Main

Industrial alkaline water electrolysers (AWE) use a concentrated KOH solution (4–7 M ~ 20–30 wt%) as electrolyte. The electrodes are separated by porous diaphragms such as Zirfon to prevent the mixing of hydrogen (H2) and oxygen (O2)1. Whereas cost-effective non-platinum group metal (PGM) materials such as nickel-based electrodes can be used as catalysts2,3, AWE has a narrow operational window. At too low currents, the amount of H2 crossed over to O2 exceeds the explosion limit4. At high currents, high ohmic resistance requires high voltages, accelerating corrosion of components5,6. Additionally, Zirfon’s low bubble point of 2 bar prevents operation at differential pressures, complicating efforts to reduce H2 compression costs7,8.

Proton exchange membrane water electrolysis uses dense membranes and typically reaches current densities of 1–2 A cm−2, and even 6 A cm−2 were reported9. However, the acidic pH necessitates the use of expensive titanium-based bipolar plates10 and platinum and iridium catalysts, which are scarce metals11. Furthermore, Nafion 212 costs around US$225 m−2 (ref. 12), and governments seek to ban fluorinated materials because their toxicity and longevity raise environmental concerns13,14.

Anion exchange membrane water electrolysis (AEMWE) uses membranes with cationic moieties and combines the advantages of AWE (cheap catalysts) and proton exchange membrane water electrolysis (differential pressure)15. However, most AEMs have a conductivity below 80 mS cm−1 (for example, FAA3: 45 mS cm−1 in 1 M KOH)16,17,18, and their quaternary ammonium groups degrade by nucleophilic substitution and Hoffman elimination19, limiting operation to diluted alkaline solutions, which improves stability but increases solubility of H2 and thus permeation20.

Acid-doped polybenzimidazole (PBI) is used in PEM fuel cells21,22 and flow batteries23,24,25, and recently, KOH-doped PBI membranes, so-called ion-solvating membranes (ISM), have been investigated for use in AWE4. KOH deprotonates the imidazole groups, and the increased hydrophilicity promotes absorption of KOH solution between the polymer chains. Depending on the doping process, conductivities of 100–200 mS cm−1 were achieved in 25 wt% KOH solution26,27,28, and an AWE with a reinforced gel meta-PBI membrane exhibited 1.7 A cm−2 at 1.8 V, 80 °C (ref. 27). However, when meta-PBI, the most well-investigated ISM, is immersed in solutions containing ≤1 M KOH, the range relevant for AEMWE, the conductivity decreases strongly. Introduction of N-alkylsulfonic acid side chains to PBI increases water and KOH absorption29, but this type of functionalization consumes imidazole NH groups, and without negative charge, the imidazole C2 position is susceptible to OH− attack30. Recently, we reported the use of sulfonated para-PBI (MS-PBI), which has 50% more ionic groups per repeat unit and thus reaches very high conductivities, even in 1 M KOH, but needs to be crosslinked to avoid dissolution31. Although this material showed excellent alkaline stability over 6 months in 1 M KOH at 80 °C, the crosslinked neutral imidazole units might act as breaking point. Furthermore, chemical crosslinking is difficult to scale up, because casting solutions could gel during membrane fabrication.

Previously, a PBI copolymer (SOPBI, a PBI that contains sulfonated and non-sulfonated diphenyl ether moieties; Fig. 1) was reported for use in PEM Fuel Cells32. In that work, the carbon atoms in the ortho positions relative to the ether groups were selectively sulfonated, and the degree of sulfonation (DOS), and thus the membrane properties, were tuned by copolymerization. Whereas phenyl ethers substituted with quaternary ammonium groups are avoided in AEM, sulfonated poly(ether ether ketone) was stable over 41 days test in 5 M NaOH at 60 °C (ref. 33). In SOPBI, the negatively charged imidazolide groups should further stabilize the ether groups.

Fig. 1: Synthesis of SOPBI membranes.
figure 1

Sulfonation of OBA, synthesis of SOPBI, its doping with KOH solution and illustration that negatively charged imidazolide ions resist OH− attack on the C2 position. aq indicates an aqueous solution.

Full size image

In this work, SOPBI properties are optimized by copolymerization of sulfonated and non-sulfonated monomers. Sulfonated units with locally high electrolyte uptake promote conductivity, whereas less sulfonated regions provide stability. Therefore, no crosslinking is needed. The absence of quaternary ammonium and neutral imidazole groups promises high alkaline stability, and in 1 M KOH solution, SOPBI demonstrated a conductivity above that of over 30 membranes that were recently reported in the literature (Supplementary Tables 4 and 5).

Membrane preparation and properties

4,4′-oxybisbenzoic acid (OBA) was sulfonated to sodium 6,6′-oxybis(3-carboxybenzenesulfonate) (SODBA), as proven by 1H-NMR (Supplementary Fig. 1). Copolymerization of SODBA with diaminobenzidine (DAB) and OBA gave a series of partially sulfonated polymers (Fig. 1). As also observed by others34, polymerizations targeting DOS > 70% give a gelled reaction mixture within 3 h of reaction.

Presumably, the bulkiness of SODBA’s sulfonic acid groups reduces the chain mobility and by this, the degree of polymerization35. Simultaneously, Friedel–Crafts-type crosslinking reactions may gain importance. 1H-NMR analysis (Supplementary Fig. 2) confirmed that polymers with DOS of 50%, 60% and 70% were successfully synthesized (50SOPBI, 60SOPBI and 70SOPBI, respectively). In the potassium form, these polymers exhibited excellent solubility in DMSO and partial solubility in DMAc. All polymers formed strong, flexible membranes, indicating that high molecular weights were obtained (>92 kg mol−1; Supplementary Table 2).

Across all KOH concentrations, 50SOPBI exhibits the lowest thickness swelling range of 7.5–63.2%. 60SOPBI swells 24.5–86.9%, and 70SOPBI 19.6–101.6% (Fig. 2a–c). Because in-plane swelling is much lower, electrolyte uptake (Supplementary Fig. 4) follows the same trend as the thickness swelling. Anisotropic swelling of PBI is known in both acidic and alkaline media and related to the preferred orientation of polymer chains, probably forming a lamellar-like structure23,36. It was reported that a freeze-dried gel PBI possesses nanofibrils, which aggregate into 28-nm-thick nano-sheets, which are oriented parallel to the plane37. Whereas non-crosslinked MS-PBI swells excessively and eventually dissolves31, SOPBI copolymers remain dimensionally stable, indicating that the strategy to have regions of high and low ion density on the polymer chain works well. Whereas 50SOPBI swells only 63% in thickness in 20 wt% KOH solution, state-of-the-art meta-PBI swells 175% (ref. 38). Presumably, the non-sulfonated OBA units are more hydrophobic than the meta-phenyl linked units and show a higher degree of π-stacking.

Fig. 2: Volume, length and thickness swelling of SOPBI membranes in KOH solutions having different concentrations and the composition of these electrolyte swollen membranes.
figure 2

a–f, Swelling of 50SOPBI (a), 60SOPBI (b) and 70SOPBI (c) membranes in KOH solutions at RT; internal KOH concentration, polymer, water and KOH weight fraction inside 50SOPBI (d), 60SOPBI (e) and 70SOPBI (f) membranes. All data points are average values over three samples. Error bars represent the standard deviation but are often smaller than the symbol that represents the data point.

Source data

Full size image

The properties of ISM depend on the amount and concentration of the absorbed electrolyte. Figure 2d–f show that the internal KOH concentration increases with the concentration of the external solution and is always higher. For 0.11% to 20 wt% KOH solutions, 50SOPBI shows an internal concentration range from 11.9 to 24.3 wt% and from 15.3% to 30.4 wt% for 60SOPBI. 70SOPBI shows lower values of 11.7% to 23.6% KOH. The reason is a strongly increased water absorption (Fig. 2f and Supplementary Fig. 4b), which results in 338% water uptake in 10 wt% KOH and excessive total uptake (Supplementary Fig. 4a). Whereas a high hydration number may improve chemical stability and conductivity of membranes39, excessive water uptake results in large swelling or even dissolution (Fig. 2c) and promotes creep under compressive forces.

As shown in Fig. 3a, the through-plane conductivity at room temperature (RT) increases with the concentration of the external KOH solution and then decreases again, similarly to the uptake data presented in Supplementary Fig. 4. Unlike meta-PBI and 5MS-PBI31, which have the highest conductivity in 25 wt% and 20 wt% KOH solution, respectively, SOPBI membranes reach their highest conductivity in 10–15 wt% KOH solution. At higher concentrations, the membranes lose flexibility and the conductivity decreases.

Fig. 3: Ion conductivity of membranes in response to electrolyte concentration and temperature and the contribution of potassium and hydroxide ions.
figure 3

a–f, Conductivity of KOH solutions43 and SOPBI at RT (a), of 50SOPBI before and after 48 h activation in 15 wt% KOH at 80 °C (b), of 50SOPBI-act at different temperatures (c), contribution of K+ and OH− to the total conductivity of 50SOPBI-act at 80 °C (d), the effect of anode concentration on the transference numbers (e) and conductivity of 50SOPBI-act in comparison to 60 literature data (f; Supplementary Tables 4 and 5), all in ≤1 M KOH, gas or water47.

Source data

Full size image

AEMWE uses electrolyte concentrations of 0.2–1 M KOH4 to mitigate ionomer oxidation at the anode and loss of interfacial contact40. The conductivity of 50SOPBI and 60SOPBI is 21 mS cm−1 and 25 mS cm−1 in 0.2 M KOH and 49 mS cm−1 and 56 mS cm−1 in 1 M KOH, respectively. For comparison, commercial FAA3-50 has a conductivity of 46 mS cm−1 in 1 M KOH17. 70SOPBI shows a conductivity of 129 mS cm−1 in 1 M KOH, which correlates with its strong swelling and water uptake (polymer weight fraction below 31%, 21% swelling in length and 61% in thickness, Fig. 2c; water uptake of 338%, Supplementary Fig. 4b) and raises concerns about its dimensional stability at elevated temperatures. Because the targeted operation temperature for ion solvating membrane water electrolysis (ISMWE) is 80 °C, the stability of all membranes was tested in 1 M KOH at 80 °C for 48 h. 70SOPBI almost completely dissolved, whereas 60SOPBI swelled 150% in thickness and 30% in length. Noticeably, 50SOPBI remained stable, with only a slight yellowish discoloration of the solution. Therefore, 50SOPBI appears to be the most promising membrane.

For use in flow batteries, meta-PBI membranes showed improved conductivity in 3 M H2SO4 when they were pre-swollen at a higher concentration23. Similarly, pre-swelling in 4 M KOH (19 wt% KOH) increased the conductivity in 2 M H2SO4 from 5 mS cm−1 to 56 mS cm−1 (ref. 41). Interactions between polymer chains are broken, and low molecular weight fractions leach out, resulting in higher solution uptake. Thus, to improve the conductivity of 50SOPBI, it was activated for 48 h at 80 °C in 15 wt% KOH (50SOPBI-act), where conductivity and volumetric swelling (Fig. 2a) are the highest. Additionally, pre-swelling may increase the degree of deprotonation. For gel PBI, X-ray diffraction data and molecular modelling suggested that only 50% of imidazole units are deprotonated in 25 wt% KOH at RT27. Supplementary Fig. 6d,e shows that 48-h activation increases the KOH weight fraction and strongly increases total uptake.

Activation increased the room temperature conductivity in 15 wt% KOH to 182 mS cm−1 (Fig. 3b). Figure 3c displays the temperature effect on the conductivity of 50SOPBI-act in 0.2 and 1 M KOH solution. The inflection point at 40 °C may indicate a change in the degree of deprotonation. This is supported by an increased membrane thickness (Supplementary Fig. 6a–c). In 1 M KOH, volume swelling increases from 94.2% at RT to 114% and 149% at 60 and 80 °C. In 2 M KOH, the volume swelling increases from 94% at RT to 114% and 140% at 60 and 80 °C. In 3 M KOH, the volume swelling is less than in 1 and 2 M KOH, from 88.4% at RT to 97.3% and 100% at 60 and 80 °C, respectively. Furthermore, the total uptake levels off above 1 M KOH (Supplementary Fig. 6d), implying that the activation set a limit to the swelling. In 0.2 M KOH and 1 M KOH, 50SOPBI-act exhibits conductivities of 83 mS cm−1 and 135 mS cm−1 at RT and 217 mS cm−1 and 358 mS cm−1 at 80 °C.

Because ISM absorbs electrolyte, the measured conductivity values are the sum of K+ and OH− conductivity (Supplementary Fig. 3a). However, OER consumes OH−, and movement of K+ ions results in a net transport of KOH from anode to cathode. Supplementary Fig. 3b shows that the transport number of K+ ions decreases as the electrolyte concentration increases. During electrolyser operation, the KOH concentration at the cathode increases, whereas that at the anode decreases. This concentration difference creates an osmotic pressure difference, and water diffuses towards the cathode. In electrolysis systems, this is countered by degassing and mixing anolyte and catholyte. In 1 M KOH at 80 °C, the conductivity of K+ and OH− ions is 151 and 207 mS cm−1 (Fig. 3d). In 2 M KOH, it is 158 and 294 mS cm−1, and in 3 M KOH, it is 122 and 411 mS cm−1, respectively. Figure 3e summarizes the effect of anode concentration on the ion transport numbers. In general, the OH− transport number decreases when the anode concentration decreases. This indicates that the membrane’s properties gradually change from anode to cathode and that the side that is in contact with the lower concentrated alkaline solution determines the overall selectivity.

As supported by Fig. 3f, we are not aware of AEMs that have higher conductivity than 50SOPBI-act. In 15 wt% KOH, a conductivity of 592 mS cm−1 at 80 °C was measured (Supplementary Fig. 5a). For comparison, the measured conductivity of hydroxide exchanged 80-μm-thick Piperion in water at 80 °C is 151 mS cm−1, and an AEMWE performance of 13.39 A cm−2 at 2 V was achieved with an optimized AEM showing a conductivity of 208 mS cm−1 in DI water at 80 °C (ref. 42). For 50SOPBI-act, the Arrhenius plot (Supplementary Fig. 5b) reveals an activation energy in 0.2 M KOH of 14.57 kJ mol−1, 1.6 times higher than that of the solution (9.08 kJ mol−1) (ref. 43). In 1 M KOH, the membranes show an activation energy of 18.76 kJ mol−1, almost twice higher than that of the solution (9.45 kJ mol−1); in 2 M KOH, the activation energy is 17.58 kJ mol−1 (1.8 times higher than for 2 M KOH solution). In 3 M KOH, the membranes show an activation energy of 19.32 kJ mol−1 (1.9 times higher than for 3 M KOH solution)43. Physically, a higher activation energy indicates that ion transport by vehicle mechanism contributes more than transport by structural diffusion, for example, because of a reduced water concentration or because narrow passages between polymer chains interrupt the hydrogen-bonded network needed for the Grotthus mechanism.

X-ray diffraction analysis of the peak at 5–10° shows that the inter-chain distances increase with the KOH concentration and correlate with the volume swelling (further discussion in Supplementary Information)23,44.

The ex situ alkaline stability of 50SOPBI-act was tested by immersing samples in 0.2, 1 and 2 M KOH solutions at 80 °C and periodically measuring conductivity at RT (Fig. 4a). In 0.2 M KOH, 50SOPBI-act showed no substantial dimensional changes and a stable conductivity of 86 ± 6 mS cm−1 for over 200 days. In 1 M KOH, conductivity increased linearly from 135.4 to around 170 mS cm−1 after 200 days (0.17 mS cm−1 d−1). In 2 M KOH, a steeper increase rate of 0.47 mS cm−1 d−1 was observed from 180 to 270 mS cm−1 for 200 days. Increased membrane thickness indicates that the conductivity increased due to increased absorption of KOH solution. Most importantly, all membranes remained visibly unchanged in 0.2, 1 and 2 M KOH solutions at 80 °C, and no discoloration of the KOH solutions was observed, implying that the polymer remained stable.

Fig. 4: Alkaline stability of 50SOPBI at 80 °C.
figure 4

a, RT conductivity versus time in 0.2, 1 and 2 M KOH for over 200 days. b, FTIR spectra of pristine 50SOPBI and aged 50SOPBI-act. c, Number average molecular weight (Mn), weight-average molecular weight (Mw) and polydispersity index (PDI) of pristine and aged membranes. * indicates that a different polymer batch was used.

Source data

Full size image

A potential attack of OH− ions would occur at the ether groups and at the imidazole C2 position of non-deprotonated benzimidazole units30. The latter reaction would open the ring and form amides and carboxylic acids. Fourier transform infrared (FTIR) spectroscopy did not show any new band in the carbonyl region (1,600–1,750 cm−1; Fig. 4b). This further supports that hydrolysis of imidazole rings is not observed as depicted in Fig. 1.

Gel permeation chromatography analysis showed that ageing did not reduce the weight-average molecular weight (Mw) of 50SOPBI (Fig. 4c and Supplementary Fig. 8). Whereas pristine 50SOPBI had a Mw of 92 kg mol−1, 50SOPBI-act immersed in 0.2 M and 1 M KOH had Mw of 111 and 99 kg mol−1, respectively, with a similar polydispersity index (PDI). PDI and Mw of pristine *50SOPBI (132 kg mol−1) were similar to those of *50SOPBI-act aged in 2 M KOH (120 kg mol−1). Because OH− attack on imidazole or phenyl ether would result in chain scission and decreased Mw, it can be deducted that 50SOPBI did not experience chain scission.

Negatively charged imidazolide groups should increase the electron density on the ether carbon atoms and by this stabilize the aromatic ether against alkaline hydrolysis. DFT calculations support that an increased activation barrier reduces the risk of nucleophilic attack by OH−. The analysis of the deformation in molecular electrostatic potential indicates a charge transfer from the benzimidazolide part to the ether part, and the results of bond-order analysis explain the transition-state destabilization in the model system containing benzimidazolide in terms of possible resonance structures. A detailed discussion is provided in the Supplementary Information.

Water electrolysis performance and durability with cell type 1

Eighty-μm-thick Piperion and 97-μm-thick 50SOPBI-act were tested in an electrolyser with 0.2, 1, 2 and 3 M KOH feed solutions at 60 and 80 °C. To achieve high performance, the cathode contained Pt/C on carbon paper, and the anode NiFe on Ni felt (cell type 1). 50SOPBI-act showed much higher performance than Piperion (Fig. 5a,b). At 80 °C and 2 V, 50SOPBI-act reached 3.8 A cm−2 and 4.8 A cm−2 in 2 M and 3 M KOH, respectively, whereas Piperion reached only 3.5 A cm−2 and 3.8 A cm−2, respectively. Piperion’s similar performance in 2 and 3 M KOH is expected because the conductivity of AEMs typically levels off at higher KOH concentrations17. The performance of the 50SOPBI-act membrane is compared with literature data in Supplementary Table 1.

Fig. 5: Electrolysis performance of cell type 1 equipped with Pt/C/Piperion/carbon paper cathode and NiFe/Ni felt anode.
figure 5

a–f, Current–voltage (iV) curves measured at 60 °C are shown for cells with 80-μm Piperion (curves for 1 M and 2 M overlap) (a), 97-μm 50SOPBI-act (b), Nafion 212 (curves for Nafion overlap) (c), 10-μm (dry state) and 40-μm (dry state) mPBI and 500-μm Zirfon; and at 80 °C for 80-μm Piperion (d), 97-μm 50SOPBI-act (e), Nafion 212 (f), 10-μm (dry state) and 40-μm (dry state) mPBI and 500-μm Zirfon.

Source data

Full size image

50SOPBI-act was also compared with the cation exchange membrane Nafion 212, a 10- and a 40-μm-thick commercial meta-PBI (mPBI) membrane activated at 80 °C (resulting in 13- and 58-μm thickness in 2 M KOH and 16- and 70-μm thickness in 3 M KOH) and 500-μm Zirfon (Fig. 5c,f). In ≥ 2 M KOH, at 60 and 80 °C, all membranes showed lower performance than 97-μm-thick 50SOPBI-act. Nafion 212 gave the lowest performance because it has high selectivity for the less conductive K+ ions45,46, which are not involved in the faradaic reactions at the electrodes. This increases ohmic resistance and locally depletes the concentration of OH− ions close to the surface of the OER catalyst. At 2 V, Nafion only reached 114 and 107 mA cm−2 at 60 °C in 2 and 3 M KOH, respectively, and 141 and 146 mA cm−2 at 80 °C. In case of the 13-μm-thick mPBI membrane, which is 7.5 times thinner than the 50SOPBI-act membrane, the performance was relatively high. At 80 °C in 2 and 3 M KOH, the current density reached 3.0 and 4.0 A cm−2. However, this is due to the membrane’s very low thickness, which sacrifices its mechanical integrity and increases hydrogen crossover, which is impractical for real applications. For durability, mPBI would need to be reinforced, but the thinnest-available polyphenylene sulfide (PPS) mesh at the moment is >60 μm thick. However, 40-μm-thick mPBI (58 μm in 2 M KOH; 70 μm in 3 M KOH) showed a much lower performance than 50SOPBI-act. At 80 °C in 2 and 3 M KOH, the current density reached only 1.5 and 1.7 A cm−2. A commercial mPBI membrane, activated at 80 °C and immersed in 2 and 3 M KOH for 2.5 months at room temperature, disintegrated (Supplementary Fig. 10), whereas 50SOPBI-act membranes that were stored at 80 °C in 2 M KOH for 15 months and in 3 M KOH for 9 months showed excellent stability. 50SOPBI-act remained flexible and did not break even when stretched by tweezers. Mechanical properties are shown in Supplementary Table 6.

Electrochemical impedance spectroscopy (EIS) analysis (Supplementary Figs. 9 and 11) reveals that 50SOPBI-act has lower high-frequency resistance (HFR) in 0.2, 1, 2 and 3 M KOH than Piperion at both 60 and 80 °C, because the thicker 50SOPBI-act has higher conductivity than Piperion.

With 0.2 and 1 M KOH, the cell with 50SOPBI-act exhibited lower performance than that with Piperion at both temperatures. This is unexpected because the conductivity of OH− exchanged Piperion in water at 80 °C (151 mS cm−1) is lower than that of 50SOPBI-act in 0.2 M KOH (216 mS cm−1) and 1 M KOH. The reason is that the cell with 50SOPBI-act has lower HFR but significantly higher polarization resistance than the cell using Piperion, similar to our previous observation for sulfonated para-PBI31. The exact reason will be investigated in future work but could be related to loss of active catalyst area by membrane entering the electrode pores, gas entrapment within the porous electrodes, catalyst poisoning, KOH depletion, oxidation of PBI in contact with anode catalyst or ion pair formation between cationic Piperion binder and anionic ISM.

Performance in a PGM-free cell

Another electrolyser with 98-µm-thick 50SOPBI-act membrane and PGM-free electrodes based on Raney Ni-type catalysts was operated with 1 and 2 M KOH electrolytes at 60 and 80 °C. The polarization curves indicate currents above 1 A cm−2 at 2 V in all conditions (Fig. 6a,b). The performance improves with increased KOH concentration, probably due to reduced ohmic and charge transfer resistance, as evidenced by EIS measurement (Supplementary Fig. 12). Presumably, higher KOH concentration improves the catalyst activity, membrane conductivity and availability of hydroxide ions. With 0.2 M feed solution, higher total cell resistance led to low performance. Increasing the temperature has a similar effect as increasing the KOH concentration. As the temperature rises from 60 to 80 °C, performance improves from 1.04 to 1.30 A cm−2 at 2 V in 1 M KOH and from 1.52 to 1.91 A cm−2 in 2 M KOH, implying that 50SOPBI-act exhibited excellent performance with PGM-free electrodes. As shown in Supplementary Fig. 12, this can be attributed to reduced ohmic and charge transfer resistance.

Fig. 6: Electrolysis performance of cell type 2 equipped with Raney Ni PGM-free electrodes, 98-µm 50SOPBI-act membranes.
figure 6

a,b, iV curves are shown for operation with 0.2, 1 and 2 M KOH solution at 60 (a) and 80 °C (b). c,d, Comparison of 98-µm 50SOPBI-act and 150-µm PPS-reinforced 50SOPBI-act in 2 M KOH at 60 (c) and 80 °C (d).

Source data

Full size image

To protect the membrane against compressive forces and sharp edges of nickel-based substrates, it was reinforced with a 60–70-µm-thick PPS textile (Supplementary Fig. 13). When a coin was pressed on the membranes, it only cut through the non-reinforced 50SOPBI membrane (Supplementary Fig. 13c). Although PPS fibres themselves are not ion conductive, the PPS-reinforced 50SOPBI-act membrane still had a high ionic conductivity of 399 mS cm−1 in 2 M KOH at 80 °C (Supplementary Fig. 14). An AEMWE cell with PPS-reinforced 50SOPBI-act showed slightly lower performance due to increased ohmic resistance (Supplementary Fig. 12c,d) but still achieved 1.3 and 1.70 A cm−2 at 2 V and 60 and 80 °C, respectively (Fig. 6c,d).

The durability of cells with PPS-reinforced 50SOPBI-act in 2 M KOH was evaluated at constant current density of 0.5 A cm−2 at 80 °C (Supplementary Fig. 15). Though the voltage increased, no failure was observed over 1,000 h. A repeat experiment demonstrated high reproducibility (Supplementary Fig. 15), and the cell was stopped after 400 h. When the cell was restarted again, voltage decreased by about 90 mV and the voltage loss rate decreased. Because the ex situ stability test showed that the membrane does not lose its conductivity (Fig. 4a), the voltage increase could stem from factors such as bubble management or loss of contact between membrane and electrode, which would not be directly a membrane issue.

H2 permeation test cell

Figure 7 and Supplementary Fig. 16 show that the H2 in oxygen level (HTO) decreases with increasing current density because the produced oxygen dilutes the crossed-over H2. With 1 M KOH, the HTO at 40 and 60 °C stays below the safe level of 2% when a cell with 50SOPBNI-act is operated at ≥100 mA cm−2, whereas a current density of 400 mA cm−2 is needed at 80 °C. Similarly, in 2 M KOH, the HTO at 40 and 60 °C is safe at current densities ≥50 mA cm−2 and safe at ≥200 mA cm−2 at 80 °C. An 80-μm-thick Piperion membrane showed a higher HTO than 50SOPBI (Fig. 7c). To reach HTO below 2% in 2 M KOH at 40, 60 and 80 °C, Piperion needs to be operated at current densities above 150, 270 and 410 mA cm−2, respectively. In summary, the 50SOPBI-act membrane can be safely operated over a wider current range than the Piperion membrane.

Fig. 7: HTO level tested in cell type 3 at different temperatures, current densities and KOH concentrations.
figure 7

a,b, HTO levels with 102-µm-thick 50SOPBI-act using 1 M KOH (a) and 2 M KOH solution (b). c, HTO levels with an 80-μm-thick Piperion membrane. Shaded area indicates a safe gas composition (50% below the lower explosion limit).

Source data

Full size image

Scaled-up operation in a 25-cm2 cell

A PPS-reinforced 50SOPBI-act was tested in a 25-cm2 electrolysis cell (Supplementary Fig. 17) with 2 M KOH feed solution at 80 °C and a current density of 500 mA cm−2 for 500 h (Fig. 8). The use of separate electrolyte chambers resulted in a net movement of KOH from anode to cathode. This increased the charge transfer resistance and thus the voltage. To minimize this effect, the electrolyte volumes were manually rebalanced every 24 h. Every five days, the electrolytes were renewed. The voltage almost fully recovered (Fig. 8 and Supplementary Fig. 18), implying that the voltage increase is not due to membrane degradation, but rather due to the KOH-depleted anolyte. Supplementary Fig. 19 reveals a negative linear correlation between voltage and anolyte concentration. Further experiments with different anode and cathode concentrations confirmed that the cell performance is strongly influenced by the anode concentration (Supplementary Fig. 20). On the basis of the voltage values after each electrolyte replacement, a voltage increase rate of 35.6 μV h−1 was observed (Fig. 8). During the first 150 h, the HFR increased by 8.7%, most probably due to activation of the cell. After that, HFR decreased 10.9%, presumably due to increased electrolyte uptake or some morphological changes over time, which is supported by Fig. 4a.

Fig. 8: Voltage and HFR during 550-h testing of a 130-μm-thick PPS-reinforced 50SOPBI-act in cell type 4.
figure 8

The test was done at 80 °C, at a constant current density of 500 mA cm−2, and with 2 M KOH feed solution (50 ml min−1); cathode was Pt/C/carbon paper, NiFe electrodeposited on Ti mesh was used as anode.

Source data

Full size image

Discussion

Sulfonated OPBI, which was originally synthesized by others for use in PEM fuel cells, can perform excellently as an ISM in alkaline electrolytes. By optimized partial sulfonation of OPBI, it was possible to control the swelling in KOH solution without the need of crosslinking. This is advantageous, because crosslinking fully sulfonated PBI membranes with dibromoxylene not only complicates the membrane casting process, but the crosslinking points are neutral imidazole groups, which can be more easily attacked by OH− than deprotonated imidazolide units.

The optimized 50SOPBI membrane with a DOS of 50% showed appropriate swelling behaviour and high conductivity over the broad range of 1 to 25 wt% KOH, that is, the new ISM serves the whole broad operational range from the AEMWE range (0.2 to 1 M KOH) over the previously not-much-investigated range of 1–4 M KOH to the AWE range (4 to 7 M KOH).

After an activation step, the 50SOPBI-act membrane reached a conductivity at room temperature of 135 mS cm−1 in 1 M KOH. At 80 °C, the conductivity reached 358 mS cm−1.

In an ex situ stability test (1 M KOH, 80 °C, 200 days), no degradation was observed. The membranes remained flexible, showed practically no discoloration of the KOH solution, and the molecular weight range was not decreased. Rather, membranes continuously absorbed KOH solution, which increased the initial conductivity at room temperature from 135 mS cm−2 to 175 mS cm−2. But this might not be seen in the application, where the membranes are tightly clamped between the electrodes.

Whereas cell operation with nickel foam, carbon felt and Ni felt electrodes did not damage the membrane, operation with catalyst-coated porous nickel sheets gave mechanical damages at the edge of the active area. This problem was tackled by reinforcing the membrane with a PPS textile, which acted as a hard stop, so that the sharp, hard edge of the nickel sheet could not damage the membrane. This was also proven in an ex situ test in which a coin was pressed onto membranes. The coin cut a disc from the non-reinforced membrane, but the PPS-reinforced membrane could not be punched through.

Performance-wise, an AEMWE cell (type 1) with 3 M KOH feed solution and 50SOPBI-act showed higher performance than a cell with commercial Piperion membrane, 4.8 A cm−2 and 3.8 A cm−2 at 2 V, respectively. In a PGM-free cell (type 2, 2 M KOH solution, 80 °C), the current density at 2 V reached 1.91 A cm−2 for SOPBI-act and 1.70 A cm−2 for PPS-reinforced 50SOPBI-act. A test for H2 crossover in cell type 3 (nickel foam electrodes) revealed that the HTO remains below the safety limit of 2% at current densities ≥50 mA cm−2 at 60 °C and 2 M KOH and ≥200 mA cm−2 at 80 °C, which allows for a broad operational capacity. In another cell (type 4, active area of 25 cm2), a 500-h test showed a stable performance with a low voltage increase of 36 μV h−1. We expect that further optimization of the cell components and operating conditions (porous transport layer, catalysts, catalysts loading, cell compression, current, current profile, temperature, pressure, electrolyte concentration, electrolyte flow rate, dry or electrolyte-fed cathode) can improve durability.

A cost calculation shows that the material costs is just US$0.84 g−1 for 50SOPBI polymer, US$1.02 g−1 or US$59 m−2 for 50SOPBI dry membrane (Supplementary Tables 7–12).

In conclusion, 50SOPBI is a very attractive membrane, and the data confirm that ISM in LA ISMWE have the potential to outperform AEM in AEMWE applications in terms of cost, performance and durability, which opens a path to reduce the cost of green H2 production.

Methods

Materials

4,4’-oxybisbenzoic acid (OBA), 20% fuming sulfuric acid and lithium chloride were purchased from Sigma Aldrich. 3,3’-diaminobenzidine (DAB) was purchased from Longyan Tianhua Biological Technology Co. Polyphosphoric acid 115% (PPA), phosphoric acid 85 wt%, sodium chloride (NaCl), dimethylsulfoxide (DMSO), potassium carbonate (K2CO3) and potassium hydroxide were obtained from Daejung Chemicals. All chemicals were used as received without further purification. PPS textile (polyphenylene sulfide industrial mesh, PPS0177/70PW: 35-µm fibre diameter, 60-µm thickness, specified working temperature up to 215 °C, 70% open area, mesh opening 177 µm) was provided by PVF Mesh & Screen Technology.

Synthesis of sodium 6,6’-oxybis(3-carboxybenzenesulfonate)

The synthesis follows a previously reported procedure34. Fifteen g (58.14 mmol) of 4,4′oxybisbenzoic acid was placed inside a 100-ml round-bottom flask equipped with argon inlet, and 45 ml 20% fuming sulfuric acid was added. Then the mixture was stirred at 110 °C for 2 h, resulting in a brown viscous solution. The solution was cooled down to RT and poured on 100 g of ground ice. When all ice was molten, 30 g NaCl were added, and a white precipitate formed. The precipitate was collected by filtration and purified by recrystallization in deionized (DI) water twice. The purified crystals were vacuum dried at 110 °C for 24 h. The yield was 20.4 g (43.93 mmol, 76%). 1H-NMR (Bruker UltraShield 400 Hz, (DMSO-d6, ppm): 8.41 (d, J = 2.1 Hz, 2H), 7.91 (dd, J = 8.4, 2.1 Hz, 2H), 6.84 (d, J = 8.7 Hz, 2H).

Synthesis of SOPBI copolymers

Exemplary, the synthesis of 50SOPBI is given. An amount of 4.634 g (21.630 mmol) DAB was placed into a 500-ml flask equipped with inert gas inlet and overhead stirrer, followed by addition of 213.5 g PPA. The mixture was heated to 110 °C for 2 h until all DAB dissolved. Then, 5 g (10.815 mmol) of SODBA was added. The mixture was rigorously stirred for 30 min followed by a slow stirring for 2 h at 140 °C. Then, 2.792 g (10.815 mmol) of OBA was added. After 30 min of rigorous stirring, the temperature was raised and kept at 160 °C for 16 h. The resulting viscous brown solution was diluted with 10-ml phosphoric acid and slowly poured into DI water, to form brown polymer fibres. The fibres were neutralized by 24-h immersion in 2 l of 0.5 M K2CO3 solution and were washed repeatedly with DI water to remove excess K2CO3. Next, the fibres were soaked in acetone for 2 h to exchange the water and afterwards dried in a vacuum oven at 100 °C for 24 h. For other copolymers, the ratios of OBA and SODBA were changed, and the amount of PPA was adjusted to reach 5.5% total monomer contents.

Membrane fabrication

LiCl at 3.5 wt% with respect to the polymer weight was dissolved in DMSO, and 8 wt% SOPBI with respect to DMSO was added. The mixture was stirred at 70 °C for 24 h. Undissolved particles were removed by centrifugation. Afterwards, the polymer solution was sonicated to remove air bubbles and stored at 30 °C. To make membranes, the solution was casted on a glass plate using a doctor blade. The wet film was kept in an oven at 80 °C for 2 h at atmospheric pressure and for 24 h in vacuum. After cooling to RT, the dry membrane was delaminated from the glass plate and stored in a zip bag.

Gel permeation chromatography

Gel permeation chromatography was conducted with a Waters 2707 autosampler equipped with a Styragel HR 3 (WAT044223) column calibrated with poly(methyl methacrylate) (PMMA; PSS, MW = 596; 1,930; 5,170; 10,900; 18,700; 41,400; 88,500; 217,000; 340,000; 608,000; 990,000 and 2,050,000 g mol−1). Signals were detected with a Waters 2414 RI detector. Before the measurement, 6-mg polymer were dissolved in 2-ml 0.05 M LiBr/NMP at 80 °C for 24 h. Then the solutions were filtered by 0.45-μm PTFE syringe filter and put into autosampler for measurement.

Conductivity and alkaline stability

A KOH-doped membrane with wet thickness twet was placed between two disc-shaped electrodes (gold or platinum) of an active area (A) of 1.767 cm2. The gap between the electrode and the membrane was filled with the respective KOH solution, and through-plane ionic conductivity was measured by EIS in the galvanostatic mode. The measurement was repeated for stacks of two and three membranes. The resistances (R) were plotted as a function of membrane thickness. The empty cell resistance (Rempty) as the total system resistance was extracted from the y-axis intercept. The conductivity of the membrane can be calculated according to following equations (1) and (2):

$$\sigma \,\left(\mathrm{conductivity},\,\mathrm{mS}\,{\mathrm{cm}}^{-1}\right)=\frac{{t}_{\mathrm{wet}}}{\mathrm{ASR}}$$
(1)
$$\mathrm{ASR}\,\left(\mathrm{area}\,\mathrm{specific}\,\mathrm{resistance},\,{\rm{\Omega }}\,{\mathrm{cm}}^{2}\right)=\left(R-{R}_{\mathrm{empty}}\right)\times A$$
(2)

Alkaline stability was characterized by measuring the conductivity at room temperature of membranes that were stored in KOH solutions at 80 °C for prolonged times.

Swelling ratio

Swelling behaviour was characterized by measuring dimensional changes for length (l) and thickness (t) between pristine dry (linitial and tinitial) membranes and doped membranes (ldoped and tdoped). Swelling ratios were calculated according to equations (3) and (4).

$$\mathrm{Length}\,\mathrm{swelling}\,\left( \% \right)=100\times \frac{{l}_{\mathrm{doped}}-{l}_{\mathrm{initial}}}{{l}_{\mathrm{initial}}}$$
(3)
$$\mathrm{Thickness}\,\mathrm{swelling}\,\left( \% \right)=100\times \frac{{t}_{\mathrm{doped}}-{t}_{\mathrm{initial}}}{{t}_{\mathrm{initial}}}$$
(4)

Water and KOH uptake

The pristine dry membranes were doped in KOH solution. After a certain time, the wet membranes were quickly wiped by tissue, and the wet weight was noted (wwet). After water evaporation at 100 °C overnight, wdry was noted. Next, the membrane samples were washed with DI water until the pH became neutral and dried at 100 °C overnight (wfinal). The total, water, KOH uptake and polymer fraction were calculated according to equations (5)–(8). The KOH-free potassium form of 50SOPBI membrane before doping with KOH and after doping and washing with water is confirmed by scanning electron microscopy coupled with energy dispersive X-ray spectroscopy (Supplementary Table 3).

$$\mathrm{Total}\,\mathrm{uptake}\,\left( \% \right)=100\times \frac{{w}_{\mathrm{wet}}-{w}_{\mathrm{final}}}{{w}_{\mathrm{final}}}$$
(5)
$$\mathrm{KOH}\,\mathrm{uptake}\,\left( \% \right)=100\times \frac{{w}_{\mathrm{dry}}-{w}_{\mathrm{final}}}{{w}_{\mathrm{final}}}$$
(6)
$$\mathrm{Water}\,\mathrm{uptake}\,\left( \% \right)=100\times \frac{{w}_{\mathrm{wet}}-{w}_{\mathrm{dry}}}{{w}_{\mathrm{final}}}$$
(7)
$$\mathrm{Polymer}\,\mathrm{fraction}\,\left( \% \right)=100\times \frac{{w}_{\mathrm{final}}}{{w}_{\mathrm{wet}}}$$
(8)

Activation of 50SOPBI membranes

To improve the conductivity of 50SOPBI, a Teflon container was filled with 15 wt% KOH solution, 50SOPBI membranes were added and the container was closed and stored at 80 °C for 48 h. After again cooling down to RT, the membrane surface was wiped with a tissue to remove excess solution, and the membrane was re-immersed in a solution with the finally targeted KOH concentration for at least 24 h and stored until use. The membranes are denoted as 50SOPBI-act.

Ion transport number

The ion transport number of K+ and OH− was conducted by modified Hittorf method by comparing the concentration change of either anode or cathode with the produced OH− at the cathode side (N). The membrane was placed in the electrolyser cell and was operated at fixed current (i) in 1, 2 and 3 M KOH solution. Periodically (t), 15-ml solution was taken from anolyte and catholyte, and the volume change was observed. The samples were later titrated with 2 M HCl to give the concentration of the anolyte and catholyte and to calculate the mol changes of the electrolyte over time. The ion transport numbers of K+ and OH− are calculated as follows:

$${{\rm{H}}}_{2}{\rm{O}}+{{\rm{e}}}^{-}\to {\mathrm{OH}}^{-}+{0.5{\rm{H}}}_{2}$$
(9)
$${\rm{Q}}\,\left(\mathrm{Amount}\,\mathrm{of}\,\mathrm{charge}\,\mathrm{involved}\,\mathrm{in}\,\mathrm{reaction}\right)=\int i{\rm{d}}t$$
(10)
$$N=\frac{Q}{nF}$$
(11)

n: number of electrons involved to produce 1 mol OH−

F: Faraday constant (96,485 C/mol)

$$\begin{array}{rcl}\begin{array}{l}{t}_{{\rm{K}}}({{\rm{K}}}^{+}{\rm{i}}{\rm{o}}{\rm{n}}\,{\rm{t}}{\rm{r}}{\rm{a}}{\rm{n}}{\rm{s}}{\rm{p}}{\rm{o}}{\rm{r}}{\rm{t}}\,{\rm{n}}{\rm{u}}{\rm{m}}{\rm{b}}{\rm{e}}{\rm{r}})\\ =\displaystyle \frac{{\rm{m}}{\rm{o}}{\rm{l}}\,{\rm{c}}{\rm{h}}{\rm{a}}{\rm{n}}{\rm{g}}{\rm{e}}\,{\rm{a}}{\rm{t}}\,{\rm{a}}{\rm{n}}{\rm{o}}{\rm{d}}{\rm{e}}}{N}=\displaystyle \frac{{\rm{i}}{\rm{n}}{\rm{i}}{\rm{t}}{\rm{i}}{\rm{a}}{\rm{l}}\,{\rm{a}}{\rm{n}}{\rm{o}}{\rm{d}}{\rm{e}}\,{\rm{m}}{\rm{o}}{\rm{l}}-{\rm{a}}{\rm{n}}{\rm{o}}{\rm{d}}{\rm{e}}\,{\rm{m}}{\rm{o}}{\rm{l}}\,{\rm{a}}{\rm{t}}\,{\rm{t}}{\rm{i}}{\rm{m}}{\rm{e}}\,t}{N}\end{array}\end{array}$$
(12)
$${t}_{{\rm{K}}}=\frac{nF\left(\mathrm{initial}\,\mathrm{anode}\,\mathrm{mol}-\mathrm{anode}\,\mathrm{mol}\,\mathrm{at}\,\mathrm{time}\,t\right)}{\int i{\rm{d}}t}$$
(13)
$${t}_{\mathrm{OH}}\left({\mathrm{OH}}^{-}\,\mathrm{ion}\,\mathrm{transport}\,\mathrm{number}\right)=1-{t}_{{\rm{K}}}$$
(14)

Electrolyser test

Cell type 1, platinum cathode and non-PGM anode

The electrodes were synthesized utilizing the catalyst-coated substrate (CCS) method. Platinum on carbon (Pt/C, Pt 47.1 wt%, Tanaka K.K.) was employed as cathode catalyst. The catalyst ink was fabricated by mixing catalyst powder, ionomer (Piperion), distilled water and isopropyl alcohol. The volume ratio of isopropanol to DI water is 10 to 1. The density of catalyst in the ink is 10 mg ml−1. The content of ionomer was 5 wt% of the catalyst. The catalyst ink was ultrasonicated for 1 h and hand sprayed onto substrates. The cathode was constructed by applying 0.5 mg cm−2 of Pt/C onto carbon paper (SGL GDL 39BB, CNL). Ionomer-free anode was prepared by electrodeposition. The electrodeposition bath consisted of DI water (~18.1 MΩ), 15 mM Ni(NO3)2·6H2O (99.999%, Sigma Aldrich) and 15 mM Fe(NO3)3·9H2O (99.95%, Sigma Aldrich). A −10 mA cm−2 amount was applied to the working electrode for 600 s under ambient conditions with a SP300 (biologic). Ti mesh was used as the counter electrode and saturated calomel electrode was used as the reference electrode. The electrode was rinsed with water and finally dried in air.

A commercial anion exchange membrane (Piperion, 80 µm) was immersed into 1.0 M KOH for at least 24 h for activation before the membrane electrode assembly fabrication. 50SOPBI membranes were activated in 15 wt% KOH at 80 °C for 48 h, denoted as 50SOPBI-act, followed by equilibration in 0.2,1, 2 and 3 M KOH for at least 24 h. All membranes were sandwiched between the anode and the cathode. Au-coated Ti and graphite bipolar plates were employed as the current collectors on the anode and cathode sides of the cell. The cell assembly was tightened to 4 N m using a torque wrench.

The electrolysis performance was measured at 60 and 80 °C using a potentiostat (HCP-803, Bio-Logic). The electrolytes (KOH solutions with concentrations of 0.2 M, 1 M, 2 M and 3 M) were circulated to the anode and the cathode with a 15 ml min−1 flow rate. The current–voltage curve of the AEMWE was obtained by linear sweep voltammetry, which ranged from 1.35 to 2.2 Vcell at a scan rate of 10 mV s−1. EIS was conducted at 1.6, 1.8 and 2 Vcell with an a.c. frequency ranging from 10 kHz to 500 mHz and an alternating voltage amplitude of 10 mV.

Cell type 2, non-PGM anode and cathode

Electrodes were produced by spraying globular gas-atomized nickel-aluminium-type alloy powders with and without Mo for anode and cathode electrodes, respectively. Al acts as a pore-forming agent. Coated porous electrodes were produced on the surface of 4-cm2 Ni mesh substrate using thermal spray technique26.

The zero-gap electrolyser cell consists of five main parts: nickel bipolar plates, Ni wire mesh as gas diffusion layers (GDL), membranes and coated Ni-based electrode as anode and coated Mo incorporated porous Ni as cathode. Activated membrane is placed between two electrodes and supported from each side by porous Ni GDLs and fixed by nickel bipolar plates from both sides. The test has been carried out in different KOH concentrations, 0.2, 1 and 2 M at two different temperatures, 60 and 70 °C, by recording polarization curves up to 2 A cm−2 with a scan rate of 10 mA s−1. EIS was performed from 100 kHz to 100 mHz to identify the losses. The operating conditions and cell hardware were kept the same for all the tests. The ohmic resistance or high-frequency resistance, appearing as the intercept of the Nyquist plots with the x axis at high frequency (left side of Nyquist plot) represents the total internal ohmic resistance of the cell, which arises mainly from solution and membrane resistance. The diameter of the semicircle is therefore equal to the charge-transfer resistance. The real axis value at the other (low frequency) intercept point indicating the overall cell resistance is the sum of the charge transfer resistance and the ohmic resistance.

Cell type 3, H2 crossover test with nickel foam electrodes

Either 0.2, 1 or 2 M KOH solution was pumped to both anode and cathode sides at a flow rate of 80 ml min−1 and partially connected via a 20 cm long ¼-inch tube located below the degassing vessels and before the pumps. The cells were operated for 4 h at each current density, with excluding data from the first 30 min for system equilibration. Measurements were initially performed at 80 °C, followed by 60 °C and 40 °C. At 16 ml min−1, nitrogen was purged into the anode degassing vessel, and the exhaust gas was passed through a drying column to remove the water vapour before reaching an electrochemical H2 sensor (Geopal GJ-EX gas detector equipped with a Membrapor H2/M-40000 sensor). Before the measurement, the sensor was calibrated with a 1,000-ppm reference gas. The results were corrected for the nitrogen fraction from the purge and by assuming zero humidity and accounting for the oxygen rate based on the faradaic current.

Cell type 4, 25-cm2 platinum cathode and non-PGM anode, else similar with cell type 1

The durability test was carried out at 500 mA cm−2 in 2 M KOH using a 25-cm2 cell with a flow rate of 50 ml min−1, and the high-frequency resistance was measured every 10 min at 10 kHz.