Membrane Accelerated Stress Test Development for Polymer Electrolyte Fuel Cell Durability Validated Using Field and Drive Cycle Testing

MEAs containing P5 and HD6 membranes were used by Ballard Power Systems in fuel cell bus stacks starting in 2002 and 2007, respectively. The P5 MEA was based on a 50 μm thick membrane with a total Pt loading of 1.05 mg/cm²; the HD6 MEA had a total Pt loading of 1 mg/cm² and a 25 μm thick membrane. The P5 stacks were operated in buses in Hamburg, Germany on different routes for approximately 3,000 hours, while the HD6 module was operated in the laboratory on an Orange County Transit Authority (OCTA) drive cycle for 6,842 hours.³² Noting that these membrane materials were between 11 and 16 years old at the time of this publication, they do not include mechanical reinforcement, thus, the chemical degradation and subsequent changes to water domain spacing and crystallinity are considered to be accurate.

ASTs were performed on Ballard P5 and HD6 MEAs using standard 50 cm² single cell hardware in a fuel cell test stand (Fuel Cell Technologies, Inc.). ASTs were also performed on MEAs containing Nafion XL (DuPont) membranes that are chemically and mechanically stabilized using Ce ions and polytetrafluoroethylene (PTFE) reinforcement, respectively.³³ Nafion XL MEAs were prepared identically by Ion Power, Inc., and had Pt contents which approached DOE automotive targets (∼0.2/0.1 mg_Pt/cm² on anode/cathode), corresponding to catalyst layer thicknesses of 6–12 μm.

After cell assembly, MEAs were conditioned using a break in protocol which consisted of cyclic operation at 0.6 V, OCV, and 0.3 V for 45, 30, and 60 seconds, respectively, for 2 hours at 80°C and 100% RH in H₂/air with stoichiometries of 1.2/2 and outlet pressures of 101/101 kPa_abs. At the time of this publication, the U.S. DRIVE Fuel Cell Tech Team has not recommended a break-in protocol for membrane testing because it is not expected to have a significant effect on membrane stability.

After conditioning, MEAs were exposed to ASTs consisting of (i) steady-state OCV hold, (ii) RH cycling in inert gases, and (iii) RH cycling at OCV. In these tests, high reactant gas flow rates were utilized in order to minimize any variation of stressors over the active area of the cell. Steady-state OCV tests were performed at 90°C and 30% inlet RH in H₂/air with gas flowrates of 0.7 and 1.67 standard liters per minute (SLPM) and outlet pressures of 250 and 200 kPa_abs, respectively.³⁴ In this test, asymmetric outlet pressures are not expected to have a significant effect on degradation rate.³⁵ The RH cycling (RHC) test was performed at 80°C in air with anode and cathode flowrates of 2 SLPM and outlet pressures of 101 kPa_abs. During the RHC AST, humidifier bottles (Fuel Cell Technologies Inc.) were switched from dry to wet (dew point = 90°C) every 2 minutes. The combined RH cycling at OCV (OCV RHC) test was identical to the RH cycling test, but performed at 90°C in H₂/air.

In addition to ASTs, Nafion XL-containing MEAs were tested in accordance with the U.S. DRIVE drive cycle protocol, which includes current cycling at 80°C in both supersaturated (dew point = 83°C) and 30% RH conditions.² Additionally, a modified version of the U.S. DRIVE drive cycle durability protocol operating with only the supersaturated portion of the cycle, called the 'wet' drive cycle, was utilized. All AST and drive cycle protocols are summarized in Table I. Additional details regarding current U.S. DOE recommended AST protocols may be found in Reference 2.

Electrochemical gas crossover

Fluoride emission measurements

Post mortem SEM

To characterize the morphological changes in fresh and degraded PEMs, small- and wide-angle X-ray scattering (SAXS/WAXS) experiments were carried out on Beamline 7.3.3 of the Advanced Light Source at Lawrence Berkeley National Laboratory. Specimens were cut from the middle of fresh and degraded MEAs and the CLs were removed using a carbon steel razor blade. To avoid the deleterious effects of CL removal with a solvent,³⁸ we have reported ionomer nanostructure data only for the membranes that were easily separated from the MEA, or that were already delaminated. Because of the high reactant gas flowrates, degradation during ASTs and the wet drive cycle is uniform across the active area. Therefore, such specimens are expected to be representative of overall PEM degradation. All scattering experiments were performed on the cathode side of each specimen. Additional anode-side measurements were also performed, when necessary.

The X-ray wavelength used was λ = 0.124 nm, with a monochromator energy resolution of E/dE of 100, and the presented patterns were collected using a 2D Dectris Pilatus 2 M CCD detector (172 μm x 172 μm pixel size). The scattering wave vector, q = 4π sin(θ/2)/λ, where θ is the scattering angle, was in the range of 0.001 to 0.04 Å⁻¹ for SAXS and 0.5 to 3 Å⁻¹ for WAXS. SAXS images for liquid-equilibrated samples were taken in-situ using custom-designed temperature-controlled liquid solution cells with X-ray transparent Kapton windows (for details, see previous studies³³). All of the experiments were carried out at 25°C and the samples were equilibrated in liquid water for at least 2 hours prior to imaging. The collected two-dimensional scattering patterns were azimuthally integrated to generate 1-D intensity profiles which were corrected for background scattering. From the SAXS data, hydrophilic-domain spacing and inter-crystalline spacing were calculated using a Gaussian fit to the peaks. The relative degree of the crystallinity, x_c, was calculated from the relative integrated area of the crystalline and amorphous peak intensities, which were determined after the deconvolution of the WAXS peak, as shown in previous work.³³

After long-term bus operation, two failed P5 MEAs and three failed HD6 MEAs were removed from their respective stacks for analysis. The P5 MEAs were operated in buses for ∼3,000 hours (4.1 months) in Hamburg, Germany, and the HD6 MEAs were stack-tested for 6,842 hours (9.4 months) in the Orange County Transit Authority (OCTA) drive cycle. Both MEAs contained significant "divot-type" damage, as shown in SEM micrographs (Figure 1a). This type of degradation is characterized by catalyst layer (CL) delamination from the PEM and local PEM thinning (Figure 1b). It is possible that CL delamination is initiated by a weakened CL-PEM interface due to chemical stress, which results in localized PEM thinning.³⁹ However, it is difficult to determine if CL delamination or PEM thinning occurs first during field operation. Regardless, local PEM thinning leaves the PEM more susceptible to mechanical stresses, which generate local cracks and tears (Figure 1c) that lead to eventual MEA failure. While both P5 and HD6 MEAs showed multiple divots near the inlets and outlets after field operation, significant global PEM thinning was not observed.⁴⁰ It should be noted that during bus operation, such MEAs typically experience temperatures of >70°C for <15% of their cycle duration with concurrent RH cycles between 85% and 100% RH.³² In contrast, the U.S. DRIVE automotive drive cycle consists of operation at 80°C with RH cycles between 30% and 100% RH.² The more benign conditions of bus operation are a contributing factor to the lack of any global thinning which is otherwise accelerated under high temperature and low RH automotive operating conditions.

Zoom In Zoom Out Reset image size

The RHC AST at 80°C resulted in no observable increase in gas crossover in either the P5, HD6, or Nafion XL MEAs after >20,000 cycles (1,333 hours) of testing (Figure 2a), which is consistent with the work of Crum and Liu.⁸ This AST was not able to differentiate between MEAs without mechanical reinforcement (P5 and HD6), and Nafion XL MEAs, which contain PTFE reinforcement that is known to substantially improve PEM mechanical durability.⁴¹ Further, the divot-type failure observed after field operation was only replicated in one location of the P5 MEA and was absent in the HD6 MEA.

Zoom In Zoom Out Reset image size

As shown in Figure 3b, this test did, however, result in local PEM tears near the cathode side of the P5 MEA, similar to those observed after field operation (Figure 1c). Here, there is a strong correlation between the location of cathode CL cracks in fresh MEAs (Figure 3a) and mechanically-induced cracks in the AST-tested MEAs (Figure 3a), which is in good agreement with the results of Singh, et al.⁴² Although the RHC AST has the ability to replicate mechanical failure via PEM cracks, it does not provide sufficient acceleration of field stresses, which does not enable it to have the ability to differentiate between PEMs with different mechanical properties on a reasonable time scale.

Zoom In Zoom Out Reset image size

The steady-state OCV AST at 30% RH and 90°C resulted in failure of the P5 and HD6 MEAs after ∼100 hours, while the Nafion XL MEA failed after ∼250 hours (Figure 2b). This test also caused significant global PEM thinning in both the P5 (Figure 3b) and Nafion XL MEAs (Figure 4) without through-thickness cracking or CL delamination. Thinning rates for different ASTs shown in Table II were calculated using previous data,⁴³ along with the SEM micrographs shown Figure 4. These results demonstrate that OCV testing at high temperature and low RH increases PEM thinning rate by over two orders of magnitude compared to field-tested MEAs. Further, the Nafion XL MEA contains Ce-based radical scavengers and reduces thinning by a factor of 3x during the OCV AST compared to the Ce-free P5 and HD6 PEMs, which is reflected from membrane thinning data shown in Table II. However, even the non-stabilized MEAs operated under field conditions in the buses show no thinning, indicating that this AST overestimates the impact of chemical degradation in the field by a factor of >100x. It should be noted that mechanical PEM failure occurred in the field before any noticeable membrane thinning occurred. Thus, it is suggested that the OCV AST does not accurately capture the magnitude of PEM degradation occurring in the wetter bus module environment.

Zoom In Zoom Out Reset image size

As shown in Figure 2c, the 90°C OCV RHC AST resulted in the failure of the HD6 and P5 MEAs after ∼100 and ∼200 hours, respectively, where PEM cracking and local PEM thinning were also observed. The Nafion XL MEA failed after ∼675 hours. As shown in Table II, during the OCV RHC AST protocols, PEM thinning rates are more similar to the field data. These results imply that the magnitude of chemical degradation during OCV RHC is significantly less than that during the constant 30% RH OCV test, alone. The reduction in chemical stress is attributed to the 2 minute hydration component of the cycle. As the MEA becomes humidified, sulfonic acid end groups in the PFSA are fully ionized and exist in the inert proton form (–SO₃⁻) which are not susceptible to hydrogen abstraction by the hydroxyl radical.⁴⁴ Such hydrogen abstraction leads to irreversible degradation of the side chain which results in complete chemical decomposition of the main chain and side chains associated with that PFSA molecule.⁴³ Presumably, during both field operation and the OCV RHC AST, the PEM is chemically degraded until mechanical failure occurs. The magnitude of this chemical degradation, however, is significantly less than that at OCV, alone.

Gas crossover measurements and SEM micrographs indicate that even beyond DOE AST targets, the RHC AST does not significantly accelerate PEM degradation and local failure. Conversely, the OCV AST introduces a mode of global PEM thinning which was not observed during field operation. The OCV RHC AST results in mechanical membrane failure that is accelerated by the chemical degradation, with the ability to distinguish between various types of membranes within a reasonable testing time. Therefore, to better capture the modes of PEM degradation and failure observed in the field, the combined chemical/mechanical OCV RHC AST is recommended. While the mechanical properties of the PEM during these ASTs were not measured directly, it is clear that their decay is significantly enhanced during the OCV RHC AST compared to field and drive cycle operation. As a result, using the OCV RHC AST, PEM stability may be assessed in hundreds of hours, compared to the latter modes of operation, which can require thousands of hours of testing.

To develop a better understanding of the changes in morphology of PEMs induced by various degradation modes, SAXS/WAXS of membranes were examined before and after durability testing. The SAXS results shown in Figure 5 indicate changes to the hydrophilic domain spacing for the different membranes and testing conditions. The broadness and the peak location of the ionomer SAXS peak change significantly after testing for both Ballard and Nafion membranes. Compared to the fresh, beginning of life (BOL) samples, two key changes are observed in the SAXS spectra of the tested/failed MEAs: (i) the ionomer peak shifts to the left indicating a larger average correlation length, which is attributed to the hydrophilic domain spacing, d-spacing, and (ii) the peak becomes broader indicating a wider distribution of d-spacing and weaker phase-separation.

Zoom In Zoom Out Reset image size

Changes to the hydrophilic domain spacing were measured after ASTs using SAXS (Figure 5). The d-spacing values for fresh MEAs are 4.5 ± 0.5 nm. The d-spacing of Ballard MEAs increases by 20% after RHC or drive cycles (Figure 5b). In these MEAs, domain spacing in both RHC and OCV RHC ASTs are more similar to the field-tested MEA than the OCV-tested MEA, which is substantially increased. This result is consistent with previous electron microscopy studies that show an increase in the spacing of sulfonic acid domains due to chemical degradation.⁴⁵ A similar trend exists for the Nafion XL MEAs (Figure 5d). The most dramatic change in Nafion XL membranes occurs after the OCV test, which increases d-spacing to 6 nm. In the RHC and OCV RHC-tested Nafion XL MEAs, d-spacing was actually decreased, which may be attributed to increased concentrations of Ce ions in the active area of the membrane, which have previously been measured during humidified operation.³⁷ Increased Ce contents may reduce d-spacing through ionomer-cation interactions and loss of local water uptake.⁴⁶ The effects of cation migration on ionomer water uptake and structure are currently under investigation by our labs.

It is worth noting that nanostructural changes in the membrane after an OCV test changes depending on the location probed, concurrent with the microscopic changes observed in SEM images such as shown in Figure 4c. While a specimen probed on the cathode side, where heavy ionomer loss occurred (denoted loss side), does not give rise to any discernable scattering signal, the ionomer from the lower-loss anode side (denoted intact side) exhibits a typical SAXS spectra, albeit with a very broad peak. In addition, the d-spacing of the membrane that failed by pinhole formation shows a relatively smaller change in d-spacing which is comparable to the d-spacing of a fresh, BOL membrane. Thus, degradation-induced changes in nanostructure of ionomers manifest themselves the most in terms of degree of phase-separation, which can be associated with the domain-network properties and degree of structural disorder. Such changes in the morphology could also imply strong variation in transport properties and structural integrity of the ionomer, which could be related to membrane performance during the tests.

Overall, degradation-induced changes vary depending on both the membrane and type of tests, yet OCV is the most severe test for both membranes in terms of its impact on the nanostructure, in accordance with previous observations. Moreover, after an OCV test, nanostructural changes in the ionomer could also vary throughout the MEA due to the non-uniform material loss and local failure. Therefore the SAXS data provides morphological evidence toward the severity of the OCV test while confirming that the RHC OCV AST closely reproduces the morphological changes caused by the field (Ballard) and drive cycle durability (Nafion XL) tests.

Changes in crystallinity were measured after ASTs using WAXS (Figure 6). Since WAXS probes the relative degree of crystallinity, the values should be compared within the same sample set to assess the changes due to ASTs. The OCV AST increases crystallinity in both cases (Ballard MEAs in Figure 6a and Nafion XL MEAs in Figure 6b), whereas OCV RHC and RHC AST tested samples are most similar to field and drive cycle.

Zoom In Zoom Out Reset image size

An interesting outcome of chemical degradation in Nafion XL after the OCV test is that the crystallinity is drastically reduced on the intact side of the membrane. It is inferred that at this location, chemical degradation destroys the structure of the PFSA molecules, which shortens the PTFE backbone length, thereby breaking the crystalline structure into amorphous one. On the loss side of the membrane, which is comprised of the PTFE reinforcement, the degree of crystallinity increases to over 70%, which is closer to the value of PTFE. Such high values of x_c were reported and attributed to the crystalline features of the reinforcement phase in a previous study.³³ These results indicate that the OCV AST test results in a significant change in the crystallinity of the ionomer which is not observed in either field or drive cycle operation. This extra degradation mode is not observed in the RHC OCV AST samples, providing further credence that the RHC OCV AST is more representative of the degradation modes present during typical PEM fuel cell operating conditions.

Under severe conditions of the OCV AST at 30% RH, while the ionomer phase of the MEA loses its characteristic features of phase-separated nanostructure, the intact reinforcement phase appears to preserve its structural integrity. Chemical degradation of the ionomer results in increased d-spacing and extreme chemical degradation can result in increased crystallinity of the ionomer. Mechanical degradation does not appear to induce any permanent change in ionomer structure (e.g., EW, backbone structure) and therefore does not affect the crystallinity of MEAs and can result in only a slight increase/decrease in d-spacing of the ionomer in PEMs, regardless of the presence of PTFE support.

Therefore, depending on the application, it is important to have an accurate combination of chemical and mechanical stressors in an AST to correctly reflect failure mechanisms in the field or a simulated drive cycle. Both SAXS and WAXS results of MEAs after various tests confirm that the morphological changes during field and drive cycle operation are most accurately reproduced by the mechanical RHC and chemical/mechanical OCV RHC ASTs rather than by a chemical OCV AST, alone.

At OCV, 2 minute wet/dry cycles (denoted 2m/2m) were shown to fail mechanically a Nafion XL PEM without significant thinning (Figure 4d), which is similar to the wet U.S. DRIVE drive cycle-tested sample in terms of thinning rate (Table II), water domain spacing (Figure 5), and crystallinity (Figure 6). During automotive operation, however, MEAs experience drier, more aggressive conditions compared to either the bus field operating conditions or the OCV RHC.

To increase the magnitude of chemical stress in the OCV RHC AST, the 2 minute cycles were adjusted to 30 and 45 seconds of wet and dry operation, respectively (denoted as 30s/45s). Figure 7a shows the changes in HFR during the two cycling protocols, which are inversely proportional to the hydration of the PEM. Changing the cycle duration from 2m/2m to 30s/45s increases the number of hygral variations per unit time by a factor of 3 (Figure 7a).

Zoom In Zoom Out Reset image size

As shown in Figure 7b, by adjusting the duration of the wet and dry intervals, the duration of the dry phase in each cycle (HFR > 0.15 Ω-cm²) is increased relative to the duration of the wet phase (HFR < 0.15 Ω-cm²). A 2m/2m cycle has approximately equal time dry and wet, whereas a 30s/45s cycle has about 75% dry time/25% wet time. This increased duration in the dry phase, along with increased temperature, leads to a higher magnitude of chemical stress, as explained above.

The gas crossover rates of Nafion XL MEAs during OCV RHC ASTs performed at 90°C with different cycle durations are shown in Figure 8. When cycle time is modified from 2m/2m to 30s/45s, AST lifetime decreases slightly from 660 to ∼600 hours. Here, the increased frequency of hygral cycles which result from reducing the cycle duration does not significantly reduce time to failure. This is consistent with previous results which suggest that MEA lifetime is more strongly influenced by the magnitude of stress and the accumulated time under stress rather than the amount or frequency of stress cycles.²⁵ And indeed, as shown in Figure 7b, in the 30s/45s cycle, the magnitude of the hygral change during each cycle is ∼25% less than the 2m/2m cycle due to the shorter duration in the wet phase. Therefore, during this cycle duration the increase in chemical stress is offset by a decrease in mechanical stress which results in a very similar lifetime compared to the 2m/2m protocol.

Zoom In Zoom Out Reset image size

Since cross-sectional SEMs and X-ray scattering data for field-tested automotive MEAs are not yet available, the U.S. DRIVE drive cycle durability protocol was utilized for validation of ASTs. This protocol is designed to replicate stressors during automotive operation. In order to replicate bus operation, this protocol was also modified by eliminating the dry regime of the cycle (called "wet" drive cycle), as described above in the Experimental section.

The cross-sectional SEM micrographs of drive cycle tested MEAs and an MEA tested using the modified 30s/45s OCV RHC AST are shown in Figure 9. From this SEM data, thinning rates were calculated and compared to the Ballard field data, as shown in Table III. Here, it is observed that the drive cycle and the 30s/45s OCV RHC AST have similar thinning rates, which are 6x faster than Ballard field-tested and wet-drive cycle thinning reflecting the drier automotive environment.

Zoom In Zoom Out Reset image size

During the drive cycle test, the average FER was 0.102 mg/cm²-h, while during the 30/45s OCV RHC test, it was 0.11 mg/cm²-h. These comparable FERs further suggest that the magnitude of chemical stress is similar between the 30s/45s OCV RHC AST and the U.S. DRIVE drive cycle durability protocol. Therefore, it is concluded that this RHC OCV AST protocol is most similar to the U.S. DRIVE drive cycle durability protocol and is most representative of the degradation MEAs experience during automotive operation. As fuel cell electric vehicles increase in prevalence, the availability of field-tested MEA data will be useful to further validate this modified protocol.