Status and perspectives of key materials for PEM electrolyzer

Proton exchange membrane water electrolyzer (PEMWE) represents a promising technology for the sustainable production of hydrogen, which is capable of efficiently coupling to intermittent electricity from renewable energy sources (e.g., solar and wind). The technology with compact stack structure has many notable advantages, including large current density, high hydrogen purity, and great conversion efficiency. However, the use of expensive electrocatalysts and construction materials leads to high hydrogen production costs and limited application. In this review, recent advances made in key materials of PEMWE are summarized. First, we present a brief overview about the basic principles, thermodynamics, and reaction kinetics of PEMWE. We then describe the cell components of PEMWE and their respective functions, as well as discuss the research status of key materials such as membrane, electrocatalysts, membrane electrode assemblies, gas diffusion layer, and bipolar plate. We also attempt to clarify the degradation mechanisms of PEMWE under a real operating environment, including catalyst degradation, membrane degradation, bipolar plate degradation, and gas diffusion layer degradation. We finally propose several future directions for developing PEMWE through devoting more efforts to the key materials.


Introduction
The pursuit of clean and renewable energy resource has been an acknowledged path to sustainable carbon neutrality [1,2].The massive scale-up of dihydrogen (H2) holds great promise for achieving the decarbonisation of various sectors, especially these hard-to-abate sectors including heavy industry, transportation, heating buildings, steel manufacture, and minerals processing.It is expected that by 2050, annual demand for hydrogen could increase to 560 million tons (i.e., tenfold that of 2015), and meet 18% of the global energy demand [3].However, over 95% of current hydrogen production depends on coal gasification and steam methane reforming (also known as grey hydrogen), contributing to global carbon emissions [4].The green transformation of hydrogen production is an essential precondition for helping facilitate the progressive decarbonization of other sectors.A critical element of the future hydrogen cycle will be water-electrolytic hydrogen production driven by renewable electricity such as photovoltaic cells, wind turbines, and hydropower (also known as green hydrogen).To this end, development of advanced electrolyzer technologies is urgently required to increase water electrolysis efficiency and reduce the cost of hydrogen.
Two kinds of industrially practiced electrolyzer technologies for water electrolysis are alkaline electrolyzer and proton exchange membrane water electrolyzer (PEMWE) [5][6][7].The alkaline water electrolyzer (AWE, Fig. 1(a)), which uses alkaline solution (typically 30% KOH solution) as the electrolyte and nickel-based materials as the electrode [8][9][10][11], is more widely used due to its mature technology and low-cost advantages.However, the alkaline electrolyzer requires long start-up preparation and presents slow response to changes in electric power load, leading to difficulty in adapting to the frequent variations of electricity from renewable energy sources (e.g., sunlight and wind).By comparison, PEMWE is a more advanced water electrolysis technology (Fig. 1(b)), which uses ultrathin proton exchange membrane (PEM) film to transport protons and isolate cathode/anode electrodes, leading to large current density (> 2 A•cm -2 ), great conversion efficiency (80%-90%) and high hydrogen purity (> 99.99%).More importantly, the high flexibility and quick response of PEMWE under highly dynamic operating conditions make the technology especially powerful for integration with intermittent renewable energy sources.In addition, anion-exchange membrane water electrolyzer (AEMWE) [12] and solid oxide electrolysis cell (SOEC) [13] are two new water electrolysis technologies at their early stages of progression.While the SOEC process requires operations at high pressure and high temperature to conduct O 2-ions (Fig. 1(c)), the AEMWE still calls for advanced heat exchange method (HEM) to create stable transport pathways of OH -ions (Fig. 1(d)).The significant advantages relative to other three water electrolysis technologies make PEMWE receive unprecedented attention for green hydrogen production.
The earliest journal publication of PEMWE can date back to 1973 by Russell et al., who have already demonstrated the remarkable performance of their early PEMWE at General Electric Company [14].In their study, the PEMWE achieves high current densities from 1 to 2 A•cm -2 at cell voltages of 1.88-2.24V and remains highly stable for more than 15,000 h.In the past 50 years, PEM technology has been commercialized by several electrolyzer manufacturers including Giner Inc., Proton OnSite, Siemens, and Hydrogenics for development of hydrogen generation.However, the large-scale practical application of commercial PEMWE is prevented by its high costs of materials, including specific catalysts and construction materials.(i) Under the highly corrosive acidic environment of PEMWE, only few precious metal-based electrocatalysts (typically Pt and IrO2 for cathode and anode, respectively) are available to show reasonable activity and stability.(ii) Titanium materials are widely used to construct the stack components of PEMWE, which increases electrolysis cost.(iii) Expensive Nafion-type membranes are used as electrolyte in PEMWE.
For the last half-century, considerable achievements have been accomplished in key materials of PEMWE to enhance performance and lower hydrogen production costs.In this review, we focus on the recent advances of key materials of PEMWE.Unlike previous review articles mainly summarizing the electrocatalyst design and synthesis [15][16][17][18][19], this review covers the progress achieved in a series of key materials including membrane, electrocatalysts, membrane electrode assembly (MEA), gas diffusion layers (GDLs), bipolar plates (BPs), and their degradation mechanisms.Such a timely review will help researchers to grasp the overall research status of PEMWE for guiding future actions.

Basic principles
The splitting of water into oxygen and hydrogen, driven by electricity, is the core reaction in PEMWE.This is expressed as Eq. ( 1): The whole electrolysis process is separated into two electrochemical half reactions, namely oxygen evolution reaction (OER, also known as water oxidation) at the anode and hydrogen evolution reaction (HER) at the cathode (Fig. 1(b)).In particular, water molecules decompose at the anode to form oxygen, protons and electrons.The protons migrate through the membrane and are eventually reduced to hydrogen at the cathode.Their reaction equations are shown as Eqs.( 2) and ( 3): ( ) ( ) Unlike aqueous electrolysis cell, PEMWE features acid membranes are solid electrolytes rather than liquid electrolytes.This solid acidic membrane allows two half reactions to happen in the separate compartments to collect highly purified hydrogen and oxygen.The usually thin membranes enable protons to travel short distances, thus reducing ohmic losses.In addition, catalysts on the membrane surface can easily transfer protons from the reaction sites to the solid electrolyte, greatly lowering mass transport limitation.Because of these properties, PEM electrolyzer does not require a strong electrolyte solution to improve electronic conductivity.Only pure water is supplied into the electrolyzer enabling a simpler system design.

Thermodynamics
The amount of energy absorbed or released by a chemical reaction under constant temperature and pressure is called reaction enthalpy (ΔH), which is given by the difference in formation enthalpies between the products and the reactants.The decomposition of water as an endothermic reaction requires an external energy drive.Since the formation enthalpy of elementary substance is zero, the reaction enthalpy of water splitting is actually equal to the formation enthalpy of water.By deriving the laws of thermodynamics, the reaction enthalpy can be expressed as Eq. ( 4): where ΔG, ΔS, and T represent free energy, entropy change, and temperature, respectively.From the point of view of energy input, reaction enthalpy can be regarded as the sum of electrical and thermal energy contributions.In the standard state, the free energy of the reaction is 237 kJ•mol -1 , while the thermal input of the entropy term is 49 kJ•mol -1 [20].Obviously, the proportion of thermal energy is relatively small, and the supply of electrical energy plays a major role (Fig. 2(a)), especially in the low-temperature operation of PEMWE (< 373.15 K).When the temperature exceeds the boiling point of water and rises further, the demand for electricity decreases and the heat compensates more.This provides the basis for the concept of high-temperature electrolysis [21].Considering that electrolysis occurs reversibly, the minimum electrical work required to drive water decomposition is related to the reversible voltage of the cell (VR) when the thermal demand is sufficient.This is calculated as Eq. ( 5): where z and F are the number of transferred electrons and Faraday constant, respectively.When an external heat source is removed, electrical energy provides all the energy consumption.At this time, the voltage of the cell is called thermoneutral voltage (Vth), which is calculated as Eq. ( 6): Due to the small thermal contribution of low temperature electrolysis, the Vth is closer to the minimum operating voltage required in practice.A simple and effective measure of performance, i.e., voltage efficiency (ε), can easily be obtained.It is defined as the ratio of the Vth to the actual cell voltage (V) at a given temperature.The expression is as Eq. ( 7): In addition, ΔG and ΔH vary not only with temperature but also with pressure.The Nernst equation clearly shows the relationship between reversible cell voltage and temperature and pressure.Under given conditions, the concentration of a gas can be replaced by the relative partial pressure, so that the Nernst equation can be expressed as Eq. ( 8): where R is molar gas constant (8.314J•mol -1 •K -1 ).At ambient pressure (p 0 = 1 bar) and T = 80 °C, the saturation pressure of water is 0.53 bar and the partial pressure of oxygen (pO 2 ) and hydrogen(pH 2 ) is 0.47 bar [22], so that the reversible cell voltage gradually decreases as the temperature increases.When the temperature is fixed, the higher operating pressure leads to a reversible voltage rise.Because the pressure term is logarithmic, the effect of pressure is relatively insignificant.

Reaction kinetics
Under operating conditions, the electrochemical process is irreversible.The current flows through the electrode, and the electrode polarizes, causing the potential to deviate from equilibrium.The magnitude of the deviation is called the overpotential (η).Thus, the actual cell voltage is equal to the sum of Vth and η.Due to the complexity of electrochemical process, η is caused by many kinetic factors (Fig. 2(b)), including activation overpotential (ηact), ohmic overpotential (ηΩ) and mass transport overpotential (ηm).
The ηact arises from the high activation energy of a certain rate-determining step, i.e., a larger voltage is required to drive electron transfer at this step.The ηact consists of the activation of the cathode (ηact,cath) and the anode (ηact,an).As the cathode uses Pt/C catalyst, its kinetics of HER is very rapid, and the most dominant contributor to the ηact is the ηact,an.When thermodynamic equilibrium is reached, the current density of anode and cathode is balanced, and the current density at this point is called exchange current density (j0), which depends on the electrode material.According to Arrhenius equation and Faraday's law, the relationship between net current density and j0 and η can be obtained as Eq. ( 9): (1 ) This is called the Butler-Volmer equation for chargetransfer kinetics, where β represents the symmetry factor or charge transfer coefficient.It can be seen that increasing the j0 or decreasing the η greatly improves the electrode kinetics of OER, thus accelerating the water splitting.For Eq. ( 9), the first and second exponential terms are the contributions of oxidation and reduction currents, respectively.When the η is high, the reduction current is small, that is, the second exponential term can be ignored, so that the Butler-Volmer equation can be simplified as Eq. ( 10): (1 ) After further deduction, the relationship between η and j can be expressed as Eq. ( 11): This is known as the Tafel equation.Based on this, the Tafel constant (b) can be calculated to evaluate the j0.Tafel slope (a) can be used to determine the rate-determining step of electrode reaction, which provides an important way to understand electrochemical mechanisms [23,24].
The ηΩ and ηm are non-Faraday losses except ηact.The ηΩ is caused by the electrical resistance of the electrode and cell assemblies and the protonic resistance of the membrane.Normally, proton transport is a major obstacle due to the highly conductive materials selected for the cell components.However, the interfacial resistances become significant when the components are passivated under high voltage and long-term operation.The effect of resistance on overpotential is usually expressed by Ohm's law as Eq. ( 12): where RΩ is the sum of all ionic and electrical resistances in the cell components.According to Ohm's law, the ηΩ increases with the increase of current density, which requires that the resistance of the whole cell be reduced as much as possible.
Similarly, the ηm is also closely related to current density.Diffusion overpotential (ηm,dff) dominates mass transport limitation at low current densities [25].Because the reaction rate at the electrode-solution interface is relatively fast, and the diffusion of ions or reactive species in the solution is relatively slow, the existence of concentration gradient enhances the polarization.The Nernst equation can be used to determine this non-equilibrium potential dominated by Fick diffusion as Eq. ( 13): ( ) where k is the reaction rate constant at equilibrium, which can be given by the concentration of reaction species at the electrode interface i ( ) c and the reference position r ( ) c , so Eq. ( 13) can be expressed as Eq. ( 14): The higher the concentration gradient, the greater the diffusive potential.In addition, when the number of species at the reaction interface increases, the mass transport limitation also increases.While at high current density, a large number of bubbles generated in a short period of time shield the active region and isolate the contact between electrode and electrolyte, reducing the catalyst utilization rate.The resulting overpotential can be called bubbles overpotential (ηm,bub).The influence of bubbles can be reduced to some extent by designing the flow rate reasonably [26].

Stack structure
In PEMWE, the core components mainly include BPs, GDL, PEM, and electrocatalysts of cathode and anode [27].Figure 3(a) shows the schematic diagram of a single cell in a PEMWE stack.The two half-cells are separated by the PEM, which transports protons during the reaction and blocks the passage of the product gas.The catalysts are directly applied to the membrane or porous transport layers.In most cell designs, the catalyst layers deposited on the membrane, forming the key component of the cells, that is, the MEA [28].Two porous transport layers (also known as GDL) are sandwiched on both sides of the MEA.The flow-field plates (also known as BPs) encapsulate the two half-cells, acting as a transfer of charge, mass, heat, and establish contact to the external power supply [29].Hydrogen and oxygen products pass through the catalyst surface, GDL, and BPs in succession to release out of the cells.Expressly, the half-cells also need to be supplemented with some sealing elements to prevent gas and water leakage.
The core components mentioned above have an important impact on the cost, performance and longevity of the PEM water electrolysis for hydrogen production.BPs and GDL make up the largest proportion of the stack cost, at 51% and 17%, respectively [30].The oxidative and corrosive operating environment limits the materials to be mainly advanced titanium-based materials, and the protective coatings such as Pt and Au are also needed [31,32].To reduce the cost of the entire electrolytic cell, the development of cheap alternative materials is necessary.Relatively speaking, PEMs and catalysts account for a smaller portion in the PEMWE stack, at 5% and 8%, respectively [30].In addition, the manufacture of MEA requires to consume 10% of the whole [30].Obviously, achieving high-efficiency water electrolysis performance while minimizing the material costs remains a huge challenge for PEMWE.

Membrane electrode assembly
MEA, integrated by PEM, cathode and anode electrocatalysts, is the heart of PEMWE, and largely determines the performance of water electrolysis.The ideal PEM should be able to meet a variety of functional requirements, including low gas permeability, excellent proton conductivity, good water absorption, low swelling ratio, outstanding chemical and mechanical stability, low cost, and high durability [33,34].Up to now, perfluorosulfonic acid (PFSA) membrane is the commonly used commercial PEM for PEMWE [35].The membrane has hydrophobic Teflon-like backbones and hydrophilic sulfonic acid side chains.Depending on the equivalent weight (EW), side-chain chemistry and length, the PFSAs can be divided into different membranes (Fig. 4(a)), such as nafion, aciplex, flemion, 3M, and short-side-chain (SSC) [36].Among them, the nafion series membranes manufactured by DuPont are a representative category.Nafion 117, 115, and 112 are the most used membranes in this series, and the different numbers represent different equivalent weights and thicknesses, which have a significant impact on the overall performance of electrolyzer [37].Ma et al. believed that the thickness of the membranes affects ionic conductivity of PEMWE, and confirmed experimentally that the thinner the membrane is, the smaller the ohmic resistances are, and the better the performance of the electrolyzer is [38].However, the use of too thin membranes may cause some problems for PEMWE, such as increased gas permeability, reduced hydrogen purity, decreased mechanical strength and durability, and potential safety hazards.It is worth noting that crossover in PEMWE is fatal [35,39].Once the gas penetrates, the hydrogen reacts with oxygen to release a large amount of heat, which will destroy the membrane and the entire stack.In view of these problems, some studies have proposed that the hydrogen permeating from the cathode side to the anode side can recombine with oxygen by using an intermediate layer containing Pt nanoparticles between the two membranes supported with catalysts, which can significantly reduce the permeation of hydrogen [40,41].In addition, nafion/graphene/ nafion PEM sandwich structure was designed by Bukola and coworkers [42].The special properties of graphene can allow protons to pass through, but block the transmission of hydrogen, which makes the sandwich structure exhibit high proton conductivity and eight times lower hydrogen crossover.
Although nafion membranes have matured as commercial membranes, they have the disadvantages of high cost and decreased proton conductivity under high temperature conditions (> 100 o C) [35].Water electrolysis at high temperatures can lower the Gibbs free energy change and improve the electrode kinetics, but the PFSA membranes will degrade under this condition.Therefore, nafion-based composite membranes and hydrocarbon membranes have received widespread attention from researchers [43].Nafion-TiO2 membrane and nafion-SiO2 membrane have been fabricated by Baglio et al. and Antonucci et al., which was proven to operate well above 100 o C [44,45].These inorganic metal oxide fillers are hygroscopic, allowing composite membranes to have better water retention and a more uniform water distribution, which results in a reduced ohmic resistance, and in turn shows better cell performance.Hydrocarbon membranes were first developed in the early 1980s.Motivated by high-temperature operation, phosphoric acid (PA)-doped polybenzimid-azole (PBI) was found to operate at more than 100 o C thanks to its high anhydrous proton conductivity and low vapor pressure, and has been widely studied by researchers [46].However, the membrane failed to achieve long-term durability, which may be caused by the poor chemical stability of PA under oxidative conditions.To sum up, the development of PEMWE is inseparable from the technology advancement of proton exchange membranes.How to strengthen the mechanical and chemical stability and reduce the cost is the direction of membrane development.It is worth noting that the safety issues caused by membranes should also be considered.
The preparation of membrane electrodes is also an important part of building an electrolyzer.At present, there are two main configurations for MEA (Fig. 4(b)), that is catalyst coated membrane configuration (CCM) and porous transport electrode configuration (PTE) [47].The PTE configuration is that catalyst deposited directly on the porous transport layers.In this case, the pore size of the PTE needs to be optimized, because small pores cause excessive mass transfer resistance, while large pores will lead to catalyst infiltration into PTE, resulting in low catalyst utilization.However, there are also studies showing that PTE configurations have better polarization behavior at high current densities above 750 mA•cm -2 , thus exhibiting better properties than CCM configurations [48].It is generally believed that CCM configurations have better contact between membrane and catalysts, which can reduce the interfacial impedance and improve the proton conductivity and durability.However, this mode may cause swelling of the membrane, so it needs to be handled more carefully.In addition, a new membrane electrode preparation method was proposed by Holzapfel et al., that is, direct membrane deposition (DMD) [49].The DMD-MEAs (Fig. 4(b)) are prepared by depositing the membrane directly onto the cathode electrode, which exhibit reduced ohmic and mass transfer losses, resulting in better electrochemical performance.It is worth noting that gas crossover and performance analysis of low catalyst-loaded DMD-MEAs will be the key issue for future research.This method provides a new idea for the preparation of MEA.
Usually, the catalyst is first prepared as a slurry, and then the CCMs can be made by transfer printing or direct coating method on membranes.The direct coating method is simple in technology, but the transfer printing avoids swelling and deformation of the membranes.The coating methods of MEA generally include spray coating (ultrasonic or aerosol), roll-to-roll (such as slot-die, gravure, and knife), and hand painting to apply the catalysts on the membrane [50,51].The ultrasonic spraying method (Fig. 5(a)) is suitable for smallscale preparation of CCM membrane in the laboratory, which can precisely control the amount of catalyst coating, and it is reliable and reproducible [52].The roll-to-roll coating methods (Fig. 5(b)), such as slot-die coating, gravure coating, and knife coating can achieve continuous coating, which improves production efficiency and reduces the costs [53].This coating method needs to be optimized for slurry composition (e.g., solvent and solid content) and processing conditions (e.g., drying temperature) to achieve a better performance comparable to lab-scale catalyst electrode [51].
With the expansion of PEMWE applications, the use of roll-to-roll coating technology to achieve mass production of membrane electrodes will be the trend in the future.It is worth mentioning that 3M has developed nanostructured thin-film (NSTF) electrodes with Ir and Pt catalysts for the cathode and anode, respectively [54,55].The NSTF catalyst is prepared by growing perylene red whiskers on a special temporary substrate microstructured catalyst transfer substrate (MCTS), and then the catalysts are deposited on it in a rod-like array structure (Fig. 5(c)) [56].Such a preparation process allows the catalyst to be deposited uniformly, and NSTF catalysts can be mass-produced by roll-to-roll.Figure 5(d) shows the performance improvement of the NSTF catalyst in three representative stages [56].Benefiting from the improved intrinsic catalytic activity of NSTF catalysts and the improved conductivity between the catalyst and the membrane, their performance has been greatly enhanced.Currently, NSTF catalysts are well developed in fuel cells and it is believed that its commercialization in electrolyzers is only a matter of time.

Gas diffusion layer
The GDL is a porous medium between the proton exchange membrane and the BPs (Fig. 6(a)) [29].The liquid/gas two-phase fluid is transported through the GDL channel to the catalyst layer, where water splits into electrons, O2, and protons at the catalyst anode.The oxygen flows back through the catalyst layer and GDL to the separator plates and then out of the cell.Electrons pass through the GDL, BPs, and external circuit to the cathode side.Meanwhile, protons travel through the PEM to the cathode, where they react with electrons to form hydrogen.The H2 then flows through the cathode GDL and leaves the cell [58,59].As long as the water flow is maintained and the electrolytic conditions are maintained, H2/O2 will be continuously produced.
High-performance GDL materials must meet the following requirements: (i) The GDL must be resistant to corrosion, due to the high overpotential of the anodic OER, the presence of oxygen, and the highly acidic environment caused by the protons generated during the water-splitting process.(ii) GDL needs to conduct electrons, so they must also have good electrical conductivity and low resistivity.(iii) The GDL must also provide mechanical support for the membrane, especially in the case of operating differential pressure, where the gas must be efficiently expelled and the water must be efficiently countercurrent to the catalytic layer [60][61][62].The GDL is usually made of carbon (e.g., carbon paper and carbon cloth) or metal materials (e.g., titanium and stainless steel) [58,60,63].Of the two materials, the carbon material can only be used as the cathode in PEMWE because of high oxidation potential of anode, as well as its low mechanical strength, which makes it difficult to operate for long periods.More interest has been generated by metal GDL due to its high conductivity, rapid production and low cost.Even at acidic and high anodic potentials, titanium is the least corrosive material and is relatively easy to form various types of porous media.Hence, titanium meshes/felts/foams/sintered powders are utilized as a GDL at the PEMWE anode.
At present, the optimization of the GDL mainly focuses on the adjustment of its pore and structure [64,65].Pore size and structure of the GDL greatly affect fluid transport.Large pores promote gas removal, but reduce electron transport efficiency and reduce the amount of water in the catalytic layer.Conversely, small pores hinder gas removal and increase mass transfer resistance.Therefore, most of the research focuses on optimizing the pore structure of the GDL to obtain good performance.Grigoriev et al. determined the optimal pore size, plate thickness, and porosity of the GDL by experimental and modeling approaches [64].The study found that the optimum sphere particle sizes ranged from 50 to 75 μm and the optimum pore sizes were between 12 and 13 μm.The porosity should be in the range of 30%-50%.Conventional titanium GDLs, including felt, woven mesh, or foam, have a fiber/foam pore morphology that results in random pore size and distribution.This random and inhomogeneous structure makes it impossible for conventional titanium GDL to accurately control liquid/gas/electron/thermal distribution.Therefore, novel GDLs with tunable and controlled pore morphologies are strongly desired.Kang et al. comprehensively studied the titanium GDL with straight-through pores and well-defined pore morphology (Fig. 6(b)) [57].Their new GDL with a pore size of 400 μm and porosity of 0.7 achieved optimum performance of 2 A•cm -2 at 80 o C and 1.66 V (Fig. 6(c)).The thin/well-tunable titanium-based GDL significantly reduced ohmic and activation losses, which electrocatalytic performance is far better than the traditional titanium felt material.And they found that porosity had a greater impact on performance than pore size.Direct visualization results of electrochemical reactions were obtained using a high-speed microscale visualization system, and bubbles were found to be generated only at the pore edge (Fig. 6(d)), explaining the effect of pore size and porosity on PEMWE performance.
At the same time, durability and degradation mitigation of the GDL are also important [66].Generally, current collector degradation can be categorized into chemical degradation and mechanical degradation.Chemical degradation is mainly caused by corrosion, while mechanical degradation is mainly caused by compression force, dissolution, and erosion by hydrothermal effect.The degradation of titanium is mainly caused by surface passivation and hydrogen embrittlement.Titanium is easy to get passivated in high potential, high humidity, and rich oxidation environment, and a layer of oxide film with low conductivity is generated, which greatly increases the contact resistance between titanium and fluid collector.Another problem is that titanium-based materials are most prone to hydrogen embrittlement.Therefore, titanium plates are often coated with precious metals, such as Au or Pt, for durability and performance requirements.However, this will also increase the cost of PEMWE.Thus, searching for the coating materials with low cost, high conductivity, high corrosion resistance, hydrogen embrittlement resistance is the unremitting pursuit.The GDL, BPs, and PEM are held together with high compression force to prevent water/gas leakage.This compression force strongly affects the performance of GDL/catalyst layer interface and GDL/BPs interface and the electrolytic cell's overall performance [67].The GDL will deform to a certain extent under this pressure, and the unsmooth surface will lead to local enhancement of current density, which will affect the mass transfer efficiency of the device and reduce the life of the device.A smooth surface for fluid collection is essential to reduce contact resistance and prevent degradation.

Bipolar plates
The BPs are multifunctional components in a PEM electrolytic cell.BPs have two basic functions: one is to electrically connect adjacent cells in a stack, and the other is to supply and remove reactants (i.e., water) and gaseous products (i.e., H2 and O2).Additional functions include mass transport and heat transfer functions [68].These functions must be maintained under high pressure, oxidation (anode), and reduction (cathode) conditions in the operating environment of the electrolytic cell.These features require BPs to be highly conductive, corrosion-resistant, impermeable, low-cost, and of sufficient mechanical strength [69].As a result of these requirements, there is not much material available for PEM cells.At present, the materials that can be used as BPs are graphite, titanium, stainless steel, etc.None of these bipolar materials has the advantage of low cost and all have various operating defects.The development of low-cost and high-performance BPs is crucial to the commercial success of PEMWE [58,70].
Graphite has previously been used in fuel cells because of its high electrical conductivity.Therefore, it was first thought of in the preparation of PEM electrolytic cells.However, graphite BPs have some problems such as poor mechanical strength, high cost, manufacturing difficulties, and high corrosion rate [68].At the anode, the high carbon corrosion decreases the thickness of the BPs, which results in an increase in electrical contact resistance between the MEA and the collector.In addition, due to the oxidation of carbon surface, the hydrophobicity of carbon BPs decreased.These effects cause the performance of the BPs to degrade quickly, resulting in a poor lifetime.Graphite plates are only suitable for cathodes.
In order to solve these problems, metal-based plates (such as titanium and coated stainless steel) have been studied recently.
Compared with graphite, titanium has excellent corrosion resistance, low initial resistivity, good mechanical strength, and light weight, making it the best plate material for PEMWE at present.However, titanium plate will also undergo passivation and corrosion phenomenon like GDL.In high potential, high humidity, and rich oxidation environment, titanium BPs surface is easy to passivation to form an oxide film.The oxide film with a low electrical conductivity greatly increases the contact resistance between the BPs and current collectors.Coating and alloy methods were developed to protect titanium plates to solve this problem.Titanium plates are usually coated with precious metals or platinum group metals to meet durability and performance requirements in high voltage and oxidation environments.For example, Jung et al. used gold-plated titanium as a BPs and observed an improvement in the performance of the electrolytic cell due to the reduced electrical contact resistance between the electrode and the plate (Fig. 7(a)) [71].The 1 μm gold coating was used as a barrier layer to inhibit the formation of passivation layer on the surface of the titanium substrate plate.However, BPs coating is quite expensive, especially when applied in large-scale electrolytic cells, which is not conducive to large-scale commercial use.Therefore, the most effective way to reduce the cost of BPs is to reduce the amount of noble metals in the coating material by improving the coating composition or preparation process.
Another way to reduce the cost of BPs is to find suitable alternative materials for titanium.Stainless steel is one of the alternatives to titanium, but stainless steel components corrode very quickly in aggressive acidic environments, so coatings are also needed to maintain reasonable service life [72].Yang et al. fabricated stainless steel plates by selective laser melting (SLM) printing and then plated surface treatment with Au [71].The BPs exhibited excellent corrosion resistance and electronic conductivity.Gago et al. produced dense titanium coatings on stainless steel substrates by vacuum plasma spraying (VPS) [31].Then Pt was deposited by physical vapor deposition (PVD) magnetron sputtering for further surface modification of Ti coating (Fig. 7(b)).The dense and robust Ti coating provides the necessary protection against pitting of stainless steel plates.Many research efforts have been made to find other low-cost titanium bipolar coatings/surface treatment solutions in PEMWE, including nitride coatings and niobium coatings.In addition to coating treatment, the design of integrated unit is also the focus of research.Yang et al. fabricated multifunctional and integrated units of PEMWE with highly complex internal structures using additive manufacturing, leading to great improved performance (Figs.7(c) and 7(d)) [73].This is the first time PEMWE components have been integrated into a single plate for water decomposition and hydrogen production.The structural innovation of the sheet has provided a radical development for the simplification of PEMWE, thus significantly reducing the number and weight of its parts, providing an opportunity for the optimal configuration of PEMWE.
The flow plate is one of the components of the BPs, and the flow channel is usually engraved on the BPs.One of its specific functions is to generate flow field uniformly distributed on the catalytic electrode.The uneven flow distribution on the surface area of the flow field plate may lead to the unbalanced use of precious catalyst materials and the device's overall efficiency is lower than expected [62,75].Therefore, the flow field plate of the electrolytic cell must be correctly designed so that the reactants (water) are evenly distributed over the catalytic reaction surface to provide a way to collect the reaction products (hydrogen and oxygen) and to provide a conductive path to the reaction site.For large-area electrolyzer, the function of flow field is particularly important, and the unreasonable design of flow field is often the main reason for the performance decline of electrolyzer.At present, researchers have designed and developed a variety of flow field structures, such as point flow fields, porous flow fields, snake flow fields, combined flow field structures and so on.In a PEMWE, the shape and geometry of the flow field directly affect the uniformity of reactant distribution and thermal management efficiency of the flow channel.Toghyani et al. studied five flow field designs for PEMWE (Figs. 8(a)-8(e)), including parallel, single path serpentine, dual path serpentine, triple path serpentine, and quadruple path serpentine [74].The results show that the single path serpentine pattern has the best performance (Fig. 8(f)), due to better distribution of hydrogen molar fraction and current density.In addition, Li et al. studied the effect of flow field on the performance of PEMWE at high temperature through experiments [76].The conclusion shows that cathodic flow field mode affects ohmic overpotential and anodic flow field mode affects activation overpotential.

Electrocatalysts of PEM electrolyzer
The commercial catalysts of PEMWEs are basically Ir-based catalyst (anode site) and Pt-based catalyst (cathode site).In order to reduce the cost of catalysts, the focus of research is to reduce the content of noble metals in catalysts or to discover new non-precious metal catalysts while maintaining excellent activity and stability.

Hydrogen evolution electrocatalysts
Due to the outstanding activity and long-time stability, Pt has been recognized as the state-of-the-art and benchmark catalyst for HER.The high costs and scarcity of Pt heavily restrict large-scale commercialization of PEMWE.Up to now, 40 wt% Pt/C has been selected as a mature commercial HER catalyst for PEMWEs, and the loading of Pt has decreased to 0.4-0.6 mg•cm -2 .Although the usage of Pt/C largely reduces the content of Pt, it is still costly when Pt/C is assembled in MEA for large-scale application.It is estimated that 1TW energy storage (TWH2) will consume 30% of annual production for Pt (54 tonnes) [77].In the past decades, numerous efforts have been devoted to reducing Pt loading or even replacing it with earth-abundant elements.
To maintain high activity with low Pt loading, one effective strategy is to reduce the size of the Pt-based catalyst dispersed in supports.Many methods have been reported to prepare small Pt clusters and single Pt atom with high dispersion on supports, such as wetness impregnation method [78][79][80], atomic layer deposition [81][82][83], and electrochemistry method [84][85][86].In these works, adopting gentle loading methods and appropriate carrier is crucial to prevent aggregation and reduce the size of Pt.Usually, the smaller size of catalysts is, the more active sites are exposed, which is in favour of improving the utilization of catalysts [87,88].Carbon-based materials are the most widely studied supports for HER because of its structural diversity, high stability, large surface area, and low cost [89,90].Very recently, Ye and coworkers successfully anchored single Pt atoms on aniline-stacked graphene (Pt SASs/AG) by a microwave reduction method (Figs.9(a)-9(c)) [91].Because of the π-π interactions between graphene and aniline molecules, graphene became hydrophilic and can be uniformly dispersed in solution.During the synthetic process in acid, PtCl6 2-can be successfully trapped by -NH3 + group of aniline by electrostatic interaction.Thus, with a Pt loading of 0.44 wt%, Pt SASs/AG exhibited an overpotential of 12 mV at 10 mA•cm -2 and a mass activity of 22,400 AgPt -1 at η = 50 mV, which is far higher than commercial Pt/C (20 wt%).
In addition to carbon materials, many novel materials have emerged as catalysts supports, such as MXenes, transition metal carbides (TMC, e.g., MoC and WC), and sulfides (TMS, e.g., MoS2 and VS2) [92][93][94].It is easy to trap Pt atoms for the existence of defect sites in these transition metal compounds, which contribute to constructing Pt nanoparticles or even single atom catalysts with high dispersion.The synergetic interaction between Pt and substrates can further improve the HER activity.For example, Zhang et al. reported enriched Mo-vacancies MXene nanosheets (Mo2TiC2Tx-VMo) synthesized by electrochemical exfoliation [92].And during the exfoliation, single Pt atoms (1.2 wt%) are successfully anchored at the Mo-defect sites, which dramatically enhance the HER performance of MXene to an overpotential of 30 mV at 10 mA•cm -2 .In another study, Zhu et al. fabricated MoS2-loaded Pt catalysts via a spontaneous photothermal redox reaction [95].Due to the existence of defects after thermal pretreatment of MoS2, single Pt atoms are uniformly anchored on the Mo-vacancies forming hybridization between Pt and S. The catalyst with low Pt loading (0.22 wt%) shows excellent HER performance with an overpotential of 44 mV at 10 mA•cm −2 .
Apart from supported catalysts, alloying is also an effective strategy to reduce Pt consumption.In 2005, Nørskov et al. clarified that the high performance of Pt originated from its moderate ability to bind hydrogen intermediates [96].This work guides us to understand that the catalytic activity is closely related to the electronic structure of catalysts surface.Then many studies are reported to optimize the electronic structure of Pt-alloys by tuning alloy compositions.A representative work was carried out by Jeff and coworkers [97].They evaluated the HER activity of 700 binary alloys via a high-throughput screening scheme with DFT calculations.Though a large number of alloys are predicted to be highly reactive for HER, only a few of them are stable in acid conditions.As a result, only BiPt alloys own Pt-like HER activity, which was also verified by experiment.However, simple binary/ternary alloys still contain high levels of Pt.Subsequently, high-entropy alloys (HEAs), composed of more metal elements (at least five components) than binary/ternary alloys, gradually garner attention to their unique properties in catalysis [98][99][100][101].For HEAs, the regulation of component is more flexible and changeable, which suggests that HEAs are a proper platform to construct low-Pt catalysts with non-precious metals, while maintaining the excellent activity [102,103].Feng et al. prepared a series of ultrasmall NiCoFePtRh HEA nanoparticles via a chemical coreduction method (Fig. 9(d)) [104].Structural characterization reveals that the HEA nanoparticles exhibit a face-centered cubic (FCC) structure.Due to the synergistic effect among the five elements, the HEAs showed a prospect for HER with a mass activity of 28.3 A•mg −1 at −0.05 V (vs.RHE) in 0.5 M H2SO4 (Fig. 9(e)).
Despite the high activities of these low-Pt catalysts, there are still two factors restricting their further development in PEMWEs.(i) The long-term stability of (ultra)low-Pt catalysts is often neglected.Although some literature claims that the catalysts have good activity, they are far from meeting the requirements of the practical application of PEM.(ii) For some advanced Pt-alloy and single Pt atom catalysts, it is challenging to achieve uniform and large-scale preparation due to the harsh synthetic conditions and high raw material costs.
Developing alternative catalysts with earth-abundant elements to replace Pt holds a great attraction.Over the decades, researchers have made many efforts to develop Pt-free catalysts for HER.Usually, the transition metal elements, used for Pt-free HER electrocatalysts, mainly include Fe, Co, Ni, Mo, W, and Cu [105].Inspired by hydrogenases and nitrogenases in nature, MoS2 firstly emerged as a promising Pt-free electrocatalyst for HER [106,107].However, for MoS2, only the edge sites exhibit active for HER and the conductivity is poor, which heavily restricts its HER activity [108].Therefore, many effective measures have been taken to increase its conductivity and the density of active sites, such as doping strategies, defect engineering, crystal phase engineering, and so on.Other transition metal sulfides (tungsten sulfides, iron sulfides, and cobaltous sulfides) have been found and used as electrocatalysts for HER.For example, our group reported several conductive metal sulfides (e.g., Ni3S2, Co9S8, and FeS) as efficient HER electrocatalysts [109].Besides transition metal sulfides, there are many other non-noble metal materials exhibit great activity for HER, such as borides (e.g., MoB2 and WB2, Figs.10(a)-10(c)) [110][111][112], phosphates (MoP, CoP, and Ni2P) [113], nitrides (e.g., Ni3N and MoN), and carbides (e.g., Mo2C, and W2C) [114,115].
To date, although many Pt-free catalysts are found, few of them can be applied to PEMWEs.The intrinsic activities of Pt-free catalysts are usually 1-2 orders of magnitude lower than that of Pt.Although many studies attempt to use high loading Pt-free catalysts to compensate for the poor intrinsic activity.However, when applied in PEMWE, the high loading Pt-free catalysts may be costlier than Pt-based catalysts.In addition, these materials primarily suffer from poor stability.Recently, Laurie and coworkers integrated CoP into an industrial PEM electrolyzer (Figs.10(d)-10(f)) [113].At the PEM working conditions, CoP could continuously test over 1,700 h at 1.86 A•cm -2 , while the dissolution of CoP brought about increase in cell potential.The activity and durability of CoP catalyst in commercial electrolyzers still have big distances compared with Pt catalyst.If catalysts deactivate in a short time, the maintenance cost of electrolyzers will be higher than the capital cost of catalysts [116].This violates the original goals of developing Pt-free catalysts.

Oxygen evolution electrocatalysts
It is generally accepted that the sluggish OER kinetics of most anodic catalysts and their inadequate long-term durability under highly corrosive operating conditions are real obstacles to PEM water electrolysis technology [58].Since OER involves multiple oxygen-containing intermediates (e.g., *OH, *O, and *OOH) [8,10,12], optimizing the adsorption of active intermediates becomes the basis for improving catalyst performance.While stability factors (acid and catalytic stability) preclude the selection of most catalysts, and the need for reasonable activity makes the selection scope narrower [15][16][17].Ir-based catalysts are the feasible choice for PEMWE due to their high activity and excellent stability.Rutile IrO2 remains the most widely commercial catalyst since it was first reported in 1973 [14].However, the OER performance of IrO2 is affected by many factors, such as particle size, crystallinity, morphology, which are usually controlled by different synthesis conditions or methods [117].Rasten et al. investigated the effect of different annealing temperatures of IrO2 on MEA [118].They found that the particle size increased and became more crystalline with increasing annealing temperature, leading to lower electrocatalytic activity but higher electronic conductivity.In this way, the total polarization of the electrode reaches an optimal value at a moderate temperature.Similarly, Tachikawa et al. studied IrO2 films prepared by sputtering and concluded that annealing linearly reduced electric double layer capacitance (EDLC) and resulted in a corresponding reduction in the number of catalytic active sites [119].In addition, a recent theoretical study has shown that the tip and corner sites of IrO2 with small size have optimized Ir-H2O interactions [120].This may support the fact that smaller nanoparticles have superior activity experimentally.Therefore, precise control synthesis of IrO2 is a key part of optimizing performance of PEMWE.
From a practical application point of view, methods that can produce significant catalyst amounts for electrode fabrication are of more interest.There are two main categories of methods: (i) Wet chemical methods, including sol-gel [121], polyol [122], and aqueous hydrolysis [123] method, are widely used in the preparation of nanomaterials because of the easy uniformity of the products and the relatively low reaction temperature.(ii) Thermal decomposition methods, such as Adams fusion [124,125] and Pechini [126] methods, are more efficient in commercial large-scale preparation due to their relatively simple operation and the high purity and small particle size of the resulting products.And the high extensibility of these methods, such as various modified Adams methods and control of reaction conditions, broadens the synthetic chemistry of IrO2.Under certain synthetic conditions (e.g., low temperature and additional additives), it is easy to obtain an amorphous Ir oxide (IrOx) [127].This catalyst has better activity than crystalline IrO2, and it is generally believed that the high activity of commercial Ir-black catalysts is also due to the formation of the IrOx layer on their surface [128].Its activity may be attributed to higher levels of bulk defects, greater number of active sites, involvement of Ir 3+ , and reactive oxygen species [129,130].However, the consensus is that the high activity leads to poor stability, and several studies have shown that its Ir dissolution rate is much faster than that of crystallized IrO2 [131,132].In addition, IrOx has worse electrical conductivity [133], resulting in severe ohmic losses at high current densities.Although conductivity is an easily overlooked criterion for evaluating catalysts, it has a significant impact on the performance of PEMWE.Because the catalyst layer is electrically in contact with the porous substrate and has a certain thickness, the contact pressure between particles in different directions can change the perpendicular and lateral conductivity, resulting in the loss of a considerable proportion of the active surface area of the poorly conductive material [118,134].Several studies have shown that heat treatment greatly increases the conductivity and stability of IrOx at the expense of decreased activity [135][136][137].The optimized IrOx shows greater application potential than the crystalline IrO2.
In order to build a larger scale electrolyzers, the cost of catalyst must be reduced due to the high cost and scarcity of Ir.The current loading of IrOx in commercial electrolyzers is 2-2.5 mgIr•cm -2 [138,139], but it is estimated that 0.05-0.1 mgIr•cm -2 of catalyst is required to achieve Gigawatt scale [116,140].However, directly reducing the loading will thin and homogenize the catalyst layer under the restriction of high packing density of IrOx, resulting in irreversible loss of performance.Allowing for better electron and proton conduction and mass transport, the optimum thickness of the catalyst layer is 4-8 μm [140].Therefore, more researches focus on the construction of catalysts with low packing density and/or low iridium content to improve iridium utilization and thus reduce the loading.
The introduction of other active or inert components into IrO2 has been considered one of the best strategies for diluting iridium.The addition of Ru is an attempt to increase the activity, since the intrinsic activity of RuO2 is much higher despite its instability [141,142].The synergistic effect of Ru-Ir is expected to enhance or basically maintain the activity of Ir sites and stabilize Ru to enable more Ru sites to function [143].Commercial Ir0.7Ru0.3O2catalysts have achieved balanced activity and stability.To further dilute iridium and improve the corrosion resistance of the catalyst, more catalytically inert but chemically stable elements are used to form solid solutions with Ir oxides.For example, Nb, Mo, Hf, Ta, Sn and so on all have wide electrochemical windows and can maintain stable oxide forms under high oxidation potential of OER and acidic conditions.Accordingly, IrNbOx [144], IrxMo1-xOy [145], IrHfxOy [146], IrxSn1-xO2 [134,147] catalyst have been proved to have higher mass activity and long-term stability (Figs.11(a) and 11(b)).However, the proportion of non-noble metal elements in solid solution catalyst is limited to 50%-60%, and further addition will significantly reduce the activity.This results from a low number of active sites and an increased resistance.Therefore, the role of solid solution IrO2 in achieving ultra-low Ir loading may be quite limited.
The development of supported catalysts is another effective strategy to reduce the catalyst loading by better dispersing IrO2 and enhancing the conductivity within the catalyst layer to improve its utilization.Similar to the requirements in PEM fuel cells [148], the ideal support materials should have the following properties: (i) Good electrical conductivity, greater than 0.1 S•cm -1 at PEMWEs operating temperatures (about 80 °C).(ii) Adequate corrosion resistance and long-term stability under high oxidation and acid conditions.(iii) Large BET surface area to better disperse the catalyst.(iv) Porous structure with sufficient porosity and pore size to allow water to enter the reaction site and transport oxygen away from the electrode.(v) Low cost, high abundance and easy access.To meet the above requirements, the most widely studied materials are metal oxides, carbides and nitrides [148,149].However, carbides and nitrides are inevitably converted to oxides over long periods of time [150], so oxides seem to be more suitable support materials at present.TiO2 has been verified as an excellent support of IrO2, and IrO2-TiO2 catalyst has been used as a commercial product in practical electrolyzers [151].However, the conductivity of TiO2 is extremely poor, which requires high Ir content (> 40 wt%) to completely cover the support to improve the conductivity [152].Obviously, this is not conducive to the further reduction of Ir loading.Therefore, many studies focus on improving the conductivity of TiO2 support.One of the most efficient methods is metal doping of Nb [153], V [154], W [155], etc.For example, Zhao et al. reported that WxTi1-xO2 can maintain the performance of electrolyzer at an Ir loading of 0.4 mg•cm -2 (Figs.11(c)-11(e)) [155].Other oxides such as SnO2 and their doping counterparts are also considered as potential candidates [156], but their ionic dissolution may be more severe than that of Ti and therefore require further evaluation [157].Although a simple physical mixing of IrO2 with other oxides can achieve some positive effects, a support that can interact beneficially with IrO2 to enhance its activity and stability is more desirable.
In recent years, considerable attention has been paid to the development of low-iridium catalysts with other compositions and structures to replace IrO2.In these catalysts, Ir sites tend to have higher intrinsic activity, resulting in mass activity far exceeding that of IrO2.Perovskite (ABO3) and pyrochlorite (A2B2O7) Ir-based oxide are two representative materials [158][159][160][161].Because they all have good compositional tunability, that is, the A and even the B sites can be replaced by one or more elements.At the same time, a variety of crystal structures greatly expand the exploration space.These factors make these materials have great potential to achieve the goal of low Ir loading.At present, the most widely studied Ir-based perovskite is SrIrO3, which is obviously unstable despite its high activity [162].In particular, Sr will undergo severe leaching, making its surface amorphous into IrOx [163].For some double-perovskites such as Ba2PrIrO6 and La2LiIrO6, complete amorphization may occur after long immersion in acidic solution (Figs. 12(a)-12(c)) [164].These catalysts are not true low-iridium catalysts because their active component is IrOx, and their significant ionic dissolubility makes them impossible to use in practice, but only as model catalysts for exploring structural evolution and mechanism.Clearly, low iridium catalysts that are resistant to chemical and electrochemical corrosion are the real candidates.For example, our group developed SrIrO3-SrTiO3 solid solution catalyst with both high activity and structural stability (Figs.12(d)-12(f)) [165,166].SrTiO3 provides a robust framework to anchor the bulk Ir sites, and slight surface reconstruction endows high activity to the surface Ir sites.However, at present, the low-iridium catalysts are still in the stage of laboratory exploration, and few of them are used in practical devices.On the one hand, these fine catalysts often require complex synthesis processes, and the gram-scale synthesis is already difficult.In addition, factors such as conductivity and dispersion that are overlooked in laboratory-level testing become key obstacles to MEA testing.Despite their attractive activity and stability, low iridium catalysts are still a long way off from practical use.
Using an Ir-free catalyst is the most direct but hardest way to achieve cost savings [17,167,168].A catalyst without iridium cannot strike a balance between activity and stability.Noble metal Ru-based oxides exhibit excellent activity, but degrade rapidly at high oxidation potential [169,170].Non-noble Mn-based oxides have better stability and lower cost, but poor activity [171].Although some catalysts such as phosphide [172] and sulfide [173] have been reported, they cannot achieve reasonable activity and stability.Non-iridium catalysts are unlikely to replace Ir-based catalysts in the foreseeable future.

Degradation mechanisms of PEM electrolyzer
Compared to other hydrogen production methods, PEM water electrolysis technology must increase the operating time to more than 10 4 -10 5 h to be economically competitive [174,175].In order to achieve this goal, it is necessary to better understand the microscopic processes that reduce the performance of electrolyzers in the long term, and to find solutions.In the past decade, researchers have made great efforts to explore the degradation mechanism of PEM fuel cells [176][177][178], however, there are far fewer reports on the degradation of PEMWEs.While the two technologies have some things in common, their differences are also significant.Notably, the intermittent and variable operating conditions of PEMWEs tend to significantly increase the degradation rate [179].

Catalyst degradation
Due to the complex operating environment of PEMWEs, such as high potential, strong acid, and frequent gas-liquid flushing, there are a lot of processes leading to catalyst degradation (Fig. 13(a)) [180], including catalyst dissolution, agglomeration and maturation of catalyst particles, passivation and corrosion of supports, exfoliation of the catalyst layer and coverage of catalyst by contaminating metals [59].Especially for low iridium content catalysts, these unfavorable microstructural evolutions are more obvious.These factors ultimately lead to the reduction of electrocatalytic active sites and the increase of the contact resistance between the catalyst layer and the GDL.
The stability of the OER electrocatalyst is crucial because the anode operates in an extreme environment with high potential and strong acidity.IrO2 and RuO2 are the most active OER electrocatalysts, however, the Pourbaix diagrams which reflect the thermodynamic stability of metals in the range of electric potential and solution pH demonstrate that both Ir and Ru tend to transform into high valence and soluble species at a high potential [181,182] (Figs.13(b) and 13(c)).It should be noted that although RuO2 shows a lower overpotential than IrO2, its dissolution rate is also much faster due to the formation of volatile RuO4 [183].In order to reduce the amount of noble metals and improve the dispersibility, carbon is often used as an electrocatalytic support due to its good electrical conductivity.Unfortunately, the oxidation of carbon at the high anodic potential of OER eventually leads to the shedding of the catalyst [184], so carbon can only be used as a HER support in PEMWE.Some researchers have used inert metal oxides such as TiO2 [185], Ta2O5 [186], Nb2O5 [187] as supports, demonstrating that the support-catalyst interaction (anchoring effect) can effectively reduce the rate of catalyst particle separation or dissolution, although high noble metal  loadings are required to provide sufficient electrical conductivity.For the cathode, although Pt is a very stable HER electrocatalyst, some researchers have observed the agglomeration and growth of Pt particles during electrolysis [174,188].The mechanisms leading to the enlargement of catalyst particles include Ostwald ripening, reprecipitation and coalescence (Fig. 13(d)) [180].A classic strategy to modulate the active sites and stability of catalysts is to modify the microscopic morphology, e.g., Pt nanotubes are less prone to Ostwald ripening and aggregation than Pt nanoparticles [189].
In the actual operation of PEMWE, due to unqualified water purification, corrosion of metal components or solubilization of catalyst components, the circulating water may introduce metal ions [59], such as Na + , Ca 2+ , Cu 2+ , Ni 2+ , and Fe 3+ .Some metal cations with a reduction potential higher than H + , such as Cu 2+ , are reduced to metallic state at the cathode, covering the catalyst particles and leading to an increase in the HER overpotential.Other type of cations such as Ca 2+ and Ti 4+ cannot be reduced to metals because the deposition potential is too low, but they may also cover the catalyst surface in the form of Ca(OH)2 or TiO2, which can cause catalysis deactivation.Fortunately, catalyst poisoning caused by metal cations is usually recoverable, because the MEA can be regenerated by acid solution (e.g., 1 mol•L -1 H2SO4) treatment [188,190], although this incurs additional operating costs.

Membrane degradation
Polymer membrane is considered to be the most degradable component in the long-term operation of PEMWE [191].Although membrane thinning may increase proton conduction, the degradation and failure of membrane can lead to exfoliation of the cathode and anode catalyst layers (Fig. 14(a)) [192], short circuits, and the risk of generating flammable hydrogen and oxygen mixtures.Mechanisms of membrane degradation typically include mechanical degradation, thermal degradation, and chemical degradation.During the first 1,000 h of PEMWE operation, failures are usually caused by mechanical degradation such as punctures, cracks, mechanical stress and pressure differences (Figs.14(b) and 14(c)) [191].During the long-term operation of PEMWE, the MEA is continuously stressed between the BPs because of the constant torque.Cracking is more likely to occur in localized stress concentration areas, such as the contact between flow field channels and membranes (Fig. 14(d)) [67].Especially when the surface of GDL is rough, the damage to the MEA will be more obvious [193].During the microscopic water cycle, lack or unevenness of water may result in uneven current distribution, which in turn results in uneven heat distribution and eventually hot spots appear on the membrane.In addition, the direct combination of H2 and O2 by osmosis is also exothermic, and these hot spots lead to accelerated degradation of the membrane [194].To prevent mechanical degradation of the membrane, the clamps should be carefully designed to provide appropriate pressure and uniform compression.Another strategy is to support the membrane with reinforcing materials to prevent creep [59].In addition, thermal management is necessary to maintain the longevity of PEMWE.
For the polymer membrane, the impurity ions in the water will also occupy the ion exchange channel because the binding force is stronger than that of H + , resulting in a sharp decrease in the conductivity.Likewise, the membrane can be regenerated by acid treatment.During electrolysis, when O2 permeates through the membrane to the cathode, the two-electron reduction process that occurs on the Pt surface not only competes with HER but also generates H2O2 [195].Worse still, H2O2 can be catalyzed by contaminating metal ions (e.g., Fe 2+ ) to generate radicals (Fenton reaction) such as HO • and HOO • , which can attack the membrane and cause severe degradation [193].In general, the extent of membrane degradation can be monitored by the concentration of F -in water [196].

BPs and GDL degradation
For anodes, titanium is currently the mainstream material for electrode plates and GDLs due to its good corrosion resistance and electrical conductivity [197].However, in harsh highpotential and strong acidic environments, titanium is inevitably passivated (Fig. 14(e)).Although the formed TiO2 protective layer avoids further corrosion, its low electrical conductivity also obviously increases the contact resistance.In addition, F -ions generated by membrane degradation will destroy the passivation layer and lead to further corrosion [195].For cathodes, titanium components exposed to saturated H2 for a long time are prone to hydrogen embrittlement [198], which refers to hydrogen penetrating into the interstitial space of metal atoms to cause cracks and lead to embrittlement and fracture.Noble metal (such as Pt [174], Au [197], and Ir [199]) coatings can effectively avoid BPs corrosion and reduce contact resistance but increase cost, so the balance between performance and overall cost must be considered.

Summary and outlook
Against the background of carbon peaking and carbon neutrality, integrating the renewable power sources with PEMWE for hydrogen production is efficient in achieving clean energy transformation.In this review, we look back on the recent progress in PEMWE.The function, development status, and degradation mechanisms of each component in PEMWE are summarized.It should be recognized that the PEM water electrolysis technology is still in the primary stage of industrial production.The challenges related to the application demands (in the aspects of cost, performance, and lifetime) are mainly associated with the development of electrocatalysts, PEM, MEA, GDL, and BPs.Therefore, the further optimization guidelines of PEMWE could be considered as follows (Fig. 15).
(i) Competent electrocatalysts.The operating conditions of PEMWE (i.e., low pH and high current density) limit the selection of electrocatalysts to rare and expensive noble metals (e.g., Pt and Ir).This hinders the large-scale application of PEM hydrogen production technology.Therefore, it is imperative to reduce the usage of Pt (≤ 0.2 mg•cm -2 ) and Ir (≤ 0.3 mg•cm -2 ) in electrocatalysts without sacrificing activity and stability.In order to balance the scaling relationship of cost, activity and stability, the design idea of electrocatalysts can be concluded as follows: (a) introducing non-noble metal components to dilute the content of noble metal in electrocatalyst; (b) developing highly conductive and large surface area support catalysts to improve iridium mass activity; (c) regulating the electronic structure and local environment of electrocatalyst to enhance its intrinsic activity; (d) revealing the catalytic/deactivated mechanism to resolve the problems of stability improvement.In recent years, numerous low platinum/iridium, even platinum-/iridium-free electrocatalysts have been developed.However, it is far from their application in PEM hydrogen production.On the one hand, the performance evaluation of these electrocatalysts is based on the three-electrode system, a lab-scale assessment unlike the operating conditions in PEMWE.The majority of the electrocatalysts cannot exhibit ampere-level current densities and year-level lifetime in PEMWE.On the other hand, the preparation technology of these alternative electrocatalysts is difficult to scale up to kilogram scale.Therefore, the performance of the electrocatalyst under PEMWE conditions and the large-scale preparation should be taken into account in further research.
(ii) Next-generation PEM.The chemical, mechanical and thermal properties of PEM are closely related to the performance and lifetime of PEMWE.In practical application, it is the properties of PEM that determine the operating conditions (e.g., temperature and pressure) of PEMWE.Take nafion membrane as an example, its maximum operating temperature is 90 o C, above which the PEM will dehydrate and degradation, leading to inactivation of PEMWE.Even though high temperature can reduce the energy barrier and accelerate the reaction kinetics of water electrolysis.The operating temperature of PEMWE is limited below 90 o C, restricting the improvement of performance.Similarly, the high pressure is favorable for hydrogen storage, but will also accelerate the crossover of gas products, leading to the reduction of current efficiency and H2 purity.Another drawback of PEM is its high cost, which originates from the fluorine in backbone structure.It follows that ideal PEM requires to possess low cost, superior chemical and mechanical stability, high proton conductivity, and low gas permeability.And the current PEM can be developed or improved in the following ways: (a) reducing the gas permeability of the membrane by adding PEM sandwich structures; (b) enhancing chemical stability and mechanical property while ensuring its hydration degree; (c) increasing its upper limit of operating temperature, and enhancing its thermal stability; (d) developing the membrane with lower cost (e.g., hydrocarbon-based membrane).
(iii) Improved MEA fabrication.The MEA consists of the PEM, cathode, and anode catalyst layers, and sometimes contains the two GDLs.It is the main battlefield of water splitting.Therefore, adjusting the fabrication of MEA can fundamentally improve the cost, performance and lifetime of PEMWE.Herein, the improvement is not limited to the modification of electrocatalyst, but the optimization of preparation technology of MEA, including ink ingredient, coating method, and hotpressing conditions.Generally speaking, the preparation process of MEA can be improved by the following ways: (a) optimizing the ink ingredient (e.g., catalyst, support composition, ionomer, and other additional loading), and developing joint dispersion techniques (e.g., ultrasonic and shear-stress dispersion) to improve the physical properties such as grain diameter, rheological property, Zeta potential, and then achieve the uniform MEA; (b) Selecting appropriate coating methods (e.g., electrochemical deposition method, ultrasonic spraying method, and transfer printing method) to enhance the mass transfer capacity of the three-phase interface on MEA, improve the utilization of noble metals, and then boost its catalytic performance; (c) reforming the current coating equipment to achieve roll-to-roll coating, which can meet the industrial demands; (d) adjusting the optimal hot-pressing temperature, pressure and duration to achieve the synergistic enhancement of activity and stability; (e) coordinating the compatibility of ink ingredient, coating method, and hot-pressing conditions to achieve the precise construction of efficient and stable MEA.
(iv) Advanced GDL.GDL is the major site for water/gas conversion and electron transport.Therefore, the optimization of GDL can be realized by improving its porosity and conductivity.Rational adjustment of the pore structure and size of the GDL will increase the electron transfer efficiency and accelerate the water/gas conversion, enabling the excellent performance of PEMWE.Titanium is the only choice for the GDL at the anode side, due to the harsh operating environment.Titanium screen mesh, sintered powder, felt, and foam are currently used as GDL in industry.It is worth noting that the major challenges for Ti-based GDL its degradation (e.g., surface passivation and hydrogen embrittlement), which will result in the increasing of contact resistance and the attenuation of mechanical strength.To prevent the GDL from degradation, it is often coated with a noble metal (e.g., Au and Pt) protective layer to meet the industrial demands of performance and lifetime.However, this will also increase the cost of PEMWE.Thus, searching for the coating materials with low cost, high conductivity, high corrosion resistance, hydrogen embrittlement resistance is the unremitting pursuit.
(v) Optimized BPs.The functions of BPs include conducting electrons, providing water flow path, separating O2 and H2, supporting electrolyzer and offering heat conduction.Therefore, the plate material demands high electrical conductivity, corrosion resistance, low gas permeability, strong mechanical strength and high thermal conductivity.Titanium with noble-metal coating (e.g., Au and Pt) is still the best choice.Introducing the noble metals significantly increases the cost.Therefore, finding the solution to reduce cost is the major difficulty for the development of BPs.The most effective way is to reduce the amount of noble metal in the coating material by improving the coating composition or preparation process.To further reduce the cost, non-noble-metal coatings (e.g., Zr and Nb coatings) are also being developed.Another direction to optimize the BPs focuses on the reasonable design of its flow field.Currently, the flow-field design of BPs mainly includes the following three types: (a) using the titanium meshes as flat separator sheet and the porous titanium sintering as GDL; (b) using the thick titanium plate with etched channels; (c) using the titanium plate which is stamped with channels.However, the above flow fields either have high processing cost, or cause stress distortion, leading to the failure to scale up for practical application.Therefore, reasonable designing the flow field of BPs is also a hot issue.

Figure 2
Figure 2 (a) Energy demand of water splitting at different temperatures.(b) Simulated cell voltage of a PEMWE.

Figure 3
Figure 3 Stack structure and key materials of PEMWE.

Figure 4
Figure 4 (a) Chemical structures of various PFSA polymer electrolyte membranes.(b) Schematic diagram of three MEA configurations in PEMWE.

Figure 6
Figure 6 (a) Schematic of thin titanium GDL functions.(b) Scanning electron microscope (SEM) images of well-defined pore GDL and traditional Ti-felt.Reproduced with permission from Ref. [57], © Elsevier B.V. 2019.(c) Performance comparison curves between different GDLs.(d) Screenshot of visualization video shows the electrochemical reaction phenomenon in one pore and schematic.(a), (c), and (d) Reproduced with permission from Ref. [29], © The Royal Society of Chemistry 2017.

Figure 7
Figure 7 (a) Manufacturing processes of Au-coated stainless steel BPs.Reproduced with permission from Ref. [71], © Elsevier B.V. 2018.(b) Scheme of Pt/Ti and Pt coatings deposited on stainless steel and photos of Pt/Ti/stainless steel and Pt/stainless steel after the corrosion tests.Reproduced with permission from Ref. [31], © Elsevier B.V. 2016.(c) Parallel flow channel, pin flow channel and pin flow channel with GDL.(d) Polarization curves results of PEMWE with different cathode plates.(c) and (d) Reproduced with permission from Ref. [73], © Elsevier B.V. 2018.

Figure 8
Figure 8 (a)-(e) Schematic of different flow fields in PEM electrolytic cell.(f) Polarization curve at the outlet channel on anode side for different flow field patterns.Reproduced with permission from Ref. [74], © Elsevier B.V. 2018.

Figure 11 (
Figure 11 (a) Schematic of the synthesis of porous IrxMo1-xOy solid solution.(b) Current densities of samples with different Mo contents at 1.5 V vs. NHE.(a) and (b) Reproduced with permission from Ref. [145], © Springer Nature 2021.(c) TEM image and (d) polarization curves for OER of 38 wt% Ir nanoparticles loaded on WxTi1-xO2.(e) Durability tests of cells at a constant current density of 1.5 A•cm -2 .(c)-(e) Reproduced with permission from Ref. [155], © IOP Publishing 2018.

Figure 15
Figure15 Future directions of key material research for PEMWE.