Cerium Migration during PEM Fuel Cell Accelerated Stress Testing

MEAs were synthesized using 100 cm² Nafion XL PEMs (DuPont, 27.5 μm thick, ∼6.0 μg/cm² ion-exchanged cerium). For fully humidified operation, a commercial Ion Power MEA was used, which had platinum loadings of 0.12 and 0.27 mg/cm² on the anode and cathode, respectively, and an ionomer to carbon (I/C) mass ratio of 1.23. For low humidity and humidity cycling ASTs, MEAs were synthesized in-house using a decal transfer method.³⁹ Electrode decals were formed by painting platinum catalyst ink (TKK, 48% Pt/C) with an I/C mass ratio of 0.75 onto 50 cm² polytetrafluoroethylene (PTFE)/fiberglass backing sheets (Saint-Gobain Performance Plastics). Platinum loading on each decal was verified at 0.2 ± 0.02 mg/cm² in air using a Thermo Scientific QUANT'X X-ray fluorescence (XRF) spectrometer with a 3.5 mm aperture, a palladium filter, an accelerating voltage of 25 kV, a tube current of 1.6 mA, and a scan duration of 20 seconds. After the desired platinum loadings were achieved, decals were hot pressed onto the PEMs for 5 minutes at 130°C and 4.8 MPa.

MEAs were assembled into 50 cm² single-cell hardware (Fuel Cell Technologies) using graphite bipolar plates, 25BC gas diffusion layers (GDLs, SGL Group), 317.5 μm polyurethane (PU) gaskets, and 25.4 μm PTFE sub-gaskets. For fully humidified operation, the active area was reduced from 50 to 25 cm² using a PU gasket. Test hardware was compressed to obtain an approximate GDL pressure of 1.5 MPa.

After assembly, cells were installed in a fuel cell test stand (Fuel Cell Technologies) in co-flow orientation with inlets high and conditioned for 16 hours at 0.6 V at 90°C and 100% RH in H₂/air with stoichiometries of 1.2/2 and outlet pressures of 101/101 kPa_abs. After conditioning, the cells were subjected to ASTs which were adopted from standard protocols.⁴⁰ Cells were operated at OCV and temperatures of 90°C with H₂/air flow rates of 0.7/1 standard liters per minute (SLPM) for 50 cm² cells and 0.35/0.83 SLPM for the 25 cm² cell, with outlet pressures of 101/101 kPa_abs. ASTs were performed at three humidity conditions: 100% RH, 30% RH, and wet/dry cycling, where the cell was exposed to fully humidified reactant gasses for 30 seconds, followed by 45 seconds of exposure to dry reactant gases, which bypassed the water reservoirs of the humidity bottles.⁴¹ During ASTs, OCV was continuously monitored using a data acquisition system on the test stand control computer.

Pressure drops (ΔPs) were measured on the anode and cathode sides of each cell under identical AST conditions using mechanical pressure gauges (U.S. Gauge) installed at gas inlets and outlets. The effective anode and cathode ΔPs during wet/dry cycling (ΔP_{wet/dry cycle}) were calculated as:

where t_wet and t_dry are the fractions of the total test duration in the wet and dry phases of the cycle, respectively, and ΔP_{100% RH} and ΔP_{0% RH} are the corresponding anode or cathode pressure drops measured at 100 and 0% RH, respectively.

At OCV, it is assumed that the molar flows of reactant gasses are conserved from inlet to outlet because no reactant gasses are consumed and water vapor does not condense to liquid water. In addition, it is assumed that the cell is isothermal and the equilibrium vapor pressure of water at 90°C remains constant. Therefore, the corresponding change in outlet RH due to ΔP was calculated as:

where P_outlet was 101 kPa_abs for all ASTs.

Every 24 to 72 hours, the cells were conditioned for 30 minutes in H₂/N₂ at 100% RH and 90°C with gas flow rates of 0.5/0.5 SLPM for 50 cm² cells and 0.25/0.25 SLPM for the 25 cm² cell and outlet pressures of 101/101 kPa_abs. Electrochemical gas crossover was then measured using linear sweep voltammetry from 0.1 to 0.5 V at a scan rate of 2 mV/s using a PAR 273 potentiostat.⁴² The crossover current density measured at 270 mV was reported.

Effluent water samples were collected from the anode and cathode gas outlets every 24 to 72 hours using a jacketed condenser system. A Dionex ICS 2100 ion chromatography (IC) system was used to analyze waters for trace concentrations of fluoride using standard test protocols.⁴³ For this technique, detection limits are around 0.025 mg/mL and the accuracy of measurements of 1 mg/mL is ±5%.⁴⁴ Fluoride standards (Fluka) were prepared at ionic concentrations of 0.5, 1, 5, 10, and 20 mg/mL in deionized (DI) water. A quadratic calibration curve was obtained with an R² of 0.999.

At each water collection interval, n, the fluoride emission (FE) from each side of the cell, measured in mg/cm², was calculated as:

where C_{F, n} is the concentration of fluoride ions, in mg/mL, V_{water, n} is the volume of water collected during that interval, in mL, and A_cell is the active area of the cell, in cm². The cumulative fluoride emission (∑ FE) at each collection interval was calculated by summing the fluoride emitted in all of the preceding collection intervals, i = 1...n:

Finally, the fluoride emission rate (FER) at each collection interval was calculated as:

where t_n is the duration of the corresponding collection interval, in hours.

A Thermo Scientific iCAP Q inductively coupled plasma mass spectrometer (ICP-MS) was used to analyze water samples for trace concentrations of cerium. Water samples collected from the 30% RH and wet/dry cycling ASTs were analyzed directly, and samples from the 100% RH AST were concentrated by a factor of ∼6 before analysis. Cerium standards (SCP Science) were prepared at 0.005, 0.02, 0.05, 0.1, 0.5, 1, and 2 μg/mL concentrations, spiked with 1 μg/mL of indium internal standard (High Purity Standards), and diluted up to 50 mL using a solution of 2% nitric acid (Fisher Optima, 70%) in DI water. A linear calibration curve was obtained with an R² of 0.999. Collected water samples were spiked with indium at 1 μg/mL and nitric acid to matrix match to the standards. Cerium emission (CE), cumulative CE (∑ CE), and cerium emission rate (CER) were calculated for each AST using Equations 3, 4, and 5, respectively, where C_F,n was replaced with the cerium concentration (C_Ce,n), in ng/mL, measured at each collection interval.

ASTs were performed until electrochemical gas crossover exceeded 20 mA/cm² or the test duration surpassed 2,000 hours. After ASTs, cells were dried for 1 hour at 90°C in 0% RH N₂/N₂ with flow rates of 0.5/0.5 SLPM for 50 cm² cells and 0.25/0.25 SLPM for the 25 cm² cell and outlet pressures of 101/101 kPa_abs. Drying in an inert N₂ environment under these conditions was previously demonstrated to have no significant impact on final cerium location after ASTs.³²

In order to conduct scanning electron microscope (SEM) evaluation of MEAs after ASTs, 2 × 1 cm specimens were cut near the outlets of the dried MEAs as shown by the green box in Figure 1. To prepare MEA cross sections for imaging, specimens were immersed in liquid nitrogen and freeze-fractured using fresh, carbon steel razor blades (SPI Supplies). The fractured surfaces were analyzed using a FEI Quanta 45 SEM with an accelerating voltage of 5 kV. Either a secondary or backscatter electron detector was used, when necessary, to obtain optimal contrast between the PEM and CLs.

Zoom In Zoom Out Reset image size

Cerium contents in the MEAs were measured using nine 2 × 1 cm specimens cut along the center of the MEAs from inlet to outlet as indicated by the yellow boxes in Figure 1. Cerium contents in the PEM and CLs were characterized in air using a Thermo Scientific QUANT'X XRF Spectrometer with a 3.5 mm aperture, an aluminum filter, an accelerating voltage of 12 kV, a tube current of 1.98 mA, and a scan duration of 1 minute. In order to prepare specimens for XRF, the GDL was manually removed from one side of the MEA specimen (Figure 2a) and the specimen was mounted on a PTFE/fiberglass backing sheet (Figure 2b). Then, the corresponding CL was wiped away using isopropyl alcohol (IPA). In order to obtain the cerium concentration in the PEM and opposite CL (C_{Ce, PEM+CL}), XRF was performed on the center of the specimen in the area marked by the red ellipse shown in Figure 2c. Next, the specimen was separated from the GDL, flipped, and remounted on the backing sheet (Figure 2d). The opposite CL was removed using IPA to expose the bare PEM (Figure 2e). The cerium content in the PEM (C_Ce,PEM) was then determined using identical XRF parameters (Figure 2f). Finally, the cerium content of the opposite CL was calculated as:

Zoom In Zoom Out Reset image size

C_Ce,PEM was analyzed in every specimen, but due to the nature of the XRF analysis technique, only one CL was analyzed in each specimen. To capture any trends in the spatial variation of CL cerium concentrations down the length of the MEA, anode and cathode CLs were analyzed in alternating order as shown in Figure 1.

Wet chemistry was used to investigate interactions between cerium ions in the MEA and carbon catalyst supports in the CLs. In chemically-stabilized MEAs, the cerium to carbon (Ce/C) mass ratio is approximately 0.059. To replicate this ratio ex situ, cerium(III) acetate (Alfa Aesar, 99.9% purity) and carbon black (Cabot VULCAN XC72) were stirred at a mass ratio of 0.1 in excess DI water at 80°C for 16 hours. After mixing, the slurry was sonicated for 20 minutes at 80°C and washed with DI water through a PTFE filter (Millipore, 25 nm pore size) under vacuum. The cerium content in the resulting filtrate was quantified using XRF with identical scan parameters as described in the Post-mortem analysis section. The mass of cerium in the filtrate (M_Ce,filtrate) was calculated as:

where C_Ce,filtrate is the areal concentration of cerium in the filtrate disk, in mg/cm², measured using XRF, and A_filtrate is the area of the resulting filtrate disk, in cm². Then, the Ce/C mass ratio in the filtrate was calculated as:

where M_filtrate is the total mass of the filtrate, in mg.

Carbon black/DI water and cerium acetate/DI water solutions were also prepared and filtered using identical procedures to verify the absence of cerium impurities in the carbon black or cellulose filter, respectively, after the filtering process.

The OCV decay rates and electrochemical gas crossover values for the three different ASTs are shown in Figure 3. 100% RH operation resulted in the most stable OCV, which decayed at a rate of 16.5 μV/hr. Over the 2,000 hour duration of the test, gas crossover did not exceed 3 mA/cm², however, only the first 1,000 hours are shown in Figure 3. 30% RH operation resulted in a substantially higher OCV decay rate of 0.19 mV/hr. At 360 hours, gas crossover surpassed 10 mA/cm², and at 456 hours, gas crossover exceeded 20 mA/cm² and the test was terminated. Wet/dry cycling resulted in the most severe OCV decay rate of 0.33 mV/hr, however, gas crossover was stable until 823 hours, when it surpassed 20 mA/cm² and the test was terminated.

Zoom In Zoom Out Reset image size

Significant PEM thinning was also observed during ASTs, as evidenced by cross sectional SEM micrographs (Figure 4). After MEA fabrication, PEM thickness was approximately 25.3 μm (Figure 4a). After 100% RH operation, PEM thickness was reduced to 24.5 μm, which corresponds to a thinning rate of 0.4 nm/hr over the 2,000 hour AST duration (Figure 4b). 30% RH operation resulted in a final PEM thickness of 20.2 μm, which corresponds to a thinning rate of 11.1 nm/hr for 456 hours (Figure 4c). Wet/dry cycling resulted in the most severe thinning. The final PEM thickness was measured at 11.8 μm after crossover failure, which corresponds to a thinning rate of 16.4 nm/hr for 823 hours (Figure 4d).

Zoom In Zoom Out Reset image size

The average FERs measured in anode and cathode effluent waters during each AST are shown in Figure 5. The total FER was calculated by summing anode and cathode FERs for each AST. 100% RH operation resulted in the lowest FERs, which implies that minimal PEM degradation occurred. In this test, the initial fluoride inventory of 3.8 mg/cm²² was only reduced by 7% after 2,000 hours, which is consistent with the small amount of thinning observed in SEM micrographs (Figure 4b). It is generally accepted that during fully humidified operation, ionomer sulfonic acid end groups are fully ionized, and exist in the inert salt form (–SO₃⁻), which are not susceptible to hydrogen abstraction by HO•.⁴⁵ Ionomer backbone unzipping is mitigated by pre-treating PEMs with elemental fluorine to remove vulnerable terminal carboxyl groups.⁴⁶ Therefore, improved stability is expected at 100% RH. The observed FERs, however, may result from other proposed side chain scission mechanisms, which cause cleavage and unzipping of the backbone and are not affected by sulfonic acid solvation.⁴⁷

Zoom In Zoom Out Reset image size

30% RH operation resulted in significantly increased FERs relative to fully humidified operation, indicating enhanced PEM degradation occurred. During low RH operation, sulfonic acid ionization is reduced in the ionomer, causing acidic protons to reside on the sulfonic acid groups (–SO₃H).⁴⁵ These groups are susceptible to hydrogen abstraction by HO•, which initiates ionomer chain scission and accelerates degradation.⁴⁵ After the 456 hour test duration, fluoride inventory was reduced by 12%, which is consistent with the moderate thinning observed in SEM micrographs (Figure 4c), however, such fluoride inventory loss and thinning are unlikely to induce crossover-related failure alone.² Therefore, we infer that the premature failure observed at 456 hours is due to the formation of a local pinhole near the edge of the active area.

Humidity (wet/dry) cycling under OCV conditions resulted in the largest FERs and a 38% reduction of fluoride inventory. These results are consistent with the severe thinning observed in SEM micrographs (Figure 4d). Under these conditions, chemical degradation progresses due to the dry component of the cycle, while the PEM simultaneously experiences enhanced mechanical stresses due to hydrothermal cycling. Due to the synergy of the chemical and mechanical degradation modes, humidity cycling results in larger FERs and PEM thinning than constant RH tests, alone.^4,7 In all ASTs, anode and cathode FERs were observed to be uniform.

These results demonstrate that OCV decay rate, PEM thinning rate, and fluoride loss are well correlated. Therefore, we confirm that the aggressiveness of PEM degradation is directly controlled by the prescribed humidification conditions. It is concluded that AST aggressiveness is wet/dry cycling >30% RH ≫ 100% RH.

Figure 6 shows the in-plane cerium area density profiles for each MEA component. During 100% RH operation, anode CL (Figure 6a) and cathode CL (Figure 6b) cerium contents remained nearly zero over the length of the scan, except at the outlet of the cathode. Figure 6c shows that cerium remained in the active area, however, a concentration gradient was observed from inlet to outlet. Additionally, points in the active area of the PEM, which are under the gasket, as indicated by the blue shaded area in Figure 6c, showed cerium area densities of ∼0.0 μg/cm². Here, migration into the neighboring, ungasketed active area is attributed to in-plane hydration variations due to compression, which will be discussed in the Average through-thickness cerium migration during ASTs section. After 30% RH operation, cerium is reduced uniformly from ∼6.0 to 3.7 μg/cm² in the active area. The in-plane profiles in both the CLs and PEM are flat (Figures 6a–6c), which indicates that cerium migration from the PEM to CLs occurs only in the through-plane direction. After humidity (wet/dry) cycling, cerium gradients are also observed from inlet to outlet in the anode and cathode CLs, as well as in the PEM (Figures 6a–6c).

Zoom In Zoom Out Reset image size

In both ASTs where the cell was exposed to 100% RH reactant gasses (100% RH and wet/dry cycling), cerium gradients were generated from PEM inlet to outlet (Figure 6c). These gradients may be explained by water flux which results from in-plane RH gradients within the MEA due to ΔP. Such gradients result in higher RH near the inlet and lower RH near the outlet.

For each AST condition, ΔPs measured on the anode and cathode sides of the cells, along with their corresponding effects on outlet RHs are reported in Table I. ΔPs are ∼10 and ∼25% of anode and cathode outlet pressures, respectively, for each AST. These values correspond to outlet RH levels which are ∼90 and ∼75% of the anode and cathode inlet RHs, respectively. Using these RH values, λ's at cell inlets and outlets were also calculated from ionomer hydration isotherms at 90°C.⁴⁵

As shown in Table I, water content (defined as λ = n_H2O/n_-SO3H) decreases significantly from inlet to outlet during 100% RH operation. Δλ is 1.9 and 5.2 on the anode and cathode side of the cell, respectively. Because of this gradient, water will equilibrate in the PEM from inlet to outlet. We believe that this water flux serves as the driving force for cerium migration toward the outlet, as experimentally observed in Figure 6c. For the 30% RH case, variations in hydration between the inlet and outlet are reduced by an order of magnitude relative to the 100% RH case. Δλ is 0.2 and 0.5 on the anode and cathode, respectively. The flat cerium profiles in the PEM at 30% RH (Figure 6c) suggest that water flux due to this hydration gradient is either small or nonexistent, and, thus, does not influence in-plane cerium migration. Similarly for the wet/dry cycling AST, the average RH over the cycle yields Δλs which are similar to the 30% RH case, yet cerium gradients are still observed from inlet to outlet (Figure 6c). Therefore, we propose that the cerium gradient forms due to exposure to fully humidified reactant gasses during the hydration component of the cycle. Previous results³² suggest that migration driven by concentration gradients or water/proton flux occurs more slowly than water evaporation, itself, which indicates that drying component of the cycle does not significantly affect the location of the cerium.

The average cerium area densities in the anode CL, cathode CL, and PEM were calculated from the in-plane profiles and are shown in Figure 7a. After 2,000 hours of OCV operation at 100% RH, the average cerium content in the active area of the PEM increased from ∼6.0 to 9.8 μg/cm², while the anode and cathode CL cerium contents increased from 0.0 to 0.95 and 4.6 μg/cm², respectively (Figure 7a). During this test, total cerium contents in the active area also increased from ∼6.0 to 15.3 μg/cm².

Zoom In Zoom Out Reset image size

After 456 hours of OCV operation at 30% RH, the average PEM cerium area density was reduced from ∼6.0 to 3.7 μg/cm², while anode and cathode CL cerium area density increased from 0.0 to 2.3 and 3.5 μg/cm², respectively (Figure 7a). During this test, total cerium in the active area also increased from ∼6.0 to 9.4 μg/cm². These results are consistent with previous through-plane cerium migration measurements performed using XRF, where a significant portion of PEM cerium migrated into the CLs during low RH operation at OCV.³¹

Wet/dry cycling for 823 hours resulted in the most significant migration of cerium out of the PEM into the CLs; only 0.99 of the initial ∼6.0 μg/cm² remained in the PEM, while the anode and cathode CLs each contained 1.9 and 3.1 μg/cm², respectively (Figure 7a). During this AST, however, total cerium content in the active area remained unchanged.

Total cerium concentration in the active area exceeded the initial concentration of ∼6.0 μg/cm² after constant-humidity operation at 100 and 30% RH (Figure 7a). PEM hydration is assumed to be 100% RH > 30% RH > wet/dry cycling during ASTs. From Figure 7a, we observe that total cerium content in the active area is correlated to hydration. The increased cerium concentrations result from significant migration from the gasketed, inactive PEM border regions into the active area.

During cell humidification, water equilibrates from the active area into the gasketed region, however, water content there always remains lower than in the active area.^48–50 This variation in water content increases the relative cerium concentration in the adjacent border, which subsequently drives its equilibration into the uncompressed, humidified active area via ionic diffusion. Water removal from the edge of the PEM exposed to ambient conditions is minimal. Therefore, we infer that water flux and subsequent cerium migration into the gasketed region are negligible after initial equilibration. Therefore, we conclude that active area hydration, which is dictated by the AST humidification condition, induces in-plane cerium migration, primarily through ionic diffusion.

In all cases, cerium migration into the cathode CL was greater than that into the anode CL (Figure 7). This phenomenon could be an artifact from conditioning the MEAs at 0.6 V before ASTs, where proton and/or water flux may have dragged cerium cations toward the cathode.^32,51 In addition, the greater cerium concentrations in the cathode CLs may be explained by net anode to cathode water flux generated by lower cathode hydration (Table I), as discussed in In-plane cerium migration profiles during ASTs section.

Figure 7b shows the cerium mass concentrations in the ionomer region of each MEA component. The mass concentrations of cerium in the PEMs were obtained by normalizing PEM cerium area densities by the final thicknesses of the PEMs after ASTs shown in Figure 4. Mass concentrations of cerium in the CLs were obtained by normalizing cerium area densities to CL ionomer volumes, which were calculated using catalyst loading and I/C mass ratio quantities listed in MEA preparation and cell assembly section. These results demonstrate that cerium concentrations in the CL ionomer significantly exceed those expected by ionic diffusion, alone, especially in the most aggressive AST conditions.

Cumulative cerium emissions during ASTs are shown in Figure 8. Cerium washout from the cell was lowest during fully humidified operation, and substantially higher for both dry operation and wet/dry cycling. At the EOT, 100% RH operation results in cumulative total (anode + cathode) cerium emissions of 6.6 ng/cm², which represents a 0.11% loss of total MEA cerium inventory after 2,000 hours. 30% RH operation results in a cumulative cerium emission of 9.5 ng/cm², or 0.16% inventory loss at 456 hours. Even without premature pinhole failure, only 0.25% of the cerium inventory would be depleted after 1,000 hours of projected operation at 30% RH. In the most aggressive wet/dry cycling AST, 17.2 ng/cm² or 0.29% of the total cerium inventory is lost at the EOT at 823 hours. Under OCV conditions, we conclude that cerium loss via washout of the effluent water is insignificant compared to the magnitude of migration within the MEA. However, cell operation under load, which results in significant water generation at the cathode, may result in greater cerium washout due to increased water flux between the CL and flow channels.

Zoom In Zoom Out Reset image size

Figure 9 reveals that the total EOT fluoride and cerium emissions are correlated. Therefore, it is proposed that the cerium measured in the effluent fuel cell water streams exists in the form of cerium cations exchanged with ionomer degradation products. While cerium-exchanged sulfonic acid groups, themselves, are not susceptible to radical attack,⁵² any side chain scission will result in complete decomposition of the main chain and side chains associated with that ionomer molecule.² As a result, neighboring, cerium-exchanged sulfonic acid groups will be expelled into effluent water streams along with HF and fluorinated polymer fragments. While total cerium washout from the MEA is negligible at OCV, this correlation provides insight into an additional mechanism for cerium migration: during ASTs, cerium is mobilized within the MEA as a result of ionomer decomposition.

Zoom In Zoom Out Reset image size

Figure 10 reveals a strong inverse correlation between FER and the remaining cerium concentration in the PEM. This trend exists for both cerium area densities (Figure 10a) and mass concentrations (Figure 10b), which were obtained by normalizing area density measurements by the final thicknesses of the PEMs after ASTs shown in Figure 4. These correlations imply a relationship between cerium migration and degradation.

Zoom In Zoom Out Reset image size

Cerium may initially partition from PEM into CLs due to concentration gradients and water/proton flux, as described in the Average through-thickness cerium migration during ASTs section. The reduced availability of radical scavengers in the PEM after migration could then lead to enhanced degradation in areas depleted of cerium. However, according to Figure 7b, the concentrations of cerium in the CLs significantly exceed expected values due to concentration gradients and water/proton flux, alone.

Therefore, an additional migration mechanism is proposed, whereby cerium is expelled from the PEM bound by a degradation product as explained in the Cerium migration due to side chain degradation section. In the PEM, the cerium to fluoride (Ce/F) ratio is ∼1/1000,² however, Figure 8 shows that the Ce/F ratio is reduced to ∼1/30,000 in the effluent water. Therefore, a large portion of the cerium mobilized by degradation is subsequently stabilized within the MEA, instead of being removed from the cell. We propose that cerium and/or cerium-exchanged ionomer degradation products interact with the surface of the carbon catalyst supports in the CLs.⁵⁴

In ex situ experiments, model fuel cell carbon compounds were mixed into cerium solutions at typical MEA concentrations in excess DI water. After filtering, it was observed that the Ce/C mass ratio in the filtrate is 0.043. This ratio is interpreted to signify that one mass unit of cerium is adsorbed on every 23 units of carbon. Cerium and carbon concentrations in the cathode CLs after ASTs (Figure 7a), correspond to Ce/C ratios between 0.029 and 0.043. Therefore, it is plausible that carbon catalyst supports have the capacity to adsorb the quantities of cerium measured in the CLs after ASTs. While the ex situ experiments do not fully replicate the chemistry of operating fuel cells, the results provide evidence for a possible mechanism for interaction between cerium and the carbon catalyst support particles. The nature of this mechanism is currently under investigation.