Quaternary Ammonium‐Free Membranes for Water Electrolysis with 1 m KOH

While anion exchange membrane (AEM) water electrolysis has many advantages, its commercialization is impeded by the low alkaline stability of most AEM, due to the fragility of quaternary ammonium groups. Ion solvating membranes (ISM) can be an alternative, but so far require high alkaline concentrations. Here, it is shown that sulfonation of polybenzimidazole results in ISM which swell strongly in 1 m KOH. Crosslinking with dibromoxylene controls the swelling, and after activation conductivities of >100 mS cm−1 can be reached in 1 m KOH. Stability in 1 m KOH at 80 °C is excellent: Conductivity remains unchanged and tensile strength and Young's modulus remains high over the test period of a half year. In an electrolyzer operating with 1 m KOH feed solution at 80 °C, a stable performance is achieved for over 500 h test without failure, suggesting that the high alkaline stability observed in ex situ tests is also achieved in the electrolyzer.


Introduction
Water electrolysis is an essential component for the energy transition from fossil fuels toward carbon-free technologies.Hydrogen produced by water electrolysis not only functions as a fuel, but also offers the possibility to store renewable energy, for example by using gas tanks, the gas grid, or caverns.
Electrolysis traditionally is done by alkaline water electrolysis (AWE). [1,2]In these systems, >25 wt.% KOH solutions are fed into the electrolysis stack, and anode and cathode are separated by a porous diaphragm that provides ionic contact between the electrodes and prevents intermixing of the evolved gases.While the high KOH concentration raises concerns for material stability, it allows the use of cheap metals for use in bipolar plates, porous transport layers, and as catalyst, like nickel or nickel alloys, etc. Drawbacks of these systems are 1) the high resistance of the separator -commercial Zirfon is 0.5 mm thick, and 2) the narrow operational capacity window, because a high minimum current is needed to prevent explosive H 2 /O 2 mixtures, and a low maximum current should not be exceeded, because the high cell resistance and relatively low catalytic activity of nickel lead to high overpotentials, which results in a rapid increase of voltage when the current is increased.
3) The porous nature of the separators prevents operation at higher differential pressures.However, an elevated cathode pressure is wanted to reduce the cost for drying and compression.4) Due to the above-mentioned points, combination of AWE systems with intermittent renewable energy sources like wind or solar is challenging.
Polymer electrolyte membrane water electrolysis (PEMWE) has been developed since the 1960s, and is now commercialized. [3]Dense polymer membranes like perfluorinated Nafion are used.Because the local pH is low, bipolar plates typically are based on expensive titanium with noble metal coatings, and catalysts are based on platinum and iridium.Since iridium is even scarcer than platinum, large implementation of PEMWE could strongly increase the costs.In addition, Nafion membranes are perfluorinated, and their use may soon be more strictly controlled by legislation. [4,5]The advantage is that PEMWE shows a large operational current window and thus can be integrated with intermittent renewable energy sources like solar and wind.Furthermore, the dense membrane allows to apply differential pressure.
A newer development is anion exchange membrane water electrolysis (AEMWE). [6,7]AEMWE combines the advantages of alkaline systems (cheap metals and catalysts), and PEMWE (use of thin, dense membranes).The drawback is the low alkaline stability of the membranes, which suffer from attack on the polymer backbone (if aromatic ether groups are present), and degradation of the quaternary ammonium groups, which are essential for the ion-conducting properties of the membranes, by nucleophilic substitution and Hofmann elimination. [6]While the initial ambition for AEMWE was to operate with pure water, operation with, e.g., 1 m KOH has the advantages of lower H 2 crossover (because solubility of H 2 decreases with increasing KOH concentration) [8] and improved electrode performance, because ionomer binders are not needed for hydroxide transport. [9]n alternative technology recently investigated is the use of ion-solvating membranes. [10,11]Such membranes can be based on polybenzimidazole (PBI).PBI is not ion conductive, but in contact with strong bases, it is deprotonated to polybenzimidazolides, and the resulting ionic structure absorbs copious amounts of water and KOH. [10]Therefore, the conductivity is based on both potassium and hydroxide conductivity, and reaches a maximum ≈25 wt.% KOH solution, where KOH solutions show a maximum conductivity. [12]If the selected PBI has no ether groups in the backbone, it should resist hydrolysis.Furthermore, these systems have no quaternary ammonium groups, and since the membranes are dense, potentially also operation at differential pressure is possible.Excellent performance of 1.7-1.8A cm −2 at 1.8 V has previously been reported. [11,13]owever, lifetime was shown to be <300 h in an AWE. [11]Potentially, lifetime could be increased if the alkalinity of the feed solution is reduced -however, at low KOH concentrations of 1 m KOH (ca. 5 wt.%),PBI's low conductivity [14] prevents operation in water electrolyzers with ≤1 m KOH. [6]n this work, we hypothesize that monosulfonated para-PBI (MS-PBI), [15] which has a higher number of groups that can be deprotonated by KOH than meta-PBI, i.e., two imidazole groups and one sulfonic acid group, will absorb more KOH solution and thus have higher conductivity.Simultaneously, Nanwen Li et al. reported that a naphthalene-based PBI derivative, functionalized with butylsulfonic acid groups, showed excellent performance in an alkaline electrolyzer with 6 m KOH feed solution. [16]he excessive swelling of MS-PBI in KOH solutions led us to the expectation that this material can also have high conductivity at low KOH concentrations of 1 m KOH, which is within the range where AEM water electrolysis cells are typically operated, and where KOH-doped mPBI has too low conductivity.Crosslinking with dibromoxylene (DBX) [17] was used to control the strong swelling in KOH solution, and stable membranes showing high conductivity similar to that of commercial AEMs in 1 m KOH were obtained.Because crosslinked sulfonated para-PBI has no ether groups and no quaternary ammonium groups, alkaline degradation is very low.

Sulfonated Para-PBI Membranes
A well-defined sulfonated homopolymer (MS-PBI) with an ion exchange capacity of 2.57 meq g −1 is obtained when sulfonated terephthalic acid is copolymerized with diaminobenzidine (Figure 1). [18]Since MS-PBI swelled excessively in 25 wt.%KOH, membranes were crosslinked with 5 or 10 wt.% DBX.The gel content, the insoluble fraction of the polymer in a good solvent, proved effective crosslinking.While 5MS-PBI lost 14% over the time of 5 days, only 4% of 10MS-PBI dissolved at 80 °C in DMAc (Figure S1, Supporting Information).In contrast to this, a non-crosslinked MS-PBI membrane fully dissolved already within 1 day at room temperature. [15]

KOH Uptake and Conductivity
As shown in Figure 2d, the swelling in aqueous KOH was highly anisotropic, with ≈2-4% shrink in the length, and 35-75% swelling in the thickness for 10MS-PBI and 97-143% for 5MS-PBI.This anisotropic behavior is well-known for PBI membranes, and may stem from a lamellar orientation of the polymer chains. [19]The absence of a matrix knee in x-ray analysis suggested that the lamellar orientation does not result in crystalline domains (Figure S10, Supporting Information).Analysis of the weight gain, shown in Figure 2a, confirmed the swelling results.While 10MS-PBI gained 106% weight in 15% KOH solution, 5MS-PBI showed the highest absorption of 150% in 20 wt.% KOH solution.For comparison, Aili et al. reported a maximum uptake of 134% in 25% KOH for non-crosslinked mPBI membranes. [20]oth membranes gained 75% in weight when immersed in 1 wt.%KOH (Figure 2a).In 5% KOH solution (≈0.9 m KOH), 5MS-PBI even showed a weight gain of 109%.For comparison, mPBI was reported to absorb only 23% at this condition [20] and the conductivity in 1 m KOH was just 0.01 mS cm −1 -≈5000 times too low for use in electrolyzers. [21]he weight gain showed a local maximum when the KOH concentration increased (Figure 2a).This behavior is typical for KOH-doped PBI.On the one hand, the amount of absorbed KOH increased continuously with increasing concentration of the doping solution (Figure 2b).On the other hand, the concentration of water in the external solution decreases with increasing KOH concentration, and the external solution's high ionic strength reduced the membrane's water uptake due to increased osmotic pressure.In sum, the water content first increases and then decreases again.When the membrane is considered as a gelled polymer solution, the water and KOH uptake directly allow to calculate a concentration of the absorbed KOH solution.Interestingly, the correlation between the KOH concentration outside and inside the membrane turned out to be highly linear and similar for both membranes (Figure 2c), with a slope close to 1.However, notably the concentration inside the membrane was ≈5 wt.% higher than that of the bulk solution across the concentration range, suggesting that the conductivity of the membrane might peak at lower concentrations than the ca.30 wt.% for KOH solution.
Conductivity of an ISM is based on that of the absorbed KOH solution, and as thus is influenced by a) the ratio of absorbed KOH solution to polymer in the swollen membrane, b) the concentration of the absorbed KOH solution (note that the conductivity of KOH solution has a maximum ≈5-7 m), and c) the morphology of the membrane.As the high weight gain in 1 and 5 wt.%KOH solution at room temperature indicated, both membrane types showed a remarkable conductivity (Figure 3).In 5 wt.%KOH (≈0.9 m), the conductivity of 10MS-PBI was 30 mS cm −1 , and that of 5MS-PBI reached even 48 mS cm −1 , which is higher than that of Fumatech's FAA3-50 membrane in slightly higher concentrated 1 m KOH (46 mS cm −1 in 1 m KOH [22] ).This makes 5MS-PBI promising for testing in AEM water electrolyzers.25] At the high KOH concentration of 25 wt.%, which is of interest for alkaline electrolysis systems, the conductivity of 5MS-PBI is similar to that of commercial mPBI, while that of 10MS-PBI is far lower, and therefore not attractive.
Figure 3b shows the effect of temperature on the conductivity in 5 wt.%KOH solution.At all temperatures, 5′MS-PBI has a higher conductivity than the more crosslinked 10′MS-PBI.While 5′MS-PBI shows the expected curve shape, 10′MS-PBI has an inflection point ≈60 °C.As shown in Figure S6 (Supporting  [14] ), crosslinked MS-PBI and Zirfon; a) at room temperature and in different KOH concentrations; b) temperature effect on conductivity in 5 wt.%KOH.5MS-PBI is filtrated before membrane casting, which may remove some not well dissolved PBI, resulting in a higher degree of crosslinking than expected; 5′MS-PBI has a lower, more exactly adjusted degree of crosslinking, because the PBI contents was analyzed before addition of the crosslinker.5″MS-PBI is similar to 5′MS-PBI, but the membrane was only dried once and solvent traces were not removed by washing with water.Data at 25 °C represents data obtained at uncontrolled ambient room temperature conditions.
Information), the membrane is fully swollen already at room temperature and shows no additional swelling when the temperature is increased, up to 60 °C.Above that temperature, the membrane continues to swell, indicating some changes in the morphology or increased plasticization ≈60 °C.Interestingly, swelling in x and y is practically constant for 5′MS-PBI and 10′MS-PBI over the whole tested temperature range (3-7%).This indicates that the membranes will not suffer from mechanical stress when the temperature of the electrolyzer is increased.In the temperature range relevant for water electrolysis, i.e., 60, 70, and 80 °C, 5′MS-PBI shows a conductivity of 122, 145, and 170 mS cm −1 , respectively.Plotting the data in an Arrhenius plot reveals an activation energy for the conduction of 13.1 kJ mol −1 , about twice that of a KOH solution, which is in the range of 7.0 kJ mol −1 , according to data from Gilliam et al. [12] [34] The difference between 5′MS-PBI and 5′'MS-PBI is that the latter one contained some solvent traces before it was doped.These trace solvent molecules prevent that the polymer chains strongly interact with each other.As a consequence, the membrane absorbs KOH solution more easily, and the conductivity increases.Activation of the 5′MS-PBI in 5 wt.%KOH at 80 °C has a similar effect, the strong absorption of KOH solution at higher temperatures forces the polymer chains apart from each other, and results in an enhanced conductivity also at room temperature.This effect was observed to remain for several weeks.Membranes washed with DMAc at 80 °C for 2 days, then with hot water to remove DMAc before doping showed an even higher conductivity.The assumption is that the non-crosslinked MS-PBI fraction was washed out, adding volume for KOH solution.

Alkaline Stability
Alkaline stability was tested by immersing membrane samples in 1 m KOH and heating to 80 °C.Non-crosslinked MS-PBI immediately lost weight (Figure 4a), and over 30 days the weight loss continued, until the samples broke into pieces.The excessive swelling of non-crosslinked MS-PBI in KOH solution resulted in dissolution, visible by a discoloration of the solution in which the membrane samples were immersed.Because of its insolubility in any solvents, crosslinked 5′MS-PBI showed much higher stability.Visually, samples aged for 6 months remained intact (Figure S3, Supporting Information), and only a slight discoloration of the solution was observed.To avoid that potential mechanical degradation by wiping the membrane surface repeatedly with a tissue reduces the weight, only the thickness change was measured for 5′MS-PBI.
Regarding the degradation mechanism, fully methylated polybenzimidazolium falls apart within a few hours in contact with 0.5 m KOH at room temperature, because the positively charged C2 position reacts with hydroxide ions, which consecutively opens the imidazole ring and breaks the polymer chain. [26]Benzimidazoles should be much more stable because deprotonation in contact with KOH solution forms negatively charged benzimidazolides, which should render reaction with hydroxides energetically unfavorable. [27]Based on experimental and theoretical data, Serhiichuk et al. reported that alkaline degradation of PBI proceeds via the fraction of neutral (i.e., not deprotonated) PBI units. [27]Therefore, as illustrated in Figure 5, the neutral N-alkylated benzimidazoles are expected to have a reactivity inbetween imidazole and imidazolium, and the neutral crosslinking sites should act as the breaking point.Another potential pathway is a nucleophilic substitution reaction on the benzylic position, which would not break the polymer backbone, but destroy the crosslinks and by this transfer 5MS-PBI into MS-PBI over time.Finally, sulfonic acid groups can be substituted by hydroxide groups -synthetically, this reaction is done at temperatures ≈300 °C in molten KOH, [28] but is discussed as a potential degradation reaction for sulfonated quinones. [29]lthough MS-PBI strongly degraded, IR analysis (Figure S4, Supporting Information) showed no significant changes.This supports that the membrane did not chemically degrade, but dissolved.Crosslinked 5′MS-PBI remained stable, but the IR spectra reveal new bands at 1687 and 1572 cm −1 , indicating the presence of amides (C═O and NH), and a broad band at 1610-1626 cm −1 , which is in the range of the N─H deformation of amines. [30]Assuming that the crosslinked, uncharged imidazole rings are the preferred breaking points, we suggest that the imidadzole rings open preferentially into the secondary amide, not into the tertiary amide.Although the tertiary amides obtained from alkaline degradation of methylated PBI have some stability and exist as an observable species, they degrade further, and membranes get brittle and break into pieces. [31,32]Similarly, tertiary amides like N-phenyl-N-benzyl-benzamide should also allow chain scission and thus lead to brittle membranes, which was not observed.Secondary amides, however, have a protonic group; because the predicted pK a of N-phenylbenzylamide is 13.3, [33] it could be deprotonated by KOH, and the resulting anionic structure should slow down chain scission.
Alkaline stability was also monitored by testing the mechanical properties over time.Because individual samples may develop defects, a large variation of the data is expected.Figure 6 shows the mechanical properties of membranes stored up to 168 days in 1 m KOH at 80 °C.Within the first week, most properties apparently decreased, but tensile strength and proportional limit stress increased again over the following 160 days, while the elongation at break constantly decreases from day 30 to day 168.In average, tensile strength, Young's modulus, and elongation at break were 54.3 ± 15.7 MPa, 1.5 ± 0.4 GPa, and 33.1 ± 11.4%, respectively, and the data range was 31-80 MPa, 1.0-2.2GPa and 21-58%, respectively.For comparison, tensile strength, Young's modulus and Elongation at break of Nafion 212 are 16, 0.16, and 261%, respectively, [35] and those of the reinforced commercial anion exchange membrane FAA3-PK-75 are 22 MPa, 0.71 GPa, and 18%, respectively. [34]In conclusion, tensile strength and Young's modulus of 5′MS-PBI are significantly larger than those of Nafion and remain at high values.This indicates that chain scission did not play a role during aging over a half year.
The alkaline stability of most AEM is significantly lower.Several AEMs broke when tested for 1000 h at 80 °C in 1 m KOH. [36]n another work, the conductivity of four commercial AEM decreased below 30 mS cm −1 within the first month. [34]In this light, the alkaline stability of 5MS-PBI membranes is excellent.A recent work showed operation at 70 °C in 1 m KOH for over 8900 h with an AEMION+ membrane, [37] which seems to be a reinforced poly(bis-arylimidazolium) membrane. [38]Presumably, reinforcing 5′MS-PBI would further enhance its robustness, and similar or even longer operation times should be achievable.

Water Electrolysis
A 99 μm thick doped 5′MS-PBI membrane was tested in an electrolyzer in 1 m KOH at 80 °C.Polarization curves and voltage over time are shown in Figure 7. Robust nickel foam electrodes without additional catalyst don't allow high performance, but avoid influences of degrading electrodes or catalyst layers.Najibah et al. reported a performance of 1.98 V at 0.25 A cm −2 in 1 m KOH at 50 °C with nickel foam electrodes for a commercial 50 μm thick FAA3 membrane, [39] while Park et al. reported a performance of 2 V at 1.42 A cm −2 at nominally same conditions, but with Pt/C and IrO 2 as cathode and anode catalysts, respectively. [23]Use of an efficient catalyst system and using a thinner membrane will strongly improve performance of cells with 5MS-PBI.Besides an initial one day break-in at 100 mA cm −2 and the recording of polarization curves or crossover-profiles, the cell was operated at constant 200 mA cm −2 .Despite some fluctuations, the cell voltage appeared stable in the first ca 300 h, after which a gradual activation seemed to have taken place, presumably because the KOH concentration increased and irreversibly swelled the membrane.
The gradual activation is reflected in the recorded impedance data (Figure S5, Supporting Information), which were measured after each polarization curve.The high frequency resistance decreased from 0.252 to 0.160 Ω cm 2 .Using the room temperature thickness of 99 μm and ca.50% thickness increase from 30 to 80 °C (Figure S6, Supporting Information), this corresponds to changes in conductivity from 59 to 93 mS cm −1 .
The hydrogen in oxygen-level (HTO) as function of current density was recorded on the second day.The by-pass valve was closed temporarily to ensure fully separated electrolyte loops while investigating the crossover, after which each current density setpoint was sampled for 1 h.The second half data of each setpoint was averaged and is shown in Figure 8a.Only points between 50 and 300 mA cm −2 are shown, since at higher current densities, the system displayed too severe drift due to increasing concentration gradients across the cell and severe electrolyte imbalances.This is exemplified in Figure S8 (Supporting Information), where the voltage starts to drift and hits the limit at the higher current density setpoints.A specific permeability of 1.28 × 10 −11 mol s −1 cm −1 bar −1 was found by linear extrapolation of the hydrogen crossover flux density to zero current, using the lower three current density points for the fit.The flux densities and fit are shown in Figure S9 (Supporting Information).The time evolution of the HTO is shown in Figure 8b.The majority of the noise-looking transients are caused by imperfect equilibration across the by-pass channel in the balance of plant of the experimental setup, which depending on the flow orientation during a sudden equilibration event can cause either relative increase or decrease of the HTO.Despite this irregularity, there are no signs of membrane failure during the duration of the test, suggesting excellent stability under the evaluated conditions.The test was terminated after 550 h to free up testing capacity for other activities.
The electrolysis tests with highly robust but poor-performing nickel foam electrodes are very well suited to demonstrate the stability of the membrane in the electrolyser, but tests with more active electrodes are needed to demonstrate the expected excellent performance of electrolysis cells with 5′MS-PBI membranes.For that purpose, a comparison with FAA3-PK-75 in a cell using Pt/C and IrO 2 as catalysts and Sustainion XB-7 ionomer binder was tried (Figure S11, Supporting Information).While EIS of the electrolysis cells revealed that the cell with 5′MS-PBI has significantly lower serial resistance than the cell with FAA3-PK-75, the cells with 5′MS-PBI had a very high polarization resistance, which strongly reduced the cell performance.The reason seems to be that the softer PBI-based membrane was more pressed into the electrode pores than FAA3-PK-75 (Figure S11d, Supporting Information), and that the quaternary ammonium-based ionomer binder could form ion pairs with benzimidazolide units, which reduces the number of mobile hydroxide ions in the interface region (Figure S11c, Supporting Information).Therefore, future work should focus on development of improved cell designs with ionomer-free electrodes and optimized cell compression.

Conclusion
While most AEM degrade in alkaline solutions and meta-PBI has practically no conductivity in 1 m KOH, quaternary ammonium (QA)-free sulfonated para-PBI showed high uptake, opening up the possibility to substitute AEM in AEMWE with ion solvat-ing membranes.Crosslinking can control the uptake.5′MS-PBI showed a conductivity of 40-60 mS cm −1 in 1 m KOH.Heating to 80 °C activated the membrane, and room temperature conductivity increased to >100 mS cm −1 .
The absence of quaternary ammonium groups resulted in excellent alkaline stability.Over 6 months at 80 °C in 1 m KOH, conductivity, tensile strength and Young's Modulus remained high, while several commercially available AEM failed within 1 month.Performance in an electrolyser (1 m KOH, 80 °C) was stable and tested without failure for 500 h.
Future work should avoid NCH 2 -crosslinking units to ensure full deprotonation and thus further enhance stability, and reinforcements should be used to counter compressive forces in the cell.Furthermore, the electrodes should be optimized for use with ISM.

Experimental Section
Materials: Polyphosphoric acid (PPA), N,N-dimethylacetamide (DMAc 99.5 wt.%) and K 2 CO 3 were obtained from Daejung Chemicals, monosodium 2-sulfoterephthalate and ,′-dibromo-p-xylene (DBX) were obtained from TCI, 3,3′-diaminobenzidine was obtained from Longyan Synthesis of MS-PBI: 50 g polyphosphoric acid (PPA) was placed into a 250 mL round bottom flask equipped with a condenser and an argon inlet.The flask was placed in an oil bath and the viscous PPA solution was allowed to stir at 100 °C for 2 h under an argon atmosphere.3,3′-diaminobenzidine (0.86 g, 4.0 mmol) and 2 sulfoterephthalic acid monosodium salt (1.10 g, 4.0 mmol) were added slowly into the PPA solution.The reaction temperature was increased from 100 to 200 °C at steps of 20 °C every 2 h.When 200 °C was reached, the temperature was held for 20 h.The hot viscous reaction mixture was poured into 1 L water.Brown fibers formed and were filtered and washed repeatedly with deionized water until the filtrate showed a neutral pH.The fibers were dried at 100 °C under vacuum for 12 h, and were then successively immersed in 10% aqueous K 2 CO 3 for 24 h, then in water, and then was dried at 100 °C under vacuum for 12 h.Gel permeation chromatography (0.05 m LiBr in N-methyl-2-pyrrolidinone) showed a M n of 25,500 g mol −1 and a M w of 81,086 g mol −1 .
Membrane Fabrication: MS-PBI was dissolved in dry DMAc at 140 °C under argon atmosphere.The solution was filtered with a 0.45 μm PTFE syringe filter, a solution of DBX in DMAc was added to reach the targeted amount of crosslinker, and so that the polymer content in the casting solution reached 3 wt.%.After stirring for 15 min, the solution was casted on a glass plate using a doctor blade.The film was initially heated to 60 °C for 2 h and thereafter further dried under vacuum at 60 °C for 24 h.The membrane was removed from the glass plate by immersing it in deionized water and then washed and dried again under vacuum at 60 °C for 24 h to remove solvent traces.For doping, the membrane was immersed in KOH solution at room temperature for 24 h.Membranes are denoted as xMS-PBI, with x indicating the amount of DBX in wt.%, based on the polymer.Because some not well-dissolved PBI can be removed in the filtration step, the real crosslinking contents is higher.For some membranes, the polymer concentration after filtration was analyzed by dropping some solution in a petri dish and evaporating the solvent.These membranes have a lower but more exact crosslinking degree, and are denoted x'MS-PBI.
Conductivity Measurements: Conductivity values of doped membranes were obtained by impedance spectroscopy.The membranes were fixed between two gaskets, leaving an active area of 1.767 cm 2 .The gap between the electrodes (gold-coated copper disc) was <1 mm and filled with the respective KOH solution used for doping.The resistance of the membrane was obtained by plotting at least three resistances of stacked membranes (e.g., 1, 2, 3 samples) against their thicknesses, assuming that the interfacial resistance between two membranes is negligible.A linear trend should be obtained, and conductivity ( m ) was calculated according to the following equation, in which slope is the slope of the linear trend (Ω/cm): Alkaline Stability: Membrane samples were weighed and then doped in 1 m KOH solution at room temperature for 1 day.Next, the temperature was raised to 80 °C and kept there for the remaining time of the experiment.To avoid absorption of CO 2 and dissolution of glass, this experiment was done in sealed polypropylene vials.Because these vials were less stable than thought and led to white precipitates in the solution, they were later changed to PTFE vials.At regular intervals, samples were characterized.For that, membranes were allowed to equilibrate for 2 h before the weight and thickness changes of the membrane samples were noted, and conductivity at room temperature was measured.
Membrane Composition and Swelling Behavior: Membrane samples were immersed for 24 h in KOH solution.They were blotted quickly with a tissue and the wet, doped weight was noted (w wet ).After drying in the vacuum at 80 °C overnight, the w dry was noted.Then KOH was removed by immersion in water until the pH remained neutral, and the membranes were again dried at 80°C in the vacuum, giving w dedoped .The weight frac- The total uptake, KOH uptake, water uptake and the concentration of the absorbed KOH solution were calculated according to Equations (5-8)  (8)   Dimensional changes during doping were analyzed by measuring the length and thickness of the pristine membranes after washing with water to remove residual solvents and drying at 80 °C (initial dimension   ) and the KOH doped membrane (d   ).Swelling values were calculated according to Equation (9).
Gel Content: Membranes samples were weighed (w initial ) and then immersed in DMAc at 80 °C without stirring, to prevent mechanical stresses.Membranes were dried at 80°C in the vacuum, and the dry weight w final was noted.Gel content was then calculated by Gel content (%) = w final w initial × 100 (10)   Electrolysis Test: For electrolysis test, the membrane was doped overnight at room temperature in 1 m KOH.Thickness after doping was measured to be 99 ± 6 μm.Just prior to assembly, the membrane was die cut to a round shape with a diameter of 56 mm.Electrodes from Ni foam were cut to a diameter of 36 mm, and afterward pressed to a thickness of ca 300 ± 20 μm.Membrane and electrodes were assembled with PTFE gaskets in a Ni coated cell house, using a pin-type flowfield.An active area of 10 cm 2 was considered on basis of the electrodes, as the rest of the membrane area was covered with PTFE gaskets.The cell was mounted in the test fixture, and heating was carried out directly in the cell house.The cell test was carried out at 80°C and ambient pressure.Electrolyte was circulated on both sides at 80 ml min −1 , in a partially separated mode, except during the hydrogen crossover investigation, where it was temporarily switched to a fully separated mode.Water was added intermittently every 1-3 days, based on faradaic consumption of water.Following a 20 h break-in at 100 mA cm −2 , polarization curves were recorded weekly, and a constant current density of 200 mA cm −2 was applied in between.Electrochemical impedance spectroscopy (EIS) was recorded potentiostatically at 1.3 V using after measuring polarization curves using a Biologic SP-240.The high frequency resistance was taken a Z re at ca 56 kHz.
Hydrogen Crossover: Hydrogen levels in the oxygen stream (HTO) were measured with an electrochemical hydrogen sensor (Geopal GJ-EX).The exhaust gas was dried over silica gel prior reaching the sensor.A constant nitrogen purging flow of 56 ml min −1 was introduced in the gas volume of the oxygen degassing vessels to facilitate a more rapid response at the hydrogen sensor.The nitrogen fraction is corrected for in the presented data, as described elsewhere. [40]Sensor calibration was carried out just prior to the test using 1%, 0.5%, and 0.1% hydrogen in air reference gasses, as well as pure nitrogen, and a polynomial fit was used to correct measurements for deviations from linearity.The hydrogen crossovercurrent density profile was recorded during the second day of testing, by holding the current density for one hour at each setpoint.

Figure 1 .
Figure 1.Synthesis and membrane fabrication.In xMS-PBI, x denotes the weight percentage of DBX to PBI in the casting solution.

Figure 2 .
Figure 2. a) Total Uptake, b) water and KOH uptake, c) concentration of the absorbed solution in the membrane, and d) thickness and length swelling.All at room temperature.

Figure 3 .
Figure 3. Conductivity of mPBI (from previous work,[14] ), crosslinked MS-PBI and Zirfon; a) at room temperature and in different KOH concentrations; b) temperature effect on conductivity in 5 wt.%KOH.5MS-PBI is filtrated before membrane casting, which may remove some not well dissolved PBI, resulting in a higher degree of crosslinking than expected; 5′MS-PBI has a lower, more exactly adjusted degree of crosslinking, because the PBI contents was analyzed before addition of the crosslinker.5″MS-PBI is similar to 5′MS-PBI, but the membrane was only dried once and solvent traces were not removed by washing with water.Data at 25 °C represents data obtained at uncontrolled ambient room temperature conditions.

Figure 4 .
Figure 4. Alkaline stability of membranes in 1 m KOH.a) weight and thickness changes and b) conductivity.Data for commercial membranes from Fumatech and Versogen are taken from previous work.[34]

Figure 5 .
Figure 5. Reactions of a) fully methylated benzimidazolium ions, b) monoalkylated benzimidazole, and c) benzimidazole with alkaline solutions; d) shows electrophilic aromatic substitution of sulfonic acid groups by hydroxide ions.In (a), C2 is positively charged, rapid reaction with OH − results in hydrolysis; in (b) C2 is not charged, and very slow reaction with OH − may occur; (c) benzimidazole is deprotonated by OH − , resulting in a negative partial charge on C2, and thus a stabilized compound.

Figure 6 .
Figure 6.Response of the mechanical properties of 5′MS-PBI to aging in 1 m KOH at 80 °C; the bars on the left and right indicate the range covered by all data points.Samples were washed with water and dried before testing.

Figure 7 .
Figure 7. Water Electrolysis in 1 m KOH, 80 °C using 300 μm Ni foam electrodes and a ca. 100 μm 5′MS-PBI membrane.a) Polarization curves recorded with weekly sampling, and b) voltage evolution over time at 200 mA cm −2 .

Figure 8 .
Figure 8. Hydrogen crossover data corrected for the N 2 fraction.a) As function of current density at fully separated electrolyte circuits, with error bars representing min and max values within the sampling interval, and b) change in the HTO as function of time, with partially separated electrolyte circuits.Since the N 2 correction calculation is based on current density setpoint, data outside the steady-state setpoints have been omitted.