Waste PET-derived BDC-ZIF-8 as an electrolyte-retaining filler for NPBI-based ion-solvating mixed-matrix membranes in alkaline water electrolysis

Ion-solvating membranes (ISMs) have emerged as promising dense separators for alkaline water electrolysis (AWE) because they enable ion transport through absorbed KOH/water phases without relying on fixed ionic groups. However, enhancing electrolyte retention while preserving membrane integrity and long-term alkaline stability remains a key challenge. In this study, mixed-ligand BDC-ZIF-8 was synthesized using benzene-1,4-dicarboxylic acid recovered from waste PET depolymerization and incorporated into a poly [2,2’-(1,4-naphthalene)-5,5′-bibenzimidazole] (NPBI) matrix to fabricate NPBI/BDC-ZIF-8 mixed-matrix membranes (MMMs). The effects of BDC-ZIF-8 loading on electrolyte uptake, dimensional stability, gas-blocking integrity, ion transport, and AWE performance were systematically investigated. The BDC-ZIF-8 filler retained a ZIF-8-like crystalline framework while exhibiting a high BET surface area of 1488 m 2 g -1 and a porous structure. Its incorporation increased the water fraction from 34% to 37% and the KOH fraction from 17% to 23%, while mitigating excessive dimensional changes after KOH uptake. The resulting MMMs maintained bubble point pressures of 6.2–6.5 bar, comparable to pristine NPBI, indicating that dense membrane integrity was preserved. In addition, the area resistance decreased from 0.067 Ω cm 2 for pristine NPBI to 0.037 Ω cm 2 for the 40 wt% MMM, leading to improved initial AWE cell performance. The NPBI/BDC-ZIF-8 20 wt% MMM also sustained 100 h single-cell operation at 0.3 A cm -2 in 6 M KOH at 80 °C, with an average cell voltage of 1.798 V. However, high BDC-ZIF-8 loadings induced greater morphological heterogeneity and less stable area resistance during prolonged ex situ alkaline exposure. These results demonstrate that waste PET-derived BDC-ZIF-8 can function as an electrolyte-retaining MOF filler for NPBI-based ISMs, while also emphasizing that filler loading must be optimized to balance electrolyte retention, ion transport, membrane morphology, and alkaline stability.

Renewable-energy-driven green hydrogen production has attracted increasing attention as a key strategy for achieving carbon-neutral energy systems (Tüysüz, 2024; Chatenet et al., 2022). Water electrolysis is one of the most representative technologies for converting electrical energy into hydrogen as a chemical energy carrier (Chatenet et al., 2022; Zeng and Zhang, 2010). Among various electrolysis technologies, alkaline water electrolysis (AWE) has been actively investigated for cost-effective large-scale hydrogen production because of its mature process basis and compatibility with non-precious-metal electrodes (Tüysüz, 2024; Zeng and Zhang, 2010; Brauns and Turek, 2020). However, AWE is typically operated in concentrated alkaline electrolytes at elevated temperatures, requiring electrolyte-contacting components, including electrodes and membranes, to maintain chemical and mechanical stability under harsh operating conditions (Tüysüz, 2024; Chatenet et al., 2022; El-Shafie, 2023). In particular, the membrane must physically separate the electrodes while enabling efficient ion transport, thereby directly affecting cell efficiency, safety, and long-term durability (Tüysüz, 2024; Brauns et al., 2021; Brauns and Turek, 2020). Therefore, the development of alkaline-stable membranes or separators with low ionic resistance, suppressed gas crossover, and sufficient mechanical durability is essential for high-performance AWE (Tüysüz, 2024; Brauns et al., 2021; Brauns and Turek, 2020). Porous diaphragms or separators are widely used in conventional AWE because their electrolyte-filled pore networks enable hydroxide ion transport. However, they generally suffer from an intrinsic trade-off between ion conduction and gas separation (Henkensmeier et al., 2024; Aili et al., 2023). Increasing porosity or electrolyte permeability can reduce ionic resistance, but it may also increase the risk of H 2 /O 2 gas crossover. Conversely, improving gas-blocking ability can restrict ion transport and increase ohmic loss (Henkensmeier et al., 2024; Brauns et al., 2021). As an alternative, anion exchange membranes (AEMs) have been proposed to promote OH − transport through fixed cationic functional groups. However, the long-term alkaline stability of these ion-conducting groups remains a critical issue under concentrated KOH and high-temperature operation (Henkensmeier et al., 2024; Kraglund et al., 2019). In this context, ion-solvating membranes (ISMs) have emerged as a promising dense membrane concept. Unlike AEMs, ISMs do not rely on fixed ionic groups; instead, ion transport occurs through absorbed KOH/water phases within a chemically stable polymer matrix (Aili et al., 2023; Kraglund et al., 2019). Nevertheless, while enhanced KOH/water retention can lower ionic resistance, excessive electrolyte uptake can cause swelling, dimensional instability, and mechanical weakening. Thus, ISMs should be designed to balance electrolyte uptake, ion transport, and membrane integrity under harsh alkaline conditions (Aili et al., 2023; Henkensmeier et al., 2024; Kraglund et al., 2019).

Polybenzimidazole (PBI)-based polymers have been widely studied as matrix materials for ISMs because their rigid aromatic heterocyclic backbones and benzimidazole groups provide excellent chemical, thermal, and mechanical stability (Kraglund et al., 2019; Hu et al., 2022b). The benzimidazole groups can interact with the KOH/water phase through hydrogen bonding and polar interactions, thereby influencing the electrolyte environment inside the membrane (Aili et al., 2016; Hu et al., 2022a). Among PBI-based polymers, poly [2,2’-(1,4-naphthalene)-5,5′-bibenzimidazole] (NPBI) has been proposed as a suitable matrix for maintaining membrane integrity under high-temperature alkaline conditions. This behavior is associated with the bulky and rigid naphthalene units, which increase backbone rigidity and help preserve dimensional stability (Hu et al., 2022b; Hu et al., 2022a). However, dense NPBI membranes still have inherent limitations in ion transport because ionic conduction is largely governed by the amount and distribution of the retained KOH/water phase. Therefore, additional structural design is required to improve electrolyte retention and reduce ionic resistance while preserving the stability of the NPBI matrix (Kraglund et al., 2019; Hu et al., 2022b; Aili et al., 2017). Introducing porous functional fillers into NPBI is a promising strategy for constructing mixed-matrix membranes (MMMs) with improved ion-solvating transport properties (Wu et al., 2016; He et al., 2017; Chen et al., 2021; Li et al., 2020).

Metal–organic frameworks (MOFs) are promising functional fillers for tuning the transport properties of polymer membranes because of their high specific surface areas, tunable pore structures, and diverse coordination environments derived from metal nodes and organic linkers (Cheng et al., 2022; He et al., 2017; Lin et al., 2018). In MOF-containing MMMs, porous fillers can help retain KOH/water within the polymer matrix, and the retained electrolyte phase can provide ion-transport pathways in ISMs. In addition, rigid MOF particles can serve as physical reinforcing domains, potentially improving dimensional and mechanical stability. However, excessive filler loading can impair film-forming ability and induce particle aggregation, polymer–filler interfacial heterogeneity, and mechanical weakening (Cheng et al., 2022; He et al., 2017; Chen et al., 2021). Among MOFs, zeolitic imidazolate frameworks (ZIFs) are notable for their high porosity and structural tunability based on Zn-imidazolate coordination frameworks. Pristine ZIF-8 has been reported to retain its crystalline structure after treatment in 8 M NaOH at 100 °C for 24 h, indicating its alkaline resistance under harsh conditions. However, the alkaline stability of MOF fillers can strongly depend on framework composition and exposure conditions (Latrach et al., 2022; Park et al., 2006; Titirici and Szilágyi, 2022). Building on the advantages of ZIF-based frameworks, this study employs mixed-ligand BDC-ZIF-8 as a functional filler to tune the pore architecture and linker-derived chemical environment of NPBI-based ISMs. BDC-ZIF-8 has been reported to retain a ZIF-like crystalline framework while forming a porous structure (Cheng et al., 2022; Latrach et al., 2022; Feng et al., 2020). Such a pore structure can provide electrolyte-retaining domains that help accommodate the KOH/water phase within the polymer matrix (Chen et al., 2021; Latrach et al., 2022; Feng et al., 2020). In addition, the BDC-derived carboxylate coordination environment can offer polar interaction sites that may influence electrolyte retention and ion-solvating behavior within the NPBI matrix (Latrach et al., 2022; Titirici and Szilágyi, 2022). Therefore, in NPBI/BDC-ZIF-8 MMMs, filler loading, dispersion state, polymer–filler interfaces, and alkaline structural stability should be systematically considered to understand the balance between initial ion transport enhancement and long-term membrane stability.

Polyethylene terephthalate (PET) contains terephthalate units in its repeating structure and can be converted into terephthalic acid (TPA), also known as benzene-1,4-dicarboxylic acid (BDC), through alkaline depolymerization (Wang et al., 2024; Nouira and Bekri-Abbes, 2025; Guo et al., 2025). Recovered BDC has been used as an organic linker for MOF synthesis, providing a route to upcycle waste PET into porous functional materials (Wang et al., 2024; Lo et al., 2016; Deleu et al., 2016). MOFs synthesized from waste PET-derived BDC have been applied to adsorption, separation, catalysis, and CO 2 capture (Wang et al., 2024; Lo et al., 2016; Chinglenthoiba et al., 2024; Waribam et al., 2023). However, the use of waste PET-derived MOFs as electrolyte-retaining fillers in NPBI-based ion-solvating MMMs for AWE has been reported in only a limited number of studies. In particular, the effects of MOF loading on KOH/water retention, ion transport, membrane integrity, and long-term resistance stability under alkaline exposure have not been systematically investigated. In this study, mixed-ligand BDC-ZIF-8 was synthesized using BDC recovered from waste PET depolymerization and incorporated into an NPBI matrix to fabricate NPBI/BDC-ZIF-8 ion-solvating MMMs. The prepared membranes were evaluated by correlating BDC-ZIF-8 loading with membrane morphology, electrolyte uptake, dimensional and mechanical stability, ion transport, and AWE cell performance. In addition, ex situ long-term area resistance measurements were compared with alkaline stability analyses of BDC-ZIF-8 to examine whether the initial reduction in ionic resistance could be maintained during prolonged alkaline exposure. This study demonstrates waste PET-derived BDC-ZIF-8 as an electrolyte-retaining MOF filler for NPBI-based ISMs and provides design considerations for MOF-filled membranes in AWE, where initial performance enhancement must be balanced with morphology control and ex situ alkaline resistance stability.

Materials

1,4-Naphthalenedicarboxylic acid (NDA, 95%) and methanesulfonic acid (MSA, 99%) were purchased from TCI. Phosphorus pentoxide (P 2 O 5, 84%), 3,3′-diaminobenzidine (DAB, 98%), and hydrochloric acid (HCl, 36.5%) were purchased from Alfa Aesar. Sodium hydroxide (NaOH, 98%) and ethanol (EtOH, 95%) were purchased from Daejung Chemicals, Korea. Deuterated dimethyl sulfoxide (DMSO-d 6, 99.8%) was obtained from Eurisotop. For the synthesis of mixed-ligand BDC-ZIF-8, zinc nitrate hexahydrate (98%) and 2-methylimidazole (2-mI, 97%) were purchased from Alfa Aesar. N,N-dimethylformamide (DMF, 99.8%) and potassium hydroxide (KOH, 85%) were purchased from Daejung Chemicals, Korea. Methanol (MeOH, 99.8%) and N,N-dimethylacetamide (DMAc, 99.8%) were purchased from Duksan Chemicals, Korea. Ni felt was purchased from Bekaert. The Ni felt had a purity of >99.99% and a porosity of 60%. Commercial Zirfon 500, specifically ZIRFON PERL UTP 500 from Agfa, was used as a reference separator. The nominal thickness and porosity of Zirfon 500 were 500 ± 50 μm and 55% ± 10%, respectively.

Synthesis of NPBI

Poly [2,2’-(1,4-naphthalene)-5,5′-bibenzimidazole] (NPBI) was synthesized by polycondensation in Eaton’s reagent based on a previously reported method (Hu et al., 2022b). Eaton’s reagent was prepared by adding P 2 O 5 (2.5 g) and MSA (25 g) to a flask. The mixture was stirred at 50 °C for 12 h until P 2 O 5 was completely dissolved. For NPBI synthesis, Eaton’s reagent (7.5 mL) and DAB (2.15 g, 10 mmol) were added to a 100 mL flask and stirred until a homogeneous solution was obtained. NDA (2.16 g, 10 mmol) was then added, corresponding to a DAB:NDA molar ratio of 1:1. The reaction mixture was heated to 140 °C and stirred for 5 h to conduct polycondensation. After the reaction, the highly viscous reaction mixture was slowly poured into 500 mL of deionized (DI) water to precipitate the polymer. The resulting coil-shaped NPBI was collected by filtration and washed three times with DI water. The polymer was then neutralized in an aqueous NaHCO 3 solution overnight to remove residual acid and washed three times with DI water. Finally, the polymer was dried in a vacuum oven at 120 °C overnight to obtain NPBI. The NPBI yield was >78%.

Recovery of BDC from waste PET

Waste PET-derived BDC was recovered by alkaline hydrolysis of waste PET. Transparent beverage PET bottles were washed with DI water and dried in an oven at 60 °C for 2 h. The dried PET bottles were cut into flakes with an approximate size of 2 × 1 cm. PET flakes (2 g) were placed in a round-bottom flask containing EtOH (60 mL) and DI water (40 mL). NaOH was then added at 15 wt% relative to the PET mass. The mixture was stirred at 400 rpm and 80 °C for 12 h under ambient pressure to depolymerize PET under alkaline conditions. After the reaction, the mixture was cooled to room temperature, and DI water (100 mL) was added. HCl (36.5%) was then added until the solution pH reached <3 to precipitate BDC from the sodium terephthalate solution. The suspension was stirred for 2 h. The precipitated white solid was collected by vacuum filtration, washed with DI water and EtOH, and dried in a vacuum oven at 100 °C for 16 h. The recovered product is referred to as BDC throughout this manuscript, following common MOF linker terminology. The BDC recovery yield was >98%.

The BDC recovery yield was calculated using Equation 1: where m BDC and m PET are the masses of recovered BDC and PET used for depolymerization, respectively. M BDC and M PET repeat unit are the molecular weights of BDC and the PET repeating unit, respectively, with values of 166 and 192 g mol -1.

Synthesis of mixed-ligand BDC-ZIF-8

Mixed-ligand BDC-ZIF-8 was synthesized based on a previously reported method, using waste PET-derived BDC as the dicarboxylate linker (Latrach et al., 2022). Zinc nitrate hexahydrate (2.009 g, 6.756 mmol), 2-methylimidazole (0.501 g, 6.102 mmol), and recovered BDC (0.101 g, 0.609 mmol) were dissolved in DMF (60 mL) under stirring until a homogeneous solution was obtained. The molar ratio of Zn:2-mI:BDC was 6.756:6.102:0.609, corresponding approximately to 11.1:10.0:1.0. The solution was transferred to a 100 mL Teflon-lined autoclave and heated at 140 °C for 24 h under solvothermal conditions. After cooling to room temperature, the solid product was collected by centrifugation at 4000 rpm for 10 min and washed three times with DMF. The collected powder was immersed in MeOH for 3 days for solvent exchange. The resulting powder was vacuum-dried at room temperature for 16 h. The obtained mixed-ligand Zn-mI-BDC-type metal–organic framework is denoted as BDC-ZIF-8 in this study. Before N 2 adsorption–desorption measurements, the BDC-ZIF-8 powder was degassed at 70 °C for 16 h under vacuum. N 2 adsorption–desorption isotherms were measured at 77 K.

Fabrication of NPBI/BDC-ZIF-8 ion-solvating mixed-matrix membranes

NPBI/BDC-ZIF-8 ion-solvating MMMs were prepared by solution casting. Solution A was prepared by dissolving NPBI (0.1 g) in DMAc (4 mL) under stirring at room temperature for 3 h. Solution B was prepared by dispersing BDC-ZIF-8 powder in DMAc (1 mL) at loadings of 10, 20, 30, 40, and 50 wt% relative to the NPBI mass. Solution B was stirred for 3 h and then sonicated for 10 min. Solution A was then added to solution B, and the resulting mixture was stirred at 80 °C for 2 h to obtain a homogeneous casting solution. The casting solution was poured into a glass dish with a diameter of 5.5 cm preheated to 80 °C. The solvent was evaporated at 80 °C for 24 h to form membranes. The membranes were further dried in a vacuum oven at 80 °C for 16 h to remove residual solvent. The average dry membrane thickness was 23 ± 2 μm. For resistivity calculations, the KOH-doped wet thickness of each membrane was measured individually. The dried membranes were immersed in 6 M KOH at 80 °C for 72 h to prepare KOH-doped ion-solvating membranes (ISMs). Pristine NPBI and NPBI/BDC-ZIF-8 membranes containing 10–40 wt% BDC-ZIF-8 were used for morphology, physicochemical, mechanical, ion transport, and alkaline water electrolysis (AWE) analyses.

Characterization

The chemical structure of NPBI was analyzed by 1 H nuclear magnetic resonance (1 H NMR) spectroscopy using an AVANCE NEO 400 MHz NMR spectrometer (Bruker). DMSO-d 6 was used as the solvent. Elemental analysis was performed using a FlashSmart elemental analyzer (Thermo Fisher Scientific) to determine the C, H, N, and S contents of commercial BDC and PET-derived BDC. Fourier-transform infrared (FT-IR) spectroscopy was performed using a Spectrum 100 FT-IR spectrometer (PerkinElmer, USA) in attenuated total reflection mode. Spectra were collected over the wavenumber range of 4000–400 cm -1 with a resolution of 16 cm -1 and 16 scans. FT-IR analysis was performed for recovered BDC, commercial BDC, BDC-ZIF-8, pristine NPBI, NPBI/BDC-ZIF-8 MMMs, and KOH-aged BDC-ZIF-8 powders. Powder X-ray diffraction (XRD) patterns were collected using a SmartLab High Resolution diffractometer (Rigaku). Cu Kα radiation with a wavelength of 1.5406 Å was used. The operating voltage and current were 45 kV and 200 mA, respectively. Diffraction patterns were recorded over a 2θ range of 5–50° at a scan rate of 5° min-1. Field-emission scanning electron microscopy (FE-SEM) images were obtained using an S-4800 microscope (Hitachi) at an accelerating voltage of 15 kV. Samples were coated with osmium before observation. Cross-sectional membrane samples were prepared by fracturing the membranes in liquid nitrogen. EDS mapping was conducted at 10 kV using a 50 mm 2 silicon drift detector (XMAX 50, HORIBA) to examine the elemental distributions across the membrane cross-sections before and after KOH doping. For KOH-doped samples, residual surface electrolyte was gently removed using a Kimtech wipe, followed by vacuum drying before analysis. N 2 adsorption–desorption measurements were conducted using a BELSORP MINI X analyzer. The BDC-ZIF-8 powder was degassed at 70 °C for 16 h under vacuum before measurement. The specific surface area and pore volume were calculated using BELMASTER software. The Brunauer–Emmett–Teller (BET) specific surface area was calculated from the adsorption isotherm. Static contact angles were measured using a Phoenix 10 apparatus (Surface Electro Optics Co., Ltd., Suwon, South Korea). DI water and 6 M KOH were used as probe liquids, and the droplet volume was 5 μL. The contact angle was recorded 5 s after droplet deposition. KOH-doped samples were measured after removing residual surface electrolyte and vacuum drying.

KOH/water weight fraction and dimensional stability

KOH/water weight fraction and dimensional stability were evaluated using membranes immersed in 6 M KOH at 80 °C for 72 h. After KOH doping, the surface electrolyte was gently removed using a Kimtech wipe, and the wet membrane mass m wet was measured. The KOH-doped membranes were then dried in a vacuum oven at 100 °C for 24 h, and the dried membrane mass m dry was measured. To obtain the dedoped membrane mass m dedoped, the KOH-doped membranes were thoroughly washed with DI water until neutral pH was reached and then dried at 120 °C for 24 h. The polymer/filler solid fraction, water fraction, and KOH fraction were calculated using Equations 2–4: where W solid, W water and W KOH represent the weight fraction of the dry solid component consisting of NPBI and BDC-ZIF-8, water and KOH in the 6 M KOH-doped membrane.

Dimensional stability was evaluated by measuring the area, thickness, and volume of the membranes before and after KOH doping. Membrane thickness was measured using a Mitutoyo 547–401 μm. Membrane area was calculated from the length and width. Membrane volume was calculated from the area and thickness. Area change, thickness change, and volume change were calculated using Equations 5–7: where A, T, and V are the membrane area, thickness, and volume, respectively. The subscripts “dry” and “wet” indicate the states before and after KOH doping, respectively. All measurements were repeated three times, and the results are reported as the mean ± standard deviation.

Gas barrier measurements

Bubble point pressure (BPP) was measured to evaluate the gas-blocking integrity of the membranes. Pristine NPBI, NPBI/BDC-ZIF-8 MMMs, and Zirfon 500 were immersed in 6 M KOH at 80 °C for 72 h before measurement. The surface electrolyte was removed using a Kimtech wipe before mounting the sample. The membrane was fixed in a bubble point pressure test cell. N 2 gas with a purity of 99.99% was introduced to one side of the membrane, while the opposite side was maintained at atmospheric pressure. The pressure was increased at a rate of 0.1 bar min -1 and monitored using a digital pressure gauge. The BPP was recorded as the pressure at which gas bubbles were first observed through the membrane in water. Each measurement was repeated three times.

Hydrogen permeability was measured at room temperature using a gas permeation system equipped with a flat-sheet permeation cell. KOH-doped wet membranes were used for the measurement. The effective membrane area was 10.2 cm 2. The measurements were used to compare the relative H 2 barrier properties of the KOH-doped wet membranes under ex situ conditions. Hydrogen permeability was calculated using Equation 8: where P H is the hydrogen permeability in mol cm -1 s -1 bar -1, J H is the hydrogen flux in mol cm -2 s -1, l is the KOH-doped wet membrane thickness in cm, and Δ p is the pressure difference across the membrane in bar. The calculated permeability was converted to Barrer using the following conversion factor:

Mechanical testing

Mechanical properties were measured using a universal testing machine (UTM, MCT-2150, Japan). Dry membranes were vacuum-dried at 80 °C for 24 h before measurement to remove residual moisture. KOH-doped membranes were measured after immersion in 6 M KOH at 80 °C for 72 h. Before measurement, excess surface electrolyte was gently removed using a Kimtech wipe. Membranes were cut into rectangular strips of 1 × 4 cm. Tensile tests were conducted at 25 °C with a crosshead speed of 10 mm min -1. Tensile strength and strain at break were obtained from stress-strain curves. Each sample was measured three times, and the results are reported as the mean ± standard deviation.

Area resistance measurements

The area resistance of the membranes was measured in 6 M KOH at 80 °C using an LCR meter (LCR-6100, GW INSTEK, Korea). KOH-doped membranes were placed between two 1 cm 2 disk-type platinum electrodes, and resistance was recorded at 10 kHz. The resistance measured with the membrane was defined as R 1, and the resistance measured without the membrane under the same electrolyte conditions was defined as R 2. Zirfon 500 was also immersed in 6 M KOH at 80 °C for 72 h and measured under the same conditions.

The area resistance was calculated using Equation 9: where A is the effective membrane area. Measurements were repeated three times, and the results are reported as the mean ± standard deviation.

Alkaline water electrolysis single-cell test

A schematic illustration of the zero-gap AWE single-cell configuration and the associated electrochemical reactions is shown in Supplementary Figure S1. Alkaline water electrolysis (AWE) performance was evaluated using a zero-gap-type single cell with an active area of 4.41 cm 2. The KOH-doped membrane was sandwiched between two bare Ni felt electrodes in a zero-gap cell configuration. A 6 M KOH solution was circulated through the cell at a flow rate of 50 mL min -1. The cell temperature was maintained at 80 °C. Polarization curves were obtained by recording the cell voltage while changing the current density stepwise. High-frequency resistance (HFR) was measured by electrochemical impedance spectroscopy (EIS). EIS was conducted over a frequency range of 100 kHz to 10 MHz with an alternating current amplitude of 9 mA.

For the 100 h single-cell operation test, the NPBI/BDC-ZIF-8 20 wt% MMM was operated at 0.3 A cm -2 in 6 M KOH at 80 °C using the same zero-gap cell configuration. After the operation, polarization and EIS measurements were conducted again under the same conditions.

Long-term ionic resistance stability test

Ex situ long-term ionic resistance stability was evaluated by monitoring the area resistance of KOH-doped membranes in 6 M KOH at 80 °C. The membrane was placed between platinum electrodes using the same setup described in Section 2.10. The KOH solution was refreshed every 3 days. The area resistance was periodically measured from 72 h of KOH immersion to 500 h. Membrane thickness was remeasured at each time point to calculate resistivity. After 3 months of alkaline exposure, selected membranes were vacuum-dried and observed by FE-SEM to examine morphological changes after prolonged KOH aging.

Alkaline stability test of BDC-ZIF-8 powder

The alkaline stability of BDC-ZIF-8 powder was evaluated by aging the powder in 6 M KOH at 80 °C. BDC-ZIF-8 powder (0.1 g) was immersed in 6 M KOH (5 mL) and aged at 80 °C for 1, 3, 7, 14, and 21 days. After each aging time, DI water (20 mL) was added to dilute the KOH solution, and the powder was collected by vacuum filtration. During filtration, the powder was washed twice with DI water (50 mL each) and then briefly washed once with EtOH (20 mL). The washed powder was dried in an oven at 80 °C and then vacuum-dried at room temperature. The pristine and KOH-aged BDC-ZIF-8 powders were analyzed by FT-IR spectroscopy under the same conditions described in Section 2.6. For FT-IR comparison, the spectra were normalized using the imidazolate ring vibration peak at 1147 cm -1. The morphology of pristine and KOH-aged BDC-ZIF-8 powders was also examined by FE-SEM after alkaline exposure.

Structural characteristics of waste PET-derived BDC-ZIF-8 and NPBI/BDC-ZIF-8 MMMs

Figure 1 and Table 1 present the chemical and structural characteristics of the NPBI polymer and BDC-ZIF-8 filler. NPBI was synthesized by the polycondensation of 3,3′-diaminobenzidine and 1,4-naphthalenedicarboxylic acid, as illustrated in Scheme 1. The chemical structure of NPBI was confirmed by 1 H NMR spectroscopy (Figure 1A). In the spectrum, the naphthalene proton signal appeared at approximately 9.4 ppm, while the phenyl proton signals were observed in the range of 7.9–8.3 ppm. These signals are consistent with the characteristic proton signals of previously reported NPBI (Geng et al., 2019; Hu et al., 2022a). BDC was recovered from waste PET through alkaline depolymerization followed by acidification, as shown in Scheme 2, and was used as the dicarboxylate linker for BDC-ZIF-8 synthesis. The mixed-ligand BDC-ZIF-8 was synthesized using waste PET-derived BDC and 2-methylimidazole. The FE-SEM image showed that BDC-ZIF-8 exhibited a hierarchical morphology consisting of plate-like or fragmented particles, rather than the rhombic dodecahedral morphology commonly observed for conventional ZIF-8 (Figure 1B). This morphological difference suggests that BDC incorporation altered the crystallization behavior of the Zn-imidazolate framework.

Elemental analysis showed that PET-derived BDC contained 57.53 wt% C and 3.71 wt% H, with no detectable N or S impurities. These values were consistent with the theoretical composition of BDC and comparable to those of commercial BDC, indicating that high-purity PET-derived BDC was successfully obtained for BDC-ZIF-8 synthesis (Supplementary Table S1). The chemical structure of BDC-ZIF-8 was examined by FT-IR spectroscopy (Figure 1C). Waste PET-derived BDC exhibited carboxylic acid and aromatic ring vibration peaks similar to those of commercial BDC, indicating that the recovered BDC retained a chemical structure comparable to that of commercial BDC. In the BDC-ZIF-8 spectrum, the carboxylic acid C=O stretching peak of free BDC near 1670 cm -1 disappeared. Instead, asymmetric and symmetric COO − stretching peaks assigned to coordinated carboxylate groups were observed at approximately 1558 and 1378 cm -1, respectively (Latrach et al., 2022). In addition, peaks near 422 and 1581 cm -1 were assigned to the Zn-N stretching vibration of the ZIF framework and the imidazolate C=N vibration, respectively (Latrach et al., 2022; Gao et al., 2023). These spectral features indicate that the carboxylate groups of BDC and the imidazolate moieties of 2-methylimidazole coordinated with Zn 2+ to form the mixed-ligand coordination framework of BDC-ZIF-8 (Mohammed et al., 2023; Latrach et al., 2022). The crystallinity of BDC-ZIF-8 was evaluated by powder XRD analysis and compared with ZIF-8 (Figure 1D). The diffraction pattern retained the main peaks of ZIF-8 at approximately 7.3°, 10.4°, and 12.7°, indicating that the basic ZIF-8-like crystalline framework was preserved after BDC incorporation (Latrach et al., 2022; Park et al., 2006). Meanwhile, additional weak peaks were observed at approximately 8.3° and 9.1°. These peaks may be attributed to secondary coordination environments induced by BDC introduction or to BDC-related crystalline domains. Therefore, the XRD results suggest that BDC-ZIF-8 retained a ZIF-8-like framework while developing additional local structural features associated with the mixed-ligand composition (Latrach et al., 2022).

The pore characteristics of BDC-ZIF-8 were investigated by N 2 adsorption–desorption measurements (Figure 1E; Table 1). The isotherm exhibited type I adsorption behavior with a hysteresis loop, suggesting the coexistence of micropores and mesopores. This pore structure is consistent with the porous characteristics reported for mixed-ligand ZIF-8 (Latrach et al., 2022; Feng et al., 2020). The BET specific surface area of BDC-ZIF-8 was 1488 m 2 g -1, indicating that BDC-ZIF-8 can serve as a high-surface-area porous filler. The high surface area and porous structure of BDC-ZIF-8 are expected to help retain KOH/water within the NPBI matrix, which is important for constructing electrolyte-retaining domains in ion-solvating MMMs. Overall, these results confirm that waste PET-derived BDC was successfully used to synthesize mixed-ligand BDC-ZIF-8. The resulting filler retained a ZIF-8-like crystalline framework while developing a high surface area, porous structure, and BDC-derived coordination environment. These structural features support its use as an electrolyte-retaining MOF filler for NPBI-based ion-solvating membranes.

Figure 2 presents digital photographs of pristine NPBI and NPBI/BDC-ZIF-8 MMMs with BDC-ZIF-8 loadings of 10–50 wt%. Pristine NPBI and the MMMs containing 10–40 wt% BDC-ZIF-8 retained their membrane form without visible cracks or fractures. In contrast, the 50 wt% BDC-ZIF-8 MMM exhibited severe shrinkage and fracture, suggesting that excessive filler loading hindered the formation of a continuous NPBI matrix. This result indicates that excessive filler loading can compromise film-forming ability and membrane continuity. Therefore, NPBI/BDC-ZIF-8 MMMs with BDC-ZIF-8 loadings of 10–40 wt%, which formed self-standing membranes, were selected for further physicochemical, mechanical, ion transport, and AWE performance analyses.

Figure 3 presents FE-SEM images showing the surface and cross-sectional morphologies of pristine NPBI and NPBI/BDC-ZIF-8 MMMs with BDC-ZIF-8 loadings of 10–40 wt%. Pristine NPBI exhibited smooth surface morphology without obvious particulate features or cracks. The cross-sectional image also showed a dense membrane structure. With increasing BDC-ZIF-8 loading, the surface roughness and cross-sectional heterogeneity gradually increased. The 10 and 20 wt% MMMs showed slightly rougher surfaces than pristine NPBI but retained continuous membrane structures without large defects or fractures. This morphology suggests that BDC-ZIF-8 was relatively well dispersed in the NPBI matrix within this loading range, enabling filler incorporation while preserving membrane continuity. In contrast, at BDC-ZIF-8 loadings of 30 and 40 wt%, particle aggregation and clustered features became more apparent on the membrane surface. The cross-sectional morphology also became more heterogeneous and rougher. These changes are likely associated with the formation of filler-rich domains and increased polymer–filler interfacial heterogeneity at higher BDC-ZIF-8 loadings. These morphology results provide an important structural basis for understanding the loading-dependent membrane properties discussed below. Moderate BDC-ZIF-8 incorporation can introduce electrolyte-retaining filler domains while maintaining membrane continuity, whereas excessive loading reduces morphological uniformity and may lead to aggregation-induced structural instability. Therefore, the morphology of NPBI/BDC-ZIF-8 MMMs is expected to play a critical role in balancing initial ion transport enhancement with mechanical integrity and long-term alkaline stability.

Cross-sectional EDS mapping was performed before KOH doping to examine the distribution of BDC-ZIF-8 within the NPBI matrix (Supplementary Figure S2). Zn signals were detected only in the BDC-ZIF-8-containing MMMs, confirming the incorporation of BDC-ZIF-8 into the NPBI matrix. The Zn signal became stronger with increasing filler loading. The 10 and 20 wt% MMMs exhibited a relatively uniform Zn distribution, whereas the 30 and 40 wt% MMMs showed more pronounced Zn-rich regions. These results suggest that higher BDC-ZIF-8 loading increased filler aggregation and morphological heterogeneity.

Electrolyte retention and dimensional stability

Figure 4 presents the FT-IR spectra of pristine NPBI, BDC-ZIF-8, and NPBI/BDC-ZIF-8 MMMs. Pristine NPBI exhibited a peak at approximately 1622 cm -1, which is assigned to the C=N stretching vibration of the benzimidazole ring. A peak near 1396 cm -1 was also observed and can be attributed to imidazole ring vibration or C–N stretching (Aili et al., 2020; Musto et al., 1993). For BDC-ZIF-8, the asymmetric and symmetric COO − stretching peaks of coordinated BDC carboxylate groups appeared at approximately 1561 and 1380 cm -1, respectively. The C=N stretching peak of the imidazolate skeleton was observed near 1581 cm -1. In addition, peaks at approximately 1310, 1180, and 1145 cm -1 were related to imidazolate ring vibrations or C-N stretching, and the Zn-N stretching vibration of the BDC-ZIF-8 framework was detected at 427 cm -1 (Latrach et al., 2022; Mohammed et al., 2023). In the NPBI/BDC-ZIF-8 MMMs, characteristic peaks of both NPBI and BDC-ZIF-8 were observed. The benzimidazole/imidazole ring vibration peaks from NPBI were retained, along with the COO −, imidazolate C=N/C-N, and Zn-N vibration peaks from BDC-ZIF-8. These spectral features confirm that BDC-ZIF-8 was successfully incorporated into the NPBI matrix without significant disruption of the main chemical structures of either the NPBI backbone or the BDC-ZIF-8 framework.

Figure 5 presents the KOH/water uptake and dimensional stability of pristine NPBI and NPBI/BDC-ZIF-8 MMMs after immersion in 6 M KOH for 3 days. In ISMs, the absorbed KOH/water phase directly participates in K + and OH − transport. Therefore, electrolyte uptake and the resulting dimensional changes are key factors that determine membrane resistance and mechanical integrity (Aili et al., 2020; Kraglund et al., 2019). In this study, porous BDC-ZIF-8 was introduced as an electrolyte-retaining MOF filler in the NPBI matrix. The filler was expected not only to increase KOH/water retention but also to help regulate swelling behavior through its rigid inorganic–organic framework. The relative mass fractions of polymer/filler solid, water, and KOH in the wet membranes were compared after KOH treatment (Figure 5A). With increasing BDC-ZIF-8 loading, the polymer/filler solid fraction gradually decreased. From pristine NPBI to the 40 wt% BDC-ZIF-8 MMM, the solid fraction decreased from 49% to 40%, whereas the water fraction increased from 34% to 37% and the KOH fraction increased from 17% to 23%. These results indicate that BDC-ZIF-8 incorporation increased KOH/water retention, thereby increasing the proportion of the ion-conducting electrolyte phase within the wet membrane. The enhanced electrolyte retention is consistent with the high BET surface area, porous structure, and polar coordination environment of BDC-ZIF-8. In addition, BDC-derived carboxylate groups and the polar coordination environment may interact with water molecules or hydrated ions, further contributing to electrolyte retention. These features support the role of BDC-ZIF-8 as an electrolyte-retaining filler that can increase the KOH/water fraction in NPBI-based ISMs.

The thickness, volume, and area changes of the membranes were evaluated after KOH treatment (Figure 5B). Pristine NPBI exhibited relatively large thickness and volume changes. In contrast, the BDC-ZIF-8-containing MMMs showed overall lower or more stabilized dimensional changes despite their increased KOH/water fractions. In general, increased electrolyte uptake can induce polymer chain swelling and dimensional changes in membranes (Hu et al., 2022b). However, in the present MMMs, rigid BDC-ZIF-8 domains may act as physical reinforcing phases within the NPBI matrix, helping suppress excessive swelling and maintain dimensional stability (He et al., 2017; Aili et al., 2020). In particular, the reduced in-plane area change suggests that BDC-ZIF-8 incorporation helps preserve lateral dimensional stability after KOH uptake. These results indicate that BDC-ZIF-8 acts as a multifunctional filler in NPBI-based ISMs by increasing KOH/water retention while helping regulate dimensional changes. In other words, BDC-ZIF-8 does not merely increase liquid uptake; it also contributes to controlling electrolyte retention and swelling behavior. This balance is important for dense ISMs operating under concentrated alkaline conditions. However, FE-SEM morphological analysis showed that high BDC-ZIF-8 loadings increased membrane heterogeneity. Therefore, enhanced electrolyte uptake should not be interpreted solely as improved membrane structure, because excessive filler loading may produce a less uniform ion-solvating phase.

DI water and 6 M KOH contact angles were measured to evaluate the wettability of the membranes (Supplementary Figure S3). For the dry membranes, the DI water contact angle decreased from 71.46° for pristine NPBI to 55.67° for the 40 wt% MMM. Similarly, the 6 M KOH contact angle decreased from 52.07° to 36.29° with increasing BDC-ZIF-8 loading. After KOH doping and vacuum drying, the DI water contact angles decreased to approximately 24°–26°, and the 6 M KOH contact angles were below 10° for all membranes. These results indicate that BDC-ZIF-8 incorporation increased membrane wettability, consistent with the increased KOH/water fractions of the MMMs.

Cross-sectional EDS mapping after KOH doping was also performed to examine the distribution of K-containing species within the membranes (Supplementary Figure S4). K signals were detected in all KOH-doped membranes, and the BDC-ZIF-8-containing MMMs retained Zn signals originating from the filler phase. These results indicate the incorporation of K-containing species after KOH doping and are consistent with the increased KOH/water fractions and increased wettability of the MMMs.

Gas-blocking integrity and mechanical properties

Figure 6 and Table 2 present the gas-blocking integrity and mechanical properties of pristine NPBI and NPBI/BDC-ZIF-8 MMMs. ISMs for AWE should maintain a dense membrane structure that can resist gas breakthrough even after KOH uptake (Aili et al., 2020; Kraglund et al., 2019). To evaluate this property, bubble point pressure (BPP) was measured at room temperature after immersion in 6 M KOH for 3 days (Figure 6A). Pristine NPBI exhibited a BPP of 6.1 ± 0.4 bar, while the NPBI/BDC-ZIF-8 MMMs showed BPP values in the range of 6.2–6.5 bar depending on the BDC-ZIF-8 loading. These results indicate that the NPBI/BDC-ZIF-8 MMMs maintained gas breakthrough resistance comparable to that of pristine NPBI. Under the same conditions, both pristine NPBI and the MMMs showed higher BPP values than commercial Zirfon 500 (Henkensmeier et al., 2024; In Lee et al., 2020), suggesting that the dense ISMs prepared in this study provided stronger resistance to gas breakthrough than the much thicker porous reference separator. BPP represents the resistance of the membrane to gas breakthrough under static pressure but does not directly quantify H 2 /O 2 crossover during AWE operation. Nevertheless, gas crossover is an important factor for product gas purity and safe AWE operation, as excessive H 2 /O 2 crossover can lead to the formation of hazardous gas mixtures (Haug et al., 2017; Barros et al., 2024; Henkensmeier et al., 2024).

The ex situ H 2 permeability of KOH-doped wet membranes was measured to compare their relative H 2 barrier properties under non-operating conditions (Figure 6B). Commercial Zirfon 500 exhibited an H 2 permeability of 89.6 Barrer, whereas pristine NPBI showed a much lower value of 0.571 Barrer. The NPBI/BDC-ZIF-8 MMMs exhibited H 2 permeability values of 0.418, 0.402, 0.395, and 0.417 Barrer for the 10, 20, 30, and 40 wt% BDC-ZIF-8 loadings, respectively. These results indicate that BDC-ZIF-8 incorporation preserved the dense gas-barrier characteristics of the NPBI-based ISMs under the tested ex situ conditions. The stress–strain behavior and mechanical properties of the dry membranes were then evaluated (Figure 6C; Table 2). Pristine NPBI showed a tensile stress of 70.20 MPa and a strain at break of 7.68%. For the BDC-ZIF-8-containing MMMs, the tensile stress increased to 72.28, 78.17, and 79.35 MPa at BDC-ZIF-8 loadings of 10, 20, and 30 wt%, respectively. In particular, the 20 and 30 wt% MMMs exhibited approximately 11.4% and 13.0% higher stress than pristine NPBI, respectively. In contrast, the strain at break decreased to 5.62, 5.02, and 4.70% for the 10, 20, and 30 wt% MMMs, respectively. This behavior can be attributed to the reinforcing effect of rigid BDC-ZIF-8 fillers in the NPBI matrix. At the same time, the rigid filler domains can restrict polymer chain mobility, leading to reduced elongation at break (He et al., 2017; Lin et al., 2018). In contrast, the tensile stress of the 40 wt% MMM markedly decreased to 38.74 MPa, and its strain at break was the lowest at 3.71%. This result is consistent with the FE-SEM morphology of the 40 wt% MMM. Aggregated filler domains and polymer–filler interfacial heterogeneity at high filler loading can induce local stress concentration during tensile deformation, resulting in mechanical weakening. Differences in mechanical properties were also observed in the KOH-doped membranes depending on membrane composition (Figure 6D; Table 2). After immersion in 6 M KOH, pristine NPBI exhibited a tensile stress of 37.81 MPa and a strain at break of 13.28%. The 10, 20, and 30 wt% BDC-ZIF-8 MMMs showed tensile stresses of 43.19, 55.97, and 48.45 MPa, respectively, which were higher than that of pristine NPBI. In particular, the 20 wt% MMM exhibited the highest tensile stress in the KOH-doped state, corresponding to a 47.87% increase compared with pristine NPBI. This result indicates that the reinforcing effect of BDC-ZIF-8 was maintained not only in the dry state but also in the KOH-doped state, which more closely reflects AWE operating conditions. However, the tensile stress of the 30 wt% MMM was lower than that of the 20 wt% MMM, and the strain at break gradually decreased over the 10–30 wt% loading range. This trend suggests that KOH-induced plasticization, restricted polymer chain mobility caused by increased filler content, and loading-induced morphological heterogeneity may have acted together (Hu et al., 2022b; Hu et al., 2022a; Cheng et al., 2022). The 40 wt% MMM also exhibited reduced mechanical properties in the KOH-doped state, with a tensile stress of 24.57 MPa and a strain at break of 5.22%. Therefore, BDC-ZIF-8 can mechanically reinforce NPBI membranes at appropriate loadings, whereas excessive loading causes mechanical weakening in both dry and KOH-doped states because of filler aggregation, interfacial heterogeneity, and stress concentration. These results suggest that 20 wt% BDC-ZIF-8 provides a more balanced mechanical reinforcement effect, while higher loadings increasingly compromise membrane uniformity and mechanical integrity.

Ion transport properties and AWE cell performance

Figure 7 presents the area resistance of pristine NPBI and NPBI/BDC-ZIF-8 MMMs measured at 80 °C after immersion in 6 M KOH for 3 days. The area resistance of pristine NPBI was 0.067 Ω cm 2. With increasing BDC-ZIF-8 loading, the area resistance decreased to 0.062, 0.046, 0.040, and 0.037 Ω cm 2 for the 10, 20, 30, and 40 wt% MMMs, respectively. In comparison, commercial Zirfon 500 exhibited a higher area resistance of 0.079 Ω cm 2 than both pristine NPBI and the NPBI/BDC-ZIF-8 MMMs under the same measurement conditions. These results show that BDC-ZIF-8 incorporation effectively reduced the initial ionic resistance of the NPBI-based ISMs. This reduction in area resistance is consistent with the increased KOH/water fraction observed with higher BDC-ZIF-8 loading. In ISMs, K + and OH − transport mainly occurs through the absorbed KOH/water phase (Kraglund et al., 2019; Hu et al., 2022a). Therefore, the electrolyte-retaining effect of BDC-ZIF-8 likely improved the ion-solvating environment within the NPBI matrix and contributed to the reduction in area resistance. In addition, the BDC-derived carboxylate groups and polar framework environment of BDC-ZIF-8 may provide interaction sites for KOH/water, further supporting electrolyte retention within the membrane (Latrach et al., 2022; Titirici and Szilágyi, 2022). These results indicate that BDC-ZIF-8 can promote initial ion-solvating transport by increasing the retained electrolyte phase in the NPBI matrix.

To examine the temperature dependence of membrane-level ion transport, the area resistance of commercial Zirfon 500, pristine NPBI, and NPBI/BDC-ZIF-8 MMMs was measured at r. t., 50, 60, 70, and 80 °C in 6 M KOH after 3 days of immersion (Supplementary Figure S7). The area resistance decreased with increasing temperature for all membranes. For example, the area resistance of pristine NPBI decreased from 0.155 Ω cm 2 at r. t. To 0.067 Ω cm 2 at 80 °C, while that of the 40 wt% MMM decreased from 0.069 to 0.037 Ω cm 2. The NPBI/BDC-ZIF-8 MMMs maintained lower area resistance than pristine NPBI and commercial Zirfon 500 over the tested temperature range. At 80 °C, the 20 and 40 wt% MMMs showed area resistance values of 0.046 and 0.037 Ω cm 2, respectively, compared with 0.067 Ω cm 2 for pristine NPBI and 0.079 Ω cm 2 for Zirfon 500.

Figure 8 presents the polarization curves and EIS Nyquist plots of AWE single cells employing pristine NPBI and NPBI/BDC-ZIF-8 MMMs. The cell tests were conducted in 6 M KOH at 80 °C using Ni felt electrodes. In the polarization curves, the current density at a given cell voltage generally increased with BDC-ZIF-8 loading (Figure 8A). At 1.8 V, the current density of the pristine NPBI cell was 0.291 A cm -2, whereas cells with 10, 20, 30, and 40 wt% BDC-ZIF-8 MMMs exhibited current densities of 0.332, 0.446, 0.575, and 0.704 A cm -2, respectively. A similar trend was observed at 2.0 V. The pristine NPBI cell showed a current density of 0.913 A cm -2, while cells with 10, 20, 30, and 40 wt% BDC-ZIF-8 MMMs reached 1.099, 1.224, 1.494, and 1.809 A cm -2, respectively. This improvement is consistent with the reduced area resistance of the BDC-ZIF-8-containing MMMs, suggesting that improved ion transport through the membrane contributed to the higher initial AWE performance. The EIS results showed a consistent trend (Figure 8B). The high-frequency resistance (HFR) of the pristine NPBI cell was 0.290 Ω cm 2. After BDC-ZIF-8 incorporation, the HFR decreased to 0.214, 0.192, 0.175, and 0.166 Ω cm 2 for the 10, 20, 30, and 40 wt% MMMs, respectively. This decrease in HFR is consistent with the reduced area resistance of the BDC-ZIF-8-containing MMMs. These results suggest that the reduced membrane-level area resistance contributed to the lower cell-level HFR and improved initial AWE performance. However, because HFR was measured in the full AWE cell, it can also include ohmic contributions from electrode contact, electrolyte resistance, and other cell components (Kraglund et al., 2019; Hu et al., 2022a). Scheme 3 schematically illustrates the proposed electrolyte-filled transport structure of KOH-doped NPBI/BDC-ZIF-8 MMMs with different BDC-ZIF-8 loadings. At low filler loading, the amount of BDC-ZIF-8 is limited, and the electrolyte-retaining domains are not sufficiently developed or connected. Therefore, the improvement in ion transport is relatively small. At intermediate loading, BDC-ZIF-8 is more effectively distributed within the NPBI matrix, providing electrolyte-retaining domains while maintaining membrane continuity and mechanical integrity. This state represents a balanced structure between ion transport enhancement and morphological stability. At high BDC-ZIF-8 loadings, particularly 30 and 40 wt%, the increased filler content enhances KOH/water retention and can form MOF-rich electrolyte-filled transport regions. This structure explains the lower initial area resistance and improved initial AWE performance obtained from the membrane resistance and single-cell measurements. However, FE-SEM morphological analysis indicated that excessive filler incorporation induced filler-rich domains and morphological heterogeneity, leading to mechanical weakening and less stable ex situ long-term resistance behavior. Therefore, the performance of NPBI/BDC-ZIF-8 MMMs is governed by a loading-dependent trade-off, in which higher filler loading improves initial electrolyte retention and ion transport, whereas excessive loading compromises morphological uniformity, mechanical integrity, and resistance stability.

The performance and membrane properties of the present NPBI/BDC-ZIF-8 MMMs were compared with those of representative AWE membranes and separators reported in the literature (Table 3). This comparison places the present dense MMMs in the context of reported porous separators and dense ion-solvating membranes.

Ex situ long-term stability and alkaline structural evolution of BDC-ZIF-8

After the 100 h operation, the single-cell performance of the NPBI/BDC-ZIF-8 20 wt% MMM was evaluated by polarization and EIS measurements in 6 M KOH at 80 °C. The polarization curve showed current densities of 0.344 and 1.103 A cm -2 at 1.8 and 2.0 V, respectively (Figure 9A). The corresponding EIS Nyquist plot gave an HFR of 0.2412 Ω cm 2 (Figure 9B), which was higher than the initial HFR of the 20 wt% MMM, 0.192 Ω cm 2, but lower than the initial HFR of pristine NPBI, 0.290 Ω cm 2. These results indicate that the 20 wt% MMM retained relatively low cell resistance after the 100 h operation, although the HFR increased compared with its initial value.

Figure 9C presents the ex situ long-term ionic resistance stability of pristine NPBI and NPBI/BDC-ZIF-8 MMMs measured in 6 M KOH at 80 °C from 72 to 500 h. This test was conducted under ex situ alkaline exposure conditions without the continuous current, gas evolution, electrode contact, or possible gas crossover associated with AWE operation. Therefore, the 500 h resistance results were used to evaluate membrane-level ionic resistance stability under alkaline exposure, rather than full AWE durability. The purpose of this test was to examine whether the initially low area resistance measured after 3 days of KOH immersion could be maintained during prolonged alkaline exposure (Kraglund et al., 2019).

Throughout the measurement period, the NPBI/BDC-ZIF-8 MMMs maintained lower area resistance than pristine NPBI. However, the resistance changes differed depending on filler loading. The 10 and 20 wt% MMMs exhibited relatively small changes in area resistance during the test and maintained low resistance more stably. In contrast, the 30 and 40 wt% MMMs showed low initial resistance, but their area resistance gradually increased with time. This result indicates that the lowest initial resistance does not necessarily correspond to the most stable resistance behavior during prolonged alkaline exposure.

The time-dependent increase in resistance observed for the 30 and 40 wt% MMMs may be related to morphological heterogeneity caused by high BDC-ZIF-8 loading. FE-SEM morphological analysis showed that the 30 and 40 wt% MMMs exhibited more pronounced surface roughness and cross-sectional heterogeneity than the 10 and 20 wt% MMMs. Such morphological heterogeneity can reduce the uniformity of the ion-solvating phase during prolonged immersion in KOH (Cheng et al., 2022; Lin et al., 2018). In addition, because high-loading MMMs contain a larger fraction of BDC-ZIF-8, any decrease in filler stability during long-term KOH exposure could have a greater effect on the electrolyte-retaining domains and resistance behavior (Titirici and Szilágyi, 2022).

These results indicate that BDC-ZIF-8 loading can reduce the initial ionic resistance of NPBI-based ISMs and help maintain lower resistance than pristine NPBI under prolonged KOH exposure. However, excessive filler loading may compromise resistance stability over time. Therefore, the long-term resistance behavior suggests that filler loading should be optimized to balance initial ion transport enhancement with morphological uniformity and alkaline durability in NPBI/BDC-ZIF-8 MMMs.

Based on its balanced mechanical properties, low cell resistance, and ex situ long-term resistance stability, the NPBI/BDC-ZIF-8 20 wt% MMM was further examined under continuous single-cell operation at 0.3 A cm -2 in 6 M KOH at 80 °C for 100 h (Figure 9D). The cell voltage remained close to 1.80 V, with an average value of 1.798 V and a maximum value of 1.804 V. Therefore, the 20 wt% MMM showed stable single-cell operation for 100 h under the tested conditions.

Figure 10 presents the alkaline stability of the BDC-ZIF-8 filler itself. BDC-ZIF-8 was immersed in 6 M KOH at 80 °C for 1, 3, 7, 14, and 21 days, and the chemical structural changes were examined by FT-IR spectroscopy. In the FT-IR spectra, the imidazolate C=N vibration peak observed in pristine BDC-ZIF-8 gradually decreased in intensity with increasing KOH immersion time. In addition, the asymmetric and symmetric COO − stretching peaks assigned to BDC-derived coordinated carboxylate groups also weakened after KOH treatment. These spectral changes suggest that the BDC–Zn coordination environment or BDC-derived local coordination structure changed over time under strongly alkaline conditions. In contrast, peaks near 1309, 1180, and 1147 cm -1, which are related to imidazolate ring vibrations or C-N stretching, were partially retained even after prolonged KOH treatment. The Zn-N vibration peak near 422 cm -1 was also observed after 21 days of immersion. However, the imidazolate-related bending vibration near 554 cm -1 and the Zn-N-C bending-related vibration near 510 cm -1 became weaker as the immersion time increased (Latrach et al., 2022; Mohammed et al., 2023). This trend implies that the local bending environment around the imidazolate linker or the Zn-N-C coordination geometry was partially altered under high alkaline conditions (Park et al., 2006; Zhang et al., 2015).

The alkaline stability results of the filler provide supporting evidence for interpreting the ex situ long-term resistance behavior of the NPBI/BDC-ZIF-8 MMMs. The 30 and 40 wt% MMMs exhibited low initial resistance, but their area resistance increased during prolonged exposure to 6 M KOH at 80 °C. Because high-loading MMMs contain a larger fraction of BDC-ZIF-8, partial structural modification of the filler during KOH exposure could more strongly affect the electrolyte-retaining domains within these membranes. In addition, the morphological heterogeneity identified by FE-SEM analysis could hinder the maintenance of uniform transport pathways during long-term immersion. Thus, the resistance increase during prolonged alkaline exposure observed in the high-loading MMMs can be attributed to the combined effects of partial alkaline structural modification of BDC-ZIF-8 and aggregation-induced morphological heterogeneity caused by excessive filler incorporation (Cheng et al., 2022; Zhang et al., 2015).

FE-SEM images were obtained to examine the membrane morphology after prolonged alkaline exposure in 6 M KOH at 80 °C for 3 months (Supplementary Figure S5). All membranes maintained continuous structures without large cracks or through-plane defects, indicating that no macroscopic membrane failure occurred during alkaline aging. However, the 30 and 40 wt% MMMs showed more pronounced surface roughness and cross-sectional heterogeneity after aging. These observations suggest that the long-term resistance increase at high BDC-ZIF-8 loading was not caused by complete membrane collapse, but was likely associated with increased morphological heterogeneity.

The alkaline stability of BDC-ZIF-8 powder was also examined by FE-SEM after exposure to 6 M KOH at 80 °C for 1–21 days (Supplementary Figure S6). Compared with pristine BDC-ZIF-8 shown in Figure 1B, the KOH-aged BDC-ZIF-8 powders retained their overall particle morphology. However, surface roughening and fragmented features became more evident with increasing exposure time. This result is consistent with the FT-IR results in Figure 10, suggesting partial alkaline modification of BDC-ZIF-8 under prolonged alkaline exposure while partially retaining Zn-N-related framework features. Further studies under practical operating conditions are needed to evaluate the applicability of these membranes in larger AWE systems.

This work investigated waste PET-derived BDC-ZIF-8 as an electrolyte-retaining MOF filler for NPBI-based ion-solvating mixed-matrix membranes in alkaline water electrolysis. Mixed-ligand BDC-ZIF-8 was synthesized using BDC recovered from waste PET depolymerization and retained a ZIF-8-like crystalline framework with a high surface area and porous structure. These features allowed BDC-ZIF-8 to regulate the KOH/water environment within the NPBI matrix. Incorporation of BDC-ZIF-8 increased the water and KOH fractions in the KOH-doped membranes, resulting in reduced area resistance and improved initial AWE cell performance. The MMMs also maintained bubble point pressure values comparable to pristine NPBI, indicating that dense membrane integrity was preserved under the tested conditions. However, the results revealed a clear loading-dependent trade-off. High BDC-ZIF-8 loadings, especially 30 and 40 wt%, provided lower initial resistance and higher initial cell performance but caused morphological heterogeneity, mechanical weakening, and less stable ex situ long-term ionic resistance behavior. In contrast, the 10 and 20 wt% MMMs offered a more balanced combination of reduced resistance, mechanical integrity, and ex situ resistance stability. In particular, the 20 wt% MMM sustained 100 h single-cell operation at 0.3 A cm -2 in 6 M KOH at 80 °C, maintaining an average cell voltage of 1.798 V. The alkaline stability analysis showed that BDC-ZIF-8 did not completely collapse in 6 M KOH at 80 °C, but partial structural modification occurred during prolonged exposure. This filler evolution, together with aggregation-induced morphological heterogeneity, likely contributed to the resistance increase observed in high-loading MMMs. Overall, this study demonstrates that waste PET-derived BDC-ZIF-8 can be used as a functional electrolyte-retaining filler for NPBI-based ISMs, while highlighting that filler loading must be optimized to balance KOH/water retention, initial ion transport, morphology uniformity, membrane integrity, and alkaline exposure resistance stability.