The Effect of Cathode Structures on Naﬁon Membrane Durability

The effect of cathode structures on the chemical stability of Naﬁon membranes is investigated. Membrane electrode assemblies (MEAs) were prepared by using Naﬁon 212 membrane and commercially available carbon supported Pt electro-catalysts. The cathode structures were controlled by the use of long side chain (LSC) and short side chain (SSC) perﬂuorosulfonic acids (PFSAs) as well as three dispersing solvents for the electrode fabrication (a water-isopropanol mixture, N-methyl-2-pyrrolidone, or glycerol). The membrane durability was evaluated by the H 2 crossover current density after a 200-hour open circuit voltage accelerated stress test. The MEA with a glycerol-processed SSC PFSA-bonded cathode exhibited a 200-fold less H 2 crossover current density than the MEA with a water-isopropanol-processed LSC PFSA-bonded cathodes. The analyzes by electrochemical impedance spectroscopy and microscopy suggest that the structural uniformity of cathodes play the most signiﬁcant role in the chemical stability of the Naﬁon membranes. This study emphasizes the importance of cathode structures on the durability of Naﬁon membranes. © The Author(s) 2014. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0151412jes] All rights reserved. not well correlated with either the OCV values at the end of the OCV test or the H 2 crossover current density These indicate that H 2 crossover change and the OCV values in the polarization curves are directly related to PEM degradation, OCV values at the end of the OCV test and the current density at low cell are inﬂuenced by other factors. H 2 crossover current density and OCV value change in the polarization curves in- dicate that with

Polymer electrolyte membranes (PEMs) are a critical cell component for polymer electrolyte membrane fuel cells (PEMFCs). The durability of PEMs are important because the life of PEMFCs is often determined by PEM failure. 1 The chemical stability of PEMs have been discussed since the 1960s, notably by engineers and scientists at GE. 2 LaConti proposed a chemical degradation mechanism for perfluorosulfonic acids (PFSAs): 3 oxygen molecules permeate through the membrane from the cathode and are reduced at the anode Pt catalyst to form hydrogen peroxide. Although PFSA membranes are stable in the presence of hydrogen peroxide, metal ions such as Fe 2+ and Cu 2+ in the catalyst layers greatly accelerate the membrane degradation rate by forming reactive oxygen species such as hydroxyl (HO·) and hydroperoxyl (HO 2 ·) radicals. [4][5][6] H 2 O 2 + M 2+ → M 3+ + ·OH + OH − [1] · OH + H 2 O 2 → H 2 O + ·OOH [2] Since the reactive oxygen species are generated in the catalyst layer, the membrane degradation is greater with a supply of H 2 and O 2 in the catalyst layers. Liu et al. reported about the substantially short lifetime for Nafion membranes when H 2 and O 2 gases were supplied to the catalyst-coated membrane, while no degradation is detected when H 2 /O 2 is decoupled to H 2 /N 2 or N 2 /O 2 . 7 Later, Burlatsky et al. suggested that an extended catalyst layer between the cathode and the membrane could consume reactive oxygen species to produce water, thereby mitigating the membrane degradation. 8 The effect of gas permeability on membrane degradation has been also demonstrated. Sethuraman et al. compared the durability of wholly aromatic membranes, BPSH-35 and Nafion 112, under accelerated stress conditions. 9 Although the BPSH-35 membranes showed poor chemical stability in ex-situ Fenton tests, the membrane electrode assemblies (MEAs) with BPSH-35 outlasted the MEAs with Nafion 112 in the open circuit voltage (OCV) and potential cycling tests under H 2 /O 2 conditions due to the low gas-crossover rates of the BPSH-35 membrane.
Because the chemical degradation of membranes is related to the formation of reactive oxygen species in the catalyst layers and their access through the membrane, electrodes, and gas diffusion layer, interactions between different MEA components become crucial for the assessment of the chemical stability of PEMs. The effects of electrocatalysts on PEM stability have been investigated. [10][11][12] The chemical degradation of Nafion membranes were reduced by 50% when Pt/C catalyst was replaced with Pt-Co/C. 10,11 Sulek et al. have also reported that Nafion lifetime could be extended by 65% when Pt/C catalyst was replaced with Pt-Ni/C. 12 The effect of the gas diffusion electrode (GDE) on membrane stability has also been investigated. Kreitmeier et al. reported that a membrane with a GDE exhibited better stability than the bare membrane when exposed to reactive oxygen species. 13 The aforementioned examples suggest that membrane degradation in fuel cells are affected not only by the chemical stability of the membrane itself but also by the interactions with other MEA components. However, the electrode structural effect on PEM durability has been largely unexplored so far.
In this paper, the impact of the cathode structure on membrane stability is investigated. The structural change of cathodes were obtained by using long side chain (LSC) PFSA Nafion (EW = 1,000 g/mol, DuPont) and short side chain (SSC) PFSA Aquivion (EW = 830 g/mol, Solvay Solexis) and three different dispersing solvents (water-isopropanol, NMP, or glycerol) for the catalyst inks. Other MEA components and test conditions were held constant. The structural differences between LSC and SSC PFSA have been studied by Kreuer et al. 14 They reported that SSC PFSA does not have any distinct differences in water and proton transport, yet provides better thermomechanical stability, which is anticipated due to the stability of the electrolyte/electrode interface. The effect of different catalyst dispersing solvents for electrode preparation has been also studied by Kim et al. 15 The electrodes prepared from water-isopropanol solvent showed numerous large scale open cracks (>100 μm) while the electrodes prepared from glycerol and N-methyl pyrrolidone (NMP) exhibited distributed microcracks (<10 μm) and a crack-free structure, respectively. Based on this information, a series of MEAs having different cathode structures were fabricated and the stability of Nafion 212 membrane was examined by measuring the H 2 crossover rate during a 200-hour OCV accelerated stress test (AST). Other relevant properties such as open circuit voltage (OCV) and polarization behaviors were examined as well. The membrane degradation process and the role of electrode structure on membrane degradation during OCV tests are discussed based on structural analyzes by scanning electron microscopy (SEM) and electrochemical impedance spectroscopy (EIS).

Experimental
Preparation of ionomer dispersion.-Ionomer dispersions were prepared by the direct dispersion technique that we developed. 16,17 First, PFSA membranes were converted to the Na + form by placing Nafion 212 in boiling 0.5 M NaOH solution for 90 min. Subsequently, the Na + form membranes were placed in de-ionized water at 80 • C for 90 min in order to remove the residual NaOH. The membranes were then dried at 60 • C for 30 min followed by dissolving in the desired solvent at an elevated temperature. Three solvent systems were used; water-isopropanol (1:1 volume ratio), NMP, and glycerol. For the water-isopropanol solution, a closed pressure vessel was used to maintain solvent composition at 210 • C for 3 hours. For NMP and glycerol, commercial vials with Teflon cap were used. The vials were heated in a convection oven at 110 • C (NMP) and 210 • C (glycerol) for 3 hours, respectively. Clear liquid solutions were obtained without any residual solid polymer. The solid content of the dispersions was 2.5 wt%.
MEA fabrication.-Carbon supported Pt catalyst (20 wt% Pt/C, BASF) was used for both anode and cathode catalyst layers. For the anode catalyst layer, the Pt/C catalyst powder (0.0625 g) was mixed with commercial 5 wt% Nafion dispersion (1 g). Glycerol (0.5 g) and 1 M tetrabutylammonium hydroxide (TBAOH) in methanol (20 μL) were added into the vial and stirred for more than 12 hours to make the standard ink slurry. TBAOH was added to convert the Nafion from the proton to the TBA + form, which may enhance the interfacial adhesion between membrane and electrode during hot pressing. 18,19 For the cathode catalyst layer, the catalyst to ionomer weight ratio was adjusted for the optimum performance. The catalyst to ionomer ratio for all electrodes, except glycerol-processed electrodes, was 5:2. The catalyst to ionomer ratio for glycerol processed electrodes was 5:1.7. To make the cathode catalyst ink slurry, the Pt/C electro-catalyst powder was mixed with the ionomer dispersion prepared by the direct dispersion technique. The ionomer dispersion consists of ionomer and the dispersing solvent without any other additives. The catalyst slurry was painted onto a 5 cm 2 decal substrate and dried at 140 • C for 1 hour. The painting/drying of catalyst layers continued (usually 4 paintings) until the Pt loading reached 0.2 mg/cm 2 . The decal-painted electrodes were then further dried overnight at 140 • C. The decal painted electrodes were then transferred onto the Na + form of Nafion 212 (EW = 1,000 g/mol, Ion Power Inc.) by hot-pressing at 700 psi pressure for 3 min. After hot-pressing, the MEAs were first placed in boiling 0.5 M H 2 SO 4 solution for one hour then deionized water for one hour in order to convert to the acid form of the MEAs. Table I shows the cathode compositions, dispersion solvents and MEA sample codes.
Cell characterizations.-The MEAs were assembled with carbon cloth gas diffusion layers (LT-2500W, E-TEK). Standard single cell hardware (5 cm 2 active area, Fuel Cell Technology Inc.) was used for OCV testing. The single cell was charged to a fuel cell station at 80 • C with hydrogen (200 sccm) and air (500 sccm) flows for the anode and cathode, respectively. The backpressure for both electrodes were The high frequency resistance (HFR) of the cell was measured while obtaining the polarization curve. In order to choose the frequency that minimizes the capacitance, a sinusoidal wave perturbation between 2 and 10 kHz was applied to the fuel cell load before obtaining polarization curves. The chosen single frequency, i.e. 3,333 Hz, was used for the HFR measurement. The same conditions were used for generating EIS, obtained under 10 mA of amplitude from 10,000 to 1 Hz of frequency at the cell current of 1, 3, 5, and 7 A. The H 2 crossover current was collected from cyclic voltammograms (CVs) which were measured from 0.06 to 0.5 V (vs Dynamic Hydrogen Electrode, DHE) with a 5 mV/s scan rate. The CVs were run with H 2 (50 sccm) and N 2 (50 sccm) flowing into the anode and cathode at a fuel cell temperature of 70 • C under fully hydrated conditions. The N 2 crossover was measured with N 2 flowing into the anode. The N 2 flow at the cathode outlet was measured using a bubble flowmeter at room temperature. The morphology of the electrodes both before and after life tests was examined by scanning electron microscopy (SEM, FEI Inspect) operated at 20 kV of accelerating voltage. Samples were prepared by cutting a section of the catalyst from the MEA, then mounting it on SEM stubs with conductive tape. examined before OCV testing ( Figure 1a). LSC-W/A exhibited a comparable performance with the LANL Standard control, suggesting that the cathode prepared from the in-house Nafion dispersion had similar characteristics to the cathode prepared from the commercial Nafion dispersion. It was noted that there was a slightly lower performance for SSC-W/A compared to LSC-W/A in the current range from 0.5 to 1.7 A/cm 2 . The Nyquist plots measured at 0.6 A/cm 2 indicate that SSC-W/A has a higher ohmic resistance (0.31 cm 2 for SSC-W/A vs. 0.28 cm 2 for LSC-W/A) yet lower charge transfer resistance compared to LSC-W/A (Figure 1b Top). The higher ohmic resistance of SSC-W/A is probably due to limited water transport in the SSC-W/A cathode that leads to only partial hydration of the membrane. The lower charge transfer resistance of SSC-W/A is possibly caused by the greater proton conductivity of the SSC binder as we observed that the proton conductivity of the SSC ionomer at 80 • C was ∼50% greater than that of the LSC ionomer. The Nyquist plots measured at 1.4 A/cm 2 show that SSC-W/A has a higher mass transport resistance at low frequency (Figure 1b Bottom). This suggests that O 2 transport in the SSC-bonded cathode is lower than that in the LSC-bonded cathode. Springer et al. reported that the low frequency impedance originates from the oxygen access limitation caused by the gas diffusion backing of the fuel cells. 21 However, in our case; the lower gas diffusion comes from slower O 2 diffusion through the cathode rather than the gas diffusion layer (GDL) since the same GDLs for both MEAs were used. Figure 2 shows the OCV test results for LSC-W/A and SSC-W/A. The OCV values gradually decreased over time, suggesting that the PEMs for both MEAs had degraded. SSC-W/A shows slightly less OCV decay than LSC-W/A, ca. the OCV decay rate: −1.8 mV/h for SSC-W/A vs. −2.6 mV/h for LSC-W/A. The reduction of the H 2 crossover current density of SSC-W/A is also less than that of LSC-W/A after 200 hours of OCV testing (Figure 2b). The changes of the H 2 crossover current density, however, show non-linear behavior: more H 2 crossover change during the last 100 hours compared to the first 100 hours. Furthermore, the increase of H 2 crossover seems to be related to i-V curve measurements taken during OCV testing. During the interval of the OCV test, we normally ran ten i-V curves in the potential range of 1.0 to 0.2 V with a step voltage of 0.05 V per 30 seconds. Without i-V curve measurements, the H 2 crossover rate increase is notably less. This suggests that cathode structural change during i-V curve measurements may accelerate the H 2 crossover. Since those MEAs showed a stable performance with multiple i-V curve measurements without OCV tests, the H 2 crossover change associated with i-V curve measurements seems to only assist the PEM failure which had been initiated by the OCV test. With ten i-V curve measurements after the first 100 hours of OCV testing, the H 2 crossover current densities of LSC-W/A and SSC-W/A after 200 hours of OCV testing reached 160 and 109 mA/cm 2 geo , respectively. 1.67 A/cm 2 (100 hours) to 0.50 A/cm 2 (200 hours). The cell current density of SSC-W/A decreases from 1.84 A/cm 2 (initial) to 1.82 A/cm 2 (100 hours) to 1.33 A/cm 2 (200 hours). These current density reductions, however, are not well correlated with either the OCV values at the end of the OCV test or the H 2 crossover current density changes. These indicate that H 2 crossover change and the OCV values in the polarization curves are directly related to PEM degradation, yet both OCV values at the end of the OCV test and the current density at low cell voltage are influenced by other factors. Both H 2 crossover current density and OCV value change in the polarization curves indicate that Nafion 212 is more stable with the SSC-W/A cathode than with the LSC-W/A cathode.

Results and Discussion
Effect of cathode dispersing solvent.-In this section, the effect of the cathode dispersing solvent on membrane durability is investigated. The initial performance of MEAs with water-isopropanol-, NMP-and glycerol-processed cathodes were compared (Figure 3). The MEAs with NMP-processed cathodes display similar initial performances to those prepared from water-isopropanol. However, the MEAs with glycerol-processed cathodes show substantially inferior performance, particularly at low cell voltage. Based on our previous microscopic analysis, the increased performance obtained by the MEAs with a water-isopropanol-or NMP-processed cathode is attributed to the easy access of gaseous oxygen to the catalyst sites through the relatively coarse electrode structure. 15 For glycerol-processed cathodes, a slightly lower ionomer composition in the catalyst layer improves the initial performance, although the performance at low cell potential does not reach the levels of other MEAs. Since the glycerol-processed cathodes showed better performance with a Pt to ionomer ratio of 5:1.7, those MEAs were tested for the PEM degradation study. Figure 4 shows the OCV test results of the MEAs prepared from three dispersing solvents. The MEAs with NMP-and glycerol-processed cathodes exhibit less OCV decay than the MEAs with water-isopropanol-processed cathodes (Figures 4a and 4b). The H 2 crossover current density of the MEAs with NMP-processed cathodes show a non-linear behavior similar to the MEAs with waterisopropanol-processed cathodes. The H 2 crossover current density for all MEAs maintained a relatively low level ca. < 5 mA/cm 2 for the first 40 hours of the OCV tests then substantially increased as the experimental time increased. At the end of the OCV tests, the H 2 crossover current densities showed a clear trend, depending on the ionomer type and cathode processing solvent: LSC-W/A (160 mA/cm 2 geo ) > SSC-W/A (109 mA/cm 2 geo ) > LSC-NMP (67 mA/cm 2 geo ) > SSC-NMP (36 mA/cm 2 geo ) > LSC-glycerol (2.3 mA/cm 2 geo ) > SSC-glycerol (0.8 mA/cm 2 geo ). The H 2 crossover values of LSC-W/A and SSCglycerol differ by a factor of 200 after 200 hours of OCV testing. Considering that the same Nafion 212 membranes were used for all of the MEAs, the data indicates that the effect of the cathode structure on PEM degradation is substantial. Figure 5 shows the H 2 /air polarization curves, HFR, and OCV values of the NMP-and glycerol-processed MEAs. The cell HFR did not significantly change after 200 hours of OCV testing. This suggests that the interfacial adhesion between Nafion 212 and the cathodes prepared from different solvents is good without adding TBAOH solution in the catalyst inks. 22,23 The OCV of the MEAs with NMP-processed cathodes decreases during OCV testing. In contrast, the OCV values of the MEAs with glycerol-processed cathodes are stable even after 200 hours of OCV testing. The OCV values from the polarization curves after 200 hours of OCV testing show a consistent trend with H 2 crossover behavior: These results indicate that the MEAs with glycerol-processed cathodes have superior PEM durability compared to the MEAs with NMP-processed cathodes. The MEAs with water-isopropanol-processed cathodes showed the least PEM durability. The current density at low cell voltage obtained from i-V curves again does not show the same behavior as OCV or H 2 crossover current density. For example, the current density of LSC-NMP, LSC-glycerol and SSC-glycerol at low cell voltage increased after 100 hours of OCV testing then decreased after an additional 100 hours of OCV testing, while the OCV and H 2 crossover current exhibited monotonic changes. This confirms that there are other major contributing factors playing a major role in fuel cell performance durability. In a previous study, it was identified that this type of cathode performance change, i.e. improved performance before dropping off, originated from a cathode structural change. 15 Note that the improving fuel cell performance has little to do with the ohmic contribution due to possible membrane thinning as shown in the iR-corrected cell potentials are shown in Figure 5. In general, glycerol-processed cathodes show more positive structural change than NMP-processed cathodes. Water-isopropanol-processed cathodes did not show a positive cathode structural change. Adding these structural effects, MEAs with glycerol-processed cathodes showed no deteriorated cell performance after 200 hours of OCV testing.

PEM degradation mechanism.-PEM Thinning vs Pinhole
Formation.-It has been reported that membrane thinning due to the PEM degradation directly causes increased H 2 crossover [24][25][26] ; this indicates that the degradation process occurs uniformly throughout the active area of the MEAs. In order to see whether the PEM degradation process occurs uniformly, we examined a cross-section of the MEAs before and after OCV testing by SEM and measured the membrane thickness change. There are no noticeable thickness variations between center and edge parts of the MEAs. Figure 6 shows that the thickness of membranes before OCV testing is 44 μm with a thickness variation of 1 μm. After 200 hours of OCV testing, the membrane thickness for all MEAs ranges from 35 to 41 μm with similar thickness variations. Since the thickness of all MEAs decreased after the AST, it may be considered that the membrane thinning causes the OCV change and H 2 crossover. However, this is not consistent with the degradation results; note that the MEAs with water-isopropanol-processed  cathodes exhibit less thickness reduction in spite of the greater PEM degradation. Furthermore, MEAs with glycerol-processed cathodes show more thickness reduction. The obvious contradiction between the thickness change and the degradation results suggests that thickness reduction may be irrelevant to PEM degradation during OCV testing and the reduced PEM thickness may be caused by cell compression or Nafion creep during OCV testing. The lack of the correlation between membrane thinning and degradation is also supported by the cell HFR after 200 hours of OCV testing (Figure 2 and Figure 5); HFR values of all tested MEAs are similar in spite of the significant differences in H 2 crossover current. If PEM degradation occurs non-uniformly, membrane failures may occur at catalyst layer edges because the catalyst edges used in many commercial fuel cells are not strong enough to resist either the sharp edges of the GDL or the stresses present in the region that is unsupported between the GDL and the gasket. 27,28 In order to determine whether pinholes were generated near the catalyst edge or the active area, the N 2 gas crossover was measured in two MEA configurations (adding an additional Teflon layer to either the center or the edge of the MEAs after the OCV AST). Both MEA configurations had the identical surface area, 2.5 cm 2 . Figures 7a and 7b show the total N 2 crossover in MEAs after 200 hours of OCV testing without the Teflon layer as a function of the inlet N 2 flow rate. LSC-W/A shows the highest N 2 crossover rate while SSC-glycerol shows the lowest N 2 crossover rate, which is consistent with the H 2 crossover current results. Figure 7c compares the N 2 crossover flow in LSC-W/A using the two MEA configurations. The N 2 crossover rates through the catalyst layer edging and the center part of the active area are 65% and 35%, respectively. This data indicates that gas crossover occurs not only through the catalyst edge but also through the center of the catalyst active area, even though the MEA edge is considered to be weaker compared to the center part of the MEA active area. The MEA edge failures reported in literature 27,28 are therefore likely due to the poorer mechanical properties of thinner PEMs, ca. 10 to 24 μm. The thicker version of PEM (Nafion 212, 50 μm thick) that was used for the OCV testing may prevent premature catalyst edge failure to a certain degree.
Electrode morphology.-If PEM degradation does not occur throughout the active surface yet in focused local areas, local hot spot and/or pinhole formation generated from the non-uniformity of cathode structures would be the major cause of OCV reduction. [29][30][31] Figure 8 shows the SEM images of cathodes prepared from LSC-W/A and SSC-W/A. With higher magnification, (Figures 8b and 8d), the morphological features for both cathodes seem to be similar. However, with lower magnification, (Figures 8a and 8c), a difference is noted: the LSC-W/A cathode has larger macro-scale cracks (40 to >100 μm) compared to the SSC-W/A cathode (∼10 μm). In literature, electrode crack formation has been discussed in various aspects. [32][33][34][35][36][37] Most reports suggest that electrode crack formations have a detrimental impact on fuel cell performance due to destroying the threephase interface, 32 a buckling deformation, 33 pin-hole formation in the membrane, 34 and/or inhibiting multi-phase transport. 35 A few papers, on the other hand, report that crack formation can also benefit the cell performance by facilitating the access of gaseous reactants to the reaction site. 36,37 It should be noted that the pinhole formation and the improved gaseous reactant access through crack formation mentioned in literature is consistent with our observations with LSC-W/A and SSC-W/A MEAs. Also, the detrimental impact of the macro-scale cracks on PEM durability seems to be consistent in agreement with our research.
Further structural analysis was performed on cathodes prepared from different dispersing solvents. Figure 9 shows the effect of dispersing solvents on the morphology of SSC-bonded cathodes. While the water-isopropanol-processed cathode (Figure 9a) has macro-scale cracks (∼10 μm), NMP-and glycerol-(Pt/C to ionomer ratio: 5:1.7) processed cathodes (Figures 9b and 9c) show crack-free and uniform electrode structures (though the cathode from glycerol has slightly better uniformity since the cathode prepared from NMP has more dark areas). The cathode prepared from glycerol (Pt/C to ionomer ratio: 5:2) (Figure 9d) also shows crack-free and uniform electrode structure, yet has a notably less porous structure. The lower porosity of SSC-glycerol (Pt to ionomer ratio = 5:2) is thought to be responsible for the poor initial polarization behavior as shown in Figure 3b. Since both NMP-and glycerol-processed cathodes have crack-free structures, pinhole formation via macro-crack formation does not necessarily explain the PEM stability difference between the SSC-NMP and SSC-glycerol.
Impedance analysis.-EIS analysis was performed for the further investigation of electrode effect on PEM degradation. Figure 10 shows the Bode plots of the MEAs before OCV testing in the frequency range of 1-10,000 Hz at 1.0 A/cm 2 . It is noted that the resistance of MEAs with glycerol-processed cathodes are higher at the low frequency range ca. 1 to 100 Hz. This explains the limited O 2 transport of the glycerol-processed cathodes. The glycerol-processed cathodes with a Pt to ionomer ratio of 5:2 show greater O 2 transport resistance as expected. Noteworthy is that the O 2 transport resistances of LSC-W/A and SSC-W/A are comparable in spite of the much larger macro-crack formation of LSC-W/A, which explains the comparable polarization shown in Figure 1a. When water-isopropanol-processed cathodes are compared with NMP-processed cathodes, the irrelevancy of micro-crack on O 2 transport becomes clearer, i.e. the similar gas transport resistance of NMP-processed cathodes without macro-crack formation. This suggests that the PEM durability does not necessarily correlate well with cathode O 2 permeability. We believe that the relatively good O 2 transport through NMP-processed cathodes is via a uniformly distributed open pore structure, possibly with much smaller length-scale.
EIS analysis during OCV testing gives further information on the structural robustness of electrodes. Figure 11 shows the Nyquist plots of the MEAs before and after OCV testing at 1.0 A/cm 2 in the frequency range of 1 to 10,000 Hz. For the MEAs with waterisopropanol-and NMP-processed cathodes, the impedance at the Figure 9. SEM images of SSC-bonded cathodes prepared from a) waterisopropanol, b) NMP, c) glycerol (Pt/C to ionomer ratio: 5:1.7), and d) glycerol (Pt/C to ionomer ratio: 5:2). low frequency range notably increased as the OCV testing proceeded. However, for the MEAs with glycerol-processed cathodes, the impedance at the low frequency range did not change much over time. The low frequency impedance of SSC-glycerol even slightly decreased after 200 hours of OCV testing, indicating that O 2 transport through the glycerol-processed cathodes did not change thereby maintaining their initial three-phase interface. Meanwhile, O 2 transport in water-isopropanol-or NMP processed-cathodes gradually changed during OCV testing. Maintaining the initial three-phase interface during OCV testing may be important for uniform cell current generations which in turn minimize PEM degradation. The three-phase interface plays a role in the cell degradation observed while obtaining i-V polarization curves. When i-V polarization curves were taken, the increased cell current and the resultant water generation at the cathode caused changes in the three-phase interface, resulting in increased H 2 crossover through the PEM. One may argue that since the current density of MEAs with glycerol-processed cathodes during i-V curve measurements was less than that of other MEAs, the overall PEM degradation of glycerol-processed cathodes was less. However, this is not the case because the H 2 crossover current density of SSC-NMP without i-V curve measurements was found to be still higher than that of SSC-glycerol with i-V curve measurements i.e. 28 mA/cm 2 vs. 0.8 mA/cm 2 . The EIS analysis cannot explain the superior PEM durability of NMP-processed cathodes compared to water-isopropanol cathodes because both showed comparable change. However this can be explained by the fact that water-isopropanol-processed cathodes have non-uniform structures from the beginning. As we discussed, the existence of macro-cracks do not necessarily impact O 2 transport (and the low frequency impedance), yet less uniform cell cur- rents are likely generated. The superior PEM durability shown in SSC-bonded cathodes over LSC-bonded cathode can be explained also by the electrode robustness aspect. As previously indicated, SSC ionomers are less likely to undergo structural change due to two major reasons. 14,38,39 First, there are more physical crosslinking between sulfonic acid groups in SSC ionomers due to the higher sulfonic acid concentration, which restrains the chain mobility and increases the softening temperature. 14,38 Second, there are more polymer chain entanglements, i.e. a lower degree of phase separation, in SSC ionomers during solvent casting due to the shorter side chain length, thus improving the morphological stability. 39 However, further studies on electrode structural differences induced from SSC and LSC ionomers are needed. In summary, our electrode characterization supports that PEM degradation is accelerated by non-uniform cathode structures and structural instability during OCV testing. Our results are consistent with the direct observation of local hot spots by Zhang et al. 40 They showed that aged MEAs after OCV testing have hot spots during oxygen and hydrogen reactions, although the majority of the membranes are still in normal condition. The relationship between electrode uniformity and PEM degradation also explains the abrupt deterioration of the MEA performance after the i-V measurements where the increased cell current and the resultant water generation at the cathode caused changes in the three-phase interface.

Conclusions
This study shows that not only the chemical stability of Nafion itself but also cathode structures strongly impact the PEM stability. Using different ionomeric binders and cathode processing solvents shows that the H 2 crossover rate in Nafion 212 can be 200-fold different after 200 hours of OCV testing. Microscopic and impedance analysis indicate that cathode structural uniformity is the major contributing factor for PEM durability. The change of the cathode three-phase interface and non-uniform current generation during OCV testing causes PEM degradation at local focused areas. The O 2 permeation rate of the cathode plays a less critical role in PEM degradation. SSC ionomeric binder and glycerol solvent allow better electrode uniformity and resistance to structural changes during OCV testing. However SSC binder and glycerol solvent limit O 2 transport to a certain degree which adversely impacts the fuel cell performance. This study emphasizes the role of cathode structure on PEM durability. Electrode structure should be considered a crucial parameter for PEM durability in addition to the chemical stability of PEMs.