Degradation of Pt-Based Cathode Catalysts Upon Voltage Cycling in Single-Cell PEM Fuel Cells Under Air or N₂ at Different Relative Humidities

The here used 5 cm² MEAs were prepared via the decal transfer method. Cathode catalyst inks were made by mixing the 47.8 wt% Pt/C catalyst (TEC10V50E, Tanaka Kikinzoku Kogyo K.K.) with a 700 EW perfluorosulfonic acid (PFSA) ionomer (24.5 wt% ionomer in water, Asahi Kasei, Japan), ultra-pure water, and 1−propanol; anode catalyst inks were prepared analogously but using a 19.8 wt% Pt/C catalyst (TEC10V20E, Tanaka Kikinzoku Kogyo K.K., Japan). The as-prepared inks had a water content of ∼12 wt%, a carbon content of 30 mg_C mL⁻¹ _ink, and an ionomer to carbon weight ratio (I/C) of 0.65. The inks were added to a 15 ml HDPE bottle filled with 25.5 g of 5 mm ZrO₂ beads (Glenn Mills, USA) as grinding medium, and then placed onto a roller mill (60 rpm) for 18 h at room temperature. Afterwards, the inks were coated onto virgin PTFE sheets using a Meyer rod and after drying at room temperature, 5 cm² electrode decals were cut. MEAs were prepared by sandwiching a 15 μm reinforced PFSA membrane (700 EW, Asahi Kasei, Japan) between an anode and a cathode decal, and further hot-pressing this assembly at 155 °C for 5 min. The catalyst loadings were determined by weighing the decal before and after decal transfer; by considering the Pt content of the catalysts and the electrode's I/C ratio of 0.65, the resulting anode/cathode loadings are 0.1/0.4 ${\rm{m}}{{\rm{g}}}_{{\rm{P}}{\rm{t}}}\,{\rm{c}}{{\rm{m}}}_{{\rm{M}}{\rm{E}}{\rm{A}}}^{-2}$ (with a maximum loading variation of ±10%).

All fuel cell testing was performed with a fully automatized, custom-made G60 test station (Greenlight Innovation Corp., Canada), equipped with a Gamry 3000 potentiostat and a Gamry 30 K booster (Gamry Instruments, USA) for electrochemical impedance measurements (EIS) and voltage cycling experiments. The MEAs were sandwiched between commercial gas diffusion layers (GDLs) with a microporous layer (MPL) and a total nominal thickness of 170 μm (H14C7 from Freudenberg Performance Materials GmbH & Co. Kg, Germany), and then assembled in between custom-made 5 cm² active area graphite flow-fields (Poco Graphite, USA) with 7 parallel channels arranged in one serpentine (0.5 mm land/channel width and 0.8 mm channel depth; for details see Ref. 19). In the final cell assembly, the GDLs were compressed by 20 ± 1% of their initial thickness using PTFE-coated fiber glass hard-stop gaskets (experimentally shown to have a maximum compression of 7% ²⁰ ), applying a torque of 12 Nm to the bolts in the endplates (Fuel Cell Technologies, USA). Cells were connected to the test station with anode and cathode in counter-flow configuration.

The MEA testing sequence is depicted in Fig. 1. It starts with a voltage-controlled activation procedure under H₂/Air (1394/3323 nccm, referenced to 0 °C/1 atm) at 80 °C/100% RH and 150 kPa_abs (all cell pressures are given as absolute pressure at the cell inlet), conducting the following sequence 10 times: 45 min at 0.6 V, 5 min at 0.95 V, and 10 min at 0.85 V. Subsequently, a recovery step with H₂/Air at 0.3 V for 2 h (40 °C/100% RH, 270 kPa_abs) was conducted, followed by a series of diagnostics tests (cyclic voltammetry, CO-stripping voltammetry, H₂ crossover measurements), a quantification of the proton conduction resistance in the cathode electrode R_{H+, cath}), differential-flow polarization curves with H₂/O₂, H₂/Air, and H₂/10% O₂ (rest N₂), as well as cathode oxygen transport resistance measurements (via the limiting current density method). Parts of this sequence are repeated after each segment of the voltage cycling AST (i.e., after 100, 1k, 3k, 10k, 20k, and 30k cycles): (i) black boxes indicate the measurements that were conducted at BOT, after each AST segment, and at EOT (30k cycles), except for the activation that is only done when starting with a fresh MEA; (ii) yellow boxes indicate measurements conducted at BOT, after 10k and 20k cycles, and at EOT; (iii) the red box indicates measurements that were only conducted at BOT and EOT. The recovery step was performed prior to each AST diagnostic to eliminate all possible sources of reversible losses (e.g., membrane dry-out).

Zoom In Zoom Out Reset image size

Cyclic voltammograms (CVs) of the cathode electrode to determine the H-adsorption/desorption charge were recorded at a scan rate of 150 mV s⁻¹ between 0.05-0.95 V_RHE at 40 °C and ambient pressure (3 cycles), whereby the Pt ECSA was evaluated from the last of 3 subsequent CVs conducted between 0.05-0.55 V_RHE, averaging the H-adsorption/desorption charge and converting it into a Pt surface area using a specific charge of 210 μC cm⁻² _Pt. For the CVs, the anode was fed with 200 nccm of fully humidified 5% H₂ (rest N₂) in order to minimize the effect of crossover H₂, and the cathode was fed with 5 nccm of dry N₂.

CO-stripping voltammograms were recorded at 40 °C and 150 kPa_abs as described previously. ¹⁷ Initially, the cathode was fed with 100 nccm of dry 10% CO (rest N₂) while polarizing it to 0.1 V_RHE; the anode was fed with fully humidified 5% H₂ (rest N₂) to minimize H₂ crossover, which has been shown to add significant experimental artifacts to the data. ²¹ This was followed by a ∼1.5 h N₂ purge of the cathode (ramping between low and high flows and fully humidified conditions) to remove any residual CO; after this, the cathode potential was scanned to 0.95 V_RHE at 100 mV s⁻¹, followed by three subsequent scans (between 0.07 and 0.95 V_RHE) which were used as a baseline and to verify that the adsorbed CO had been completely oxidized. The integration between the first anodic scan and the second one (baseline) was used to obtain the Pt ECSA at fully humidified conditions, using a specific charge of 420 μC cm_Pt ⁻².

The H₂ crossover was measured as previously described in the literature. ²² The measurements were carried out at 80 °C with fully humidified H₂/N₂ mixtures at the anode (600 nccm) and with dry N₂ (150 nccm) at the cathode, initially setting anode/cathode pressures to 170 kPa_abs. Starting with 60% H₂ (rest N₂), the cell was left to equilibrate for 120 s; then the cell voltage was set to 0.35 V for 120 s, whereby the oxidative current due to H₂ crossover was averaged over the last 30 s. This measurement sequence was repeated with four more H₂ concentrations down to 10% H₂ (rest N₂). A linear fit of these five data points provides the H₂ crossover current density at 100% H₂ (from the slope) as well as the MEA shorting current density (i_short, from the y-axis intercept). As anode and cathode pressures were set to be equal, this corresponds to the diffusion-driven H₂ permeation, which we will denote as i_{H2, diff}. Repeating these experiments with the same anode pressure (i.e., 170 kPa_abs) but a decreased cathode pressure of 140 kPa_abs, the resulting H₂ crossover current density obtained from the slope corresponds to the sum of the diffusion- and convection-driven H₂ crossover current density (i_{H2, diff} + i_{H2, conv}). If the latter is the same as the value of i_{H2, diff}, an increase of the H₂ crossover over the course of an AST would indicate membrane thinning; on the other hand, if over the course of an AST the value of (i_{H2, diff} + i_{H2, conv}) becomes significantly larger than i_{H2, diff}, this would indicate the formation of membrane pinholes.

The proton conduction resistance across the cathode electrode (R_{H+, cath}) was determined from potentiostatic electrochemical impedance spectroscopy (PEIS) measurements recorded at 80 °C with H₂ and N₂ flows of 1000 nccm at anode and cathode, respectively, humidified at various RH values. The cell voltage was set to 0.2 V, and the impedance spectrum was acquired with a voltage perturbation of 3.5 mV rms, varying the frequency from either 500 kHz (operation with the Gamry 3000 potentiostat) or 300 kHz (operation with the potentiostat and the Gamry 30 K booster) and a low frequency of 0.2 Hz (recording 20 points per decade). The selected RH values were 30%, 50%, 70%, and 100%, and to ensure reproducibility three consecutive measurements were carried out at each RH; for these measurements, the cell pressure was adjusted to result in H₂/N₂ partial pressure of ∼220 kPa. The obtained Nyquist plots were fitted to a transmission line model (TLM) as described in the literature. ²³

Differential-flow polarization curves were recorded in current-controlled mode at 80 °C/100% RH and 170 kPa_abs. Constant flows of 2000 nccm of H₂ on the anode were used, while 5000 nccm of either O₂, air, or 10% O₂ (rest N₂) where used on the cathode (at this operating conditions, the maximum pressure drop across anode and cathode were 2 kPa and 22 kPa, respectively). Prior to recording a polarization curve, the cell potential was set to 0.75 V for 15 min; followed by a sequence of current-controlled steps from low to high current density. Each of the applied currents was held for 5 min, and the measured cell voltages were averaged over the last 30 s; the OCV was recorded at the end of this sequence. The ORR kinetic parameters of the cathode catalyst (namely, mass and specific activity) were determined from H₂/O₂ polarization curves, corrected by: (i) the ohmic resistance loss (i.e., i_geo · R_HFR), with R_HFR obtained from galvanostatic electrochemical impedance spectroscopy (GEIS) measurements at each current step (from 100 kHz to 10 Hz, with a current perturbation of 10% of the DC current, restricted to min/max values of 0.1/3.0 A); (ii) the effective cathode proton conduction resistance (${R}_{\text{H}+,\,\text{cath}}^{\text{eff}}$) according to the method presented by Neyerlin et al., ²⁴ with the R_{H+, cath} value [in ${\rm{m}}{\rm{\Omega }}\,{\rm{c}}{{\rm{m}}}_{{\rm{M}}{\rm{E}}{\rm{A}}}^{2}$] obtained from the above described PEIS measurements corrected by a current density-dependent utilization factor.

Limiting current density measurements were performed to quantify the O₂ transport resistance (${R}_{{\text{O}}_{2}}^{\text{total}}$) in the cathode electrode, as described in the literature. ^17,20,25 For this, voltage-controlled polarization curves were recorded at 80 °C/100% RH and differential flows of 2000 nccm of H₂ at the anode and 5000 nccm of O₂/N₂ mixtures at the cathode, whereby the dry mole fraction of O₂ was varied from 0.5% to 24%. Limiting current densities were determined at cell voltages of 0.3 V, 0.15 V, 0.10 V, and 0.05 V (each held for 2 min). To separate the pressure-dependent (${R}_{{\text{O}}_{2}}^{\text{PD}}$) and the pressure-independent (${R}_{{\text{O}}_{2}}^{\text{PI}}$) components of ${R}_{{\text{O}}_{2}}^{\text{total}},$ the polarization curves were measured at 170, 270, 350,and 500 kPa_abs. ²⁰

The voltage cycling experiments conducted in this work were conducted either with H₂/N₂ (anode/cathode) or with H₂/Air (anode/cathode) flows, which will further on be referred to as H₂/N₂ V-cycling and H₂/Air V-cycling, respectively. As only very low currents are flowing during the voltage cycling under H₂/N₂ (i.e., capacitive and H₂ crossover currents), the applied cell voltage is essentially identical with the cathode potential measured versusthe reversible hydrogen electrode (RHE) potential, and temperature/RH are constant along (i.e., in-plane) and normal to the MEA (i.e., through-plane, across the electrode thickness). In the here applied H₂/N₂ V-cycling protocol, the cell voltage (and thus the cathode potential) was modulated according to a square-wave profile, with an UPL of 1.0 V_RHE and a LPL of 0.6 V_RHE, whereby the dwell times at each potential were set to 8 s and the potential transient was less than 0.1 s (i.e., >4 V s⁻¹). H₂/N₂ flows of 200/75 nccm (anode/cathode) were used during the voltage cycling segments of the overall test procedure (see Fig. 1), with the cell being operated at ambient pressure and 80 °C, selecting different gas inlet RH values (30, 50, 70, and 100%).

In order to compare the Pt ECSA loss and the MEA performance degradation when conducting voltage cycling experiments under either H₂/N₂ or H₂/Air at 80 °C and a given cell RH, obviously the same voltage profile must be applied. While this is easily achieved with regards to the profile shape/duration, the question about the proper correspondence between the UPL and LPL values of the cell voltage (E_cell) versus that of the cathode potential (E_cath) must be addressed. In contrast to H₂/N₂ V-cycling, the current densities at the lower potential limit are orders of magnitude larger during H₂/Air V-cycling, so that the cathode potential in this case is higher than the cell voltage at the LPL, to a good approximation described by the HFR-corrected cell voltage (E_{cell, HFR−free}), ignoring the potential drop across the cathode electrode:

$$\begin{eqnarray}&&{E}_{\mathrm{cell},\,\mathrm{HFR}-\mathrm{free}}={E}_{\mathrm{cell}}+{i}_{\mathrm{geo}}{\rm{\cdot }}{R}_{\mathrm{HFR}}{\rm{\approx }}{E}_{\mathrm{cath}}\end{eqnarray} \tag{ 1 }$$

Since the Pt ECSA loss during voltage cycling is a strong function of the cathode potential rather than the cell voltage, a meaningful comparison between the Pt ECSA loss over the course of H₂/N₂ versus H₂/Air V-cycling ASTs requires that the cell voltage must be continuously adapted to maintain a constant cathode potential, particularly at the high current density at the LPL in H₂/Air V-cycling. In order to achieve this, a potentiostatic electrochemical impedance spectroscopy (PEIS) analysis was carried out at the beginning and at the end of each AST sequence at the same operating conditions at which the H₂/Air V-cycling was conducted, determining the R_HFR value at a cell voltage of 0.6 V (with a voltage perturbation of 5 mV rms, conducted between 1-300 kHz and with 20 points per decade). From the obtained Nyquist plots, R_HFR was extracted from the x-intercepts. The R_HFR value obtained before starting a new AST segment, multiplied by a factor between 0.85 and 0.95 (i.e., 85% to 95% iR-compensation) was used for the positive feedback iR-compensation control mode of the Gamry potentiostat (Gamry Frameworks 3.23). This way, the lower potential limit of the cell voltage was continuously adapted, so that the desired cathode LPL of 0.6 V_RHE (i.e., E_{cell, HFR−free} ≈ 0.6 V) could be closely maintained throughout the AST.

Using the above approach, H₂/Air V-cycling ASTs at 80 °C were conducted with cathode UPL and LPL values of 1.0 V_RHE and very close to 0.6 V_RHE, respectively (square-wave, with 8 s at each potential, with >4 V s⁻¹ transients), controlled via a Gamry 30 K Booster. The gas flow rates were also set to 2000 nccm of H₂ and 5000 nccm of air, corresponding to a minimum stoichiometry of 18 for the largest measured current density, so that RH gradients along (in-plane) the MEA are negligible; the inlet RH values of the gas feeds were either 50, 70, or 100%.

The consequence of the iR-compensation mode on the cell voltage and the calculated HFR-free cell voltage (acc. to Eq. 1), and roughly corresponding to E_cath) for H₂/Air V-cycling ASTs conducted at 100% and 70% RH are shown in Figs. 2a and 2b, respectively. Here, the green lines depict the actually applied cell voltages (E_cell) set by the iR-compensation mode, the orange lines depict the resulting current densities (i_geo), and the red lines depict the calculated HFR-free cell voltage (E_{cell, HFR−free}) determined by Eq. 1. Clearly, the iR-compensation makes no difference in the UPL values due to the essentially zero currents. On the contrary, large variations between E_{cell, HFR−free} (red line, Fig. 2) and E_cell (green line, Fig. 2) are seen at LPL. However, as desired, the value of E_{cell, HFR−free} that roughly corresponds to E_cath remains very close to the desired cathode LPL of 0.6 V_RHE at both 100% RH (within ∼10 mV, see Fig. 2a) and 70% RH (within ∼25 mV, see Fig. 2b); this difference becomes smaller towards the end of the 30k voltage cycles due to the decrease in current density over the AST. This demonstrates that the selected iR-compensation mode achieves a reasonably close agreement between the desired and the actual cathode LPL. This is critical when trying to compare H₂/N₂ V- versus H₂/Air V-cycling, since it has been shown that an increase in the LPL leads to a decrease in the Pt ECSA loss during voltage cycling, particularly if it approaches the onset potential for platinum oxide formation (above 0.7 V_RHE). ²⁶ Furthermore, Stariha et al. have determined differences arising from a LPL value that is not constant over the course of a cycling AST under H₂/Air voltage cycling AST. ²⁷

Zoom In Zoom Out Reset image size

In summary, using the iR-compensation mode, the LPL of the cathode electrode can be closely controlled to 0.6 V_RHE (within ∼20-30 mV), so that the here conducted H₂/N₂ and H₂/Air V-cycling ASTs have very similar cathode LPL values (owing to the negligible current at 1.0 V, the cathode UPL values are identical). In addition, the high stoichiometric flows during our H₂/Air V-cycling ASTs ensure the absence of temperature and RH gradients along (in-plane) the MEA, which have previously been shown to result in inhomogeneous degradation. ²⁸ Therefore, any differences arising between these tests would be due to differences that arise from membrane degradation and/or cyclic variations of the local temperature/RH at/across the cathode electrode.

Pristine and aged MEA cross-sections were prepared with a cross section polisher (Jeol IB-19530CP, Japan). For this, the GDLs were carefully removed from the aged MEAs and small sections of an MEA were punched out and placed between two copper foils. This assembly was then milled for a total of 4 h (3 h at an acceleration voltage of 6 kV and 1 h at 4 kV) at −90 °C to reduce MEA damage and to improve spatial resolution. The SEM images were taken with a JEOL SEM (JSMIT 200, Japan) at an acceleration voltage of 15 kV with an SE (secondary electron) detector. From each MEA, two samples were taken, one at the anode inlet and one at the anode outlet, and from each sample five images were obtained. The thicknesses of the catalyst layers and of the membrane were determined using imageJ, whereby the areas of selected segments of the electrodes and the membrane were determined by integration of the images and then divided by their respective lengths, yielding an average thickness and its standard deviation. One MEA per AST was taken, together with a BOT sample that was subjected to a modified activation procedure (i.e., assembling the MEA in a cell, and then holding it at 80 °C/100% RH under 1000 nccm N₂ flows anode and cathode at ambient pressure for 24 h).

Figure 3 shows the Pt ECSA versus the cycle number of V-cycling ASTs conducted at 80 °C as a function of inlet RH and cathode gas composition. Here, the Pt ECSA at BOT, after 10k and 20k cycles, and at EOT (30k cycles) was determined from CO-stripping voltammetry (ECSA_CO), while at 100, 1k, and 3k cycles the ECSA was obtained from the CV-based H-adsorption/desorption area (ECSA_H-UPD) multiplied by the BOT ratio of ECSA_CO/ECSA_H-UPD (∼1.1).

Zoom In Zoom Out Reset image size

As observed in the H₂/Air V-cycling study by Kneer et al. ¹⁵ and in the H₂/N₂ V-cycling study by Harzer et al., ¹⁷ Fig. 3 shows an initially rather rapid ECSA loss over the early stages of the AST, amounting to ∼20% over the first 100 cycles, independent of the RH during the voltage cycling AST. This fast initial degradation is generally rationalized by considering that very small Pt nanoparticles are particularly prone to dissolution and coalescence, so that they are lost in the early stages of voltage cycling, concomitant with a loss of Pt ECSA. This is consistent with the observation by Makharia et al. ³ that heat-treated Pt/C catalysts, exhibiting larger Pt particle sizes, have a reduced ECSA loss rate when normalized to the catalyst's ECSA at BOT. In a later study by Shao-Horn et al., ¹⁴ the authors emphasized that the particle size distribution has to be considered when analyzing the voltage cycling stability of Pt/C catalysts in order to account for the effect of smaller Pt nanoparticles.

While the effect of RH is relatively minor over the first 100 cycles, over the further course of the V-cycling ASTs the ECSA loss is the stronger the higher the RH, which had also been observed in the H₂/Air V-cycling experiments by Kneer et al. ¹⁵ and in the H₂/N₂ and H₂/Air V-cycling experiments by Bi et al. ²⁹ Consistent with the latter study and with that by Uchimura et al., ³⁰ no differences in ECSA loss between H₂/N₂ and H₂/Air V-cycling are observed in our study (compare open and solid symbols in Fig. 3). Here, it should be mentioned that no complete H₂/Air V-cycling test could be conducted at 50% RH due to the premature formation of membrane pinholes, which we ascribe to the cyclic mechanical stress of the membrane caused by the local RH differences between the UPL (quasi zero current) and at the LPL (very high current densities, see Fig. 2) during voltage cycling under H₂/Air ²² (further discussed in the next section). The identical ECSA loss when cycling under H₂/N₂ and H₂/Air suggests that for reasonably well-controlled cathode UPL and LPL values, the current density during cycling (low in the former and high in the latter) does not play a significant role, at least as far as the ECSA loss is concerned.

As stated above, the higher ECSA retention at low RH has already been reported in several studies. ^15,26,29,31 In general, the lower ECSA loss at low RH has been ascribed to two effects:

• (1)  
  The low water activity (${a}_{{\text{H}}_{2}\text{O}}\equiv {p}_{\text{H}2\text{O}}/{p}_{\text{H}2\text{O}}^{\text{sat}}\,\equiv \,\text{RH}$) at low RH shifts the onset of the platinum oxide formation to higher potentials, ³² so that the change of platinum oxide coverage between a given LPL and UPL, considered to trigger platinum dissolution, is less at low RH. The effect of RH on the platinum oxide coverage in the context of voltage-cycling ASTs has been analyzed by Bi et al., ²⁹ who however concluded that it is minor compared to the effect of RH on platinum diffusivity.
• (2)  
  At low RH, the water content of the ionomer in the cathode catalyst layer is reduced, which substantially decreases the diffusivity of dissolved platinum ions, so that the loss of platinum into the membrane phase (via the reaction with crossover H₂) is reduced. The latter leads to the well-known Pt-band formation in the membrane, which is observed after voltage cycling ASTs ^1,2,29 and extended OCV holds. ³³ Values for the Pt mass loss from the cathode electrode into the membrane phase over extended V-cycling at 100% RH were reported in several studies: ∼20%-25% after 10k voltage cycles (0.6-1.0 V) under H₂/N₂ at 80 °C (for an overall ECSA loss of ∼60%), ¹ ∼15% after 30k voltage cycles (0.6-0.95 V) under H₂/N₂ at 80 °C (for an overall ECSA loss of ∼75%), ² and ∼30% after 3.3k voltage cycles (0.87-1.2 V) under H₂/Air at 60 °C (for an overall ECSA loss of ∼75%). ²⁹ The latter study also examined the Pt mass loss into the membrane phase at a lower RH of 50%, where it was reduced to ∼10% (3.3k voltage cycles (0.87-1.2 V) under H₂/Air at 60 °C, with an overall ECSA loss of ∼50%), indicating the importance of RH on this mechanism.

To our knowledge, there are only two publications with experimental values for the platinum diffusion coefficient (D_Pt) in a PFSA ionomer, namely for a Nafion^® membrane equilibrated in liquid water/electrolyte at room temperature: 0.14 × 10⁻¹¹ m² s^{−1 34} (measured for (PtCl₆)²⁻ ions) and 4 × 10⁻¹¹ m² s^{−1 35} (determined by online mass spectrometry). As outlined in our previous work, ³⁶ these can be used to estimate the effective Pt diffusion coefficients (D_Pt,eff) in the ionomer phase of the cathode electrode by considering its ionomer volume fraction of ∼20 vol% and the ∼5-fold increase of cationic diffusion coefficients from 25 to 80 °C (measured for Ce³⁺ and Co²⁺ ions), an estimate that yields similar values of (0.14-4) × 10⁻¹¹ m² s⁻¹ (i.e., the volume fraction and the temperature gain factor roughly compensate each other). For the here used 8 s hold time (t_hold) at UPL/LPL, these D_Pt,eff values would result in a diffusion length (δ) of ∼3-18 μm (based on δ ≈ (D_Pt,eff × t_hold)^0.5), which is on the same order as the cathode electrode thickness of ∼9 μm (determined from the later shown SEM cross-sections). This suggests that a significant fraction of the platinum ions released during voltage cycling at 80 °C/100% RH would in principle be able to reach the membrane phase, where they could redeposit via reaction with crossover H₂. However, as Pt ions diffusing towards the membrane can also deposit within the cathode electrode on larger Pt particles on the carbon support via Ostwald ripening, only a fraction of the dissolved Pt ions are expected to reach the membrane phase where they would redeposit, as suggested by the above discussed studies, which report an overall loss of Pt into the membrane phase of only ∼15%-30% at 100% RH. ^1,2,29

While there are no data on the platinum diffusion coefficient at low RH, Coms and LaLonde reported a ∼36-fold decrease of the diffusion coefficient of Ce³⁺ in a Nafion^® NR211 membrane between the liquid water saturated membrane and the membrane at 40% RH ³⁷ a factor which may serve as a rough estimate also for the platinum ion diffusion coefficient. Applying this factor to the above range of D_Pt,eff values at 80 °C/100% RH, the diffusion length of the platinum ions during the 8 s hold time would equate to δ ≈ 0.6-3 μm at 80 °C/40% RH, which indeed would indicate that the mechanism of platinum redeposition in the membrane phase would become less effective at very low RH values (no data are shown for lower RH values, but the decrease of the Ce³⁺ diffusion coefficient with decreasing RH becomes the stronger the lower the RH). This suggests that the significantly lower ECSA loss observed during voltage cycling at 30% RH (see blue symbols in Fig. 3) is at least in part due to the much reduced effective platinum diffusion coefficient, which substantially reduces the Pt loss into the membrane phase, in agreement with the modeling-based conclusions drawn by Bi et al. ²⁹ That this assumption is reasonable can be seen by comparing the ECSA at EOT for V-cycling under H₂/N₂ at 100% RH ($\sim 5\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$; red solid symbols) with that at 30% RH ($\sim 21\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$; blue solid symbols) shown in Fig. 3. If the Pt loss into the membrane phase at 100% RH would amount to ∼15%-30% as reported in the literature, ^1,2,29 this would correspond to $\sim 8-16\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$ based on an initial ECSA of $\sim 55\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1},$ so that a complete suppression of Pt diffusion into the membrane phase at 30% RH would suggest an ECSA loss at 30% RH of $\sim 13-21\,{{\rm{m}}}_{{\rm{Pt}}}^{2}\,{{\rm{g}}}_{{\rm{Pt}}}^{-1},$ which is reasonably close to the observed value of $\sim 21\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$ shown in Fig. 3. In summary, while we do not claim that the diffusion of Pt ions into the membrane phase is completely suppressed at 30% RH, this is likely a significant contribution to the substantially lower ECSA loss at 30% versus 100% RH.

One aspect to discuss in view of the data shown in Fig. 3 is the essentially identical ECSA loss for H₂/N₂ and H₂/Air V-cycling (solid and open symbols in Fig. 3). For our experimental procedure, the cathode UPL/LPL values are identical at the UPL and very similar at the LPL, while the RH gradient along the flow-field is at most 10% during the H₂/Air V-cycling due to the high reactant stoichiometry (calculated for an air stoichiometry at the highest measured current density and the pressure drop in the flow-field). However, in the case of H₂/Air V-cycling, the temperature gradient between the MEA and the flow-field is significant at high current densities, and the resulting higher MEA temperature would lead to a lower RH at the MEA compared to the reactant RH in the flow-field. In fact, it has been estimated on the basis of thermal conductivity data for a similar Freudenberg GDL as used here that the MEA temperature is ∼6 °C higher than the flow-field temperature at a current density of ∼3.0 ${\rm{A}}\,{\rm{c}}{{\rm{m}}}_{{\rm{M}}{\rm{E}}{\rm{A}}}^{-2}$ and 0.6 V; ³⁸ this would translate into a local RH at the membrane of ∼80% RH at the LPL for the H₂/Air V-cycling at 80 °C/100% RH and of 40% RH at the LPL for the H₂/Air V-cycling at 80 °C/50% RH. In this case, the local RH at the membrane between UPL and LPL would vary by ∼20% in the former case (i.e., between ∼100 and ∼80% RH) and by ∼10% in the latter case (i.e., between ∼50 and ∼40% RH). While these RH fluctuations are not very large, they are the likely explanation for the perhaps slightly lower ECSA losses of the H₂/Air versus the H₂/N₂ V-cycling ASTs, particularly for the AST conducted at 100% RH (open versus solid symbols in Fig. 3).

While the ECSA loss is essentially identical for the H₂/N₂ and H₂/Air V-cycling ASTs, operation under H₂/Air might additionally lead to the chemical degradation of the membrane and the ionomer in the cathode catalyst layer, ¹³ while the above discussed local RH variation at the membrane between low and high current density (at the UPL and LPL, respectively) may also lead to mechanical membrane degradation. ²² This will be examined in the following, conducting H₂ crossover measurements as well as measurements of the proton conduction resistance of the cathode electrode (R_{H+, cath}) over the course of the V-cycling ASTs.

The H₂ crossover current densities were measured at two different conditions at 80 °C/100% RH: (i) at equal anode and cathode pressures of 170 kPa_abs to determine the diffusion-controlled H₂ crossover current density through the membrane (i_{H2, diff}), which would increase with decreasing membrane thickness and would thus be a measure for membrane thinning due to chemical degradation (shown in Fig. 4a); (ii) at a reduced cathode pressure, so that the anode pressure is higher than the cathode pressure (170 versus 140 kPa_abs, respectively; shown in Fig. 4b), which in case of membrane pinholes would lead to an increased H₂ crossover current density due to an additional convection-driven H₂ flux (i_{H2, diff} + i_{H2, conv}). For the H₂/N₂ V-cycling ASTs, the i_{H2, diff} values remain constant over 30k cycles (see solid symbols in Fig. 4a) and the H₂ crossover current densities also remain at the same value when the cathode pressure is reduced (data not shown), indicating that there is neither membrane thinning nor pinhole formation over the course of the H₂/N₂ V-cycling ASTs, irrespective of the RH condition. The same behavior is observed for the H₂/Air V-cycling AST conducted at 100% RH (see open red symbols in Fig. 4a). However, for the H₂/Air V-cycling ASTs conducted at 70 and 50% RH, the i_{H2, diff} values increase substantially after 30k and 20k cycles, respectively (open yellow and green symbols in Fig. 4a), accompanied by an even stronger increase of the i_{H2, diff} + i_{H2, conv} values (see Fig. 4b). This suggests that in addition to a possible thinning caused by chemical degradation, the membranes have also formed pinholes at these points in the ASTs (note that in the case of significant pinhole formation, measurements at nominally identical anode and cathode pressure can also show an increased H₂ crossover, even in the absence of an overall membrane thinning). This must be due to the fact that the chemical degradation of PFSA membranes strongly increases with decreasing RH, ^39,40 so that the combination of chemical degradation and mechanical stress due to the above discussed local RH cycles leads to membrane pinhole formation. For this reason, the H₂/Air V-cycling AST at 50% RH had to be terminated after 20k cycles due to excessive H₂ crossover.

Zoom In Zoom Out Reset image size

In order to evaluate whether there has also been any degradation of the ionomer in the cathode catalyst layer and an associated increase of its proton conduction resistance (R_{H+, cath}) after the various V-cycling ASTs, the values of R_{H+, cath} were determined at BOT as well as at EOT (i.e., after the 30k voltage cycles shown in Fig. 3). Figure 5 shows the R_{H+, cath} values obtained at 80 °C/50% RH (hatched bars) and 80 °C/100% RH (solid bars). Compared to the BOT R_{H+, cath} values under these conditions (gray bars in Fig. 5, no significant changes can be observed after 30k H₂/N₂ V-cycling ASTs (EOT) conducted at 30%-100% RH (colored bars, middle segment). The same can be observed for the EOT R_{H+, cath} values after the H₂/Air V-cycling ASTs, even though previous reports suggest that extended operation under H₂/Air can lead to ionomer morphological changes ⁴¹ and dehydration, ⁴² and that dry OCV periods can induce ionomer chemical degradation. ¹³ Therefore, an irreversible mechanical/chemical degradation of the ionomer in the cathode catalyst layers over the course of the here conducted voltage cycling ASTs can be excluded, even though ∼65%-93% of the ECSA were lost (see Fig. 3); a constant R_{H+, cath} over similarly large ECSA losses during H₂/N₂ V-cycling ASTs was also observed by Della Bella et al. ³⁶ For the present study, however, it should be noted that the R_{H+, cath} measurements were carried out after a recovery step (see Fig. 1), designed to recover reversible degradation effects, ^43,44 which may mask the possible presence of reversible R_{H+, cath} changes (e.g., by ionomer dehydration). Nevertheless, the data shown in Fig. 5 indicate that no irreversible losses in the cathode proton conduction resistance have occurred over the here conducted H₂/N₂ and H₂/Air V-cycling ASTs, even under very low RH conditions.

Zoom In Zoom Out Reset image size

In order to determine whether the above discussed increase of the H₂ crossover currents over the course of the H₂/Air V-cycling ASTs is accompanied by membrane thinning, cross-sectional SEM images of an MEA at BOT and of MEAs after 30k H₂/Air V-cycles (EOT) at either 80 °C/70% RH or 80 °C/100% RH were analyzed. Furthermore, the images were analyzed to determine whether there was any thinning of the cathode electrodes due to carbon-support corrosion at the high cathode UPL of 1.0 V_RHE. While in our previous work for identically conducted V-cycling tests under H₂/N₂ between 0.6-1.0 V_RHE at 80 °C/100% RH no cathode thinning was observed after 30 cycles, ¹⁷ voltage cycling under H₂/Air with UPLs of 0.95 V or less were reported to exhibit significant carbon-support corrosion and noticeable cathode thinning. ⁴⁵

Figure 6a shows an SEM cross-section image of a BOT MEA subjected to a modified conditioning sequence (see Experimental section), while Figs. 6b and 6c show images of MEAs after 30k H₂/Air V-cycles at either 80 °C/70% RH or 80 °C/100% RH, respectively. Here, the centrally located membrane reinforcement (bright feature), the extension of the membrane above/below the reinforcement (dark gray regions), and the cathode (top) and anode (bottom) cathode catalyst layers can be clearly distinguished. The average thicknesses of the cathode, anode, and the membrane determined from these and additional images are presented in Table I. This analysis shows that there is no thinning of the cathode catalyst layer; i.e., that 30k voltage cycles to an UPL of 1.0 V_RHE are not accompanied by a significant carbon-support corrosion even under H₂/Air (as mentioned above, none was observed for an identical AST protocol under H₂/N₂ ¹⁷ ), suggesting that the presence of oxygen does not significantly enhance the carbon oxidation kinetics. Table I also shows that no membrane thinning occurs over the course of the H₂/Air V-cycling ASTs (note that the membrane thickness of ∼13 μm, measured in vacuum, is less than the 15 μm specification from the supplier, measured at ambient air, which is due to the differences in water content). Therefore, the observed increase of the i_{H2, diff} values of the MEAs cycled under H₂/Air at 80 °C/70% RH (yellow open symbols in Fig. 4a) must be caused entirely by the presence of membrane pinholes.

Zoom In Zoom Out Reset image size

Another feature that can be observed in the cross-sectional SEM images is the presence of a Pt-band close to the cathode/membrane interface. This has been previously observed in experiments for MEAs subjected to extended OCV holds at different H₂/O₂ concentrations at anode/cathode ³³ or to extended voltage cycling (under H₂/Air, H₂/O₂, or H₂/N₂). ^1,29 Zhang et al. showed that the location of the Pt-band depends predominantly on the partial pressures of H₂ and O₂ in the cell, which in the case of H₂/Air operation resulted in a Pt-band in the membrane that was located ∼20% of the membrane thickness away from the cathode electrode, ³³ consistent with the Pt-band location observed in Figs. 6b and 6c.

When more closely inspecting the cathode catalyst layer in the SEM images, one can discern the presence of a "platinum depletion zone" near the membrane/cathode interface for the MEA cycled at 80 °C/70% RH (Fig. 6b), indicated by the gray-scale gradient in the electrode (brighter regions corresponding a higher platinum density) that is absent in the image of the BOT MEA (Fig. 6a). This suggests that near the membrane/cathode interface the platinum loss by Pt ion diffusion and redeposition in the membrane is a dominant loss mechanism, as first shown by Ferreira et al., ¹ leading to a Pt depletion zone in the cathode electrode that was also observed in H₂/N₂ V-cycling ASTs by Della Bella et al. ³⁶ and modelled for V-cycling ASTs by Jahnke et al. ⁴⁶ The extension of the Pt depletion zone further into the cathode electrode can be noted in the SEM image of the MEA cycled at 80 °C/100% RH (Fig. 6c), which we ascribe to the enhanced Pt ion diffusivity at high RH (see above discussion). In summary, the observations with regards to the Pt distribution across the MEA, to carbon-support corrosion, and to membrane thinning for MEAs subjected to H₂/Air V-cycling are very similar to those reported for MEAs subjected to H₂/N₂ V-cycling under otherwise identical conditions. ^17,36

Differential-flow H₂/O₂ performance curves at 80 °C/100% RH and 170 kPa_abs for BOT MEAs and MEAs after 30k (EOT) H₂/N₂ V-cycling ASTs are depicted in Fig. 7a. The BOT performance (black symbols/line) of all tested MEAs is very consistent, showing a maximum standard deviation of ±8 mV at 2.0 A cm⁻² _MEA (based on 14 independently measured MEAs). At EOT, the H₂/O₂ performance follows the same trend as the absolute ECSA value at EOT (see Fig. 3), with the poorest performance for the MEAs that had been subjected to H₂/N₂ V-cycling at 80 °C/100% RH (red symbols/line) and have a remaining ECSA of only $\sim 4\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$.

Zoom In Zoom Out Reset image size

In order to examine the relationship between the H₂/O₂ performance and the cathode ECSA, we conducted the analysis approach presented by Owejan et al. ⁴⁷ and Harzer et al., ¹⁷ assuming negligible overpotentials for the hydrogen oxidation reaction (HOR) and negligible O₂ transport resistance losses. In this case, the cell voltage corrected by the voltage losses caused by the ohmic resistance (i_geo · R_HFR) and by the effective proton conduction resistance in the cathode catalyst layer (i_geo ${\rm{\cdot }}{R}_{\text{H}+,\text{cath}}^{\text{eff}},$ acc. to Neyerlin et al. ²⁴ ) should be proportional to the ORR overpotential described by simple Tafel kinetics (see Neyerlin et al. ⁴⁸ ). The thus corrected cell voltage (${E}_{\left({\text{R}}_{\text{HFR}}+{{\rm{R}}}_{{\rm{H}}+,\,\mathrm{cath}}^{\mathrm{eff}}\right)-\mathrm{free}}$) should then be proportional to the logarithm of the geometric current density divided by the Pt loading (${L}_{\text{Pt}}$) and the Pt ECSA, which corresponds to the specific, i.e., Pt surface area normalized current density (${i}_{\text{spec}}$):

$$\begin{eqnarray}&&\begin{array}{c}{E}_{\left({{\rm{R}}}_{\mathrm{HFR}}{+{\rm{R}}}_{{\rm{H}}+,\,\mathrm{cath}}^{\mathrm{eff}}\right)-\mathrm{free}}\equiv {E}_{\mathrm{cell}}+{i}_{\mathrm{geo}}\cdot {{\rm{R}}}_{\mathrm{HFR}}\\ +\,{i}_{\mathrm{geo}}\cdot {R}_{{\rm{H}}+,\,\mathrm{cath}}^{\mathrm{eff}}\propto TS\cdot \mathrm{log}\left[\frac{{i}_{\mathrm{geo}}}{{i}_{0}\cdot {L}_{\mathrm{Pt}}\cdot ECSA}\right]\\ \propto \,TS\cdot \mathrm{log}\left[\frac{1}{{i}_{0}}\cdot {i}_{\mathrm{spec}}\right]\end{array}\end{eqnarray} \tag{ 2 }$$

Here, TS is the ORR Tafel slope (in mV dec⁻¹), ${i}_{0}$ is the ORR exchange current density (in ${\rm{A}}\,{\mathrm{cm}}_{\mathrm{Pt}}^{-2}$), and the term ${L}_{\text{Pt}}{\rm{\cdot }}{ECSA}$ corresponds to the cathode electrode roughness factor (rf, which generally is given in units of ${\rm{c}}{{\rm{m}}}_{{\rm{P}}{\rm{t}}}^{2}\,{\rm{c}}{{\rm{m}}}_{{\rm{M}}{\rm{E}}{\rm{A}}}^{-2}$; note that the geometric current density has to also be corrected for the H₂ crossover current density (i_{H2, diff}) and the shorting current density (i_short).

The analysis according to Eq. 2 is shown in Fig. 7b, plotting ${E}_{\left({\text{R}}_{\text{HFR}}+{\text{R}}_{\text{H}+,\text{cath}}^{\text{eff}}\right)-\text{free}}$ versus the logarithm of ${i}_{\text{spec}}.$ The data for the BOT MEAs (black symbols/line) closely follow a Tafel behavior, with an apparent Tafel slope of ∼78 ± 1 mV dec⁻¹, close to the theoretical slope of 70 mV dec⁻¹ (see gray line). For the lower specific current densities, this is also true for the EOT MEAs that had undergone 30k cycles of the H₂/N₂ V-cycling AST, even though they are offset by ∼20 mV to higher potentials, which we ascribe to an increased exchange current density of larger (i.e., Ostwald ripened) Pt particles. At higher specific current densities, starting at ∼0.05 ${\rm{A}}\,{\rm{c}}{{\rm{m}}}_{{\rm{P}}{\rm{t}}}^{-2}$, the EOT MEA data in Fig. 7b deviate from Tafel behavior. The same phenomenon, setting in at essentially the same specific current density, has also been reported for the BOT performance of MEAs with varying cathode Pt loadings for a given catalyst ⁴⁷ (thus varying electrode rf while keeping the same catalyst ECSA) as well as for H₂/N₂ V-cycling ASTs at 80 °C/100% RH with different voltage profiles and different UPL/LPL values (thus varying electrode rf and catalyst ECSA). ¹⁷ This deviation from Tafel behavior has been ascribed to the so-called Pt-specific oxygen transport resistances which become the more pronounced the lower the electrode roughness factor becomes. However, the striking observation from Fig. 7b is that the data of all EOT MEAs fall onto the same "master" curve, indicating that the H₂/O₂ performance loss seen in Fig. 7a only depends on the specific current density and thus on the cathode ECSA, no matter at which RH condition the H₂/N₂ V-cycling ASTs were conducted.

As shown above, the only difference between H₂/N₂ and H₂/Air V-cycling ASTs so far has been that the latter lead to membrane pinhole formation (Fig. 4a), while for both AST types the ECSA loss is essentially identical for a given RH condition (see Fig. 3) and R_{H+, cath} remains constant over all ASTs (see Fig. 5). This suggests that also the H₂/O₂ performance might be unaffected by the AST type, which indeed is more or less the case, as shown in Fig. 8a for the EOT MEAs after H₂/N₂ and H₂/Air V-cycling ASTs conducted at both 80 °C/70% RH (yellow symbols/lines) and 80 °C/100% RH (red symbols/lines). The ∼25 mV lower EOT performance at 2.0 ${\rm{A}}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$ of the MEAs subjected to the H₂/Air V-cycling AST at 80 °C/100% RH compared to those cycled under H₂/N₂ is likely within the error of the measurement for such strongly degraded MEAs (only $\sim 4\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$ remaining ECSA, see Fig. 3), where the deviation between two repeat measurements is up to ∼30 mV (corresponding to the width of the error bars). This is supported by the fact that one would also expect less chemical membrane/ionomer degradation in H₂/Air at 100% RH compared to 70% RH, whereby the H₂/O₂ performance curves for both the H₂/N₂ and the H₂/Air V-cycling ASTs are identical. However, it cannot be excluded that there are minor differences in the Pt distribution across the cathode electrode between the two types of ASTs. As a matter of fact, one would expect a wider Pt depletion zone in the H₂/N₂ case where the Pt-band is at the very membrane/cathode interface, ^1,2,29,33 which should actually lead to a larger ECSA loss (as the Pt-band location marks the point where the crossover H₂ concentration is expected to be zero, Pt ions need to diffuse a shorter distance in the H₂/N₂ case), which is not observed within the error of the measurement (see Fig. 3). The Tafel analysis in Fig. 8b (i.e., acc. to Eq. 2) leads to the same conclusions, namely that the H₂/O₂ performance of the EOT MEAs is essentially independent of whether the MEAs were subjected to H₂/N₂ and H₂/Air V-cycling ASTs and that the H₂/O₂ performance depends predominantly on the cathode ECSA. In this case, however, ${E}_{\left({\text{R}}_{\text{HFR}}+{\text{R}}_{\text{H}+,\,\text{cath}}^{\text{eff}}\right)-\text{free}}$ at 2.0 ${\rm{A}}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$ is now ∼50 mV lower for the EOT MEAs cycled under H₂/Air at 80 °C/100% RH compared to those cycled under H₂/N₂, which we ascribe to the inaccuracy of the ECSA measurements (thus an inaccuracy of the value of ${i}_{\text{spec}}$) when the ECSA is only $\sim 4\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$, even when using CO-stripping voltammetry. ³⁶

Zoom In Zoom Out Reset image size

In order to quantify the evolution of the cathode catalyst ORR activity over the course of the V-cycling ASTs, the ORR activity of the cathode catalyst was determined from differential-flow H₂/O₂ polarization curves at 80 °C/100% RH and 170 kPa_abs at each AST segment. For this, ${E}_{\left({\text{R}}_{\text{HFR}}+{\text{R}}_{\text{H}+,\,\text{cath}}^{\text{eff}}\right)-\text{free}}$ was plotted versus the logarithm of the current density (corrected by the H₂ crossover and MEA shorting current densities) and fitted by a linear regression line between 50 and 500 $\mathrm{mA}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$ (kinetically limited region), which was extrapolated to determine the current density at an ${E}_{\left({\text{R}}_{\text{HFR}}+{\text{R}}_{\text{H}+,\,\text{cath}}^{\text{eff}}\right)-\text{free}}$ value of 0.9 V. Dividing this current density by the platinum loading then yields the ORR mass activity of the cathode catalyst at a resistance-corrected voltage of 0.9 V and at H₂/O₂ partial pressures of 120 kPa_abs (${i}_{\text{mass}}^{0.9\text{V}}$), ¹⁷ which is plotted versus AST cycle number in Fig. 9a. For clarity, only the ${i}_{\text{mass}}^{0.9\text{V}}$ values for H₂/N₂ and H₂/Air ASTs conducted at the same conditions, namely 80 °C/70% RH (yellow symbols/lines) and 80 °C/100% RH (red symbols/lines) are shown. The BOT ${i}_{\text{mass}}^{0.9\text{V}}$ values are in very good agreement with a previous study with the same type of catalyst (blue lines in Fig. 4 of Ref. 17). As one would expect on the basis of Fig. 3, the decrease of the ORR mass activity over the course of the V-cycling ASTs is faster at higher RH values (see Fig. 9a). Here it should be noted that a minor change of our characterization sequence (see Fig. 1) compared to that by Harzer et al. ¹⁷ eliminated the initial increase of ${i}_{\text{mass}}^{0.9\text{V}}$ over the first 100 AST cycles that was observed in their study, namely by acquiring the H₂/O₂ polarization curve more closely after the recovery step. A comparison of Fig. 9a with Fig. 3 shows that while ${i}_{\text{mass}}^{0.9\text{V}}$ decreases with decreasing Pt ECSA, there is no significant difference between H₂/N₂ (solid symbols) and H₂/Air (open symbols) V-cycling ASTs, indicating essentially identical degradation mechanisms for both ASTs.

Zoom In Zoom Out Reset image size

By dividing the ${i}_{\text{mass}}^{0.9\text{V}}$ data in Fig. 9a by the ECSA data in Fig. 3, the specific ORR activity (${i}_{\text{spec}}^{0.9\text{V}}$) was obtained and is shown in Fig. 9b. The ${i}_{\text{spec}}^{0.9\text{V}}$ values increase gradually from ∼0.2 at BOT to ∼0.3 over the first 3000 cycles for all four ASTs. Afterwards, the ${i}_{\text{spec}}^{0.9\text{V}}$ values for the MEAs subjected to V-cycling ASTs at 100% RH increase more rapidly, reaching a ∼2-fold higher value after 30 k cycles compared to those cycled at 70% RH. The ${i}_{\text{spec}}^{0.9\text{V}}$ differences between the H₂/N₂ and the H₂/Air cycled MEAs at 10 k and 20 k are not reflected in the ${i}_{\text{mass}}^{0.9\text{V}}$ values (Fig. 9a) and are likely an experimental artefact, as ${i}_{\text{spec}}^{0.9\text{V}}$ is obtained by dividing ${i}_{\text{mass}}^{0.9\text{V}}$ by the ECSA, which both have larger errors when the ECSA and thus the current density are very low. In general, however, the observed increase of the ${i}_{\text{mass}}^{0.9\text{V}}$ values with decreasing ECSA is what one would expect for an increase of the Pt particle size ⁴⁹ that occurs over the course of V-cycling ASTs. ^1,7

Figure 10 shows the BOT differential-flow H₂/Air (2000/5000 nccm) performance of all tested MEAs at 80 °C/100% RH and 170 kPa_abs (black symbols/line). After 30 k (EOT) cycles of the H₂/N₂ (solid symbols) and H₂/Air (open symbols) V-cycling ASTs conducted at 80 °C and different RH conditions (marked in the figure), the trend in the H₂/Air performance losses follows the above observed losses of the Pt ECSA (Fig. 3 and of the H₂/O₂ performance losses (Fig. 8a), indicating again a small difference between the H₂/Air performance losses for MEAs subjected to H₂/N₂ or H₂/Air V-cycling ASTs (as discussed for the H₂/O₂ polarization curves, the differences between the MEAs cycled under H₂/N₂ versus H₂/Air at 80 °C/100% RH (red symbols/lines) are likely insignificant for such strongly degraded MEAs). However, the H₂/Air performance losses are substantially larger than those measured under H₂/O₂: for example, for the V-cycling ASTs conducted at 80 °C/70% RH (with an EOT ECSA of $\sim 12\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$), the EOT H₂/Air versus H₂/O₂ performance losses are ∼150 mV (Fig. 10) versus ∼70 mV (Fig. 8a), respectively, and for ASTs conducted at 80 °C/100% RH (with an EOT ECSA of $\sim 4\,{{\rm{m}}}_{\mathrm{Pt}}^{2}\,{{\rm{g}}}_{\mathrm{Pt}}^{-1}$), the EOT H₂/Air versus H₂/O₂ performance losses are ∼400 mV versus ∼150 mV; respectively. This can be ascribed to the additional effect of the oxygen mass-transport resistance that is most relevant under H₂/Air and that increases with decreasing cathode electrode roughness factor (rf; in units of ${\mathrm{cm}}_{\mathrm{Pt}}^{2}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$), i.e., with decreasing ECSA when identical cathode platinum loadings are used. ^17,36,47 Consistent with this, the performance losses get even larger for differential-flow H₂/10%O₂ (rest N₂) performance curves (not shown), as reported also by Harzer et al. ¹⁷

Zoom In Zoom Out Reset image size

To determine the oxygen transport resistance (${R}_{{\text{O}}_{2}}^{\text{total}}$), limiting current density measurements were conducted with differential flows of H₂ and O₂/N₂ mixtures at 80 °C/100% RH and pressures between 170-500 kPa_abs, following previously described procedures. ³⁶ By measuring limiting current densities at different pressures, ${R}_{{\text{O}}_{2}}^{\text{total}}$ can be deconvoluted into a pressure dependent component (${R}_{{\text{O}}_{2}}^{\text{PD}}$) that represents the Fickian O₂ transport resistance in the flow-field and in the GDL substrate and in a pressure independent component (${R}_{{\text{O}}_{2}}^{\text{PI}}$) that represents the non-Fickian O₂ transport resistance in the very small pores in the MPL (${R}_{{\text{O}}_{2}}^{\text{MPL}}$) and the cathode catalyst layer (${R}_{{\text{O}}_{2}}^{\text{CCL}}$) as well as the cathode rf-dependent local oxygen transport resistance in the region close to the Pt surface (${R}_{{\text{O}}_{2}}^{\text{Pt}}$). ⁵⁰ Figure 11 shows the measured ${R}_{{\text{O}}_{2}}^{\text{total}}$ values at BOT (left-most bars of each group of bars) as well after 10k, 20k, and 30k cycles (subsequent bars of each group of bars) of the different V-cycling ASTs (marked by the x-axis labels), whereby the colored segments of the bars correspond to ${R}_{{\text{O}}_{2}}^{\text{PD}}$ and the upper gray parts of the bars correspond to ${R}_{{\text{O}}_{2}}^{\text{PI}}.$ The BOT values of ${R}_{{\text{O}}_{2}}^{\text{total}}$ (0.68 ± 0.06 s cm⁻¹) agree within ∼10%, and the BOT values of ${R}_{{\text{O}}_{2}}^{\text{PD}}$ (0.36 ± 0.05 s cm⁻¹) and ${R}_{{\text{O}}_{2}}^{\text{PI}}$ (0.32 ± 0.02 s cm⁻¹) for each set of two MEAs are also reasonably similar, even though their quantification is more error prone (note that for clarity the BOT values were evaluated separately for each set of two MEAs). Furthermore, as one would expect, the ${R}_{{\text{O}}_{2}}^{\text{PD}}$ values (colored bars), representing the O₂ transport resistance in the GDL and the flow-field, remain reasonably constant over the course of the V-cycling ASTs, consistent with the expectation that the GDL should not degrade in these tests.

Zoom In Zoom Out Reset image size

On the other hand, while the ${R}_{{\text{O}}_{2}}^{\text{PI}}$ values also remain roughly constant over the course of the V-cycling AST at 80 °C/30% RH (left-most gray bars in Fig. 11), they increase more strongly for the V-cycling ASTs when they are conducted at increasingly higher RH. This can be ascribed to the more pronounced Pt ECSA loss as the RH during the V-cycling ASTs is increased (see Fig. 3), which leads to an increase of ${R}_{{\text{O}}_{2}}^{\text{PI}},$ an observation that was also reported by others over the course of V-cycling ASTs. ^17,36,51 The increase of ${R}_{{\text{O}}_{2}}^{\text{PI}}$ with decreasing cathode rf had first been observed by Greszler et al., ⁵² who varied the cathode electrode roughness factor by varying the Pt cathode loading. They, as well as Kongkanand and Mathias, ⁵⁰ showed that the relationship between ${R}_{{\text{O}}_{2}}^{\text{PI}}$ and the cathode roughness factor can be described by Eq. 3, predicting an ${R}_{{\text{O}}_{2}}^{\text{PI}}$ increase with decreasing roughness factor due to the growing effect of the local oxygen transport resistance (${R}_{{\text{O}}_{2}}^{\text{Pt}}$):

$$\begin{eqnarray}&&{R}_{{\text{O}}_{2}}^{\text{PI}}{\rm{\approx }}{R}_{{\text{O}}_{2}}^{\text{MPL}}+{R}_{{\text{O}}_{2}}^{\text{CCL}}+\frac{{R}_{{\text{O}}_{2}}^{\text{Pt}}}{{rf}}\end{eqnarray} \tag{ 3 }$$

Zoom In Zoom Out Reset image size

In terms of the total O₂ transport resistance, Eq. 3 can be rewritten as:

$$\begin{eqnarray}&&{R}_{{\text{O}}_{2}}^{\text{total}}{\rm{\approx }}{R}_{{\text{O}}_{2}}^{\text{PD}}+{R}_{{\text{O}}_{2}}^{\text{MPL}}+{R}_{{\text{O}}_{2}}^{\text{CCL}}+\frac{{R}_{{\text{O}}_{2}}^{\text{Pt}}}{{rf}}\end{eqnarray} \tag{ 4 }$$

Here, the first three terms on the right-hand-side are considered to be approximately independent of the cathode electrode roughness factor, so that a plot of ${R}_{{\text{O}}_{2}}^{\text{total}}$ versus 1/rf should lead a straight line with a slope corresponding to ${R}_{{\text{O}}_{2}}^{\text{Pt}}.$ As seen in Fig. 12, ${R}_{{\text{O}}_{2}}^{\text{total}}$ over the course of the V-cycling ASTs (data taken from Fig. 11) versus the inverse of the cathode electrode roughness factor (calculated from the MEA loadings and the Pt ECSA values given in Fig. 3) can indeed be described reasonably well by a linear regression line (black dashed line in Fig. 12), independent of whether the MEAs were subjected to H₂/N₂ or H₂/Air V-cycling ASTs. Also added to the plot for comparison are the ${R}_{{\text{O}}_{2}}^{\text{total}}$ data of Harzer et al. ¹⁷ (blue stars; note that these are not included in the regression analysis), which were obtained over the course of a H₂/N₂ V-cycling AST at 80 °C/100% RH, using the same cycling protocol and the same cathode catalyst type/loading. According to Eq. 4, the slope of this line corresponds to ${R}_{{\text{O}}_{2}}^{\text{Pt}},$ which based on the linear regression analysis has a value of 13.4 ± 0.8 s cm⁻¹ at 80 °C/100% RH. This may be compared to other studies that used MEAs with cathode layers made with similar Vulcan carbon supported Pt catalysts, namely with ∼12 s cm⁻¹ (at 80 °C/62% RH) reported by Greszler et al., ⁵² with ∼10 s cm⁻¹ (at 80 °C/∼70% RH) reported by Yarlagadda et al., ⁵³ and with ∼8.5 s cm⁻¹ (80 °C/95% RH) reported by Della Bella et al. ³⁶ The rather high value obtained in the present study is likely due to having conducted the limiting current density measurements at 100% RH, which within the experimental accuracy of the dewpoint control may have resulted in the presence of liquid water in the cathode electrodes and thus to larger oxygen transport resistances. However, the overall conclusion that we can draw from Figs. 10–12 is that the loss of H₂/Air performance and the increase of oxygen transport resistances are essentially identical for the here conducted H₂/N₂ and H₂/Air V-cycling ASTs and predominantly depend on the cathode roughness factor.

Similar to the voltage loss analysis carried out for the differential-flow H₂/O₂ polarization curves shown in Figs. 7b and 8b, an analogous analysis was applied to the differential-flow H₂/Air polarization curves (see Fig. 10) over the course of the V-cycling ASTs. This is done by adding a term to Eq. 5 that considers the effect of ${R}_{{\text{O}}_{2}}^{\text{total}}$ on the H₂/Air voltage loss ($\unicode{x02206}{E}_{\text{O}2-\text{tx}}$):

$$\begin{eqnarray*}&&{E}_{\left({\text{R}}_{\text{HFR}}+{\text{R}}_{\text{H}+,\,\text{cath}}^{\text{eff}}+{\text{R}}_{{\text{O}}_{2}}^{\text{total}}\right)-\text{free}}={E}_{\text{cell}}+{i}_{\text{geo}}{\rm{\cdot }}{R}_{\text{HFR}}\end{eqnarray*}$$

$$\begin{eqnarray}&&+\,{i}_{\text{geo}}{\rm{\cdot }}{R}_{\text{H}+,\,\text{cath}}^{\text{eff}}+\unicode{x02206}{E}_{\text{O}2-\text{tx}}\,\propto \,{TS}{\rm{\cdot }}\mathrm{log}\left[\tfrac{1}{{i}_{0}}{\rm{\cdot }}{i}_{\text{spec}}\right]\end{eqnarray} \tag{ 5 }$$

As for the H₂/O₂ curves, R_HFR was measured directly during the acquisition of the H₂/Air polarization curves, ${R}_{\text{H}+,\,\text{cath}}^{\text{eff}}$ was determined from the R_{H+, cath} values shown in Fig. 5 (acc. to Neyerlin et al. ²⁴ ), and $\unicode{x02206}{E}_{\text{O}2-\text{tx}}$ was determined from the ${R}_{{\text{O}}_{2}}^{\text{total}}$ values shown in Fig. 11, using the following relationship given by Zihrul et al.: ¹⁸

$$\begin{eqnarray*}&&\unicode{x02206}{E}_{\text{O}2-\text{tx}}=\frac{R{\rm{\cdot }}\text{T}}{F}{\rm{\cdot }}\left(\frac{1}{4}+\frac{{\gamma }}{{\alpha }}\right){\rm{\cdot }}\end{eqnarray*}$$

$$\begin{eqnarray}&&\mathrm{ln}\left(\frac{{p}_{\text{O}2,\text{channel}}\text{}-\text{}\frac{R{\rm{\cdot }}T}{4F}\,{\rm{\cdot }}\,{R}_{{\text{O}}_{2}}^{\text{total}}\,{\rm{\cdot }}{\,i}_{\text{geo}}}{{p}_{\text{O}2,\text{channel}}}\right)\end{eqnarray} \tag{ 6 }$$

Here, R is the gas constant (8.314 J mol^-1 K^-1), T is the temperature in Kelvin, F is the Faraday constant, ${\gamma }$ is the ORR reaction order with respect to the oxygen partial pressure (using ${\gamma }$ = 0.54, as reported by Neyerlin et al. ⁴⁸ ), α is the effective transfer coefficient for the ORR (using a value of α = 1, corresponding to a Tafel slope of 70 mV dec^-1), and p_O2,channel is the oxygen partial pressure in the flow-field channel. Figure 13 shows the thus derived ${E}_{\left({\text{R}}_{\text{HFR}}+{\text{R}}_{\text{H}+,\,\text{cath}}^{\text{eff}}+{\text{R}}_{{\text{O}}_{2}}^{\text{total}}\right)-\text{free}}$ values for the H₂/Air polarization curves shown in Fig. 10 plotted versus the logarithm of ${i}_{\text{spec}}$ (based on the ECSA data in Fig. 3 and corrected by H₂ crossover and shorting currents).

Analogous to the transport resistance corrected H₂/O₂ polarization curves (Figs. 7b and 8b), the transport resistance corrected H₂/Air polarization curves (Fig. 13) essentially overlap for all MEAs subjected to the different V-cycling ASTs, independent of the RH condition and of whether the ASTs were conducted under H₂/N₂ or H₂/Air. Thus, consistent with the transport resistance corrected H₂/O₂ polarization curves, all of the transport resistance corrected H₂/Air polarization curves follow a single "master" curve, indicating again that the ECSA loss uniquely describes the H₂/Air performance when corrected for the HFR, the cathode proton conduction resistance, and the overall oxygen transport resistance. However, in the absence of any other transport resistances and for simple Tafel kinetics, the ${E}_{\left({\text{R}}_{\text{HFR}}+{\text{R}}_{\text{H}+,\,\text{cath}}^{\text{eff}}+{\text{R}}_{{\text{O}}_{2}}^{\text{total}}\right)-\text{free}}$ versus log(${i}_{\text{spec}}$) data should follow a straight line (a Tafel line with an ideal slope of 70 mV dec^-1 is marked by the gray line in Fig. 13), which is only the case up to ∼0.01 ${\rm{A}}\,{\rm{c}}{{\rm{m}}}_{{\rm{P}}{\rm{t}}}^{-2}$, indicating the presence of other, unassigned voltage losses. This phenomenon was first reported by Owejan et al. ⁴⁷ or MEAs with very low cathode Pt loadings and then later on by Harzer et al. ¹⁷ and Zihrul et al. ¹⁸ for MEAs aged over the course of V-cycling ASTs (in both reports, deviations from a straight Tafel line were observed at specific current densities higher than ∼0.02 ${\rm{A}}\,{\rm{c}}{{\rm{m}}}_{{\rm{P}}{\rm{t}}}^{-2}$). The origin of these unassigned losses are still under debate, but there is growing evidence that they may at least in part be due to an incorrect assessment of the ${R}_{{\text{O}}_{2}}^{\text{total}}$ values by the commonly used steady-state limiting current (SLC) approach ²⁵ (also used in the present study). In contrast, for the more recently developed transient limiting current (TLC) method first proposed by Göbel et al., ⁵⁴ the unassigned mass transport resistances are substantially minimized or even eliminated, ^54,55 as the TLC method is believed to determine ${R}_{{\text{O}}_{2}}^{\text{total}}$ under conditions that more closely reflect those during the acquisition of a polarization curve. Nevertheless, even though the here used SLC method results in large unassigned voltage losses at higher specific current densities (i.e., a deviation from a straight Tafel line beyond ∼0.01 ${\rm{A}}\,{\mathrm{cm}}_{\mathrm{Pt}}^{-2}$), the resistance corrected H₂/Air polarization curves all clearly scale with ${i}_{\text{spec}},$ independent of the AST conditions.

Zoom In Zoom Out Reset image size

Based on the above observations, one would expect that the degradation of the differential-flow H₂/Air cell voltage over the course of the H₂/N₂ and H₂/Air V-cycling ASTs at different RH conditions also only depends on the cathode electrode roughness factor (or on the cathode ECSA, since all MEAs have identical Pt cathode loadings of 0.4 ${\mathrm{mg}}_{\mathrm{Pt}}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$). This is indeed confirmed in Fig. 14, where the differential-flow H₂/Air cell voltages (recorded at 80 °C/100% RH and 170 kPa_abs) are plotted for four different current densities (0.2, 0.5, 1.0, and 2.0 ${\rm{A}}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$) versus the cathode electrode roughness factor at BOT (black diamonds) as well as after 10k, 20k, and 30k voltage cycles (differently colored symbols). Quite clearly, for each current density, the data of all the H₂/N₂ (solid symbols) and H₂/Air (open symbols) V-cycling ASTs follow an identical trend (marked by the gray dashed guide-to-the-eye lines), indicating that the H₂/Air performance is a unique function of the cathode electrode roughness factor, independent of the RH condition of the V-cycling ASTs and of whether they were conducted under H₂/N₂ or H₂/Air. Such a phenomenon was first reported by Della Bella et al., ³⁶ who showed that the H₂/Air cell voltages at a given current density over the course of H₂/N₂ V-cycling ASTs (conducted at 80 °C/100% RH) with different UPLs and different UPL hold times followed an identical trend line, suggesting that within their examined parameter space (UPLs from 0.85-1.0 V_RHE and hold times from 1-8 s) the performance loss of the MEAs would only depend on the cathode electrode roughness factor (or ECSA, as all their MEAs had the same Pt cathode loadings of 0.1 ${\mathrm{mg}}_{\mathrm{Pt}}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$). At the low current density of 0.2 ${\rm{A}}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$(see uppermost data set in Fig. 14), the H₂/Air performance losses appear to be solely due to ORR kinetic losses, as a 10-fold loss in cathode electrode roughness factor results in a cell voltage loss of ∼70 mV, as one would expect based on simple Tafel kinetics; the analogous behavior was also observed by Della Bella et al. ³⁶ On the other hand, at a large current density of 2.0 ${\rm{A}}\,{\mathrm{cm}}_{\mathrm{MEA}}^{-2}$ (lowermost data set in Fig. 14), the H₂/Air performance decreases much more rapidly with decreasing rf once the cathode electrode rf goes below ∼80 , where local oxygen transport resistances becoming increasingly, consistent again with the observations by Della Bella et al. ³⁶

Zoom In Zoom Out Reset image size

In summary, the findings reported by the latter authors (Fig. 12 in Ref. 36) in conjunction with the data shown in Fig. 14 suggest that for V-cycling ASTs with a UPL of ≤1.0 V_RHE, conducted either under H₂/N₂ between 30%-100% RH or under H₂/Air between 70%-100% RH, the H₂/Air performance only depends on the cathode electrode roughness factor that is reached after a given number of voltage cycles (note: the number of cycles over which a given cathode electrode rf value is reached decreases with increasing RH and increasing UPL). In other words, chemical/mechanical degradation of the ionomer in the cathode electrode that would be expected under H₂/Air conditions (particularly at the lower RH) does not seem to affect the performance loss, nor is there any evidence that carbon-support corrosion would be enhanced for H₂/Air compared to H₂/N₂ V-cycling ASTs.