A brief introduction of electrode fabrication for proton exchange membrane water electrolyzers

Proton exchange membrane water electrolyzer (PEMWE) is a major enabler of green hydrogen production. The development of water electrolyzers is a vital step in driving the progress of a hydrogen-based economy. The system inside the electrolyzer is a zero-gap cell featuring low ohmic resistance and boosted mass transport, leading to higher energy efficiency and minimized capital cost. Besides, utilizing PEM in the electrolyzer for sustainable hydrogen production enables the system to perform with many advantages, including superior energy efficiency, higher hydrogen purity, and high flexibility. Therefore, as PEM electrolyzers continue to evolve, sustainable hydrogen production on a larger scale will be realized in the near future. This review summarizes the status quo of PEM water electrolyzers in the past four years. We will start with a brief introduction of the core of a water electrolyzer, namely the membrane electrode assembly (MEA), which will be followed by an introduction of fabrication methods of MEA, including CCM methods, catalyst-coated electrode methods, and other innovative fabrication methods. Next, we will summarize recent attempts to modify electrodes and membranes in MEAs to promote the performance of PEMWE. Subsequently, catalyst development for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) in MEA is discussed, highlighting novel HER/OER catalysts and strategies to reduce the content of noble metals. Lastly, conclusion and perspectives are provided to present a blueprint to inspire the future development of PEMWE.


Introduction
In response to the increasingly urgent need for energy sustainability and growing environmental problems, significant amount of resources has been allocated to the research and development of renewable energy storage and conversion. Hydrogen, a promising clean fuel with a high specific energy (140 MJ kg −1 ), has an important role in achieving carbon neutrality envisioned by the international community. Thus, particular attention has been paid to replace conventional hydrogen production via steam methane reforming (SMR) at high temperature and under high pressure that emits flue gas with a carbon-free alternative.
Hydrogen production from water splitting powered by electricity has emerged as a prospective alternative to the fossil-fuels-based technology that enables hydrogen synthesis in milder conditions without producing environmentally detrimental by-products [1]. The two half-reactions in the electrochemical water splitting (or water electrolysis) are the cathodic hydrogen evolution reaction (HER) and the anodic oxygen evolution reaction (OER). The currently well-known set-ups for water electrolysis include solid oxide electrolysis cell, polymer electrolyte membrane electrolysis cell (PEMEC), anion exchange membrane electrolysis cell, and alkaline electrolysis cell [2,3], the most prominent of which is PEMEC as its life cycle assessment, when limited by the reaction kinetics at the electrode. Hence, the incorporation of electrocatalysts in PEMWE is vital. By and large, there are two common categories of methods of incorporating electrocatalysts, namely, CCE and CCM. The former is the most robust preparation strategy involving coating catalyst on the surface of GDLs using brush or spray coating and the latter is more exclusively used to deposit two CLs on both sides of the membrane, either directly via spray coating or indirectly by transferring the catalyst to the membrane surface through a decal process [28]. In this section, common MEA fabrication strategies for PEMWE, followed by some innovative approaches, will be introduced to serve as a cornerstone for further discussion on the performance of MEA for hydrogen production.  [28], Copyright (2020), with permission from Elsevier. (b) In-situ visualization of hydrogen evolution reaction in the conventional CCM-based electrode and novel thin GDE. Reprinted from [41], Copyright (2018), with permission from Elsevier. (c) Plots of mass activity at 1.6 V and effective Pt loading versus Pt film thickness on TT-LGDLs based on data from [41]. (d) RCL values of R2R-coated GDEs with different I/C ratios and spray-coated GDE with (w/OL) and without overlayer (w/o OL). Reprinted from [45], Copyright (2021), with permission from Elsevier. electrolyzer fabricated using CCM. In figure 2(b), the novel thin GDE exhibited more HER active sites compared to the thicker conventional CLs in conventional CCM set-up due to the larger electrical resistance caused by the lack of uniformity of catalyst distribution in the latter. With novel thin GDE at ultra-low catalyst loading, catalyst utilization and Pt catalyst mass activity increased, resulting in more hydrogen bubble generation occurring at the rim of the observation hole. At low CL thickness of 15 nm and low effective Pt loading of 0.032 mg cm −2 , its mass activity could reach as high as 58 times that of the conventional CCM at 1.6 V, 80 • C and 1 atm (figure 2(c)).
To date, many techniques are available to manufacture GDE [42,43], and hence, an effective comparison between them is essential in providing meaningful insights for further improvement in GDE fabrication. A recent study has made a comparison between CCM by dry spraying and GDEs by airbrush, screen printing, doctor blade, drop casting, and inkjet printing to investigate the difference in CL microstructures, mass transfers and charge transfers [44]. In the study, ionomer coverage on Pt/C decreased in the following order: airbrush > screen printing > dry spraying > drop casting > inkjet printing > doctor blade. This result was consistent with the higher electrochemically (EC) active surface area (ECSA) values of GDEs fabricated by airbrush and screen printing, demonstrating higher catalyst utilization at higher ionomer coverage. Upon considering charge and mass transport properties, the three best MEAs were obtained by airbrush, screen printing and ink jet methods as they showed optimal balance of properties such as ohmic resistance, porosity, and covered Pt. Particularly, GDE fabricated by screen printing exhibited a high fraction of ionomer covered Pt/C and good charge and mass transport properties, delivering the best performance at high currents.
To enable the mass production of MEAs, a continuous production process is needed. Considering the relatively robust mechanical properties of the diffusion media and the low tendency of membrane swelling during CL coating, Mauger et al applied the continuous roll-to-roll (R2R) process in CCE-MEA manufacturing, where the CL was coated using a slot die with a slot gap of 250 µm and a coating width of 8 cm, followed by drying at 80 • C in the air flotation ovens in the coating line [45]. By adjusting the ionomer-to-carbon ratio (I/C) in the coating ink and drying parameters, a CL with an ionomer-rich top surface was prepared, eliminating the ionomer overlayer that is typical in spray-coated GDEs and reducing the production cost. Moreover, the protonic sheet resistance of the CL (R CL, consisting of the bulk CL resistance and the catalyst-membrane interfacial resistance) of R2R-coated GDEs were lower than that of spray-coated GDEs with an ionomer overlayer. While a slight decrease in R CL with increasing I/C ratio was observed (figure 2(d)) resulting from the increased proton conductivity, the decreased resistance did not improve their performance significantly, possibly due to the excessive filling of the catalytic pore space and hence, 0.9 I/C was considered optimal in forming a good interface while preventing inhibition of gas diffusivity.

Catalyst-coated membrane (CCM)
Coating the CLs on the membrane is conducive for close contact between the catalyst and electrolyte membrane to form a strong interfacial connection, which usually improves MEA performance while reducing catalyst loading [46,47]. Hence, a CCM is usually used to fabricate a MEA that outperforms CCE-MEAs owing to its low contact resistance [48]. In a comparative study, Moghaddam and Easton fabricated both CCM-MEA and CCE-MEA by spray coating and demonstrated that CCM-MEA generally outperformed CCE-MEA in fuel cells owing to the more coherent surface structure and lower charge transfer resistance [49]. Similarly, Bühler et al showed that CCM-MEA had better kinetics than the CCE-MEA in the PEM water electrolysis despite exhibiting an inferior polarization behavior and inferior performance reproducibility to CCE-MEA for current densities above 750 mA cm −2 [50].
Notwithstanding the possibility of direct coating of a membrane, the decal process, involving an indirect transfer of CL to the membrane, is the representative coating approach for CCM that is most suited for MEA mass production in the manner shown in figure 3(a). The catalyst ink is first prepared by mixing the catalyst, an ionomer, and suitable solvents, followed by coating the ink mixture on the decal substrate (e.g. Kapton film or Teflon film) using a doctor blade to spread the ink uniformly on the substrate [51]. After drying (which can affect the formation of effective three-phase boundary during operation), the CL is cut and sent for hot-pressing to tighten the adhesion of the anode, the cathode, and the membrane. In the end, the decal substrate is peeled off from the electrodes. Despite being the most favorable for mass production, several shortcomings of this process need to be resolved, namely [51,52]: (1) mitigating the incomplete transfer of the catalyst to the membrane; (2) optimizing the temperature and pressure during hot-pressing as the ionomer and membrane may not withstand high temperatures; (3) delicate design of the catalyst ink with suitable composition to improve catalyst utilization; and (4) engineering the microstructure of the CL to expedite the transport of the gaseous product during electrolysis.
The property of the decal substrate is the crux for an efficient transfer of the CL; thus its material selection is important. From a study involving polypropylene (PP), low-density polyethylene, silicone coated polyethylene terephthalate, polytetrafluoroethylene (PTFE), reinforced-PTFE, and Kapton films, Akella et al implied that PP is the most suitable material for decal substrates [53]. PP was shown to possess superior thermal stability during hot-pressing and higher hydrophobicity that checked the coagulation of the catalyst ink slurry cast. Importantly, 100% catalyst transfer to the Nafion membrane was observed with PP.
Low-temperature decal transfer was also developed to further reduce the cost of the decal process and further enhance the MEA performance by tuning of properties of the membrane material. For example, Kyeong et al synthesized new aliphatic backbone-containing sulfonated poly (arylene ether sulfone)s (HexX), which was suitable for fabricating MEAs via low-temperature decal transfer [54]. Introducing the hexyl groups between biphenyls improved the flexibility of the backbone, which reduced the glass transition  [51], Copyright (2011), with permission from Elsevier. (b) Single-cell performance of Hex-BP-2 in 1000 cycles of the wet-dry cycling test and (c) linear sweep voltammetry (LSV) curves of Hex-BP-2 before and after 1000 cycles of the wet-dry cycling test. Reprinted from [54], Copyright (2022), with permission from Elsevier. (d) Schematic diagram of large-scale R2R catalyst coating. Reprinted from [55], Copyright (2020), with permission from Elsevier. temperature (T g ) of the membrane, facilitating the easy transfer of the CL from the substrate to the membrane due to the rubber-like state of the polymer near T g . With further blending of HexX with biphenyl-containing sulfonated poly (arylene ether sulfone)s (BPSXs) to form a blend membrane (Hex-BP), the stability of MEA fabricated from Hex-BP-2 (consisting of 50.93 mol% Hex and 49.07 mol% BP) was greatly enhanced to withstand over 1000 wet-dry cycles ( figure 3(b)). Besides, low H 2 crossover current densities of 0.12 and 0.13 mA cm −2 were maintained for Hex-BP-2 before and after 1000 wet-dry cycles respectively (figure 3(c)).
Although membrane swelling is effectively eliminated by direct application of a dry CL on the membrane, an additional step of transferring CL from the decal substrate to a membrane is required, which is not desirable for mass production. To this end, the use of a R2R process for large-scale CCM-MEA fabrication was studied by Park et al to replace the decal process [55]. A large-area PEMWE MEA is produced by directly coating the IrO 2 catalyst on the membrane through a R2R process, where sheets of membrane film were spliced to create a long web to cover the path of the R2R system and a slot die was used for continuous coating process, followed by in-line drying in the hot-air dryer (figure 3(d)). The production throughput was increased by over 500 times at a low temperature in contrast with the laboratory-scale spray coating. The performance of the resultant water electrolyzer is also comparable to that with spray-coated CCMs, with a cell voltage of 1.91 V at a current density of 2 A cm −2 .

Other fabrication methods
The common MEA fabrication methods show the tremendous potential in the current development of MEA manufacturing. Notably, MEAs prepared by the CCM methods have lower interfacial resistance between the membrane and CL than those prepared by the CCE methods, but CCMs can easily form cracks in the CL due to the vaporization of the solvent in the catalyst ink as well as the swelling of the membrane [56,57]. In view of the limitation of traditional methods, novel MEA fabrication methods are desired to overcome these challenges.
Notably, the direct membrane deposition (DMD) is first proposed by Klingele et al for fuel cell preparation, in which dispersed polymer electrolyte was directly deposited on the CLs of the anodic and cathodic electrodes [58]. Stähler et al first applied the DMD method in fabricating PEMWE MEA, where the cathodic CL, a membrane with 20 ± 2 µm, and the anodic CL were successively slot-die-coated layer by layer on each other [59]. This allowed free selection of the PTL in the MEA, and simultaneously, the formation, properties, and composition of each layer can be controlled. As a result, the reduced interfacial resistance was ascribed to good contact between different layers. However, the thinness of the membrane and the quick corrosion of carbon are set to challenge industrial operation of this PEMWE MEA.
To further improve the DMD approach, Holzapfel et al prepared DMD-MEA with a titanium substrate at the anode side and suitable membrane thicknesses of 60 µm, 200 µm, and 230 µm for realistic operation, in which the membrane was asymmetrically deposited solely on the cathodic substrate instead of using the symmetric DMD design (figure 4(a)) [60]. In this method, PTFE frames were used to center the MEA and the configuration is shown in figure 4(b). When compared with CCM and porous transport electrode (PTE)-type MEAs, DMD-MEA exhibited improvement in the electrochemical performance (figure 4(c)), with membrane thickness of 60 µ m being ideal owing to the reduced kinetic, ohmic, and mass transport losses.
However, in the conventional DMD process, the catalyst particles can fill the pores of microporous layers during catalyst coating in GDE fabrication, which will influence the surface morphology of the CL, further affecting the polymer morphology after dispersing the ionomer on GDE. A novel approach was then proposed by Yang et al, where Nafion ionomers were directly coated on both cathode and anode CLs, followed by hot pressing of both ionomer-coated CLs and attaching GDLs on both CLs [61]. In this MEA, crack formation on the CL surface was mitigated (figure 4(d)). Besides, expanded PTFE (ePTFE) film was attached on the ionomer layer of ionomer-coated cathodic CL before hot pressing with the ionomer-coated anodic CL to provide meliorative mechanical properties and stability of the resultant MEA. This enhancement was verified by Xing et al when ePTFE was added to one of the ionomer-coated GDEs (figure 4(e)) [57]. In addition, Shang et al impregnated a porous ePTFE matrix with crosslinkable poly (phenylenesulfonic acid) (cPPSA) to form a composite cPPSA-ePTFE membrane, enhancing the proton conductivity of the membrane [62]. The thickness of the ePTFE-reinforced membrane deposited on the electrode will be a vital consideration in future research to reduce interfacial resistance.
CCM and CCE methods are still two dominating classes of fabrication methods for MEA production in mainstream studies despite the tremendous endeavors made in enhancing these approaches and even replacing them. By exploring the main contributors to the elevated performance brought by various innovations and working towards fully addressing the thorny challenges in both material synthesis and device fabrication in the industrial setting, more opportunities to upscale MEA fabrication can be created.

Electrode and membrane modifications
The fabrication procedure of the MEA is influential to the performance of water electrolysis. By delicate design of the fabrication procedures, researchers emphasize more on maximizing the contact area between different layers and finding ways to promote mass and charge transfer by tuning interlayer interactions. However, electrodes and membranes can be creatively modified to further optimize the property of MEAs for  [61]. Copyright (2020) American Chemical Society. (e) Schematic illustration of fabrication of reinforced MEA by combining the ePTFE and DMD approaches. Reprinted with permission from [57]. Copyright (2022) American Chemical Society. efficient water electrolysis. In this section, modification strategies that have been employed on electrode and membrane to ameliorate the PEMWE performance during recent years will be reviewed and discussed.

Electrode modification
Modifying the morphology of the electrode can help to establish an efficient charge transport pathway and reduce the loading of catalysts. For instance, by changing the morphology to a 1D nanostructure, a large aspect ratio of the nanostructure facilitated the improvement of electric conductivity [63]. Xie et al fabricated an integrated HER electrode design comprising in-situ grown Pt nanowires (PtNWs) on the thin Ti LGDLs (TT-LGDLs), which was chemically synthesized on Ti substrates via a surfactant-free process [64]. Compared with the conventional GDE (figure 5(a)), the electrode thickness was successfully reduced from ∼300 µm to ∼25 µm, and the application of PtNW GDE improved the catalyst utilization by exposing more catalyst surface to contact with electrolyte to increase the concentration of TPBs. The smooth LGDL surface was completely covered by a CL with a small loading of 0.045 mg cm −2 and this morphology was maintained after a stability test of 10 h at 100 mA cm −2 . Moreover, a faster gas bubble removal rate was observed in PtNM/Ti electrode due to the increased surface roughness, thus reducing bubble coverage. As a result, this electrode with catalyst loading of 0.2 mg Pt cm −2 reached the lowest cell voltage of 1.643 V and high cell efficiency of 90.08% at 1 A cm −2 in the PEM electrolyzer cell. A similar morphology modification strategy was employed by Kang et al, in which a novel 2D-patterned electrode (figure 5(b)) exhibited boosted catalyst utilization and mass activity [65]. This 2D-pattern electrode consisting of CL stripes (with the optimal set-up comprising 1.0 mm catalyst stripes and 0.3 mm gap) was synthesized using a shadow mask on the Nafion membrane, followed by spraying of the catalyst ink. In this system, the edge effect on the anode was induced by the membrane property, proton conduction path, internal potential distribution, and catalytic sites distribution to assist the transport of generated protons in OER to a further region to reach the cathode HER active sites. The optimal 2D-pattern electrode saved more than 21% of the anodic noble metal material while maintaining the same performance of the system where the anode fully covered the membrane.
As mentioned, OER will restrict the OWS owing to its sluggish kinetics. In the meantime, the co-production of hydrogen and oxygen as well as the use of PEM restrict the direct utilization of sporadic and fluctuating renewable energy, causing the formation of a hazardous H 2 /O 2 mixture [66]. Although PEM is required to prevent gas crossover, its high cost is the main consideration, and hence, finding a way to decouple the OER and HER becomes significant, through which even the use of membrane can be eliminated. To this end, several solid-state redox mediators have been reported up to now, such as pyrene-4,5,9,10-tetraone [67], polyaniline [68], nickel hydroxide [69], and polytriphenylamine [70]. By incorporating them as auxiliary electrodes, the ion exchange between the anode and cathode is mediated, enabling the production of hydrogen and oxygen at different time and space [71]. For example, Liang et al create an innovative and low-cost hydrated Turnbull's blue analog (TBA) electrode, which served as a solid-state redox mediator (figure 5(c)) [66]. The CuFe-TBA was synthesized by a facile co-precipitation method. There were two steps in the decoupled water electrolysis, the first of which involve oxygen production and cathodic CuFe-TBA reduction, followed by hydrogen production associated with the anodic oxidation of reduced CuFe-TBA. As a result, the redox mediator showed a reversible capacity of 80 mAh g −1 at a current density of 0.5 A g −1 and achieved high-rate performance with a respectable capacity of 42.7 mAh g −1 at a high current density of 120 A g −1 (figure 5(d)). They fabricated a PEM-less electrolyzer assembled from a Pt-coated Ti mesh electrode, a RuO 2 /IrO 2 -coated Ti mesh electrode, and a CuFe-TBA film electrode, which was investigated in 0.5 M H 2 SO 4 by chronopotentiometry. The OER process gave an average cell voltage of around 0.68 V at an applied current of 5 mA for step 1, while step 2 exhibited a cell voltage of 0.9 V, summing up to obtain an overall voltage of 1.58 V (figure 5(e)). Moreover, the cell voltage of step 1 and step 2 remained stable after a 170 h cycling test, indicating the superior durability of this novel electrode.
Since OER kinetics is the main obstructor in water electrolysis, improving mass transport and cost efficiency in the PTLs on the anode is thus indispensable in removing the barrier to better MEA performance. By varying the thickness and porosity of fiber-sintered titanium PTL on the anode, Peng et al studied the impact of bulk and interface transport properties of PTLs on PEMWE performance at ultra-low Ir loading (0.05 mg Ir cm −2 ) [72]. They demonstrated that the optimal kinetics was reached in the PTLs with intermediate porosity and interfacial area due to the balance between oxygen removal and water accessibility, reducing the resistance for electrons to transport through the interface. Besides, the mass transport resistance (or overpotential) can be reduced by increasing PTL porosity and reducing through-plane tortuosity. Furthermore, the influence of PTL in different forms was studied by Bühler et al, where the anodic porous fiber-sintered PTL outperformed the powder-sintered one for loading above 1.0 mg IrO2 cm −2 owing to lower ohmic loss [73].
In developing modified PTLs, Stiber et al produced novel PTLs by diffusion bonding of a Ti porous sintered layer (PSL) on a Ti expanded metal mesh (PSL/mesh-PTL), which eliminated the use of flow field in the bipolar plates [74]. The PSL that possessed a uniform pore size distribution was fully bonded to the mesh-PTL. This new PTL allowed efficient gas/water management at high current densities. The high liquid and gas permeability contributed to ameliorated performance at a high current density of 6 A cm −2 under extreme conditions of either 90 • C or 90 bar H 2 output. At a nominal load of 4 A cm −2 , higher efficiency was also observed for PSL/mesh-PTL compared to the common mesh-PTL that exhibited significant energy loss due to mass transport.
Owing to the corrosive environment at the anode, the PTL typically consists of titanium at the anode. However, the protective passive layer covered on the Ti surface (oxide/hydroxide) leads to high surface contact resistance originating from its lack of electrical conductivity [75,76]. This results in loss of energy, which is exacerbated during the intensive operation at large current densities [77,78]. To circumvent the issue of excessive formation of the passivation layer, a common way is to treat the surface by coating with precious metal (e.g. Pt) as a protective layer through magnetron sputtering or alloying [79][80][81][82], but this method increases the overall cost of the electrode. To this end, Bystron et al proposed an alternative method of reducing the extent of passivation based on the chemical etching of Ti in the HCl aid and investigated the influence of appropriate treatment of Ti surface on the surface contact resistance and the cell performance [77]. The surface of Ti was modified by HCl etching to form a Ti hydride underlayer to grant the metal surface improved resistance to excessive passivation. More importantly, the surface contact resistance was reduced after the treatment, with its tendency to increase suppressed even during water electrolysis for more than 100 h at 1 A cm −2 . Besides, the passivation can also be suppressed by depositing a suitable catalyst that also improves the electrochemical performance. For instance, Doan et al applied spray coating and thermal treatment, and a thin layer of IrO 2 /TiO 2 catalyst was formed on the surface of Ti PTL to prevent surface passivation [76]. Simultaneously, the performance of PEMWE was enhanced as the cell ohmic resistance was reduced due to the prevention of TiO 2 formation.

Membrane modification
The well-established perfluorosulfonated acid (PFSA) membranes represented by the Nafion membranes are characterized by high proton conductivity, excellent mechanical strength, and good chemical stability [83]. However, there are efficiency and safety issues (especially with hydrogen flammability limit of 4% H 2 concentration in O 2 at ambient conditions) since the PEM allows significant permeation of gaseous products, especially when the thickness of the membrane is decreased to reduce cell resistance [84,85]. Moreover, degradation of the membrane will largely deteriorate the cell performance and durability because the membrane is prone to chemical attack by trace radical species such as hydroxyl radicals in the presence of H 2 , O 2 , and Pt [86]. The degradation of membranes can then cause kinetic deactivation and irreversible resistive losses that are one order of magnitude higher compared to the kinetic losses [87]. Hence, developing modification strategies of enhancing the lifetime and mitigating the gas crossover issue is instructive to further promote the performance of PEMWE.
In addition to PFSAs, hydrocarbon membranes are promising candidates since they are fluorine-free and have less gas crossover and lower production costs [88,89]. Nevertheless, they are less stable because the carbon-hydrogen bonds are weaker than the carbon-fluorine bonds of PFSA [90]. Klose et al prepared a fully hydrocarbon-based MEA with sulfonated poly (phenylene sulfone) (sPPS) acting as both membrane material and ionomer binder in the electrode for water electrolysis, where sPPS had an equivalent weight of 360 g eq −1 and ion exchange capacity of 2.78 meq g −1 [91]. The sPPS-MEA reached a higher current density of 3.48 ± 0.03 A cm −2 at 1.8 V (figure 6(a)), outperforming the MEA using Nafion 115 with the same catalyst loading (1.5 A cm −2 at 1.8 V), which is the result of the 60% lower high-frequency resistance (HFR) (161 ± 7 mΩ cm 2 ) for sPPS. With hydrogen crossover current density below 0.3 mA cm −2 , pure sPPS-membranes exhibited only a third of hydrogen crossover current density of Nafion 115 membranes in a fully humidified surrogate test, enabling safe operation at lower current densities and resulting in a larger predicted operational current density range of sPPS-MEA (figure 6(b)). However, the durability of sPPS-MEA hindered its commercial adoptability due to membrane softening caused by high water content and degradation caused by increased crossover. Likewise, Park et al developed a hydrocarbon-based membrane and ionomer for PEMWE, using sulfonated poly (arylene ether sulfone) with a degree of sulfonation of 50 mol.% (SPAES50) synthesized via condensation polymerization [92]. They evaluated the influence of different membrane thicknesses (20,30, and 40 µm) and ionomer contents (5, 10, 20, 30 wt%). The 5 wt% and 10 wt% ionomer contents in the anode and the cathode, respectively, gave the best PEMWE performance owing to the optimal charge transfer resistance since the cell performance was largely influenced by the blockage of the active sites by the excessive ionomer content. Moreover, the hydrogen crossover issue was resolved in a SPAES50 membrane with a thickness of 20 µm as smaller channel size reduced hydrogen permeation, indicating its potential in replacing the thicker Nafion 115 with a similar hydrogen crossover. By optimizing the membrane properties, the current density reached 1069 mA cm −2 at 1.6 V, exceeding those of Nafion-based PEM with thicknesses of 25 µm and 125 µm.
Moreover, there are more novel membrane materials emerging in recent years, such as adding an appropriate amount of nanofillers into the initial polymer material to make composite membranes to enhance the production performance [93][94][95][96]. For example, graphene-based nanofiller is very versatile, being a popular reinforcing agent in enhancing mechanical properties, thermal stability, electrical conductivity, and gas barrier property of polymer matrices in a wide range of applications [97]. Mahdi et al prepared a Nafion/sulfonated graphene oxide (SGO) nanocomposite membrane for deposition of the catalyst with low loading [98]. The nanocomposite membrane was synthesized by solution casting, followed by coating Pt nanoparticles on the surface (N.SGO.Pt) through the electroless deposition method by varying the mass ratio of sulfanilic acid/GO and Pt salt solution concentration. The cross-sectional image (figures 6(c) and (d)) suggested the successful deposition of Pt catalyst on pure Nafion membrane and optimal N.SGO 1 .Pt 5.2 nanocomposite membrane with slightly higher Pt layer thickness. This is because the high level of ion exchange site, -SO 3 H groups, introduced by SGO, provided more nucleation sites for Pt 2+ ions to adsorb and grow into Pt nanoparticles, thereby leading to higher penetration of Pt nanoparticles into the membrane and better distribution on the membrane surface.
In addition, the use of MXene in PEMWE is noteworthy since MXene as a rising star of 2D material featured for its excellent conductivity, high strength, and hydrophilic nature [99,100]. Considering p-type semiconductor Cu 2 O is a potential filler with superhydrophilic nature to enhance the ionic conductivity, MXene coupled with Cu 2 O was prepared by Waribam et al to serve as filler in the sulfonated polyether ether ketone (SPEEK), a hydrocarbon-based polymer, to make a novel PEM for water electrolysis [99]. This hybrid MXene-Cu 2 O/SPEEK membrane was fabricated by a simple solution casting method to disperse the MXene-Cu 2 O composites into the SPEEK matrix (figure 6(e)). In detail, MXene was synthesized by HF-etching the Ti 3 AlC 2 MAX, followed by adding copper precursors to obtain MXene-Cu 2 O composites. Through constructing a hybrid membrane, the proton conductivity of the membrane was enhanced compared to the pristine PEEK membrane, with the highest value of 0.0105 S cm −1 achieved in 4%MXene-Cu 2 O/SPEE (figure 6(f)), which was due to the uniform distribution of composite for the formation of hydrogen bonds, bringing about higher H 2 evolution volumetric flow rate. The formation of hydrogen bonding between the SPEEK polymer and MXene-Cu 2 O composite was also the key to reducing the swelling area in the membrane.
Besides material selection of the membrane, the selection of membrane fabrication method, which relies on the polymer chosen and the desired structure [101], is also critical in improving PEMWE efficiency by reducing interfacial resistance, degradation and cost. For instance, Siracusano et al compared the Aquivion membranes with an equivalent weight of 980 g eq −1 and a thickness of 90 µm manufactured using extrusion and casting methods [102]. While both had low gas permeability at current density of 1-4 A cm −2 and a differential pressure of 20 bar and their respective MEAs exhibited a voltage efficiency of over 80% (versus thermoneutral potential) at high current densities (3-4 A cm −2 ) and high temperatures (80 • C-90 • C), the recast membrane showed smaller surface roughness, increasing the contact area between the ionic cluster of the membrane and CLs. Hence, in the activation region of the polarization curves, the recast membrane slightly outperformed the extrusion membrane. However, the recast membrane had higher polarization resistance but lower series resistance at 1.8 V and in low-temperature operating conditions. Generally, attempts have been made to optimize the physicochemical and mechanical properties of both PEM and the electrodes through various modification strategies and tuning of the operating conditions. While novel membranes are introduced to replace the conventional PEM, modifications of the electrode discussed here are mostly focused on enhancing the interaction between PTL and CL and tuning of the properties of PTLs or GDLs. In particular, regulating the porosity, durability, and conductivity of PTLs is overriding since it is the primary location for mass and electron transport.

Catalysts for PEM water electrolysis
Currently, the cost of catalysts in the PEMWE system is relatively small, accounting for around 5% [11,103,104]. Nonetheless, as the system scales up to the MW or GW-based stack power, the contribution of the catalyst in the production cost becomes more significant while the contribution of balance-of-plant cost is expected to be minor [103]. As such, the transition from the gold-standard platinum group metal (PGM) catalysts to PGM-free or PGM with noticeably reduced loading is pressing. The industrial benchmark loadings of Pt/C catalyst for HER and IrO 2 for OER are 0.5-1.0 mg Pt cm −2 and 2 mg Ir cm −2 , respectively [105]. The amount of platinum loading can be feasibly reduced (by an order of magnitude) without significantly reducing the performance due to the high kinetics of HER on the cathode side in the acidic environment, but it is difficult to reduce the loading of iridium catalyst without significantly reduce the efficiency at the same current density since the anode is usually the kinetic bottleneck [106,107]. Numerous efforts have been devoted to reducing the catalyst loading as well as employing catalyst other than PGMs in the MEA system to obtain a feasible result under practical device conditions, which will be discussed in this section.

Catalysts for HER
Improving HER catalysis is typically achieved through either reduction in PGM loading or replacement of PGMs with either transition metal-based catalysts or molecular catalysts. Reduction in PGM loading can be achieved by optimizing CL thickness and employing suitable support material, whereas transition metal-based catalysts are improved by morphological control, employing porous shell structures, sulfidation, modification and electrodeposition.

Optimizing PGM content
Many synthesis strategies have been deployed to minimize Pt catalyst loading, such as atomic layer deposition, electrodeposition method, sputter deposition method, modified thin-film methods, wet chemistry method, and annealing [108][109][110]. For example, Wang et al successfully reduced Pt catalyst loading to only 0.003 79 mg cm −2 by using direct electron beam (e-beam) evaporation deposition on Nafion membrane, forming a highly uniform Pt thin film with low surface roughness and an optimal thickness of 6 nm that was evenly distributed on the membrane [111]. Its current density reached 500 mA cm −2 at 1.64 V under 80 • C, which was competitive with Pt/C-based MEA. Its low thickness was sufficient to lower the mass transport loss, thus reaching a balance between maximizing the amount of active sites and preventing blockage of proton diffusion channels in the membrane by the deposited thin film. When the temperature was increased from 20 • C to 80 • C, the catalytic activation barrier could be further reduced while increasing proton transportation rate, dramatically increasing the current density from 483 to 855 mA cm −2 at 1.75 V.
HER on PGM catalysts can also be improved by improving the carbon support, particularly through doping. Kumar and Himabindu reported a MEA containing boron-doped carbon nanoparticles supported 30 wt% Pd as the cathode catalyst synthesized by chemical vapor deposition method, which exhibited a high stability of up to 500 h at 1 A cm −2 under a temperature of 80 • C [112]. B doping improved the conductivity of the carbon support by increasing the density of charge carriers. The cell voltage of the fabricated MEA decreased while the cell efficiency increased (up to 72%) at higher temperatures up to 80 • C owing to the promoted HER kinetics at a higher temperature, thus delivering similar performance to 30 wt% Pt supported on carbon black and having higher cost efficiency than Pt.

Transition metal-based catalysts
One way to replace PGM is by using corrosion-resistant noble metals such as Ag. A robust catalyst that exhibits insignificant degradation under high operational cell voltage can be produced by controlling the morphology and subsequently, the exposed facets of the catalyst, proton availability and H atom coverage can be tuned to significantly improve HER efficiency. Mo et al prepared a series of silver catalysts with different morphologies, namely nanocubes, nanowires, and nanospheres [113]. While the Ag catalyst was inferior to the Pt/C catalyst over a potential range of −1.0 to −2.0 V in the electrolyzer, the Ag nanocubes delivered the highest current density (3.5 A cm −2 ) when a more negative potential of −2.5 V was applied, exhibiting a potential for upscaling. More H atoms were readily absorbed on the metal surface at a more negative applied potential, causing a change in the rate-determining step (RDS) from the Volmer step to the Tafel step. The Ag-H bonds at the Ag nanocube surface were weaker than Pt-H bonds, favoring H migration and the H 2 formation, thereby increasing the H 2 productivity.
Transition metal-based catalysts, including transition metal chalcogenides, phosphides, nitrides, and alloys, are also promising candidates for HER catalysis, having demonstrated decent HER activity in the acidic environment [114]. Besides, early transition metals are more abundant (with around 2-3 orders of magnitude higher estimated concentration in Earth's crust) and cheaper than PGMs [106,115]. On account of unfilled d orbitals in transition metals, their properties are prodigiously different from other metals [116]. Feng et al prepared ultrafine tungsten carbide (WC) nanocrystals encapsulated in porous N-doped carbon nanosphere (NC) through a facile and scalable one-step pyrolysis method to obtain WC@NC [117]. As demonstrated in SEM images (figures 7(a) and (b)), WC@NC had a uniform, grape-like, monodisperse nanosphere morphology. Besides, the WC nanoparticles were wrapped in N-doped carbon shells which was critical in confining the WC nanoparticles, thus preventing active ingredients from leaching in the harsh operating environment while improving electrical conductivity. This structure was favorable in optimizing the H-binding energy, thus boosting its HER activity and stability in both acid and alkaline environments when tested in the half-cell system. In the single-cell test for performance evaluation, a CCM was prepared for MEA with WC@NC and IrO 2 acting as the cathode and anode, respectively. The best performance was achieved at 80 • C, with a voltage of 2 V delivering a current density of 0.78 A cm −2 ( figure 7(c)). However, low durability was observed at over 12.5 h in this MEA system, which might be due to metallic cation poisoning, CL detachment, and catalyst dissolution. Similarly, the significance of incorporating carbon materials in transition metal-based HER electrocatalysts was also demonstrated by Morozan et al, where combining FeMoS with carbon nanotubes resulted in higher HER activity, exhibiting its suitability as a cathode catalyst in MEA for PEM electrolyzer single cell operation [118].
Additionally, Kim et al have prepared a transition metal oxysulfide (Ni y Co 1−y O x S z ) cathode, which was directly grown on the carbon paper (CP) [37]. A two-step electrodeposition coupled with anion exchange methods was applied to control the morphology and composition of the catalyst, where the Ni y Co 1−y O x /CP was first synthesized with shapes varying from sphere to fern-like shape, followed by a sulfidation process to partially replace the O atoms with S to obtain Ni y Co 1−y O x S z /CP catalyst. Both the plots of intrinsic activity normalized by ECSA as functions of S surface composition and proportion of S atoms having higher binding energies displayed 'volcano plot' behaviors (figures 7(e) and (f)), in which Ni 0.78 Co 0.22 O x S 0.25 /CP outperformed other catalysts in terms of intrinsic activity, emphasizing the critical role of S composition in influencing its intrinsic HER activity. In PEMWE, Ni 0.64 Co 0.36 O x S 0.28 /CP was determined to be a suitable cathode catalyst in MEA for practical operation owing to the lowest overpotential obtained from the preceding three-electrode cell system, with IrO 2 /CP as the anode. At the cell voltage of 2.0 V, the PEMWE reached a current density of 0.72 A cm −2 . The relationship between S atoms with higher binding energy and the HER intrinsic activity is in agreement with another study, in which an amorphous MoS x catalyst was electrodeposited on the CP as a cathode in the electrolyzer [120]. In this system, the impacts of deposition potential and deposition times were also uncovered, wherein the deposition time showed more significant influence, especially in the crack and crevice formation at highly negative deposition potential in excessive deposition durations, causing undesired aggregation of catalyst on the surface.
Furthermore, appropriate modification strategies can be applied to TM-based catalysts to further optimize the catalytic activity and rectify their existing shortcomings such as unsatisfactory conductivity and unregulated electronic structure [116]. For example, though MoS 2 has a free energy of hydrogen adsorption (∆G H ) similar to that of Pt due to the active edge sites for HER, the basal planes of MoS 2 are EC inactive [121,122]. Mo et al decorated single transition metal atoms (TM = Fe, Co, Ni, Cu) on the basal planes of single-layer MoS 2 , thus creating more active sites by activating the inert basal plane [119]. The single TM atoms were doped in the position of Mo atop sites. Among all the TM-modified systems, the tetrahedral anchored Co in the absence of the Co-Mo bond induced a proper downshift of the conduction band so as to endow the S sites the with highest HER activity. As a result, the Co-S MoS 2 exhibited the best catalytic performance among other studied transition metal doped systems when tested in a PEMWE with Nafion 115 membrane as the PEM and IrO 2 as the anode catalyst ( figure 7(d)) at an applied cell potential of −1.0 V to −2.4 V. An even higher performance of molybdenum sulfide-based HER catalyst was achieved by Holzapfel et al, where [Mo 3 S 13 ] 2− nanoclusters were supported on nitrogen-doped carbon nanotubes (Mo 3 S 13 -NCNT) through a self-assembly process attributed to the high affinity of Mo atoms to N atoms with higher electronegativity in CNTs [123]. This catalyst delivered a superior current density of 4 A cm −2 at a cell voltage of 2.36 V (figure 8(a)) due to the EC induced activation process of the catalyst at high current density. The electrochemical performance in MEA enhanced dramatically during the cell activation process ( figure 8(b)), particularly when applying Nafion 212 membrane with a smaller thickness that was able to lower ohmic resistance, resulting in a significant increase in the current density by 908 mA cm −2 at 2.2 V. Moreover, there was only a small degradation rate of 83 µV h −1 ascribed to the increased HFR-free cell voltage during a stability test of 100 h at 1 A cm −2 , exhibiting a stable polarization behavior.
It is noteworthy that the electrodeposition method is broadly used in fabricating the catalyst for use in the electrode because of its inherent benefit of achieving high catalyst utilization [109,125]. For example, Choi et al prepared a CuNiMo ternary alloy by rapid electrodeposition, which allowed the direct and simple fabrication of a catalyst on CP for use as a PTL in PEMWE [126]. Cu 44.4 Ni 46 Mo 9.6 was the optimal ternary alloy with increased surface area, in particular, showing a double layer capacitance of 125 mF cm −2 , much higher than that of the single metal and binary alloy catalysts studied. The PEMWE single cell system using Cu 44.4 Ni 46 Mo 9.6 exhibited no degradation after operation for 48 h. In terms of the cost, the cost efficiency of the MEA using Cu 44.4 Ni 46 Mo 9.6 was at least 190% higher than those of previously reported MEA using PGM catalysts. In this system, the synergy between the components in Cu 44.4 Ni 46 Mo 9.6 made a significant change in the active surface area, intrinsic activity, and durability, rendering it a cost-effective ternary catalyst for future commercial application. Similarly, the synergic effect among metals in the alloy and even with the metallic support has also been discovered by Kim et al, who synthesized a transition metal alloy NiMo catalyst supported by Cu foam deposited on CP (NiMo/CF/CP) through electrodeposition [127]. The CF with a highly roughened surface served as a support for NiMo electrocatalysts in order to obtain a large electrochemical surface area and enhanced the mass transfer. With CF as support, the overpotential was reduced from 297.9 mV to 68.7 mV at −10 mA cm −2 , signifying the importance of metallic support in this study. The PEMWE using NiMo/CF/CP as the cathode achieved a remarkable current density of ∼2.0 A cm −3 at 2.0 V, but it was lower than that using the Pt/C as the cathode catalyst. On top of that, electrodeposition is also capable to control the size and distribution of the catalyst on the conductive support by adjusting the operating parameters. Yoon et al reported an acid-durable cobalt phosphide (Co-P) catalyst synthesized via pulse electrodeposition on CP, where the P content was controlled by adjusting the applied dissolution potential [128]. The acid durability of the best-performing Co-P catalyst (Co-P-0.3, synthesized using dissolution potential of −0.3 V SCE ) was evaluated in 0.5 M H 2 SO 4 through an accelerated degradation test (ADT) comprising 500 cyclic voltammetry (CV) cycles. The recorded linear sweep voltammetry (LSV) curves for 500 ADT cycles suggested that the performance of the Co-P-0.3 catalyst remained relatively stable with a non-negligible deviation found after 500 CV cycles marked by an increase in overpotential of only 43.73 mV and the lack of change in the surface morphology.
Although electrodeposition is widely applied, limitations such as aggregation issues and susceptibility of the formed porous structure to degradation under cell operation conditions still exist [125], discouraging the industrial adaptation of this approach. Therefore, exploring the scalability of this method would be profound, and fortunately, works have been done in this aspect. For instance, Wei et al proposed a scalable and facile electrodeposition method in ethaline-based deep eutectic solvent employing a sacrificial anode set-up to synthesize NiFe-based electrocatalysts for the investigation of industrial water electrolysis [129]. In this method, different catalytic electrodes were prepared by simply adjusting the composition of the electrolyte, and simultaneously, the energy consumption, the potential system decomposition, and the contamination of the anode could be reduced by introducing Fe anodes as Fe sources.

Molecular catalysts
In addition to heterogeneous catalysts, the potential of molecular catalysts is also tapped because of the tunability of their catalytic activity and selectivity through modifying geometry of their ligand structure [130], contributing to realizing an applicable MEA under practical conditions. For instance, the remarkable stability of a HER catalyst in a PEM water electrolyzer was realized by Bellini et al, in which a dinuclear Ru diazadiene olefin complex, [Ru 2 (OTf)(µ-H)(Me 2 dad)(dbcot) 2 ], supported on carbon black was employed as the cathode to produce hydrogen (28 L H2 min −1 g Ru−1 ) with a current density of 400 mA cm −2 at 1.9 V and 80 • C [131]. An excellent turnover frequency of 7800 mol H2 mol catalyst −1 h −1 was maintained over a week of operation at 0.2 A cm −2 and a temperature of 80 • C, showing the predominant stability of this device. This could be traced back to the inherent ligand framework comprising exclusively stable C-C, C-N, and C-H bonds without any sensitive components that caused catalyst deactivation.

Hydrogen evolution-oxidation bifunctionality
The advanced design of MEA can be aimed at integrating fuel cell and electrolyzer into a single cell, for which the voltage reversibility of the integrated cell has to be improved to prevent induction of a large positive potential on the hydrogen electrode and severe system degradation [132][133][134]. This design is particularly useful in practical operations when there is a sudden power outage during water electrolysis or fuel starvation in the fuel cell. Lee et al designed a highly crystalline Ir-based alloy electrocatalyst to improve the kinetics of OER, HER, and hydrogen oxidation reactions (HORs) simultaneously, in which the carbon-supported IrNi alloy nanoparticles with high crystallinity (IrNi/C-HT) were synthesized [124]. Under OER operation, the crystalline nanoparticles generated an atomically thin IrNiO x layer on IrNi/C-HT, the number of d-band holes of which experienced a significant increase during OER, resulting in excellent OER catalytic activity. Under HER/HOR operation, this thin IrO x layer was reversibly transformed into a metallic surface, exhibiting high catalytic activity for HER/HOR. Therefore, the key to the reversibility is the thin IrNiO x layer, which was generated via dissolution and re-deposition mechanism (figure 8(c)), where the metallic surface was formed after IrNiO x dissolution under HER condition and re-deposition of the dissolved Ir ions occurred when metallic surface served as a substrate to accelerate the reduction of Ir ions to metal under HER/HOR condition, promoted by the adsorbed hydrogen. In the MEA system, the IrNi/C-HT catalyst exhibited high reversibility and maintained its bifunctional catalytic activity for HER/OER and HOR/OER even after 10 polarity conversions (up to 240 min) ( figure 8(d)).
Summarily, the search for alternative HER catalysts is on the spotlight in recent years (table 1), especially through developing transition metal-based electrocatalysts, which is more appealing than manipulating the content of Pt by adding support and applying well designed synthesis methods. However, most of the transition metal-based catalysts investigated in MEA are still inferior to the benchmark noble-metal catalyst, implying that the recent novel catalyst is yet to be capable of realizing industrial standard water electrolysis. The benchmark Pt/C performance is more frequently exceeded by other PGMs, and hence, the reliance of the technology on PGMs is not expected to be eliminated in near future.

Catalysts for OER
Minke et al have demonstrated that the scarcity of iridium element makes its current demand unsustainable for a commercial PEMWE market with a production rate of about 7 t a −1 [107]. It is suggested that the loading of iridium catalyst in PEMWE cells needs to be dramatically reduced to a target loading of 0.05 g kW −1 by 2035. Enhancing OER catalysis can be done by incorporating suitable support material and employing better catalysts.

Support materials
In reducing Ir loading, employing suitable support is the most used strategy that enhances the coordination between catalyst and support as well as electrical conductivity by utilizing conductive materials as the substrate. Carbon-based materials are commonly used as support due to their large surface area, good electric conductivity, and porous structure [137,138]. However, the materials used at the anode of PEMWE must withstand an acidic (0 < pH < 1) and oxidizing environment (with anodic potential above 1.5 V versus reversible hydrogen electrode, RHE) [139]. As carbon-based materials are prone to carbon corrosion and performance deterioration in OER, developing alternative support materials is necessary.
Corrosion-resistant TiO 2 support can effectively lower the loading of precious metal catalysts while endowing the system with stable catalytic activity during PEMWE operation due to its surface that is favorable for catalyst dispersion, requiring only minimum amount of IrO 2 nanoparticles to form a conductive network [140,141]. Regmi et al reported a core-shell design by depositing a conformal layer of Pt nanoparticles on the TiO 2 core by photoreduction to prepare conductive layer-coated supports (CCSs), followed by dispersing Ir catalyst on the CCSs using wet impregnation method [142]. This semiconductor-assisted photon-driven reduction facilitated the formation of evenly distributed Pt nanoparticles on the TiO 2 surface with strong inter-particle interaction to prevent the aggregation of nanoparticles when coated with the Ir catalyst in the subsequent process. By inserting the Pt interlayer, the direct interaction between Ir and Ti was avoided, resulting in high stability of the Ir surface. Even after annealing, Ir still retained its metallic character due to efficient charge transfer from Ir to Pt. As a result, the conductivity of the catalyst with significantly decreased Ir loading (39 wt%)   two-pronged approach increased the electrical conductivity and the amount of surface active sites to promote OER kinetics. The oxygen vacancies behaved as charged species and facilitated the charge transfer process. With a low Ir mass loading, MEA single cell integrated with the optimal IrO 2 /TNO anode presented a voltage of 1.832 V at 1 A cm −2 , which is lower than that with unsupported IrO 2 (1.858 V at 1 A cm −2 ) and exhibited good stability during the 100 h operation at 1 A cm −2 in a single cell. Another group of suitable support materials comprise of two-dimensional (2D) nanomaterials possessing unique geometric and electronic properties owing to their large surface area with abundant active sites, atomic-scale thickness, excellent mechanical properties, and tunable surface chemistry, thus having the promise of serving as excellent substrates to stabilize catalysts [145,146]. For example, graphitic carbon nitride has a 2D graphite-like structure with the tri-s-triazine unit as the building block. It is a metal-free and cost-effective substrate for OER catalysts to confine the catalyst through strong electronic interaction induced by C or N atoms, simultaneously decreasing the catalyst loading and maximizing the exposure of active sites [147,148]. Wang et al introduced N defect into the g-C 3 N 4 (N-CN) that acted as a catalyst substrate for IrO 2 nanoparticles [149]. N defects in CN regulated the extra electrons to the closest C atoms along with increased π-electron delocalization in a conjugated network of CN, thereby supplying more electrons to IrO 2 . The change in the electronic structure of CN was verified by the UV-Vis diffuse reflection spectra, where the light absorption was enhanced for N-CN relative to CN (figure 9(a)) stemming from the boosted π-electron delocalization in N-CN. Besides, the porous structure with larger pore size distribution was demonstrated by N-CN through the N 2 adsorption-desorption isotherms ( figure 9(b)), leading to a BET specific surface area (348.17 m 2 g −1 ) that is almost 13 times larger than CN to provide more loading sites. Owing to the strong electronic interaction between IrO 2 and N-CN, IrO 2 /N-CN presented the optimal OER potential of 1.778 V and good durability for 300 h at current density of 1.6 A cm −2 . Furthermore, theoretical calculations revealed the superiority of IrO 2 /N-CN in OER kinetics and indicated that the energy barrier of the RDS of IrO 2 /CN ( * OH → * O, 1.42 eV) could be reduced in IrO 2 /N-CN ( figure 9(c)). The density of state profile further verified the formation of Ir-N bonds and strong electronic interaction in IrO 2 /N-CN from a larger overlap between occupied N p orbitals and Ir d orbitals, contributing to weakened adsorption of oxygen intermediates to improve the overall kinetics.
Recent studies have demonstrated that stabilizing the Ir-based catalyst on the one-dimensional (1D) rather than 2D nanomaterial supports is also viable in boosting the electrochemical performance [150]. The allure of 1D nanomaterials is from the increased surface area for more loading sites, the abundance of open spaces among the adjacent 1D nanomaterials to promote mass accessibility, and efficient oriented charge transport [151]. For instance, Jiang et al designed an ordered array nanostructured electrode with defective Ir film coated on WO x nanorods support (Ir@WO x NRs) [152]. WO x nanorods were grown on W foil through hydrothermal synthesis and Ir catalyst was electrodeposited on the surfaces of WO x NRs, followed by preparation of MEA via decal transfer process ( figure 10(a)). By increasing CV deposition to 100 cycles, significant aggregation of Ir led to the formation of a defective Ir film ( figure 10(b)). The optimal Ir coating (Ir@WO x NRs-100) had a mass loading of 0.14 mg Ir cm −2 to obtain a better performance of 2.2 A cm −2 at 2.0 V and brilliant stability of 1030 h at 0.5 A cm −2 , superior to that of commercial Ir black CCM. Not only could the WO x support with appropriate porosity and ordered nanorod nanostructure stabilize the dispersion of Ir and provide vertical electron transport channels, but it also promoted water conservation and transportation. In addition, 1D nanostructures can strengthen the interaction between the particles in the CL. Hegge et al designed a hybrid structure combining IrO x nanofibers with a conventional IrO x nanoparticles CL and successfully reduced the Ir loading to 0.2 mg Ir cm −2 , exceeding the performance of MEAs with anodes of only either IrO x nanofibers or IrO x nanoparticles [153]. The IrO x nanofibers interlayers formed a good electrical connection throughout the region and facilitated the distribution of electrons over the CL, strengthening the connectivity of the particles instead of leaving electrically disconnected regions in the CL with only nanoparticles (figures 10(c) and (d)). This was also evidenced by in-plane conductivity due to the obviously lower sheet resistance for the hybrid CL, compared with the non-hybrid counterparts with the same IrO x loading. Although the 1D support exhibits superiority in various aspects, the density and morphology of the nanostructure should also be taken note of to avoid overloading. Hrbek et al reported a fiber-like structure of a CeO x layer serving as catalyst support in a PEMWE, which was deposited on PEM through reactive magnetron sputtering to form CCM, together with simultaneous plasma etching of the PEM [154]. The working pressure during the treatment was optimized at 0.4 Pa, balancing between the level of structural porosity and avoiding fiber shortening and densification at high pressure that would disrupt the electron pathway and cause poor dispersion. The optimized support with could efficiently stabilize a thin-film Ir catalyst, with combined Pt + Ir loading of 220 µg cm −2 .

Catalyst materials
It is worth noting that the oxidation state of Ir can correlate with the OER activity. For example, Kim et al successfully loaded the Ir catalyst onto a highly roughened dendritic Au support grown on a CP substrate, and the Ir metal loading was controlled at a microgram scale (Ir/Au/CP) by the number of Ir electrochemical deposition pulse [155]. With increasing Ir coverage, the Ir 4f 7/2 peak binding energy and Ir 3+ + Ir 4+ + Ir >4+ peak area ratio continuously declined (figure 11(a)) due to the increased surface density of the Ir-Au interface and the presence of electron density gradient throughout Ir islands in the direction normal to the Ir-Au interface in view of electronic depletion at the Ir-Au interface. In addition, the optimal Ir-based GDE (30Ir/Au/CP, where the number of deposition pulse is 30) was EC and thermally (TC) oxidized to change the Ir electronic structure by forming IrO x , aiming at adjusting the balance between activity and stability. The relationship between specific activity and Ir 3+ /(Ir 3+ + Ir 4+ ) ratio was thereupon established ( figure 11(b)), further demonstrating the influence of the oxidation state of Ir on OER activity, which decreases in the order of EC-IrO x /Au/CP > 30Ir/Au/CP > TC-IrO x /Au/CP. With lower contribution from ohmic overpotential and mass-transfer losses even in the high current density region (figure 11(c)), a high Ir mass activity of 440.5 A mg Ir −1 at 1.9 V was achieved using 30Ir/Au/CP. Besides, Cha et al discovered the relationship between the surface oxidation of Ir and the dissolution rate by investigating IrO x /Ti 4 O 7 catalyst with various iridium/support ratios [156]. With a higher Ir 4+ /[Ir 3+ + Ir 4+ ] ratio, less iridium degradation was observed under OER conditions, suggesting better corrosion resistance and good stability. A similar conclusion was also drawn by Saveleva et al, where a diminution in Ir dissolution rate and slower establishment of a steady state were observed in oxidized Ir nanoparticles supported on Sb-doped SnO 2 aerogel compared with the unsupported catalyst [157]. This could be attributed to the decrease in the formation of highly unstable Ir 3+ since the dissolution process tended to occur through the formation of intermediates in the oxidation state Ir 3+ . At this point, it can be summarized that a trade-off of Ir 3+ and Ir 4+ can enhance the intrinsic OER activity and stability.
Modifications focusing on the OER catalyst itself is also an indispensable part of boosting its performance in MEA. Strategies employed to regulate the structure-performance relationship of noble-metal-based catalysts include defect engineering, strain engineering, and hybrid engineering [158]. Huang et al utilized Co-hexamethylenetetramine metal-organic framework as the precursor and a fast-quenching method to synthesize tensile-strained RuO 2 nanorods growing on antimony-tin oxide (ATO) particles (s-RuO 2 /ATO) [35]. The tensile strain induced by the fast-quenching process altered the electronic state of RuO 2 . The x-ray diffraction peaks of s-RuO 2 revealed the expanded lattice parameters relative to the normal RuO 2 , exhibiting the tensile strains within the nanorods. Moreover, in the Fourier-transformed extended x-ray absorption fine structure (FT-EXAFS) ( figure 11(d)), the longer radial distance was observed for the first and second shells of s-RuO 2 /ATO than that of n-RuO 2 /ATO (without quenching process). Besides, a shift of E g peak of  [152]. Copyright (2021) American Chemical Society. Schematic illustrations of (c) catalyst with low loading that is electronically disconnected from the PTL and (d) hybrid IrOx anode comprising a low-loaded CL and a nanofiber interlayer. Reprinted with permission from [153]. Copyright (2020) American Chemical Society. s-RuO 2 /ATO to a lower wavenumber in the Raman spectrum (figure 11(e)) further verified the presence of tensile strain. As a result, the energy barrier of the limiting step of OER was reduced and the absorption energies of oxygen intermediates were also weakened on the tensile-strained RuO 2 (figures 11(f) and (g)). Impressively, the resultant PEMWE required only 1.51 V to reach a current density of 1 A cm −2 .
Another modification strategy is constructing a composite catalyst such as IrO 2 -RuO 2 composites. The durability of IrO 2 exceeds that of RuO 2 while RuO 2 has higher OER activity, thus combining these two components can obtain a highly robust OER catalyst in the MEA anode but the resulting MEA is still costly [159,160]. Nevertheless, the cost of the MEA can be reduced when noble-metal-free components are used to dilute the content of noble metal in composite electrocatalysts [161]. For instance, Kaya et al investigated the magnetized IrO 2 -Fe 3 O 4 as an anode catalyst in MEA that exhibited a four-fold increase in MEA current Figure 11. (a) Ir 4f7/2 peak binding energy and Ir 3+ + Ir 4+ + Ir >4+ peak area ratio as functions of Ir coverage, (b) the relationship between specific activity and Ir 3+ /(Ir 3+ + Ir 4+ ) ratio, and (c) overpotential subdivisions of cell overpotential based on PEMWE polarization curve in a single cell comprising 30Ir/Au/CP anode. Reprinted from [155], Copyright (2020), with permission from Elsevier. (d) The Ru K-edge FT-EXAFS spectra of n-RuO2/ATO, s-RuO2, and s-RuO2/ATO after phase corrections, (e) the Raman spectra of n-RuO2/ATO, s-RuO2/ATO, commercial RuO2, and ATO, and the free energy diagrams of (f) s-RuO2 and (g) RuO2 calculated on the coordinatively unsaturated sites. Reproduced from [35]. CC BY 4.0. © 2022 The Authors. Advanced Science published by Wiley-VCH GmbH. density in comparison with the non-magnetized and pristine IrO 2 catalyst. The loading of IrO 2 was significantly reduced in the composite CL by replacing 20% of IrO 2 with the more abundant and cheaper Fe 3 O 4 while retaining high electrochemical performance [162]. Moreover, a ternary composite can be constructed to take the advantage of the synergy between more components. By applying IrO 2 -RuO 2 -TaO x coating on Ti felts PTL via a thermal deposition method, the precious metal loading was decreased to 1 mg cm −2 and achieved a performance of 1.836 V at 2 A cm −2 in a single cell, owing to combination of the high stability of IrO 2 and TaO x in the acidic environment and the high OER catalytic activity of RuO 2 [163].
OER activity of a catalyst can also be optimized by tuning its morphology or architecture. The morphology-controlled nanostructures are able to provide an unsaturated coordination environment for a more reactive reaction [164]. For example, Park et al fabricated an IrO 2 inverse-opal MEA as an anode for OER via the decal-transfer method [165]. The inverse-opal structure was synthesized by utilizing self-assembled polystyrene (PS) beads on a Ni-sputtered fluorine-doped tin oxide substrate. Thereinto, PS was utilized as a template to construct micropores in the electrode, followed by infiltration of IrO 2 through pulse electrodeposition while adjusting the number of cycles to attain the optimal structure (with the optimal number of cycles being 140 k) with high porosity (figures 12(a) and (b)). By reducing the catalyst loading to 0.02 mg cm −2 , inverse-opal MEA outperformed the conventional MEA by delivering 870 mA cm −2 at 1.6 V, which is 2.5 times that of the latter. Also, there was no decline in the performance of the novel MEA with micro-pores in the high-voltage region (while it was obvious for the conventional MEA) due to promoted mass transport for reactants and products, which mitigated active site blockage. Besides, the novel MEA had lower ohmic resistance and charge-transfer resistance due to the ordered and interconnected pores to promote charge transfer.
In addition, another way to beneficially modify the architecture of the catalyst is by synthesizing core-shell catalysts with unique physical and chemical properties. Several advantages can be obtained from a core-shell structure, including introduction of interfacial strain that tunes the electronic structure, modified atomic vicinity that affects the charge transfer, increased surface area, providing protection to the unstable core by isolating it from the external environment, and facilitating the full utilization of the surface material [166,167]. On this ground, the core-shell structure is expected to effectively regulate the intrinsic catalytic activity and stability of the resultant material, simultaneously lowering the noble metal loading. For example, Zheng et al synthesized a core-shell electrocatalyst with an IrRu x core and Ir-rich shell (IrRu x @Ir) via a CO-induced phase-segregation strategy [168]. The electronic interaction between the core and the shell induced efficient charge transfer from Ir to IrRu x , thus tuning the electronic structure of the shell and increasing the oxidation state of surface Ir sites to improve the OER activity. The Ir shell protected the inner Ru atom against the dissolution and retained the structural stability during the durability test in the acidic environment. Similarly, Lv et al designed a self-assembled RuO 2 @IrO x core-shell heterostructure nanocomposite, in which the Ir-based shell also effectively protected the RuO 2 core from dissolution, resulting in a cell potential of 1.683 V at 1 A cm −2 and with enhanced catalyst durability [36]. In-depth exploration of the core-shell structure further shows its promise in producing stable catalyst ink, which is often the most overlooked parameter for fabrication of large-area electrodes with homogeneous catalyst deposition. Pham et al synthesized IrO 2 -coated TiO 2 core-shell microparticles (IrO 2 @TiO 2 ) as an OER catalyst, and deposited them on Ti PTL [169]. There were two steps in the synthesis process based on surface charge modification indicated by zeta potentials of TiO 2 particle in ethanol solution (figure 12(c)): (1) formation of H 2 IrCl 6 shell on TiO 2 core, which depended on the surface charge of TiO 2 that was converted from negative to positive by adding acetic acid and H 2 IrCl 6 for better attachment of [IrCl 6 ] 2− anions, followed by continuously growing H 2 IrCl 6 shells around the TiO 2 particles upon solvent evaporation; and (2) conversion of H 2 IrCl 6 shell into an IrO 2 layer via pyrolysis at 500 • C for 30 min in air. The stability of the catalyst ink was reflected in back-scattering signal intensity that was influenced by the ink destabilization ( figure 12(d)). The back-scattering signal intensity of IrO 2 @TiO 2 was relatively even throughout the height of the ink compared with that of mixed IrO 2 + TiO 2 powders that would have undergone sedimentation, manifesting the high stability of IrO 2 @TiO 2 ink due to surface changes by IrO 2 coverage.
Impressively, competitive durability of the OER catalyst could be achieved even with low-iridium PEMWE MEA. Möckl et al investigated the durability of a novel iridium catalyst with a low iridium packing density (derived from amorphous hydrous iridium oxide deposited as thin film on low surface area TiO 2 ) in a ten-cell water electrolyzer short stack, where the first five cells utilized conventional catalyst with iridium loading of 2 mg Ir cm −2 and the last five cells applied the novel catalyst with a lower loading of 0.25 mg Ir cm −2 [170]. The MEA was tested for 3700 h under ambient pressure and 60 • C by load current density cycling between 0.2, 1.0, and 2.0 A cm −2 . Although the low-loading catalyst exhibited 30-fold higher beginning-of-life mass-activity, there was a pronounced rise in the cell voltage during the first 1000 h for stabilization, but less significant during the rest of the time for both types of catalysts. This was attributed to the change in the oxidation state of the catalyst surface as it converts into the less active IrO 2 . Similar  [165], Copyright (2019), with permission from Elsevier. (c) Schematic illustration of synthesizing IrO2@TiO2 catalyst, in which the surface charge was manifested in zeta potential of the particle in ethanol solution, and (d) the ink stability test of IrO2@TiO2 and IrO2 + TiO2 catalyst based on the evolution of backscattering spectra. Reproduced with permission from [169]. © 2020 The Authors. Published by Elsevier B.V. (e) Durability test at 1 and 4 A cm −2 under 80 • C for bare and integrated PtCo catalyst-based MEAs. Reprinted from [171], Copyright (2020), with permission from Elsevier. durability trend was observed by Pantò et al on an MEA with recombination catalyst-based anode (IrRuO x + PtCo) and Pt/C cathode that enabled an even lower PGM loading of 0.6 mg MEA PGM cm −2 [171]. The test was carried out for 3500 h, with 100 h at 1 A cm −2 and 3400 h at 4 A cm −2 (figure 12(e)), and the cell voltage was found to have increased rapidly at the beginning, which was ascribed to the accumulation of the gases in the micropores of catalysts in addition to the modification of the oxidation state at the anode catalyst surface. During the test, the novel MEA exhibited lower degradation and delivered a cell voltage that was 30 mV lower than that without PtCo (bare MEA), thus successfully reaching a voltage efficiency of ∼80% at 4 A cm −2 .
Summarily, most studies are still focused on optimizing noble metal catalyst (RuO 2 and IrO 2 ) to achieve high OER activity at the anode of a MEA, since they are highly active and acid-durable OER catalysts. As they are precious metals, their scarcity and high material cost will inhibit widespread PEMWE adoption. Therefore, great efforts have been devoted to developing appropriate support materials to minimize the loading of precious metals in the electrode while retaining the electrochemical performance of the current commercial OER catalysts (table 2). Future exploration on novel catalysts to replace the utilization of noble metal-based OER catalysts in PEMWE MEA is necessary.  Figure 13. An overview of catalyst improvement works discussed in this review.

Conclusion and perspectives
Water electrolysis is the most suitable technology to produce hydrogen sustainably without producing any undesirable by-products. To drive the industrial development of water electrolysis further, scientists are seeking a way to improve the catalysts and the membranes as well as their integration into MEA, the core of a water electrolyzer. An electrolyzer utilizing PEM in MEA produces an acidic environment for hydrogen production. Compared with other commercially available electrolyzers, PEMWE can usually present higher performance, but the cost and the corrosive reaction environment are inhibiting its widespread commercial adoption.
To discuss PEMWE and its challenges in detail, different fabrication methods to incorporate electrocatalysts into MEA have been first introduced in this review, whereby CCM and CCE are the most common approaches that have potentials for industrial application. The importance of facilitating favorable catalyst-membrane interaction through the fabrication process has been highlighted. The innovative modification strategies applied to electrodes and membranes to ameliorate the performance of PEMWE have been discussed, focusing on efforts to improve charge and mass transport efficiencies and improving the lifetime of the membranes. In addition, recent encouraging progress in the development of HER and OER catalysts, summarized in figure 13, has been reviewed to provide a better comprehension on rational design of high-performance catalysts used in MEAs. In particular, various synthesis strategies to produce thin films of Pt catalysts and replacing Pt have been the focus of studies on HER catalysis, whereas OER catalysis studies are dominated by the improvement of catalyst supports and the structural modification of precious-metal-based catalysts.
Despite the brilliant achievements in PEMWE, there are still challenges to be addressed. Firstly, in addition to the innovative and facile synthesis methods developed to improve the device performance and manufacturability, the environmentally friendly synthesis methods designed for environmental sustainability are easily overlooked. Green chemistry plays a pivotal role in environmental management, whereby the chemical process is designed in a way that reduces or eliminates the use or generation of hazardous materials, thus speeding up the progress toward sustainable development goals [173,174]. As mentioned, precious-metal-based electrocatalysts, especially those containing Ir, are still widely used for OER anode in PEMWE. Common catalyst synthesis methods such as hydrothermal method, electrodeposition, and sol-gel method lack both good morphology control and environmental sustainability. Therefore, it would be rewarding to explore ways that fulfill both these requirements. Faustini et al prepared an Ir-containing oxide anode catalyst (Ir 0.7 Ru 0.3 O 2 ) with a hierarchical ultraporous architecture comprising a nanoneedle network assembled into microporous micrometric particles, which enabled high accessibility of reactants to the catalytically active surface during PEMWE operation [34]. They also proposed a spray-drying fabrication process using water as a solvent to assist the evaporation-induced self-assembly, which is cost-effective, waste-free, and easily scalable. This approach conferred both the ability to tune the chemical composition of the catalysts and the ability to introduce porosity using templates.
Secondly, to accommodate the industrial operation, a PEMWE should be able to withstand the harsh conditions of high current density and high pressure in order to reduce operational cost [11]. Although many papers have claimed the high performance of their PEMWE catalysts at ampere-level (1-4 A cm −2 ), it is still difficult to compare the performance of a single cell with other reported systems since the parameters in other electrolyzers may be different, such as testing condition (e.g. different current densities), membrane type, and catalyst mass loading. A few studies introduced the Ir-mass-normalized power density metric (which is obtained by dividing the cell power density by noble metal loading) to eliminate the influence of other testing factors with a focus on catalyst utilization [103,169]. Besides, the deviation also lies in the geometrical area of the working electrode used to normalize the current density because it differs from the actual catalytic surface area. Hence, it is necessary to establish a series of standardized criteria for comparing the performance of MEAs for water electrolysis.
Thirdly, the majority of the studies have dived into developing efficient electrodes and optimizing the operating parameters, which can deliver minimal overpotential and consume less power at near industrial-scale current density by the experimental practices. On one hand, developing advanced PEMs, GDLs, and electrocatalysts will continue to be the hotspot of future research, which will place emphasis on the cost and lifetime of their materials, accompanied by specific requirements for the reduced gas permeability of PEM, sufficiently robust coating of GDL, and highly catalytic activity of electrocatalysts. On the other hand, optimization done by experimental approaches is challenging from the viewpoints of cost, time, and technical restrictions. As such, involving artificial intelligence and data science in PEMWE development for comprehensive investigation is desirable for accurately optimizing electrolyzer design and predicting the best structure of potential catalysts based on algorithms. For example, machine learning can guide the design of a PEMWE electrolyzer by fast and easy simulation of the hydrogen production rate and cell current density of the electrolyzer by correlating with various design parameters to obtain the optimal result to guide the subsequent experiments [175]. Hence, assistance by machine learning offers the best approach to achieving high performance at the shortest possible time.
With the help of advance manufacturing technologies, more efficient fabrication of electrolyzers can be realized. Three-dimensional (3D) printing, the well-known subset of additive manufacturing, is the most widely used technology for expeditious prototyping through layer-by-layer fabrication of an object based on its 3D model data [176]. Therefore, it offers flexibility to produce microscale structures in a programmable and facile fashion, with a well-controlled design in shape, size, and porosity [177]. Ignited by this technology, 3D-printable MEA becomes appealing. To our delight, some studies have demonstrated the successful application of 3D printing technology to manufacture the components in MEA [178]. For example, Yang et al fabricated 3D-printable bipolar plates using fused deposition modeling without any post-processing, which provided a rapid synthesis route for the mechanical support of the electrolyzer [179]. Besides, 3D printing is also useful in fabricating heterogeneous catalysts, especially for industrial catalysts. For instance, robocasting can be used to fabricate structured monoliths and powder bed 3D printing methods can be explored to produce catalysts of various shapes [180,181]. Hence, the emerging 3D printing technology showcases its potential in MEA manufacturing for water electrolysis in a more precise and efficient manner while retaining the overall cell performance.
Finally, a good MEA is a kernel of realizing the high efficiency of electrolytic cells industrially, which is the unremitting pursuit of researchers. Hence, the collaboration between industry and academia to improve the performance and encourage widespread adoption of the state-of-the-art PEMWE technology is crucial. In improving industrial water electrolysis, multidisciplinary cooperation between engineering and sciences can provide farsighted solutions.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.