Durability of Newly Developed Polyphenylene-Based Ionomer Membranes in Polymer Electrolyte Fuel Cells: Accelerated Stress Evaluation

The chemical durability of the hydrocarbon (HC) polymer electrolyte membrane, sulfonated poly(phenylene) quinquephenylene (SPP-QP), is evaluated at 90°C and 160 kPaG pressure of hydrogen and air supplying anode and cathode, respectively, under open circuit voltage (OCV) conditions as an accelerated stress test in a polymer electrolyte fuel cell (PEFC). To evaluate the degree of deterioration of the membranes, exhaust water is collected from both electrodes periodically during the tests, and the chemical species are analyzed by ion chromatography (IC). The SPP-QP membrane-based cell with appropriate gaskets and gas diffusion layers (GDL) shows the highest durability, in comparison with several other cells, and exhibits a high OCV for more than 1000 h and the lowest emission rate of sulfate (30 μg cm−2, 2.6% loss) accumulated over 1000 h. We conclude that the simple hydrophilic structure and hydrophobic structure of the SPP-QP membrane, consisting solely of phenylene groups, leads to remarkably high intrinsic chemical stability and stability during the OCV stress evaluation; however, it also makes the membranes brittle. We also suggest that this negative aspect of the SPP-QP membrane might be mitigated by use of appropriate cell components such as gaskets and GDLs.

Fuel cell vehicles (FCVs) using polymer electrolyte fuel cells (PE-FCs) have been attracting much attention due to their low levels of exhaust emissions and the general transition from gasoline-fueled vehicles to electromotive vehicles. For further proliferation of FCVs, reductions of cost and increases in durability are required. Perfluorosulfonic acid (PFSA) membranes such as Nafion have been widely used as the electrolyte membrane for PEFCs due to their high proton conductivity and mechanical stability. 1,2 However, their fully fluorinated chemical structure leads to high cost, low glass transition temperature and high gas permeability, 3 which make it difficult to meet the target value of FCV duration about 5000 h of driving time. 4 In recent years, to address these problems, reinforced PFSA materials and alternatives to PFSA membranes have been researched.
In the case of PFSA membranes, expanded polytetrafluoroethelene (ePTFE) is used as a reinforcement material to increase the dimensional stability during humidity cycling. [5][6][7][8] Specific examples of composite membranes are Nafion XL 9 and GORE-SELECT membranes, 10 which are manufactured by DuPont and Gore, respectively.
During operation, reactive radical species are generated and attack functional groups in the PFSA membranes. 2,3,11 To mitigate this type of degradation, radical scavengers, e.g., CeO 2 , MnO 2 , and ZnO, are also used in PFSA membranes. [12][13][14] In actual FCVs, selected types of composite PFSA membranes with radical scavengers are being used. However, these are high in cost because of the use of the costly PTFE mechanical support layers, in addition to the costly PFSA ionomer.
Hydrocarbon membranes are promising alternative materials. The chemical structure of aromatic hydrocarbon (HC) membranes leads to higher glass transition temperatures and lower gas permeability than those of PFSA membranes. 15 Despite these advantages, they have several disadvantages, including lower proton conductivity and lower chemical stability. Recently, there have been improvements in the HC membrane proton conductivities, which are comparable or higher than those of PFSA membranes. [16][17][18][19][20] The improvement of chemical stabil-ity while maintaining high proton conductivity is one of the most important challenges for HC membrane development at present.
In the case of HC membranes, electron-rich ether and thioether linkages around sulfonic acid groups are known to be susceptible to attack by oxidative radicals. 21 Crosslinking is also a promising, effective approach for improving stability. [22][23][24] In our group, reinforced materials that include a fluorine-free mechanical support layer have been investigated for HC membranes. 25 We have also carried out research on new HC membrane structures over the past decade, including sulfonated polyimide (SPI), 26,27 sulfonated poly(arylene ether), 28,29 and sulfonated poly(arylene ether sulfone ketone) (SPESK) 30,31 membranes.
In our previous study, 32 we investigated a sulfonated benzophenone poly(arylene ether ketone) semiblock copolymer (SPK-bl-1, SPK) 33 and a phenylene poly(arylene ether ketone) semiblock copolymer (SPP-bl-1, SPP) 34 membrane under accelerated open circuit-voltage (OCV) conditions appropriate for FCVs. The results indicate that a simple hydrophilic structure that does not include ketone groups leads to increased chemical stability versus radical attack decomposition. On the other hand, an SPP membrane quickly decomposed when subjected to a Fenton's reagent test, which may be attributed to the presence of heteroatoms in the hydrophobic structure. Based on these results, we developed a new hydrocarbon membrane that consists solely of phenylene groups, which is designated a sulfonated poly(phenylene) quinquephenylene (SPP-QP, Figure 1). 35   a Initial sulfonic acid concentration as a reference for the sulfate concentrations detected in the exhaust water after durability testing.
In the present research, we focus on the durability of the SPP-QP membrane, in comparison with previous results for HC membranes (SPP and SPK) under identical accelerated OCV conditions. 32 During the durability evaluations, the exhaust water from both sides of the cells was collected and then analyzed by ion chromatography (IC). After the durability evaluations, the post-test membrane electrode assemblies (MEAs) were analyzed with nuclear magnetic resonance (NMR) spectroscopy. These results were compared with those for the initial MEAs. From the results, the durability of the membranes is discussed in detail. We also studied approaches to mitigate the mechanical stress by use of GDLs and gaskets in order to evaluate the chemical stability of the membranes under accelerated conditions in MEAs.

Experimental
MEA.-The SPP-QP membrane was synthesized according to the method as previously described. 35 The membranes were prepared by solution casting onto a flat glass and were dried in air. The thickness of the membranes was 25 μm. The properties of the membranes are listed in Table I. Catalyst-coated membranes (CCMs) were prepared in the same manner as previously described. 32 A commercial carbon black catalyst (Pt/CB, TEC10E50E, Tanaka Kikinzoku Kogyo, K. K.) was used for the anode, and a commercial platinum-loaded graphitized CB catalyst (Pt/GCB, TEC10EA50E-HT, Tanaka Kikinzoku Kogyo, K. K.) was used for the cathode. The mass ratio of Nafion binder to the carbon support (N/C) was adjusted to 0.70. These pastes were sprayed on both sides of the membranes by use of a Pulse-Swirl-Spray (PSS, Nordson) apparatus. The geometric electrode area was 29.2 cm 2 , and the Pt loading amount of the catalyst layer was 0.50 ± 0.05 mg Pt cm −2 for both anode and cathode. The CCMs were hotpressed at 140°C and 10 kgf cm −2 for 3 min, and then they were assembled with two gas diffusion layers (GDLs, 25BCH, SGL Carbon Group, and soft GDLs, which were experimental GDLs prepared by Panasonic) 36,37 and mounted into a spring-loaded Japan Automobile Research Institute (JARI) cell, which has serpentine flow channels in both the anode and the cathode carbon separators, with subgaskets (PPS, 25 μm) and gaskets (150 μm, PEN substrate and silicone rubber adhesive film on both sides: Si/PEN/Si, KUREHA ELASTOMER Co., Ltd., and PTFE, NICHIAS Co.) for both sides. The soft GDL was prepared as previously described. 36 The soft GDL was prepared from acetylene black, graphite, PTFE and surfactant solution. The shear stiffness and bending stiffness of the soft GDL were more than 10 times lower than those of the SGL GDL, due to the absence of PTFE-bonded hard carbon fibers. The sheet thickness and porosity were adjusted to ca. 400 μm and ca. 50%, respectively. In the case of SGL 25BCH, the sheet thickness was 230 μm, which included the microporous layer (MPL, ca. 100 μm, consisting of carbon and PTFE); the latter was impregnated into approximately half the thickness of the carbon paper, and the total porosity was ca. 76%. We have designated these combinations of MEA components as cell types 1, 2 and 3 cells, as listed in Table II. The membrane designations, e.g., "SPP-QP type #" (see Figure 2), simply refer to the same membrane tested in different cell configurations.
Membrane and gasket stress-strain curves.-Tensile tests of the membranes and gaskets were conducted with a Shimadzu AGS-J 500N universal test machine equipped with a temperature and humiditycontrollable chamber. The tests followed the DIN-53504-S3 standard (dumbbell shape of 35 × 6 mm (total) and 12 × 2 mm (test area)). After equilibrating the membrane samples for at least 3 h under 80°C and 60% RH conditions, the stress-strain curves were obtained at a stretching rate of 10 mm min −1 . For the gasket samples, the curves were obtained at room temperature and humidity conditions (25°C, ca. 30% RH).
Fuel cell operation.-In this study, we selected the same accelerated OCV stress-evaluation protocol that was provided by Honda R&D Co., Ltd., as described in a previous report. 32 Hydrogen (76% RH, anode) and air (86% RH, cathode) were supplied at 90°C and 160 kPaG pressure. The flow rates were 200 mL min −1 for both electrodes. During the tests, exhaust water from both electrodes was collected.

Measurement of cell performance parameters during
operation.-At the initial point and every 200 h during durability testing, cyclic voltammetry (CV) was carried out with a potentiostat (PGST30 Autolab System, Eco-Chemie) to evaluate the  electrochemical surface area (ECA) of Pt and peak shifts in the high potential region induced by adsorbates from the decomposition of the membranes on Pt in the cathode catalyst layer at 90°C and ambient pressure. Prior to the measurements, hydrogen (100 mL min −1 ) and nitrogen (150 mL min −1 ) humidified at 76% RH and 86% RH were supplied to the anode and cathode, respectively. The cathode potential was scanned between 0.06 and 1.0 V at 20 mV s −1 , with a H 2 flow rate of 50 mL min −1 at the anode and the cathode N 2 flow halted just prior to the measurement. The ECA values were determined from the hydrogen adsorption charge in the negative-going potential scan, referenced to Q H°= 210 μC cm −2 , which is a standard value for polycrystalline platinum. The crossover hydrogen from the anode to the cathode was measured by use of linear sweep voltammetry (LSV) at 90°C and ambient pressure in a cell purged with hydrogen (100 mL min -1 ) at the anode and nitrogen (150 mL min −1 ) humidified at 76% RH and 86% RH at the cathode. The cell potential was swept from 0.15 to 0.6 V at a sweep rate of 0.5 mV s −1 . The current-potential (I-E) polarization curves of the MEAs were measured by supplying hydrogen and oxygen/air at 90°C and ambient pressure. The flow rates of all gases were controlled by mass flow controllers. The utilizations of the reactant gases were 70% for hydrogen and 40% for oxygen/air. Hydrogen gas and oxygen/air were pre-humidified at 76% RH and 86% RH, respectively. The cell potentials were measured galvanostatically as a function of current density by use of an electronic load (PLZ-664WA, Kikusui Denshi) controlled by a measurement system (FCE-1, Panasonic Production Technology). The resistances were measured at 1 kHz under load by use of a digital ac milliohmmeter (Model 3566, Tsuruga Denki) and were used as the cell ohmic resistances for IR correction purposes. The I-E curves were galvanostatically measured under steady-state operation, with a measurement time of 5 min for each point.
Post-test analysis of the MEAs.-After the durability evaluations, the MEAs were carefully disassembled from the cells for post-test analyses. The MEAs were analyzed by means of a helium gas leak test in order to check for cracks or pinholes in the CCM. In this test, the MEA was attached to the helium leak inspection apparatus. Helium gas was supplied to one side of the MEA, and the amount of helium leakage through the membrane was detected on the other side of the MEA by a helium leak detector (HELEN M-212LD, CANON ANELVA Co., Ltd.). In the plane of the MEA, the electrode area of 29.2 cm 2 was divided into 25 (5 × 5) regions of 1 cm 2 each, and the helium leak rate was detected in each region. 1 H NMR spectra were obtained on a JEOL RESONANCE JNM-ECA400 instrument using deuterated dimethyl sulfoxide (DMSO-d 6 ) as a solvent and tetramethylsilane (TMS) as an internal reference.
Exhaust water was collected from both electrodes periodically during the tests, and the chemical species were analyzed by ion chromatography (IC). The concentrations of sulfate per geometric area (29.2 cm 2 ) were analyzed with a Thermo Fisher Scientific ICS-2100 system with an Ion Pac AS18 column and 4 mM ∼80 mM potassium hydroxide (KOH) aqueous solution as eluent.

Durability of the cell under accelerated OCV stress evaluation.-
The cell voltages were continuously measured during the accelerated OCV stress evaluation, and the hydrogen leak current density and the composition of the water that was drained from each cell were measured every 200 h by use of LSV and IC, respectively, as shown in Figures 2a-2c. The cell voltage for the SPP-QP type 1 cell decreased significantly after 700 h. In the case of SPP-QP type 2 and type 3 cells, the cell voltages maintained high values for over 1000 h, as well as those for the SPP cell and SPK cell (Figure 2a). The hydrogen leak current density also behaved nearly the same for each of the cells, except for the SPP-QP type 1 cell, which reached ca. 100 times higher H 2 leakage at 800 h than those of the others (Figure 2b). The amount of emitted sulfate ions for the SPP-QP type 1 cell reached 59 μg cm −2 (geometric area 29.2 cm 2 , 5.1% loss) at 800 h, which was lower than those of the SPP and SPK cells. Additionally, those for the SPP-QP type 2 and type 3 cells were 28 and 32 μg cm −2 , (2.4% and 2.8% loss) respectively ( Figure 2c). The emitted concentration values of sulfate decreased in the order SPK » SPP > SPP-QP type 1 > SPP-QP type 2 ≈ SPP-QP type 3. This was thought to be attributable to the intrinsic chemical stabilities of the membranes themselves. To check these, the intrinsic chemical stabilities of the SPP-QP, SPP and SPK membranes were evaluated with Fenton's reagent. For this test, membranes samples with dimensions of 5 cm × 5 cm × 25 μm were used, with a mass of ca. 100 mg; these were immersed in an aqueous solution containing 3% H 2 O 2 and 2 parts per million (ppm) Fe 2+ at 80°C for 1 hour. The residual M n and M w values were 99% and 99% for SPP-QP, 43% and 35% for SPP, and only 11% and 10% for SPK, respectively, and ion exchange capacity (IEC) of SPP-QP hardly changed after test (Table III). These results indicated that the SPP-QP membranes had much higher intrinsic chemical stability compared to those of SPP and SPK membranes, which was reflected in the results of the accelerated OCV stress evaluation.
CVs were measured initially and every 200 h for all cells (Figures 3a-3e). The Pt surface oxidation/reduction peaks changed remarkably with time in the case of the SPK cell: the onset potentials of the oxidation peaks moved from 0.6 V to 0.8 V, and large oxidation currents were observed at E > 0.8 V (Figure 3a). In the case of the SPP cell, after 800 h, the oxidation peak at 0.8 V became smaller, and the oxidation current over 0.9 V increased slightly (Figure 3b). In contrast, the CV peaks for the SPP-QP cells hardly changed over 1000 h for the full potential range (Figures 3d, 3e). Even the SPP-QP type 1 cell maintained a similar shape for the CV at 600 h (Figure 3c). The CVs after 800 h for the SPP-QP type 1 cell were not obtainable due to large H 2 cross-over. Additionally, after each performance evaluation, we found that the CVs for the SPK and SPP membranes recovered to nearly the same shapes as those of the initial CVs. We consider the reason to be as follows. Our previous research reported that the changes that occurred in the CVs could mainly be attributed to the adsorption of membrane degradation products on the catalyst; these products accumulated during the OCV durability test and were then washed out from the MEA with the water that was produced during the interim performance evaluation. 38 Therefore, the results from the CV measurements also indicate that the decomposition of the SPP-QP membrane was much milder than that of the SPP membrane and by far milder that of the SPK membrane, because the amounts of adsorbates in the SPP-QP cell were smaller than those in the SPP and SPK cells. Hence, we conclude that the durability of the membranes increased in the order SPK < SPP < SPP-QP, according to the single cell evaluation. After the accelerated OCV stress evaluation, the cells were disassembled and checked by visual observation. We were able to observe a pinhole at the edge of the CL for the SPP-QP membrane using the type 1 cell (red circle in bottom right corner), as shown in Figure 4a. The helium leakage for the SPP-QP type 1 cell also showed a high leakage rate in the edge region of the CCM after the test (Figure 4b). Ishikawa et al. have reported that the stress was much higher at edge sites of CCMs due to the configuration of MEAs and cells, in which there is a large degree of membrane shrinkage and swelling, in comparison with other sites. 37 They also reported that the use of soft GDLs was able to decrease the mechanical stress from fastening the cells, which improved the mechanical strength of the membranes in the cells 37 ( Figure 5).
From the viewpoint of the membranes themselves, the stress-strain curves differed significantly between the SPP-QP and SPP, SPK membranes (Figure 6a). SPP-QP exhibited higher stress strength, but, on the other hand, for strain aspect, it was much lower in comparison with the other membranes, which means that the SPP-QP membrane is more brittle. We consider that this difference is quite significant for the mechanical stability in the cells.
Since the soft GDL was chosen as an MEA component to reduce the stress at edge sites, we also considered modifying the gaskets. Figure 6b shows stress-strain curves for a PTFE gasket and a Si/PEN/Si gasket, which exhibited very different Young's moduli (0.3 GPa, 2.0 GPa, respectively) and elongation (strain at break point Figure 5. Schematic cross-sectional images and diagrams at the edge configuration of MEAs and their membrane degradation mechanisms for type 1 (a) basic configuration, which has no edge modifications, and type 2 (b) configuration, which was modified by the use of a sub-gasket and a soft GDL. = 360%, 33%, respectively). These values indicate that the PTFE gasket is much softer and easier to stretch than the Si/PEN/Si gasket. Thus, we selected the PTFE gasket in place of the Si/PEN/Si gasket with the SGL GDL and expected that it would bring about nearly same effect as that of the soft GDL.
In order to examine for these effects, we constructed a new type of cell, designated a type 2 cell, using the soft GDL with the Si/PEN/Si gasket, and a type 3 cell, using the PTFE gasket with the SGL GDL, as listed in Table II.
During accelerated OCV stress evaluation, the voltages (Figure 2a), the amounts of emitted sulfate ion from the IC (Figure 2c), and the shapes of the CVs (Figures 3d, 3e) all indicated that both types 2 and 3 cells using SPP-QP possessed higher stability than those of others. The voltages were over 0.9 V during the evaluation period, the sulfate emissions were about one-third as high as that of the SPP cell and one tenth as high as that of the SPK cell, and, significantly, the shapes of the CVs hardly changed. These results indicate high chemical stability for the SPP-QP membrane in the cells during accelerated OCV stress evaluation. The effect of the soft GDL in the cells decreased the mechanical stress during shrinkage and swelling of the membrane. The PTFE gasket also brought about decreased mechanical stress in the cells. After the durability evaluation, the PTFE gasket was able to peel away easily from the membrane, while the Si/PEN/Si gasket had adhered strongly to the membrane. In the case of the PTFE gasket, the mechanical stress might have decreased by slippage of the membrane. We also consider that there is a need to investigate the effect of the PTFE gasket on the mitigation of mechanical stress in more detail in future work.  Degradation site analysis of the molecular structure.-These SPP-QP membranes used for all cell types were analyzed by NMR in order to clarify the molecular structural changes before and after the accelerated OCV stress evaluation. Figure 7 shows the initial NMR spectra of the SPP-QP membrane and those after the stress evaluation. The NMR spectral peaks for the SPP-QP membranes for all cells were hardly changed before and after the stress evaluation. The peak-area ratios of the hydrophilic moiety and hydrophobic moieties, and the IEC values calculated from them, were also hardly changed (Table IV). The initial ratio of the integrated areas for hydrophilic to hydrophobic structure was 34:66, and the corresponding IEC value for the SPP-QP membrane was 2.9 meq g −1 . These values were maintained essentially without change after the accelerated OCV stress evaluation, with values of 33:67 and 2.9 meq g -1 , respectively. In contrast, our previous research reported changes of the peaks for the hydrophilic moieties and IEC values for SPP and SPK membranes that were decreased significantly after the stress evaluation. 32 From these NMR analyses, the chemical structure of the SPP-QP membrane was found to have a higher stability versus the accelerated OCV stress evaluation in comparison with those for the SPP and SPK membranes. The water-soluble oxidative degradation by-products were washed away during cell operation. In the IC results, sulfate ion was detected in the SPP-QP membrane cell, which means that the degradation of the SPP-QP membrane occurred to a measurable extent, but it was lower than that of the other HC membranes. The difference might be attributed to the simple hydrophilic structure and hydrophobic structure, made up by phenylene groups in the SPP-QP membrane whose chemical structure hardly underwent radical attack, resulting in a membrane that saw less chemical decomposition.

Conclusions
Our newly developed polyphenylene-based HC membrane, SPP-QP, was evaluated under accelerated stress conditions at OCV. The results for the cell using the SPP-QP membrane with soft GDL, type 2, and PTFE gaskets, type 3, showed the highest durability in comparison with other cells and exhibited high OCV values for more than 1000 h and the lowest emission rate of sulfate emission (30 μg cm −2 , 2.6% loss) accumulated over 1000 h. On the other hand, with the normal MEA configuration, type 1, the SPP-QP membrane was ruptured at an edge site after 800 h in accelerated OCV evaluation due to the brittleness of the SPP-QP membrane. The simple hydrophilic and hydrophobic structure, made up solely of phenylene groups in the SPP-QP membrane, led to remarkably high chemical stability; however, it also made the membrane brittle. We conclude that the SPP-QP membrane has high intrinsic durability, based on the OCV stress evaluation, in comparison with the other HC membranes. The next stage of new development for HC membranes will require more flexible structures while maintaining the high chemical stability of the SPP-QP membrane. We also suggest that this negative aspect of the SPP-QP membrane is able to be mitigated by the use of modified cell components such as appropriate gaskets and GDLs.