisted method was used to prepare the ultrasmall (sub-10 nm) composition-tunable Rh-Ir alloy NPs [67], which displayed a significant enhancement of acidic OER performance to pure Ir NPs. Theoretical studies demonstrated that the introduction of foreign Rh atoms could regulate the electronic structure and strain, which could weaken the adsorption strength of oxygen-containing intermediates. The mass activity of Rh 22 Ir 78 NPs could reach 1.17 A mg −1 Ir at the overpotential of 300 mV and achieve a threefold increase compared with pure Ir NPs. Ru-based alloys also provided opportunities for optimizing the intrinsic OER activity and improving the stability.
For example, combinatorial synthesis and high throughput screening method were used to prepare a series of bimetallic Ru-M (M = Pd, Ir, Cu, Co, Re, Cr, Ni) alloys [98]. The Ru-Co, Ru-Cu, and Ru-Ir exhibited distinctly improved OER activities compared to pure Ru catalyst. Liu and co-workers [68] reported a simple hydrothermal method to synthesize ultrasmall IrRu nanoclusters supported on amorphous tellurium (Te) NPs (IrRu@Te), which was high conductive and acid-stable. This IrRu@ Te catalyst displayed distinguished catalytic performance and decent stability, only needing an overpotential of 220 mV to deliver 10 mA cm −2 , benefitting from the strong electronic coupling between IrRu nanoclusters and Te, which could effectively reduce the kinetic barrier for OER and prevent the dissolution of IrRu nanoclusters.
In addition to preparing bimetallic alloys, designing ternary or multi-metallic alloys is a plausible approach to altering the electronic environment and increasing the number of surface active sites [69,73,99]. A co-reduction polyol method was used to prepare transition-metal (TM) doped RuIr nanocrystals catalysts [69], the introduction of different TMs has a significant effect on the catalytic performance (Fig. 8d-h). To deliver a current density of 10 mA cm −2 , the Co-RuIr only required a low overpotential of 235 mV and achieved a long-term chronopotentiometry measurement at 10 mA cm −2 . Various multi-metallic-based hollow NPs such as nanoframes and nanocages have been designed and acted as ideal catalysts [70][71][72], owing to their highly exposed accessible active sites and readily tunable components. A cubic Pt 39 Ir 10 Pd 11 nanocage catalyst was prepared and applied as a bi-functional catalyst toward oxygen reduction reaction (ORR) and OER in acidic media [70], with an average edge length of 12.3 nm and porous walls of 1.0 nm. This catalyst exhibited an unexpected activity and durability ascribed to the highly open structure and electron coupling. Park et al. prepared a robust IrNiCu catalyst with a double-layered nanoframe (DNF) structure. The unique porous morphology endowed it with remarkable activity and stability for acidic OER. This work provided a simple synthetic method for engineering high-active catalysts with complex nanoframe structures.
Metal oxide catalysts Compared to metal catalysts, metal oxide catalysts are more active and stable under acidic and oxidizing OER conditions. Hartig-Weiss et al. [100] reported an Ir oxide catalyst supported on high surface area and high-conductivity antimony-doped tin oxide, which displayed a very compelling OER activity and reduced the usage of Ir metal, benefitting from the uniform catalyst dispersion and strong metal-support interaction (SMSI). The oxidized Ir quantum dots supported on the surface of graphdiyne (IrO x QD/GDY) are synthesized and showed incredibly acidic OER performance and robust stability [74], which only need a small overpotential of 236 mV to reach a current density of 10 mA cm −2 . In this work, GDY could effectively improve the electrical conductivity, facilitate the charge transfer, easy gas releasing and protect the Ir catalysts from corrosion. The early transition metals have a relatively high resistance to corrosion in strong acids [101], thus, the introduction of early transition metals may have a significant effect on the acidic OER performance. An early transition metal hafnium-modified IrO x (IrHf x O y ) was successfully prepared and displayed intriguing OER activity in both alkaline and acid electrolytes [75]. In order to improve the OER performance of RuO x , a lowcost rutile-structured chromium-ruthenium oxide catalyst (Cr 0.6 Ru 0.4 O 2 ) was prepared through pyrolysis of Cr-based MOF [76], which unfolded an ultra-low overpotential of 178 mV at 10 mA cm −2 , as shown in Fig. 9. Combining XANES results and DFT calculations, they found that the presence of electron transfer from Ru to Cr, and the high valence Ru has the plausible ability for the OER. The introduction of Cr could tailor the electron structure of the RuO 2 phase and has a profound effect on optimizing the durability and activity. It is well known that RuO 2 has relatively higher acidic OER activity derived from its appropriate binding ability of oxygencontained intermediates, but poor stability. Many reports  [65]. d,e HAADF-STEM images of Co-RuIr nanocrystal. f OER LSV curves of Co-RuIr, RuIr, RuO 2 , and IrO 2 in 0.1 M HClO 4 solution. g The long-term continuous overall water splitting for Co-RuIr/Co-RuIr at a current density of 10 mA cm −2 . h Schematic of catalytic mechanism on Co-RuIr catalyst in acidic media [69] have demonstrated that the unsatisfactory stability is ascribed to the over-oxidation of Ru-based materials and the formation of the soluble Ru x>4 derivatives [77,102]. In order to improve the stability, Zhang and co-workers [77] prepared a bimetallic-doped RuO x (W 0.2 Er 0.1 Ru 0.7 O 2-δ ) catalyst as shown in Fig. 10. They found that the introduction of W and Er elements could modulate the electron structure of RuO 2 and distinctly increase the formation energy of oxygen vacancy, resulting in an excellent durability of up to 500 h in 0.5 M H 2 SO 4 , as well as unrivalled OER performance. Because the amorphous/crystalline (a/c) heterostructure usually unravels unique electronic synergy and exceptionally catalytic performance, Zhang et al. reported a sodium-decorated a/c-RuO 2 electrocatalyst [78]. This a/c-RuO 2 catalyst displayed a high resistance to acid corrosion and excellent OER activities in all pH environments. Since the relatively higher activity of RuO 2 and much higher stability of IrO 2 , preparing IrRu bimetallic oxide or core/shell structure is a promising strategy to form a catalyst with high activity and stability. A highly active and stable RuIrO x nanonetcage catalyst was reported by Li and co-workers [79] through a dispersing-etching-holing method using MOF as the template. This open nano-netcage structure with abundant mesoporous could simultaneously improve the atom utilization of Ru/Ir and increase the ECSA. With the unique morphology and component, this RuIrO x catalyst displayed incredibly startling activity with a low overpotential of 233 mV for reaching 10 mA cm −2 .
Perovskite oxides have also received much attention in acidic OER due to the low usage of precious metals, compositional flexibility, and structural stability [103,104]. A pseudocubic SrCo 0.9 Ir 0.1 O 3-δ perovskite catalyst with IrO 6 octahedra was designed and prepared [105]. This SrCo 0.9 Ir 0.1 O 3-δ catalyst displayed a high turnover frequency of 2.56 ± 0.15 s −1 at 1.5 V, which was two orders of magnitude higher than that in the IrO 2 catalyst. They found that the structural reconstruction of SrCo 0.9 Ir 0.1 O 3-δ was generated during the anodic oxidation process and accompanied by the Sr and Co elements leaching. Therefore, the catalytic species are cornershared and under-coordinated IrO x octahedrons derived from SrCo 0.9 Ir 0.1 O 3-δ . The alkaline OER is catalyzed by SrRuO 3 thin film with relatively higher activity [106], but the performance is poor in acidic OER. Based on this, a Na-doped SrRuO 3 (Sr 1-x Na x RuO 3 ) was synthesized by a wet-chemistry procedure [102], which displayed both improved activity and stability in an acidic solution. The enhanced OER mechanism is demonstrated based on XANES and theoretical calculations, Na incorporation increased the valence state of Ru, weakened Ru-adsorbate bonds, and improved the dissolution potentials.

Other key materials for PEM water electrolysis
As shown in Fig. 11, expect catalysts, other key materials for PME water electrolysis are membrane electrode assembly (MEA), current collector, and bipolar plate. The photograph of the PEM water electrolysis cell assembly is presented in Fig. 11a.

Membrane electrode assembly (MEA)
The membrane electrode assemblies are made up of proton exchange membrane, ionomer solution, and catalysts.
Proton exchange membrane is also called polymer exchange membrane, which plays a critical role in separating product gases, transporting protons, and supporting catalyst layers [12]. The commercial membrane must have advantages of high mechanical strength and proton conductivity, outstanding oxidative and thermal stability. At present, these membranes are mainly focused on perfluorosulfonic acid polymer membranes, such as Nafion, Fumapem, and Flemion. However, the most commonly used are Nafion membranes (Nafion 115, 117, and 212)   [12] in PEM water electrolysers because of the advantages of high proton conductivity, high durability at higher current densities (2 A cm −2 ), but the high price and complex disposal when disused restrict its development. Research efforts have been devoted to reducing the cost of these polymer membranes and simultaneously optimizing the ion exchange characteristics and stability. In order to improve the performance of PEM and reduce the cost, the enhanced composite scheme can be used to improve the mechanical properties of PEM, which is beneficial to reduce the thickness of the membrane. Moreover, the membrane resistance and electrolytic energy can be reduced by increasing the ionic conductivity of the film, which is beneficial to improving the overall performance of the electrolyser. According to the different catalyst layer supports, membrane electrode preparation methods consist of the catalyst-coating substrate (CCS) method and catalyst-coating membrane (CCM) method. The CCS method directly coats the catalyst active components on the gas diffusion layer, while the CCM method directly coats the catalyst active components on both sides of the proton exchange membrane, which is the biggest difference between the two manufacturing processes. Compared with the CCS method, the CCM method has a higher catalyst utilization rate and greatly reduces the proton transfer resistance between the membrane and the catalytic layer. CCM is the most common method to prepare MEAs. In this process, the homogeneous catalyst slurry, which was performed by sonication of a mixture of catalysts, ionomer solution, and solvent, is directly coated on the membrane and hot-pressed at high pressure [107]. The sputtering method, spraying method, electrochemical deposition and decal method are usually used in the CCS and CCM methods.

Current collector
Current collectors are porous media placed between the MEA and bipolar plate at both electrode sides. In PEM water electrolysers, water is usually input on the anode side of the cell, where the OER takes place. The water diffuses through the channels in the separation plate and through the current collectors. The major functions are electric conduction between the electrode and the bipolar plate, as well as transport of liquid/gas from the reaction site with as low as activation, thermal, ohmic, and fluidic losses [108]. The current collector should have energetic corrosion resistance and super electrical conductivity because of the strong acidic media and strongly oxidative environments. Additionally, the pore size and structure of the current collector are important, because the produced gases in the PEM water electrolysis process should be effectively expelled and water must be reaching the catalytic sites of the electrode surface [109]. The porous titanium plates are the most used current collectors for both sides of MEA due to their decent conductivity, excellent mechanical stability, and corrosion resistance.

Bipolar plate
Bipolar plate is an essential component in the PEM electrolyser stack, reducing the cost of bipolar plates is the key to controlling the cost of electrolysers. In the harsh working environment of the anode of the PEM electrolytic cell, if the bipolar plate is corroded, it will lead to the leaching of metal ions, thereby contaminating the PEM system. Therefore, the commonly used bipolar plate protection measure is to prepare a layer of anti-corrosion coating on the surface. The bipolar plates play multifunctional roles, such as separators, conduct heat and current between single cells in a stack, heat and water management [110]. Thus, the bipolar plate must have excellent strength, high thermal conductivity, low permeability, and high shock durability. Graphite has been previously used as a bipolar plate due to its excellent electrical conductivity. Nevertheless, there are many problems including the high corrosion rates, depressed mechanical strength, and exorbitant cost. Currently, the most commonly used bipolar plates are titanium and coated stainless steel [111,112]. But there are still some challenges, such as hydrogen embrittlement, passivation, and corrosion [113]. Therefore, exploiting cost-effective and stable bipolar plates is still under challenge.

Conclusions and prospects
In summary, hydrogen is considered as an environmentally benign alternative to traditional fossil fuels due to its high energy density and carbon-free pollution. AWE, SOE, and PEM water electrolysis are the main electrolysis technologies for hydrogen production. PEM water electrolysis is hailed as the most desired technology for high purity hydrogen production and self-consistent with volatility of renewable energies. From the thorough overview of the recent development of acidic HER and OER electrocatalysts for water splitting, there still exist many obstacles to exploiting and studying high-efficiency electrocatalysts. These are summarized as follows: 1) Pt and Pt-based catalysts are still the most popular materials for acidic HER. Numerous studies have made great processes in improving the activity and stability by using nonprecious metals, but the performance still fails to meet the requirement of practical electrolysers. In order to optimize the catalytic performance, compositional modulation, downsizing, and tailoring the electronic structure are effective strategies to increase the intrinsic activity and density of catalytically active sites. 2) OER catalyst for PEM is mainly focused on IrO 2 with high stability. However, the relatively low activity and the exorbitant cost severely restrain the practical applications. Compared with IrO 2 , although Rubased catalysts exhibited high activity and low price, the over-oxidation in acidic electrolytes resulted in poor stability. More attention needs to be paid to optimizing the activity of IrO 2 and improving the stability of Ru-based catalysts.
3) The catalytic processes in PEM electrolyser, especially for acidic OER, still have no well-established reaction mechanism. To discover the reaction mechanism on catalytic sites and monitor the dynamic evolution of the catalyst during the reaction process, advanced operando/in situ technologies are needed. For example, operando XAFS, Raman, XRD, and FTIR spectra, could explore the morphology, composition, and crystal structure change, as well as identification of real active sites, during the long-term electrolysis. In addition, the theoretical calculation is a critical tool to investigate the reaction pathways and unravel the catalysis mechanism. 4) The remarkable catalyst for industrial application needs higher current densities while maintaining a low operating voltage. Such commercial PEM water electrolysis requires a long operation time at high current densities of 2-3 A cm −2 for high energy conversion efficiency. However, such higher current densities usually need relatively high operating voltage, most catalysts are unstable during the OER process in acidic media. Much more effort should be devoted to exploring the functional links between stability and activity.