Uniform nanoporous zirconia composite membrane enabling high-performance alkaline water electrolysis Authors: Zhipeng Xu, Zhihao Lin, Daohui He, Jingjing Yin, Yangke Pan, Liang Guo, Junbin Liao, Huimin Ruan, Congjie Gao, Jiangnan Shen, Young Moo Lee Published: 2026 ABSTRACT Zirconia-based membranes typically face a trade-off between ionic transport and gas barrier performance, limiting their effectiveness in alkaline water electrolysis, where low area resistance, high bubble-point pressure, and long-term stability are essential. Here we show a method for fabricating a nanoporous composite membrane. Yttria-stabilized zirconia nanoparticles are uniformly dispersed into a sol-state polybenzimidazole matrix via a one-pot sol-gel process. The resulting membrane exhibits a sponge–like, uniform nanoporous morphology with a high porosity of ~85%. Yet, it maintains a higher bubble-point pressure (>25 bar). Strong interfacial interactions exist between yttria-stabilized zirconia nanoparticles, phosphoric acid, and polybenzimidazole molecular chains. The synergistic multi-pathway structure facilitates continuous hydroxide ion migration. In alkaline water electrolysis, the membrane delivers a high current density of 13.1 A cm–2 at 2.0 V, and operates effectively over a wide range of alkaline electrolyte concentrations. Reinforcement with a polyphenylene sulfide mesh enables stable operation for 7000 h, and the membrane performs reliably in a large-area single-cell stack. This work introduces a scalable route to gel-state ceramic–polymer membranes for high-efficiency hydrogen production. Introduction As global efforts intensify toward achieving carbon neutrality, hydrogen has emerged as a pivotal clean energy carrier due to its high gravimetric energy density, zero carbon emissions, and broad applicability in energy storage, transportation, and industrial sectors. However, the widespread deployment of hydrogen energy is constrained by the efficiency and cost of green hydrogen production technologies. Among available methods, water electrolysis powered by renewable electricity offers a direct, carbon-free route to hydrogen generation1–3. Currently, alkaline water electrolysis (AWE) and proton exchange membrane (PEM) electrolysis represent the leading technologies. AWE is well-established and cost-effective for large-scale production. Its electrolytic efficiency remains limited by sluggish reaction kinetics and material constraints. Consequently, the development of advanced separator membranes that combine high ionic conductivity with robust chemical and mechanical stability is central to improving AWE performance4,5. Recent progress in anion exchange membranes (AEMs) has propelled anion exchange membrane water electrolysis toward using non-platinum group metal (PGM) catalysts, offering a cost-effective alternative to PEMWE6–8. Yet, achieving long-term durability under strongly alkaline and oxidative environments remains a major challenge. Polymer-based membranes with fixed cationic groups often suffer from backbone degradation or ion-exchange group instability9. In contrast, alkali-doped inorganic-polymer composites, such as ZrO2/polysulfone (PSU) systems, demonstrate greater chemical robustness10. Polybenzimidazole (PBI), owing to its amphoteric imidazole groups, is a promising ionic conductor and has been extensively applied in high-temperature PEM fuel cells11. KOH-doped PBI membranes have shown improved alkaline conductivity and chemical stability; however, maintaining structural integrity during long-term operation remains an issue, as swelling and thinning are often observed12,13. With the continuous development of AWE membrane technology, the traditional asbestos membrane is gradually being replaced by PPS fabric and composite membranes. Those that have been successfully commercialized include the Zirfon® series. At present, research on composite membranes has made progress in the composition of double-sided coatings, such as hydrophilic particles14–16 and alkali-resistant polymers17,18. For instance, Liao et al.19 have prepared a thin and flexible porous composite membrane composed of PSU and nano-ZrO2, along with a polypropylene (PP) non-woven fabric support layer through a combination of casting and phase inversion strategies, achieving a high bubble point pressure (24 bar) and a low surface resistance (0.14 Ω cm2). Nonetheless, the resulting membranes exhibited nonuniform pore distributions with many finger-like pores inside. Similarly, Li et al.17,20 prepared ZrO2-polymer composites via solvent casting and non-solvent-induced phase separation, which often led to the barrel effect, where gas crossover and ionic resistance compete detrimentally. Due to the inherent limitations of conventional non-solvent induced phase separation (NIPS) fabrication methods, such as weak interfacial bonding between the organic and inorganic phases and the tendency of particles to agglomerate, the resulting microstructure often exhibits heterogeneity. Therefore, developing a simple yet efficient fabrication approach that can, at the microscale, effectively synergize the optimization of rapid ion transport and efficient gas barrier properties remains a significant challenge in this field21–23. To overcome the limitations of conventional zirconia-based composite membranes prepared via the NIPS method, we developed a one-step sol-gel strategy to fabricate gel-state zirconia composite membranes. In this approach, phosphoric acid (PA) acts as a molecular bridge, enabling uniform dispersion of yttria-stabilized zirconia nanoparticles (YSZNPs) within the polymer matrix, promoting strong interfacial coupling, and simultaneously inducing in situ hydrophilic modification of the inorganic phase. This coordination promotes the formation of a spongy internal structure with uniformly sized nanopores and an asymmetric architecture featuring a relatively dense surface skin. This ingeniously engineered structure results in a high bubble point pressure (>25 bar) despite the membrane’s high porosity (up to 85%), and establishes multiple, efficient OH− transport pathways. The incorporation of a PPS mesh further reinforces the composite, endowing it with favorable mechanical strength. As a result, the membrane achieves a low area resistance (0.055 Ω cm2), favorable gas barrier properties, and long-term operational stability. Results Preparation and structure This study reports a composite membrane in which YSZNPs are dispersed in a PBI sol matrix that simultaneously functions as both a binder and an electrolyte host. In AWE, the resulting ternary polymer/H2O/KOH system can co-absorb K⁺ and OH⁻ ions. The fabrication process is outlined in Fig. 1. Previous studies have shown that incorporating YSZNPs enhances membrane conductivity, while yttria doping improves alkaline stability, thereby boosting electrochemical performance24. Traditionally, methods for incorporating ZrO2 into polymer mainly focused on dispersing inorganic powders in solutions to improve interfacial compatibility25. ZrO2-based composite membranes for AWE have been fabricated via dip-coating and hot-pressing to improve hydrophilicity, suppress gas permeability, and reduce ohmic resistance. As shown in the SEM images of Zirfon® UPT220 and UPT500 (Fig. S1), both membranes exhibit micron-scale pores on the surface and across the cross-section, with ZrO2 particles distributed in localized clusters. In contrast, YSZNPs were introduced in situ during the preparation of the PBI sol in polyphosphoric acid (PPA). PA was used to achieve uniform dispersion of the nanoparticles and anchor YSZNPs onto the PBI nanofibers. The cast slurry is doctor-bladed onto glass and exposed to air, allowing PPA hydrolysis and yielding a PA form of the membrane. Subsequent immersion in deionized water followed by KOH treatment induced deprotonation. This one-pot, low-cost route facilitates scalable membrane fabrication under mild hydrolysis conditions.Fig. 1Fabrication of gel-state zirconia–PBI composite membranes.a Schematic illustration of the fabrication of G-PBI membranes. b In situ sol-gel preparation of YSZs-n composite membranes within the PBI matrix. Fabrication of gel-state zirconia–PBI composite membranes. a Schematic illustration of the fabrication of G-PBI membranes. b In situ sol-gel preparation of YSZs-n composite membranes within the PBI matrix. The synthesis of high-molecular-weight polymers and the membrane structure were tailored by adjusting the weight ratio of inorganic particles to tetraamine, the volume of PA and dicarboxylic acid monomers during PBI synthesis (Fig. S2 and Table S1), and the sol-gel phase inversion process. Here, the composite membranes are denoted as YSZs-n, where n represents the mass fraction of YSZNPs in the total solid content. Nanoparticle dispersion in the system was achieved via PA assistance (Fig. 2a). Three types of membranes with varying compositions were prepared. Cryo-scanning electron microscopy (Cryo-SEM) imaging (Figs. 2b–e and S3) revealed that YSZNPs (~50 nm) were uniformly dispersed within the PBI matrix, as further confirmed by energy dispersive X-ray spectroscopy (EDS) elemental mapping under cryogenic conditions (Fig. S4). Compared with the gel-state G-PBI membrane without YSZNPs, which exhibits more irregular pores, the YSZ-filled membranes show narrower pore-size distributions and uniform sponge-like nanoporous architecture. The membranes are free of through-pores (Fig. 2e) and possess a relatively dense outer skin layer (Figs. S3a and S5), which retains the nanoparticles in the interior. Unlike the conventional finger-like morphology, the sponge-like architecture forms continuous pathways for electrolyte transport while suppressing gas crossover19. Even at relatively high inorganic filler loadings, the method prevents internal agglomeration. The surface of the membrane with YSZNPs added became rougher, the side in contact with air exhibited a maze-like folded pattern and contained nanoscale pores (Fig. 2f), whereas the side in contact with the glass substrate was relatively denser (Figs. S6 and S7). Overall, the membrane presents an asymmetric structure. Among membranes with varying compositions, the surface roughness of the dried membranes increases with increasing YSZNPs content. (Fig. S8).Fig. 2Structural and interfacial characteristics of the composite membranes.a Schematic diagram of the PA bridge between PBI and YSZNPs. Cryo-SEM cross-sectional images of wet membranes: b YSZs-70, c YSZs-40, d G-PBI, and e YSZs-70 (low magnification); f Cryo-SEM of the upper epidermal layer of wet YSZs-70 membrane (air-contact side). g XRD patterns of dried membranes and ZrO2 powder. h FT-IR spectra of characteristic regions and i the magnified spectra. j Zr 3d XPS spectra of dried YSZs-n/PA (after extensive PA removal), YSZs-n/KOH (after KOH treatment) and ZrO2. Structural and interfacial characteristics of the composite membranes. a Schematic diagram of the PA bridge between PBI and YSZNPs. Cryo-SEM cross-sectional images of wet membranes: b YSZs-70, c YSZs-40, d G-PBI, and e YSZs-70 (low magnification); f Cryo-SEM of the upper epidermal layer of wet YSZs-70 membrane (air-contact side). g XRD patterns of dried membranes and ZrO2 powder. h FT-IR spectra of characteristic regions and i the magnified spectra. j Zr 3d XPS spectra of dried YSZs-n/PA (after extensive PA removal), YSZs-n/KOH (after KOH treatment) and ZrO2. During composite formation, YSZNPs undergo chemical etching in hot PA. Previous studies have reported that ZrO2 etched in 20% hydrofluoric acid (HF) at 60 °C develops surface cracks and irregularities that enhance hydrophilicity26. In hot PA, the surface of the YSZNPs underwent slight etching, resulting in increased surface roughness, while the internal lattice structure remains unchanged (Figs. S9 and S10). X-ray diffraction (XRD) patterns show characteristic ZrO2 reflections at 2θ ≈ 28.2°, 31.5°, and 34.2°, corresponding to the (–111), (111), and (002) planes (Figs. 2g and S11). Notably, all peaks shifted slightly to higher angles, indicating strong interfacial interactions. Fourier transform infrared (FT-IR) spectra (Fig. 2h) further show that the Zr–O stretching band shifts to higher wavenumbers (∼505 cm−1), reflecting reduced bond electron density from electrostatic interactions. In the YSZ-containing samples, the P–O stretching band (1050–1150 cm−1) also shifted to higher frequencies (Fig. 2i), corroborating the strong interfacial interactions between the YSZ particle surfaces and PA-doped PBI. After KOH doping, the intensity of this peak weakens (Fig. S12). At the YSZ surface, Zr–OH (and minor Y–OH) groups undergo ligand exchange with phosphate species (H2PO4−/HPO42−) to form inner-sphere complexes with covalent Zr–O–P linkages. Upon heat treatment, a thin zirconium phosphate layer may form27. The oxygen atoms in the P=O double bond and imidazole rings can also coordinate with Zr centers. In the polymer, the PBI provides basic nitrogen sites capable of proton acceptance. When PA-doped PBI contacts PA-adsorbed YSZ, the PA molecules act as molecular bridges, binding the protonated PBI chains while coordinating with the YSZ surfaces. This bridging facilitates partial penetration of PBI chains into surface irregularities or interparticle pores, thereby enhancing interfacial entanglement and adhesion. Consequently, YSZNPs remained uniformly dispersed throughout the PBI sol during gelation without sedimentation, effectively resolving the long-standing issue of nanoparticle settling in conventional zirconia membranes23. During the PPA-to-PA transformation, differences in solubility between PBI and the surrounding medium induced phase separation. Strong hydrogen bonding between PBI chains promotes aggregation and formation of a physically crosslinked network. Macroscopically, nanoparticles are embedded within entangled PBI nanofibers bound by PA, yielding a sponge-like nanoporous morphology. As shown in Fig. S13, even at a YSZNPs content of 85 wt% (YSZs-85), a defect-free composite membrane with a uniform nanoporous structure can still be obtained. To elucidate the interfacial chemistry, X-ray photoelectron spectroscopy (XPS) was performed on pristine ZrO2, YSZs-n/PA (rinsed with deionized water), and YSZs-n/KOH (after KOH soaking) (Figs. 2j and S14–20). The Zr 3d peak in YSZs-n/PA shifts by +1.2 eV to higher binding energy relative to ZrO2 and YSZs-n/KOH, indicating electron transfer from Zr to P within the YSZs-n/PA composite. A peak at ~191 eV (P 2s) confirms the presence of phosphate, which decreases in intensity after KOH treatment, evidencing dissolution of phosphate under alkaline conditions. The Y2O3 stabilizer introduces abundant oxygen vacancies that can mediate electron transfer and enhance radical scavenging capacity28. PA molecules can attack Zr–O–Zr linkages or react with surface OH⁻ groups to generate zirconium phosphate. Upon KOH treatment, the phosphate dissolves, exposing fresh Zr4⁺ sites that rapidly hydrolyze to Zr–OH groups, increasing surface energy and improving wettability. Further insights were gained from molecular dynamics (MD) simulations, which evaluated the radial distribution function (RDF) and coordination number (CN) of hydrogen bonds between the imidazole rings of PBI molecular chains and PA molecules in YSZs-70/PA and G-PBI/PA models (Figs. S21, 22, and Supplementary Data 1). The simulation results reveal well-defined structural order. Notably, the incorporation of YSZNPs weakened the acid–base interactions between PBI and PA. The self-diffusion coefficient (D) of PA molecules, obtained from time-dependent mean squared displacement (MSD) analysis, was significantly smaller in YSZz-70 than in G-PBI, suggesting that strong YSZ–PA interactions (electrostatic attraction and hydrogen bonding) constrain PA mobility. PA thus acts as a molecular bridge, linking imidazole rings to YSZ surfaces and contributing to interfacial stability. Physical properties of composite membranes The physical properties of the composite membranes were systematically evaluated. Compared with the commonly used PSU polymer, the PBI-based framework exhibits higher intrinsic thermal stability, as confirmed by thermogravimetric analysis (TGA, Figs. 3a and S48). The dried PBI membrane shows rapid mass loss below 250 °C due to the release of bound water, followed by decomposition of the imidazole rings and aromatic backbone above ~550 °C. Increasing the ZrO2 content further enhances high-temperature stability, with the YSZs-70 sample showing the smallest overall mass loss at elevated temperatures. Good wettability, which is essential for effective electrolyte uptake and OH⁻ transport29,30, was evaluated via static and dynamic water contact-angle measurements (Fig. 3b, c). After KOH treatment, the commercial UPT220 retained the highest contact angle (64.4°), whereas all composite membranes exhibited markedly enhanced hydrophilicity. This improvement arises from the polar imidazole groups within the PBI matrix and the highly hydrophilic ZrO2 particles generated during hot-PA treatment. The increased surface roughness of the membrane and its open nanoscale pores further enhance hydrophilicity. Consequently, the water contact angle of YSZs-70 decreased to 11.7°, significantly lower than that of G-PBI (18.3°). Such pronounced hydrophilicity helps to suppress gas-bubble adhesion on the membrane surface during water electrolysis, thereby improving operational stability31.Fig. 3Physicochemical and mechanical properties of the membranes.a TGA profile of the membranes. b Static and c dynamic water contact angles of KOH-pretreated membranes. d Bubble point pressure (BPP) is measured using ethanol as the wetting liquid. e Membrane porosity and f pore size distribution of wet membranes, extracted from the Cryo-SEM image in Fig. 2 using ImageJ software. g Tensile stress-strain curves of KOH-doped membranes. h Creep compliance of wet membranes (25 °C, 100% RH). i Photographs of the membranes and image of the YSZs-70 membrane after twisting and folding. All error bars indicate the s.d. of at least three independent measurements. Physicochemical and mechanical properties of the membranes. a TGA profile of the membranes. b Static and c dynamic water contact angles of KOH-pretreated membranes. d Bubble point pressure (BPP) is measured using ethanol as the wetting liquid. e Membrane porosity and f pore size distribution of wet membranes, extracted from the Cryo-SEM image in Fig. 2 using ImageJ software. g Tensile stress-strain curves of KOH-doped membranes. h Creep compliance of wet membranes (25 °C, 100% RH). i Photographs of the membranes and image of the YSZs-70 membrane after twisting and folding. All error bars indicate the s.d. of at least three independent measurements. Owing to their nanoscale porosity, uniform, defect-free morphology, and asymmetric dense skin layer structure, the prepared membrane exhibited a high bubble-point pressure (BPP; Fig. 3d), reaching the upper limit of the pore size analyzer (>25 bar), far exceeding that of the UPT220 separator (2.8 bar). This high BPP effectively prevents gas crossover during electrolysis. Despite containing up to 70 wt% YSZ nanoparticles, the sol-gel derived composite membrane maintains a high porosity of approximately 85% and an average pore size of 28 nm (Fig. 3e, f). Compared with other separators (Fig. S23), the pore size distribution is much narrower and more homogeneous. Owing to the soft, gel-like nature of the PBI–YSZ network, the mechanical strength of the membrane is moderate (Figs. 3g and S24a). Nevertheless, unlike conventional brittle zirconia membranes that require a PPS mesh for support, the YSZs-70 composite membrane is flexible enough to be freely twisted or folded without mechanical failure (Fig. 3i). Even at a high YSZNPs content of 85 wt.% (Figs. S24b and S25), comparable to that of commercial Zirfon® series membranes, the composite membrane remains intact—an uncommon feature for zirconia-based systems. In contrast, the PSU-85 membrane prepared via the NIPS method fails to maintain structural integrity. Under compressive stress during electrolysis, membranes typically undergo irreversible creep thinning, which degrades gas barrier performance. Creep-compression testing was employed to accelerate pore collapse and evaluate structural stability. As shown in Fig. 3h, compared with G-PBI, the composite membranes exhibit lower creep modulus and a slower creep rate, which can be attributed to their higher solid-phase content and enhanced interfacial reinforcement. The YSZs-70 sample shows better creep resistance with a creep rate of 32.5 × 10−12 Pa−1 s−1 (Table S2). Cryo‑SEM images after creep testing (Fig. S26) show that the cross‑section of the G‑PBI membrane becomes dense due to complete collapse, whereas the YSZs‑70 membrane retains a portion of its nanoporous structure. This enhanced structural stability mitigates long-term deformation and gas crossover, thereby ensuring safe and stable AWE operation. KOH absorption and ionic conductivity After complete hydrolysis, the membranes were thoroughly washed with water to remove residual, strongly bound PA. In a 6 M KOH solution, water loss occurred within the membrane; however, increasing the solid content mitigated shrinkage (Fig. S27). The G-PBI membrane, with its large internal pores, undergoes severe dimensional contraction. As a representative ion-solvating membrane (ISM), PBI-based membranes experience deprotonation upon immersion in KOH solution, enabling electrolyte uptake (Fig. 4a). The interconnected three-dimensional gel network facilitates efficient OH⁻ transport. Notably, the internal KOH concentration is consistently higher than the external solution, whereas that of YSZs-70 is slightly lower (34.6 wt%). At lower external KOH concentrations (1 M, Fig. S28), this trend reverses, with YSZs-70 exhibiting a relatively higher internal concentration, indicating strong OH⁻ absorption by YSZNPs32. The nanoporous channels formed between YSZNPs generate a strong capillary force that locally concentrates hydroxide ions. To maintain electroneutrality, the negatively charged surfaces are compensated by a layer of K⁺ cations, forming a Helmholtz layer at the interface (Fig. S29), leading to ion enrichment in these regions33. Concurrently, extensive deprotonation of the PBI backbone occurs, while the active sites on YSZNPs approach saturation34.Fig. 4Ionic transport behavior and alkaline stability.a Weight change of membranes upon absorption of KOH at different concentrations. b Ionic conductivity of the membranes doped with 6 M KOH. c Ionic conductivity of the YSZs-70 membrane doped with KOH solutions of varying concentrations. d Schematic diagram of ion transport pathways. e Conductivity retention of membrane during immersion in 6 M KOH over time at room temperature (25 °C). All error bars indicate the s.d. of at least three independent measurements. Ionic transport behavior and alkaline stability. a Weight change of membranes upon absorption of KOH at different concentrations. b Ionic conductivity of the membranes doped with 6 M KOH. c Ionic conductivity of the YSZs-70 membrane doped with KOH solutions of varying concentrations. d Schematic diagram of ion transport pathways. e Conductivity retention of membrane during immersion in 6 M KOH over time at room temperature (25 °C). All error bars indicate the s.d. of at least three independent measurements. In G-PBI, OH⁻ migration mainly proceeds through interactions with the imidazole functional groups and along adsorbed-water channels. The temperature dependence of ionic conductivity for all membranes is shown in Fig. 4b. At 80 °C, the YSZs-70 membrane equilibrated in 6 M KOH exhibits a high ion conductivity of 930 mS cm−1, exceeding that of G-PBI (839 mS cm−1). After activation, a conductivity of 311 mS cm−1 is achieved in 1 M KOH (Fig. 4c). Arrhenius analysis reveals only a minor difference in activation energy at 6 M KOH (Figs. S30–S33), with YSZNP-containing membranes showing a slight reduction. Owing to the presence of through-going nanopores in both the skin layer and bulk, ion transport behavior closely resembles that in bulk KOH solution. Accordingly, the ion transference numbers of the porous membranes approach those of KOH aqueous solutions at corresponding concentrations (Fig. S34). In contrast, the relatively dense skin layer of the G-PBI membrane limits continuous free-water channels35,36. At a YSZNP content of 70 wt%, the ion-transport mechanism transitions from polymer- and aqueous-phase dominated conduction to a composite conduction mode. (Fig. 4d). Ion migration proceeds along nanoscale YSZ–PBI interfacial regions, surface hydroxyl networks, adsorbed water phases, and deprotonated PBI sites. Nevertheless, long-range ion transport remains predominantly governed by liquid electrolyte within interconnected pores, following a Grotthuss-type mechanism. Ex situ alkali stability was evaluated by immersing membranes in 6 M KOH and periodically measuring their ionic conductivity. Membranes with higher YSZNPs contents retained conductivity more effectively over time (Figs. 4e, S35 and S36). DFT calculations indicate that in strong alkali, PBI degradation proceeds primarily through imidazole rings opening and main-chain scission37,38. In the composite membranes, OH⁻ preferentially adsorbs on ZrO2 surfaces, forming an interfacial buffer that sequesters hydroxide from the polymer backbone. This mechanism mitigates localized alkaline attacks by reducing the flux of free OH⁻ to vulnerable imidazole sites. MD simulations further elucidated this behavior. RDF and CN analysis (Figs. S37, 38 and Supplementary Data 1) show that in G-PBI, water molecules are randomly distributed (CN < 1), whereas in YSZs-70, the vicinity of the imidazole rings exhibits a mild repulsion towards water. MSD analysis reveals significantly higher diffusion coefficients for both H2O and OH⁻ in the composite system, indicating reduced spatial confinement and weaker interaction constraints. These results suggest a less ordered local structure, which accelerates ion and water transport, thereby enhancing overall conductivity. The ionic conductivity of the YSZs-70 membrane is higher than that of several state-of-the-art AEMs and ISMs13,35,39–42. Electrochemical performance and durability of AWE To evaluate the device-level performance of the composite membranes, we conducted AWE in 6 M KOH at 80 °C (Fig. 5a). Catalyst-free nickel foam was used for both anode and cathode to isolate membrane-related effects and eliminate artifacts arising from catalyst activity or electrode degradation. Under these conditions, the YSZs-70 membrane delivered >1.0 A cm−2 at 2.0 V (Fig. 5b), surpassing all other membrane samples as well as the commercial UPT220 zirconia benchmark (Figs. S39–S42). Electrochemical impedance spectroscopy (EIS) attributed the performance of YSZs-70 to reduced ohmic and charge-transfer losses. At 500 mA cm−2, the ohmic area-specific resistance was 65 mΩ cm2 at 80 °C (Fig. 5c), significantly lower than that of the slightly thinner UPT220 (132 mΩ cm2). Hydrogen permeability measurements under both isobaric and differential pressure conditions (Fig. S43, Tables S3 and S4) show that all composite membranes exhibit lower hydrogen permeability than UPT220, demonstrating better gas barrier properties compared to zirconia-based composite membranes prepared via the conventional NIPS method.Fig. 5Electrochemical performance in AWE.a Schematic illustration of the alkaline water electrolyzer set-up. b Polarization curves of different membranes using Ni foam/Ni foam as electrodes in 6 M KOH at 80 °C. c EIS spectra of the membranes at 500 mA cm−2. d Polarization curves of the YSZs-70 membrane with NiFe-LDH/PtRuC electrodes in 6 M KOH at various temperatures. e Polarization curves of the YSZs-70 membrane with NiFe-LDH/PtRuC electrodes in 1 M KOH at various temperatures. f Polarization curves of various membranes with NiFe-LDH/PtRuC electrodes in 6 M KOH at 80 °C. g EIS spectra of the membranes at 500 mA cm−2. h In situ durability of the YSZs-70 membrane in 6 M KOH at 60 °C under a constant current density of 500 mA cm−2. i EIS spectra before and after the durability test. Electrochemical performance in AWE. a Schematic illustration of the alkaline water electrolyzer set-up. b Polarization curves of different membranes using Ni foam/Ni foam as electrodes in 6 M KOH at 80 °C. c EIS spectra of the membranes at 500 mA cm−2. d Polarization curves of the YSZs-70 membrane with NiFe-LDH/PtRuC electrodes in 6 M KOH at various temperatures. e Polarization curves of the YSZs-70 membrane with NiFe-LDH/PtRuC electrodes in 1 M KOH at various temperatures. f Polarization curves of various membranes with NiFe-LDH/PtRuC electrodes in 6 M KOH at 80 °C. g EIS spectra of the membranes at 500 mA cm−2. h In situ durability of the YSZs-70 membrane in 6 M KOH at 60 °C under a constant current density of 500 mA cm−2. i EIS spectra before and after the durability test. Since high-performance AWE is enabled by active catalysts, concentrated alkaline electrolytes, and elevated operating temperatures, we further evaluated membranes using PtRu/C as the cathode and NiFe-LDH as the anode in 6 M KOH. At 100 °C, the YSZs-70 membrane achieved a current density of 13.1 A cm−2 at 2.0 V (Fig. 5d) with a low internal resistance of 26 mΩ cm2 (Fig. S44). When the temperature was reduced to 80 °C, the membrane still delivered a performance of 10.5 A cm−2 @ 2.0 V; even at 10 °C, a performance of 2.0 V @ 2.3 A cm−2 was maintained (Fig. S45). YSZs-40 exhibited slightly lower performance than that of YSZs-70 but displayed a similar trend (Fig. S46). By contrast, the Zirfon® UPT220 membrane reached only 3.7 A cm−2 at 2.0 V in 6 M KOH at 80 °C (Figs. 5f, S47 and S48). When operated in 1 M KOH solution (Fig. 5e), YSZs-70 still achieved a current density of 4.9 A cm−2 at 2.0 V and 80 °C, comparable to the commercial AEMs (PAP-TP-85, A40; Figs. S49–S52). This demonstrates a broadened operational window for alkaline concentration, whereas traditional PBI-type ISMs exhibit strong dependence on high alkaline loading35. Long-term durability was assessed under a constant current of 500 mA cm−2 in 6 M KOH at 60 °C (Fig. 5h). Over approximately 1000 h of continuous operation, the cell voltage increased only marginally, corresponding to a low degradation rate of ~4.3 μV h−1. Notably, refreshing the KOH solution nearly fully restored the voltage, suggesting that electrolyte degradation, interfacial changes, and bubble accumulation—rather than membrane aging—are responsible for the minor voltage rise. EIS measurements before and after operation (Fig. 5i) confirm negligible membrane degradation. In contrast, the G-PBI membrane exhibits shrinkage and significant thickness loss during cycling (Fig. S53), in addition to oxidative degradation of the polymer, accompanied by a drop in voltage. These observations indicate an increased risk of membrane perforation and gas leakage during operation when using pure gel-state polymer membranes. Enhanced composite diaphragm To further enhance the mechanical robustness of the membrane (Fig. 6a), a commonly used 100 μm PPS mesh was employed as a reinforcing support to fabricate the YSZs-PPS-70 nanoporous composite membrane (Figs. S54–S58). For comparison, a YSZs-PPS-85 membrane, with a composition similar to previously reported zirconia-based membranes (85 wt% YSZNPs and 15 wt% polymer)19,21, was also prepared. This membrane exhibits a BPP exceeding 25 bar, along with high porosity and low area resistance (0.058 Ω cm2, YSZs-PPS-70 is 0.055 Ω cm2, Figs. S59–S62), both of which are approximately half that of UPT220. The PPS framework protects the membrane from compressive stress imposed by the nickel substrate and effectively prevents mechanical failures, such as membrane puncture. A Cryo-SEM image of the YSZs-PPS-70 membrane is shown in Fig. 6b. Because PPS fibers are non-conductive, the incorporation of the support leads to a moderate decrease in overall ionic conductivity; nevertheless, the membrane maintains a high conductivity of 756 mS cm−1 (Fig. S63) and has a lower hydrogen permeability coefficient (Fig. S64, Tables S3 and S4). AWE tests were conducted using nickel foam electrodes. Although the increased ohmic resistance results in slightly reduced performance, the YSZs-PPS-70 membrane still achieves 2.5 A cm−2 at 2.3 V in 6 M KOH (Figs. 6c and S65), with an ohmic resistance of 94 mΩ cm2—lower than UPT220. When paired with NiFe-LDH and PtRu/C electrodes, the system delivers >7.5 A cm−2 at 2.0 V in 6 M KOH at 100 °C (Fig. 6d), accompanied by a low internal resistance of 43 mΩ cm2 (Fig. 6e). The water electrolysis performance of YSZs-PPS-85 is comparable to that of YSZs-PPS-70 (Figs. S66 and S67). As expected, the YSZs-PPS-70 membrane shows the highest hydrogen purity of 99.993% at 1.0 A cm−2 (Fig. S68), higher than UPT220 (99.983%). A higher bubble point pressure corresponds to a smaller maximum pore size and the absence of through-pores, thereby effectively suppressing gas crossover between the anode and cathode. Meanwhile, hydrogen purity is also strongly influenced by the actual thickness of the MEA and the operating current density.Fig. 6Performance of reinforced membranes in the laboratory and scaled-up cells.a Schematic diagram of the YSZs-PPS-70 membrane preparation process. b Cross-sectional structure of cryo-SEM near PPS fibers (local extrusion induced during cutting); c polarization curves YSZs-PPS-70 membrane using Ni foam/Ni foam electrodes in 6 M KOH, and the inset shows EIS spectra at 500 mA cm−2. d Temperature-dependent polarization curves of the YSZs-PPS-70 membrane with NiFe-LDH/PtRuC electrodes in 6 M KOH. e EIS spectra at different temperatures at a current density of 500 mA cm−2. f In situ durability of the YSZs-PPS-70 membrane in 6 M KOH at 60 °C under a constant current density of 500 mA cm−2, and the insets display photographs of the anode end plate and YSZs-PPS-70 membrane after 5000 h of operation. g EIS spectra before and after the durability test. h Photograph of the 64 cm2 single-cell electrolyzer; i polarization curves the YSZs-PPS-70 membrane measured in 64 cm2 electrolyzer using Ni foam/Ni foam electrodes in 6 M KOH, and the inset shows EIS spectra at 500 mA cm−2. j In situ durability of the YSZs-PPS-70 membrane in the 64 cm2 electrolyzer. Performance of reinforced membranes in the laboratory and scaled-up cells. a Schematic diagram of the YSZs-PPS-70 membrane preparation process. b Cross-sectional structure of cryo-SEM near PPS fibers (local extrusion induced during cutting); c polarization curves YSZs-PPS-70 membrane using Ni foam/Ni foam electrodes in 6 M KOH, and the inset shows EIS spectra at 500 mA cm−2. d Temperature-dependent polarization curves of the YSZs-PPS-70 membrane with NiFe-LDH/PtRuC electrodes in 6 M KOH. e EIS spectra at different temperatures at a current density of 500 mA cm−2. f In situ durability of the YSZs-PPS-70 membrane in 6 M KOH at 60 °C under a constant current density of 500 mA cm−2, and the insets display photographs of the anode end plate and YSZs-PPS-70 membrane after 5000 h of operation. g EIS spectra before and after the durability test. h Photograph of the 64 cm2 single-cell electrolyzer; i polarization curves the YSZs-PPS-70 membrane measured in 64 cm2 electrolyzer using Ni foam/Ni foam electrodes in 6 M KOH, and the inset shows EIS spectra at 500 mA cm−2. j In situ durability of the YSZs-PPS-70 membrane in the 64 cm2 electrolyzer. Long-term durability was further evaluated at a constant current of 500 mA cm−2 (60 °C, 6 M KOH), as shown in Fig. 6f. Pronounced voltage fluctuations occurred during the first 5000 h, accompanied by oxidation-induced blackening of the anode flow channels. After replacing the cell hardware, the operating voltage returned to its initial value. Over the entire ~7000 h test period, the voltage decay rate was 13.8 μV h−1. The ohmic internal resistance increased by 52 mΩ cm² over the first 5000 h, followed by an additional rise of 14 mΩ cm² after cell replacement (Figs. 6g and S69). Finally, a large-area (64 cm2) YSZs-PPS-70 membrane was assembled into a single-cell stack (Figs. 6h, i,S70 and S71) to assess scale-up feasibility. At 500 mA cm−2, the stack operated at approximately 2.14 V in 6 M KOH at 80 °C using Ni foam electrodes, which is higher than the small-area (4 cm2) cell under identical conditions (Fig. 6C). Such performance degradation upon scaling is a well-recognized challenge in zero-gap water electrolysis and primarily arises from cell engineering factors rather than intrinsic limitations of the membrane material43. Specifically, scaling up to 64 cm2 significantly increases the difficulty of achieving uniform mechanical compression (leading to increased contact resistance, Fig. S70b), efficient bubble removal, and homogeneous electrolyte distribution across the enlarged flow field. The 1-cell stack also exhibited satisfactory durability (Fig. 6j): at 32 A (60 °C, 6 M KOH), it sustains continuous operation for 200 h with a voltage decay rate of 92.4 μV h−1 and shows no signs of failure, demonstrating strong practical potential for large-scale AWE. Future work focusing on the optimization of large-area devices will be crucial for translating the high intrinsic performance of such composite membranes into practical industrial-scale electrolyzers. To further assess economic feasibility, the cost of the raw materials is summarized (Table S5). This approach employs low-cost inorganic ceramic materials as the primary component, comparable to Zirfon®. In comparison with AEM, the membrane offers clear advantages in material cost and environmental compatibility during fabrication. The benefits of such low-cost, highly stable composite membranes have also been demonstrated in other electrochemical systems, including carbon dioxide and ammonia electrolysis44–46. Discussion In this work, we developed a high-performance composite membrane for AWE by integrating YSZNPs with a PBI matrix. Using a sol-gel route, PA served as a molecular bridge to uniformly disperse YSZNPs and establish strong interfacial bonding within the polymer, yielding a composite membrane featuring a highly uniform nanoporous structure. The resulting membrane combines several key advantages—including high porosities (~85%), high bubble-point pressure (>25 bar), and low area resistance. In the AWE tests, the composite membrane exhibits excellent electrochemical performance and long-term stability. Further optimization may involve using polybenzimidazoles with enhanced alkaline stability tolerance—such as poly(2,2′-(m-mesitylene)-5,5′-bibenzimidazole) (mes-PBI)47, poly(2,2′-(1,4-naphthalene)-5,5′-benzimidazole) (NPBI)48 and other sterically hindered structures—to further strengthen the chemical stability of the polymer backbone. Overall, this study provides an effective materials-design strategy to simultaneously advance conductivity, gas barrier property, and long-term durability in zirconia-based composite membranes for AWE. Methods Materials 3,3′-diaminobenzidine (DAB, 99%, Energy Chemical), terephthalic acid (TPA, 99%, Energy Chemical), phosphoric acid (PA, 85%, Energy Chemical), polyphosphoric acid (PPA, 115%, Energy Chemical), and potassium hydroxide (95%, Greagent). Yttria-stabilized zirconia nanoparticles (YSZNPs, 50 nm, Suzhou Yuante New Material Co., Ltd). PPS mesh (100 μm, 120-mesh, Zhejiang Zheng Ben New Material Co., Ltd). Nickel foam (Suzhou keshenghe, ≥99.99%, 110 ppi, 0.5 mm). Zirfon® UPT220 and Zirfon® UPT500 separators were provided by Agfa. All chemicals were used as received without further purification. Preparation of G-PBI membrane G-PBI was synthesized following a previously reported procedure49. One hundred sixty grams of PPA, 1.66 g of TPA, and 2.14 g of DAB were added to a three-necked flask under a nitrogen atmosphere. The mixture was heated to 140 °C and stirred until fully dissolved. The temperature was then raised to 190 °C, causing the mixture to thicken and turn dark brown. The viscous PBI solution was cast onto a glass plate and spread into a uniform membrane using a doctor blade. The membrane was hydrolyzed at room temperature (55% relative humidity) to a final thickness of 280 ± 20 μm. To remove residual PA, the membrane was soaked in deionized water, with the water replaced every 12 h, until the solution reached neutral pH. Finally, the membrane was immersed in KOH solutions of different concentrations at room temperature for at least 3 days to complete doping and deprotonation. Preparation of YSZs-n membrane The preparation process of the YSZs-70 composite membrane was similar to that of G-PBI, with the exception that the amount of PPA was reduced to 120 g. After the reaction was complete, the system was cooled to 140 °C. Next, 9 g of YSZNPs were dispersed in PA solution with a total volume of 35 mL and stirred until a uniform suspension was obtained. This dispersion was then added to the G-PBI reaction solution, which caused the viscosity of the mixture to decrease. Vigorous stirring was performed for 1 h to ensure homogeneity. The remaining treatment steps were identical to those for the G-PBI membrane. Preparation of YSZs-PPS-70 membrane The above steps are similar; 9 g of YSZNPs is dispersed in a PA solution with a total volume of 40 ml. Pour the aforementioned PPA mixture onto a glass plate preheated to 60 °C. After casting the first layer, attach the PPS mesh to the first YSZs-70 layer and let it stand for 5 min. Next, cast the next layer on top of the mesh, then leave the composite structure to stand for hydrolysis. The final membrane thickness is 320 ± 20 μm. Preparation of YSZs-PPS-85 membrane The above steps are similar. Then, 22 g of YSZNPs is dispersed in a PA solution with a total volume of 50 mL. The preparation method of the composite membrane is consistent with that of the YSZ-PPS-70 membrane. Preparation of PSU-85 membrane The PSU-85 composite membrane was prepared via the conventional NIPS method. The composite membrane is free of the PPS mesh reinforcement layer. Two grams of PSU polymer was added to 10 g of NMP solution and stirred at 60 °C until fully dissolved. After cooling to room temperature, 11.2 g of YSZNPs was added and mixed uniformly. The casting solution was then cast onto a clean glass plate using a doctor blade. Subsequently, the glass plate with the cast film was immersed in deionized water for phase inversion until the composite membrane formed. The final thickness of the membrane is approximately 280 ± 20 μm. Membrane characterization The chemical structure and elemental composition of the fabricated composite membranes are systematically characterized by X-ray diffraction (XRD, Empyrean Nano diffractometer), FT-IR (Nicolet is 50 + IN10), and XPS (Thermo Kalpha spectrometer). Cryo-scanning electron microscopy (Cryo-SEM, Hitachi Regulus 8220) coupled with energy-dispersive X-ray spectroscopy (EDX) is utilized to observe the microscopic morphology and determine the elemental distribution of the polymer matrix. Thermogravimetric analysis (TGA) was performed using a Netzsch thermal analyzer. Bubble point pressure (BPP) was measured using a capillary flow porometer (POROMETER Porolux 500). Static water contact angles were measured using a Dataphysics OCA20 goniometer and analyzed by using Washburn method. Analyze mechanical properties and compression creep using the QM-1190 universal testing machine. Additional details are provided in the Supplementary Information. Porosity Porosity was determined by immersing the membranes in deionized water at 25 °C for 100 h to obtain the wet mass and wet volume. Samples were then dried at 100 °C for 12 h to determine the dry mass. Porosity was calculated using the following equation:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Porosity}\left(\%\right)=\frac{{M}_{{wet}}-{M}_{{dry}}}{{V}_{{wet}}\times {\rho }_{{water}}}\times 100$$\end{document}Porosity%=Mwet−MdryVwet×ρwater×100where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${M}_{{{{\rm{wet}}}}}$$\end{document}Mwet is the mass of the membrane after immersion in water, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${M}_{{{{\rm{dry}}}}}$$\end{document}Mdry is the mass of the dried membrane, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{{{{\rm{wet}}}}}$$\end{document}Vwet is the volume of the wet membrane, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rho }_{{{\mathrm{water}}}}$$\end{document}ρwater is the density of water. Alkaline electrolyte uptake The membrane samples were immersed in KOH solutions of different concentrations at 25 °C for at least 3 days. The KOH content and water content of the membrane were determined by the gravimetric method as described in previous work, with the calculations shown as follows:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${KOH}\,{Uptake}(\%)=\frac{{w}_{{KOH}+{Water}+{Membrane}}{-}{w}_{{Membrane}}}{{w}_{{Membrane}}}\times 100$$\end{document}KOHUptake(%)=wKOH+Water+Membrane−wMembranewMembrane×1002\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${w}_{{Membrane}}(\%)=\frac{{w}_{{Membrane}}}{{w}_{{KOH}+{Water}+{Membrane}}}\times 100$$\end{document}wMembrane(%)=wMembranewKOH+Water+Membrane×1003\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${w}_{{water}}\left(\%\right)=\frac{{w}_{{KOH}+{Water}+{Membrane}}{-}{w}_{{KOH}+{Membrane}}}{{w}_{{KOH}+{Water}+{Membrane}}}\times 100$$\end{document}wwater%=wKOH+Water+Membrane−wKOH+MembranewKOH+Water+Membrane×1004\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${w}_{{KOH}}\left(\%\right)=\frac{{w}_{{KOH}+{Membrane}}{-}{w}_{{Membrane}}}{{w}_{{KOH}+{Water}+{Membrane}}}\times 100$$\end{document}wKOH%=wKOH+Membrane−wMembranewKOH+Water+Membrane×100 Among them, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${w}_{{{\mathrm{KOH}}}+{{\mathrm{Water}}}+{{\mathrm{Membrane}}}}$$\end{document}wKOH+Water+Membrane refers to the weight of the membrane sample after carefully wiping off the excess solution on its surface; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${w}_{{{\mathrm{KOH}}}+{{\mathrm{Membrane}}}}$$\end{document}wKOH+Membrane is the weight of the sample after vacuum drying at 100 °C for 12 h; and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${w}_{{{\mathrm{Membrane}}}}$$\end{document}wMembrane represents the weight of the membrane sample after extensive washing in deionized water until the solution reaches a neutral pH, followed by drying at 100 °C for 12 h. Ionic conductivity and activation energy Ionic conductivity was measured using the two-probe AC impedance method. After removing excess surface electrolyte, the membrane was placed between two electrodes of an electrochemical workstation (IviumStat, Netherlands). The membranes sealed the KOH-doped sample to prevent dehydration. The membrane’s impedance was tested using AC impedance with a frequency range of 1Hz–100 kHz. The conductivity σ (mS cm−1) was calculated using the following formula:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sigma=\frac{l}{{RA}}$$\end{document}σ=lRAwhere l (cm) refers to the distance between the electrodes, and A (cm2) is the effective cross-sectional area of the membrane, respectively. R (Ω) values were obtained from the Nyquist plots. The activation energies of the PA-doped membranes were calculated according to the Arrhenius equation:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${ln}\sigma={ln}{\sigma }_{0}-\frac{{E}_{a}}{{RT}}$$\end{document}lnσ=lnσ0−EaRTwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${E}_{a}$$\end{document}Ea represents the activation energy, measured in kJ mol−1, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R$$\end{document}R stands for the gas constant, with a value of 8.314 J mol−1 K–1, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T$$\end{document}T signifies the absolute temperature in Kelvin (K). Area resistance and ion transference At 25 °C, the area resistance of the membrane was measured in a customized H-type cell. The composite membrane was mounted between two chambers equipped with Pt electrodes, and the gap was filled with a 30 wt% KOH solution. An electrochemical workstation was used to measure the resistance. The area resistance was calculated as follows:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$AR=A\times (R-{R}_{{blank}})$$\end{document}AR=A×(R−Rblank)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R$$\end{document}R (Ω) is the resistance of the electrodes with the membrane, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${R}_{{blank}}$$\end{document}Rblank (Ω) is the resistance of the electrodes without the membrane and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A$$\end{document}A (cm2) represents the available cross-sectional membrane area. The tested membrane is sandwiched between potassium hydroxide solutions with different concentrations to assemble a concentration cell. Selective ion transport through the membrane induces the generation of transmembrane potential. The ion transference number can be calculated by measuring this potential in combination with the Nernst equation. Detailed information is provided in the Supplementary Information. Alkaline stability Membrane samples were immersed in a 6 M KOH solution at room temperature, and the membranes intended for long-term storage were placed in sealed PTFE vials. The ionic conductivity was measured periodically to characterize the alkaline stability. Hydrogen permeation and purity test A single cell was assembled using the membrane (pre-immersed in 6 M KOH solution for 12 h), with a Pt/C (40 wt%, Johnson Matthey Hispec 4000, 0.5 mg cm–2 Pt) as both anode and cathode. Catalyst ink was prepared by dispersing catalyst powder, ionomer (Nafion D520, 5 wt%), distilled water and isopropanol, with the ionomer content in the catalytic layer being 20 wt%. Before testing, both the cathode and anode chambers were purged with humidified N2 for 1 h, followed by humidified H2 on the cathode for 30 min. After backpressure stabilization of the two chambers for 30 min, the hydrogen crossover limiting current of the membrane under different temperature conditions was obtained via linear sweep voltammetry (LSV). The range was 0–0.8 V, and the scanning rate was 5 mV s−1. Hydrogen permeability was calculated using the following equation:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$K=\frac{{{{\rm{i}}}}}{2{FP}}$$\end{document}K=i2FPwhere K refers to the hydrogen permeability (mol cm−2 s−1 kPa−1), F represents the Faraday constant (96,485 C mol−1), and P stands for the hydrogen partial pressure difference (kPa). The purity of hydrogen was analyzed using a Fuli GC 9790II gas chromatography system equipped with a thermal conductivity detector, with high-purity argon as the carrier gas. Gas composition was monitored in situ and in real time. Hydrogen gas produced during electrolysis was collected from the cathode gas chamber at various current densities. All gas chambers were carefully sealed during operation. The outlet gas passed through a back pressure valve, was dried in a drying bottle, and subsequently introduced into the gas chromatograph for quantitative analysis. Computational methods Molecular models were built with Materials Studio. The models were constructed in a cubic simulation box using the Amorphous Cell module and the COMPASSIII force field. All models were established under periodic boundary conditions at 298 K. The models were further optimized by MD simulations using the Forcite module. The Perdew–Burke–Ernzerhof (PBE) functional within the generalized gradient approximation was used to describe the exchange-correlation energy of the system. After obtaining the model density, the final equilibration of the packing models was carried out with the following sequence of simulation steps. Before the dynamic simulation, 100 annealing cycles with NVT–MD annealing at 1 atm, using a time step of 1.0 fs, where the initial temperature was 433 K and the middle-cycle temperature was 633 K; Then, 2000 ps MD in the NPT ensemble at 1 atm and 353 K was carried out to compress the cell to a reasonable density. The precise Nose–Hoover temperature control mode and Berendsen pressure control mode are used in the dynamic process. The simulation trajectories are finally extracted for subsequent data analysis. Related configurations have been provided as Supplementary Data 1. AWE test The membrane electrode assembly (MEA) was prepared using the catalyst-coated substrate method. The anode was made with NiFe-LDH based on a previously reported method50. For NiFe-LDH synthesis, 0.350 g Ni(NO3)2·6H2O, 0.162 g Fe(NO3)3·9H2O, and 0.32 g urea were fully dissolved in 30 mL deionized water under vigorous magnetic stirring. A pre-cleaned 2 × 2 cm nickel foam (NF) was immersed in the above homogeneous solution. The mixture was sealed in a Teflon-lined autoclave and hydrothermally treated at 120 °C for 12 h. The obtained sample was rinsed thoroughly with deionized water and vacuum-dried at 25 °C for 12 h. The cathode was a PtRu/C catalyst (TANAKA TEC86E86) supported on carbon paper (SGL 28BC). The ionomer (Nafion D520) in the catalyst layer made up 20% by mass, and the Pt loading was set at 0.5 mg cm−2. The zero-gap electrolysis cell consisted of a pure nickel plate, gaskets, and the MEA, with temperature controlled by an external heater. AWE tests were conducted using the DH7002A + DH7101B electrochemical workstation (DongHua Analytical), with a circulating KOH solution at 20 mL min−1 on both sides of the cell. Polarization curves without iR correction were obtained via LSV with a scan rate of 10 mV s−1 and a voltage range of 1.25 V to 2.0 V. EIS was performed at a constant current density of 500 mA cm−2 (DH7002A, frequency range 0.1–100 kHz). The in situ durability of the membrane was tested using the LANHE CT2001A battery testing system (electrodes: NF/NF, current density: 500 mA cm−2, electrolyte: 6 M KOH, temperature: 60 °C). All tests were conducted at ambient pressure. For the single-cell stack (64 cm2), titanium plates were used, and constant current testing was conducted using a high-power DC power supply (JIAMAN DC POWER). Durability testing was carried out in 6 M KOH solution at 60 °C with a current density of 500 mA cm−2. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Transparent Peer Review File Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Transparent Peer Review File Source data Source Data Source Data