Influence of Ionomer Content in IrO₂/TiO₂ Electrodes on PEM Water Electrolyzer Performance

MEAs with an active area of 5 cm² were prepared by a decal transfer method. For the hydrogen cathode electrodes, platinum supported on Vulcan XC72 carbon (46.7 wt% Pt/C; TEC10V50E from Tanaka, Japan) was used as catalyst. For the oxygen anode, IrO₂ supported on TiO₂ (IrO₂/TiO₂ with 75 wt% iridium; Elyst Ir75 0480 from Umicore, Germany) was used. Catalyst inks were prepared from catalyst powder, 2-propanol (purity ≥ 99.9%, from Sigma Aldrich), de-ionized (DI) water (18 MΩ cm) and Nafion ionomer solution (20 wt% ionomer; D2021 from IonPower, USA). ZrO₂ grinding balls (5 mm diameter) were added and the components were mixed for 24 hours using a roller mill to achieve a homogenous suspension. The ink was then coated onto ETFE foil (25 μm thick, FP361025 from Goodfellow, UK) utilizing a Mayer-rod coating machine. After drying, the electrodes were hot-pressed onto a Nafion 212 membrane (50 μm thick; from Quintech, Germany) for 3 min at 155°C at a pressure of 2.5 MPa. The catalyst loading was determined by weighing the ETFE decals before and after the decal transfer step, using a microbalance (± 1 μg; from Mettler Toledo, Germany). For the hydrogen cathode electrodes the loading was 0.35 ± 0.05 mg_Pt cm_MEA⁻² and the ionomer to carbon weight ratio was fixed at 0.6/1. The ionomer content of the oxygen anode electrodes was varied between 2.2 and 28.0 wt% relative to the total weight of the electrode. The anode catalyst loading was 2.00 ± 0.25 mg_Ir cm_MEA⁻² for the electrodes with an ionomer content between 3.9–28.0 wt% and 1.46 ± 0.10 mg_Ir cm_MEA⁻² for the sample with 2.2 wt% ionomer. Sintered titanium (from Mott Corporation, USA) with a porosity of ∼50% and a thickness of 280 ± 10 μm as well as a carbon fiber paper (TGP-H-120T from Toray, no MPL, 20 wt% PTFE) with a thickness of 370 ± 10 μm were used as porous transport layers (PTL) at the anode and at the cathode, respectively. The MEA was placed between the PTLs and sealed with virgin PTFE foil. Sealings with an appropriate thickness were chosen to achieve a 25% compression of the carbon PTL. (under the applied compression, the titanium PTL can be considered incompressible).

Cross-sectional scanning electron microscopy (SEM) measurements were performed with a JEOL JCM-6000Plus NeoScope scanning electron microscope at an accelerating voltage of 15 kV to determine electrode thickness. MEA cross-sections were prepared by cryo-fracturing MEAs in liquid nitrogen. 5 values for the electrode thickness were measured at three different locations to account for inhomogeneities in the electrode thickness. High resolution SEM images were taken with a JEOL JSM-7500F scanning electron microscope with an accelerating voltage of 1 kV. The Brunauer-Emmet-Teller (BET) surface area was determined by adsorption of liquid nitrogen on the catalyst powder with a Quantachrome Autosorb-iQ after outgasing for 6 hours at 200°C.

Tests were performed on an automated test station from Greenlight Innovation equipped with a potentiostat and booster (Reference 3000 and 30 A booster, Gamry). All measurements were done at 80°C cell temperature; deionized (DI) water was pre-heated to 80°C and fed to the anode side of the electrolysis cell at a rate of 5 ml min⁻¹. Polarization curves were recorded at hydrogen pressures between 1–30 bar_a absolute pressure. The oxygen side was always kept at ambient pressure. For the high pressure tests, the product gas at the anode side of the cell was diluted with nitrogen (100 nccm) to prevent the development of an explosive gas mixture which otherwise could be formed by the permeation of H₂ through the membrane into the anode compartment. After a warm-up step under N₂ atmosphere to reach the desired temperature, the cell was conditioned by ramping the current to 1 A cm⁻² and holding this value for 30 min. Subsequently, galvanostatic polarization curves were recorded at current densities starting at 0.01 and increasing to 6 A cm⁻². Each current step was held for 5 min, before recording the cell voltage to ensure stabilization. A slight improvement in cell performance was typically observed for the first two polarization curves, while it remained constant afterwards. Consequently, the first two polarization curves can be regarded as an additional conditioning step and were not included in the data analysis shown here. AC impedance measurements in a range of 10 Hz–20 kHz with current perturbations of ±200 mA were carried out at each current step to determine the high frequency resistance (HFR), which was obtained from the high-frequency intercept of the Nyquist plot with the real axis. CVs of the IrO₂/TiO₂ anode electrode were recorded at a scan rate of 50 mV s⁻¹ at 80°C. The anode working electrode was flushed with H₂O at a flow rate of 5 ml min⁻¹, while the cathode counter electrode was purged with dry H₂ at 50 ml min⁻¹.

To identify the contribution of different kinds of voltage losses, polarization curves were first corrected by the HFR, which is considered to be the sum of the membrane resistance R_memb and the electronic resistance (sum of contact resistances between flow-fields and PTLs and bulk PTL resistances), R_el. To validate this correlation, R_el and R_memb were measured separately: R_el as described in the next section and R_memb by a conductivity measurement. For the latter, a piece of Nafion 212 membrane was assembled in an in-plane conductivity cell (BekkTech, USA) and placed in liquid water at 80°C. The conductivity was determined from the HFR measured via AC impedance spectroscopy after immersing the membrane in water for more than 20 h to ensure that the membrane is fully hydrated.³⁰

Fig. 3a shows an exemplary SEM image of an MEA cross-section determined after the electrochemical measurements, employed to quantify electrode thickness. All Pt/C electrodes with a loading of 0.35 ± 0.05 mg_Pt cm⁻² (corresponding to a carbon loading of 0.40 ± 0.06 mg_c cm⁻²) had a thickness of 10 ± 1 μm. This is in agreement with the expected electrode thickness of 28 ± 2 μm for a carbon loading of 1 mg_c cm⁻².²⁶ The thickness of the IrO₂/TiO₂ electrodes ranged from 8–12 μm, depending on the exact catalyst loading. From the catalyst loading, L_cat, the anode electrode thickness, t_an, and the average catalyst density, ρ_cat = 9.5 g cm⁻², the catalyst volume fraction, V_cat, can be calculated:

Zoom In Zoom Out Reset image size

The average catalyst density was calculated based on the assumption that the catalyst consists of iridum(IV)oxide and titanium(IV)oxide. With the weight percentages of the components (86.9 wt% IrO₂ and 13.1 wt% TiO₂) and the densitiy of the materials (11.7 g cm⁻³ for IrO₂ and 4.23 g cm⁻³ for TiO₂ (rutile)) the average catalyst density can be calculated. As shown in Fig. 3b, V_cat is about 28% for all IrO₂/TiO₂ based anode electrodes (s. red diamonds in Fig. 3b), yielding an electrode void volume of 72% in the absence of ionomer (i.e., for 0 wt% ionomer). The empty pores are partially filled by the ionomer which is incorporated into the electrode structure; its volume fraction in the electrode can then be calculated by:

where L_ion is the loading of ionomer in the electrode, ρ_ion = 2.1 g cm⁻² is the ionomer density, and V_{ion, dry} corresponds to the volume of the dry ionomer. Swelling of the ionomer under operating conditions (80°C, liquid water) must be considered when calculating the actual ionomer volume fraction, V_{ion, wet}, which is about a factor of 1.8 higher compared to the dry ionomer volume.²⁷ The ionomer volume fraction in equilibrium with liquid water at 80°C, V_{ion, wet}, is shown by the blue circles in Fig. 3b, while the remaining void volume in the electrode, V_void, is shown by the green triangles in Fig. 3b. Error bars represent one standard deviation based on the variation of the measured electrode thickness. Note that the same analysis was applied previously to PEM fuel cell MEAs.³¹

For an ionomer content increasing from 2.2–28.0 wt% relative to the total electrode weight, the wet ionomer volume fraction in the electrode increases from 5–84%. The remaining void volume is obtained by subtraction of V_cat and V_{ion, wet} from the total electrode volume, i.e., from 100%. Hence, for an ionomer content of 20–25 wt%, the electrode void volume will be essentially completely filled with ionomer. Finally, using the BET surface area of the IrO₂/TiO₂ catalyst of 31 m² g_cat⁻¹, the effective ionomer film thickness on the catalyst surface can be calculated assuming a uniform ionomer film over the total catalyst surface, which then results in nominal ionomer film thicknesses increasing from ≈0.6 to ≈11 nm as the ionomer content increases from 2.2 to 28.0 wt%. The assumption of a homogeneous ionomer film is only valid if the pore size in the electrode is much larger than the nominal film thickness. From the inset in Fig. 3a it can be seen that the pores in the IrO₂/TiO₂ electrode range from ≈10 to ≈100 nm. Even smaller pores might be present as well, but could not be identified from the SEM image. Consequently, the assumption of a uniform ionomer film thickness might not be applicable, especially for samples with a high ionomer content.

Steady-state polarization curves at 80°C and ambient pressure are shown in Fig. 4a for MEAs with an anode ionomer content of 3.9, 11.6, and 28.0 wt%. The best performance is obtained for the intermediate ionomer content of 11.6 wt%, with V_ion,wet ≈ V_void ≈ 35% (s. Fig. 3b). Here, a cell voltage of 1.57 V is measured at a current density of 1 A cm⁻², and even at 6 A cm⁻² the cell voltage stays clearly below 2 V (s. blue circles in Fig. 4a). For the sample with 3.9 wt% ionomer (s. red diamonds in Fig. 4a), the performance decreases only slightly, while for the sample with 28 wt% ionomer (s. green squares in Fig. 4a), the cell voltage is significantly higher, which is partly due to a higher HFR (s. below). In Fig. 4b, the cell voltages corrected by the HFR (iR-free) are plotted for all MEAs. Again, the best performance can be observed for the sample with 11.6 wt% ionomer (s. blue circles in Fig. 4b), while the cell voltage increases when the ionomer content either decreases or increases with respect to this value.

Zoom In Zoom Out Reset image size

For high current densities, the amount of heat produced at the MEA can reach up to ≈3 W cm⁻² and since the temperature is not measured directly at the membrane but in the flow-fields, a significant difference between the actual MEA temperature and the measured value can be expected, estimated to be ≈6°C at 6 A cm⁻². Consequently, all further analysis is carried out only for current densities up to 3 A cm⁻², where the maximum heat production is <1 W cm⁻², resulting in a reasonably small temperature difference of less than 2°C.

A more detailed analysis of the influence of ionomer content on cell performance is given in Fig. 5, where the iR-free cell voltages at three fixed current densities (0.1 / 1.0 / 3.0 A cm⁻²) are plotted against ionomer content. A minimum iR–free cell voltage is observed at 11.6 wt% for all current densities (Fig. 5a), increasing for both lower and higher ionomer content as could already be seen in Fig. 4b. However, the maximum difference in iR-free cell voltage is only ≈50 mV at 3 A cm⁻². Fig. 5b shows the HFR for the different current densities vs. the ionomer content. It can be seen that the HFR is almost constant for all MEAs with an ionomer content up to ≈16 wt%, with a value of 52.5 ± 1.5 mΩ cm². Considering that the HFR should be the sum of R_memb and R_el, this HFR-value can be compared with the independently measured values for R_memb and R_el. In the applied cell configuration with an average compressive force of 1.7 MPa on the flow-field lands (s. Fig. 2a), R_el is ≈12 mΩ cm² as described earlier. For the Nafion 212 membrane, a conductivity of 142 mS cm⁻¹ was measured at 80°C in liquid water, which is in agreement with the values reported in literature.^32–36 The thickness of the membrane, assembled in the cell, is between ≈58–77 μm. The lower value is obtained if the volume expansion due to exposure to water occurs isotropically, whereas the higher value was calculated according to Liu et al., making the assumption that the volume expansion only occurs in the through-plane direction.²⁵ This leads to a membrane resistance R_memb of ≈41–54 mΩ cm². The sum of R_el and R_memb therefore is predicted to range in between 53–66 mΩ cm². Comparing these values to the measured HFR (52.5 ± 1.5 mΩ cm²), the agreement between measured and predicted HFR is very good, if one assumes an isotropic volume expansion of the membrane.

Zoom In Zoom Out Reset image size

At an ionomer content of ≈20 wt%, the HFR increases significantly to ≈63 mΩ cm² and stays at this value for even higher ionomer contents. The sudden increase of the HFR coincides quite well with the predicted complete filling of the electrode void volume for an ionomer content of approximately >20 wt% (s. green triangles in Fig. 3b). This suggests that in the case when the ionomer volume exceeds the void volume between the catalyst particles, an electronically insulating film of residual ionomer forms at the electrode/PTL interface, leading to a higher contact resistance and thus to a higher HFR. The increase in iR-free cell voltage in this case is likely related to an additional electronic insulation of parts of the catalyst particles, leading to higher kinetic overpotentials and thus higher iR-free cell voltages.

To ensure reproducibility of the results, three different MEAs with the optimum ionomer content of 11.6 wt% were tested. The anode catalyst loading was between 2.00–2.20 mg_Ir cm_MEA⁻² and the cathode catalyst loading was 0.35 ± 0.05 mg_Pt cm_MEA⁻² for the three samples. For current densities up to 3 A cm⁻² the electrochemical measurements showed excellent reproducibility, with differences in cell voltage of less than 10 mV, Tafel Slopes between 45–47 mV dec⁻¹ and HFRs of 52–55 mΩ cm².

High pressure electrolysis tests were carried out with the MEA with the above found optimal anode ionomer content of 11.6 wt%. Polarization curves were recorded at differential pressure conditions, with a cathode pressure p_cath of 3, 10, and 30 bar_a, while the anode pressure was kept at 1 bar_a. Before and after the differential pressure tests, a polarization curve was recorded at ambient pressure conditions (p_cath = p_an= 1 bar_a) to ensure that no degradation had occurred during the differential pressure tests. The recorded cell voltages before and after the differential pressure tests were identical within a range of ±2 mV. The polarization curves at different hydrogen pressures are shown in Fig. 6. The cell voltage increases with increasing operating pressure, and at the highest shown current density of 3 A cm⁻², an increase of the cathode pressure from ambient pressure to 30 bar_a results in an increase in cell voltage of <50 mV (s. Fig. 6a), which is at least qualitatively consistent with other data in the literature.¹ Since the HFR is essentially independent of pressure (data not shown), a similar gain in E_iR–free is observed (s. Fig. 6b). The fact that the differences between the curves in Fig. 6 are decreasing with increasing current density is addressed in more detail in the Discussion section.

Zoom In Zoom Out Reset image size

In the Discussion section, we will now analyze the various voltage loss terms as a function of the ionomer content in the anode electrode of the MEA, seeking to explain the origin of the increasing cell voltage at ionomer contents above and below the experimentally determined optimum value. In addition, we will provide an explanation for the reported decrease in mass transport resistance with increasing operating pressure.³

For the calculation of η_HER, the Butler-Volmer equation can be linearized due to the fast reaction kinetics.¹⁰ The overpotential can then be defined as:³¹

where

With an exchange current density of i_{0, HER} = 250 mA cm_metal⁻² at 80°C,¹² a cathode catalyst loading of L_Pt = 0.35 mg_Pt cm⁻², and an electrochemically active surface area of A_{Pt, el} = 60 m² g_Pt^{−1 31} for the catalyst used in this study, R_{K, HER} equates to 0.29 mΩ cm², which results in an η_HER value of less than 1 mV even at a current density of 3 A cm⁻². Thus, the kinetic overpotential for the HER, η_HER, can be neglected in the further analysis.

On the other hand, the effective proton transport resistance for the hydrogen cathode can be calculated following the approach described by Gu et al.:³¹

where

Here, the sheet resistance for proton transport in a Pt/C electrode, , can be calculated from the reported sheet resistivity of ≈25 Ω cm for a Pt/Vulcan electrode with an I/C-ratio of 0.6/1 at 80°C and a relative humidity of 122% (i.e., in the presence of liquid water)²⁵ and the electrode thickness of ≈10 μm (s. Experimental section), equating to a proton conduction sheet resistance of ≈ 25 mΩ cm². Together with the above determined charge transfer resistance (R_{K, HER} ≈ 0.29 mΩ cm²), this yields a β–value of ≈9 (s. Eq. 9). Thus, the effective proton transport resistance, , calculated by Eq. 8 is ≈2.5 mΩ cm², which would result in very small voltage loss of ≈7 mV at 3 A cm⁻².

The overpotential for the OER can be determined from a Tafel fit of the data in Fig. 4, as shown in Fig. 7, where the iR-free cell voltage is plotted on a logarithmic current scale. The Tafel slope was determined in the 10–100 mA cm⁻² region, where the behavior is approximately linear and the effects of proton and mass transport resistances can be neglected. The Tafel slopes are between 45–50 mV dec⁻¹ (see Table I), which is reasonably consistent with the values of 40–56 mV dec⁻¹ reported by Matsumoto and Sato for sputtered and thermally prepared IrO₂,³⁹ and with 40–45 mV dec⁻¹ reported by Reier et al.⁴⁰ for amorphous iridium oxide on a titanium substrate. In analogy to the quantification of the activity for the oxygen reduction reaction (ORR) in PEM fuel cells, where the activity of different catalysts is benchmarked at an iR–free cell voltage of 0.9 V, due to the negligible transport resistances at the low current densities at this voltage,⁴¹ we here propose to quantify the activity of OER catalysts at an iR-free cell voltage of 1.45 V. At this potential, the current density is large enough to neglect the effect of ohmic shorting, but still small enough to largely exclude the influence of mass transport resistances. The respective values for the current density at 1.45 V_iR–free and for the mass-specific current densities (in units of A g_Ir⁻¹) are given in Table I.

Zoom In Zoom Out Reset image size

Considering that the anode and cathode catalyst loadings are essentially identical for the three MEAs shown in Fig. 7, one would expect a superposition of the Tafel lines in the kinetically controlled region, i.e., at low current densities. One explanation for the clearly higher value of η_OER for the MEA with an anode ionomer content of 28 wt% (s. green triangles in Fig. 7) might be a lower electrochemically active surface area (ECSA) compared to the other samples. For oxide based catalysts, the number of active sites available for the OER is typically related to the voltammetric charge q obtained from integration of a cyclic voltammogram (CV),^42–44 assuming that the number of active sites for the OER is proportional to the voltammetric charge. The CVs for the samples with 3.9, 11.6, and 28.0 wt% ionomer recorded at a scan rate of 50 mV s⁻¹ are shown in Fig. 8, whereby the voltammetric current is normalized by the mass of iridium. Quite clearly, the voltammetric charge for the samples with 3.9 and 11.6 wt% ionomer is very similar, while it is significantly lower for an ionomer content of 28.0 wt%. The values of the mass normalized voltammetric charge or mass specific charge, q^*, obtained from an integration of the CVs between 0.05–1.00 V are shown for all examined MEAs in the last column of Table I. The substantially lower mass specific capacity, q^*, of the MEA with 28 wt% ionomer (155 C g_Ir⁻¹; s. Table I) compared to the 3.9 and 11.6 wt% ionomer containing MEAs (294 C g_Ir⁻¹; s. Table I) can only be explained by a lower catalyst utilization in the former electrode. This, we believe, is due to the electronic insulation of parts of the catalyst by the ionomer, which occurs when the electrode void volume approaches zero (s. green triangles in Fig. 3b). This is in agreement with the findings of Xu et al., who also observed a decrease in the mass specific capacity at high ionomer content.²⁹ Assuming that q^* is directly proportional to the active sites for the OER, the differences in η_OER in the Tafel region would be related to:

With a Tafel slope of b ≈ 47 mV dec⁻¹ and the values for q*₁ = 294 C g_Ir⁻¹ and q*₂ = 155 C g_Ir⁻¹ for the MEAs with 11.6 wt% and 28.0 wt% ionomer, respectively, the predicted value for Δη_OER is ≈13 mV, which is in excellent agreement with the shift of the iR-free cell voltage observed in Fig. 7. Consequently, when normalizing the mass specific activity (evaluated at 1.45 V_iR–free; s. Table I) of the IrO₂/TiO₂ catalysts by their mass specific capacity, the resulting capacity normalized activity averaged over all MEAs is 138 ± 6 mA C⁻¹, i.e., essentially identical for all MEAs tested here. In summary, part of the lower performance at high ionomer content is due to the electronic insulation of some of the catalyst particles by the ionomer. Therefore, the effect of ionomer void volume filling must be considered when conducting kinetic experiments.

Zoom In Zoom Out Reset image size

To quantify the overpotential due to proton transport in the anode, first the electrode sheet resistance for proton conduction has to be determined, which in principle can be calculated from:²⁵

where t_an is the electrode thickness, σ is the conductivity of the ionomer, V_{ion, wet} is the ionomer volume fraction in the liquid water equilibrated electrode, and τ is the apparent tortuosity of the ionomer phase in the electrode. The electrode thickness and ionomer volume fraction were determined from the cross-sectional SEM images as shown before (s. Fig. 3). The conductivity of the ionomer is assumed to be the same as that for the Nafion 212 membrane (142 mS cm⁻¹ at 80°C in contact with liquid water) because of the essentially identical equivalent weight (EW). The apparent tortuosity was determined previously for a Pt/C electrode and turned out to be between 0.7–1.5 for a temperature of 80°C and a relative humidity of 122%.²⁷ In a first approximation, we assume that the ionomer phase tortuosity of the IrO₂/TiO₂ electrode is in the same range as that obtained for the Pt/C catalyst. The so calculated minimum and maximum values for are shown in Table I.

From the estimated sheet resistances for proton conduction, (s. Table I), the effective proton conduction resistance in the anode electrode, , can be calculated from:⁴⁵

where ζ is a correction factor which accounts for the effect of a reduced catalyst utilization, and which is a function of the dimensionless ratio of the proton conduction sheet resistance over the kinetic resistance:⁴⁵

where i is the current density, the proton transport resistance and b the Tafel slope.

From the above analysis, it is now possible to quantify all of the voltage loss terms on the right-hand side of Eq. 3, with the exception of the mass transport term, η_mt (last term on the right-hand side of Eq. 3). The magnitude of the latter can be estimated by subtracting all the known voltage loss terms from the reversible cell voltage, as will be shown in the following section.

The fraction of the total overpotential which is not due to ohmic (determined by the HFR) and OER kinetic losses (the HER kinetic losses are only on the order of 1 mV; see above) can be determined from the difference between the Tafel line (s. dashed lines in Fig. 7) and the iR-free cell voltage (s. Fig. 4b). This remaining overpotential for the samples with 3.9, 11.6 and 28.0 wt% ionomer is plotted in Fig. 9 (red diamonds), along with the voltage loss contributions from proton conduction resistance in the anode and cathode electrodes. The additional undefined mass transport losses remaining after the subtraction of the proton conduction related voltage losses in the cathode electrode (green areas in Fig. 9) and in the anode electrode (blue areas in Fig. 9) are visualized by the red area in Fig. 9.

Zoom In Zoom Out Reset image size

Let us first examine the analysis for the MEA with the anode electrode with a low ionomer content of 3.9 wt% shown in Fig. 9a. For this MEA, the anode void volume is ≈65% (s. Fig. 3b), so that one would expect little voltage losses due to O₂ and H₂O transport resistances. This can be examined by summing up the contributions from the proton conduction induced voltage losses in the cathode (green area in Fig. 9a) and in the anode (blue area in Fig. 9a), whereby the uncertainty in the latter is indicated by the blue dashed lines in Fig. 9a. Indeed, as expected, the unassigned voltage losses range between only ≈5 and ≈20 mV at 3 A cm⁻², as is indicated by the voltage difference between the blue dashed lines and the red line in Fig. 9a. As we will explain in the next section, these losses are likely due to an overpotential caused by the pressure buildup of gaseous hydrogen in the cathode electrode, which we estimate to be on the order of ≈20 mV at 3 A cm⁻².

For the MEA with the optimum ionomer content of 11.6 wt% in the anode, the unassigned voltage losses at 3 A cm⁻² range from ≈20 to ≈30 mV (s. difference between the blue dashed lines and the red line in Fig. 9b). For this MEA, the void volume in the electrode (≈35%, s. Fig. 3b) is only roughly one half of that the above discussed MEA with an ionomer content of 3.9 wt%, so that an additional O₂ and/or H₂O transport resistance induced voltage loss of ≈10–15 mV seems reasonable.

Finally, the unassigned voltage losses for the MEA with 28.0 wt% ionomer in the anode electrode range between ≈40–45 mV at 3 A cm⁻² (s. difference between the blue dashed lines and the red line in Fig. 9c). While approximately 20 mV may be caused by the above mentioned additional losses at the cathode, we believe that the remaining losses are largely due to oxygen transport resistances originating from the strong diffusion barrier imposed by a completely ionomer filled anode electrode (s. Fig. 3b). In this case it is conceivable that the local O₂ pressure at the electrode increases, which in turn would increase the reversible cell potential according to the Nernst Equation by ΔE_rev, equating to a mass transport induced overpotential, η_mt:

where and are the partial pressures of O₂ in the electrode and in the flow-field channel, respectively. For a perfectly crack-free ionomer filling of the anode electrode, the produced O₂ must be removed by permeation through the ionomer phase. The required O₂ partial pressure in the catalyst layer, , to support O₂ permeation at a rate equivalent to the O₂ generation rate can then be calculated by:⁴⁶

where t_eff is the effective permeation path, represents the oxygen generation rate (i.e., ), and is the permeability of O₂ in the Nafion ionomer. As a first approximation, t_eff is assumed to be half of the anode electrode thickness (i.e.). For , a value of is estimated for our electrode, based on the permeability of Nafion 212 at the given operating conditions (80°C, liquid water).⁴⁶ With these parameters, a pressure gradient Δp of up to ≈200 bar would be required at 3 A cm⁻², resulting in an overpotential η_mt of ≈45 mV (based on Eq. 14, with 0.53 bar_a O₂ in the flow-field channel at ambient pressure and 80°C). On the other hand, the expected loss due to an O₂ partial pressure gradient is more on the order of ≈20–25 mV (s. above), which would correspond to a much lower value of Δp ≈ 10 bar. For this reason, we believe that the main path for oxygen removal from the electrode is not permeation but convective transport through cracks or pinholes in the ionomer layer within the electrode, even for electrodes with nominally zero void volume.

Mass transport on the H₂ side could play a role for all MEAs in the case of water flooding of the cathode catalyst layer, as illustrated in Fig. 10. At high current densities, the transport of H₂O to the cathode by electro-osmotic drag can become quite significant, with drag coefficients under electrolyzer operating conditions in the range of 2.5–3.2.⁴⁷ The H₂O dragged to the cathode will be removed by a combination of pressure-driven back-diffusion through the membrane to the anode and water expulsion into the cathode flow-field by the evolving H₂ gas. If the contact angle inside the cathode catalyst layer is <90° (i.e., if the cathode electrode is hydrophilic), the H₂ pressure inside the cathode electrode must reach the capillary pressure, p_cap, in order to push the water out of the electrode pores. The capillary pressure can be estimated by:

with the surface tension of water , the capillary radius r_cap, and the water contact angle in the electrode θ. The surface tension of water, , at 80°C is 0.0626 N m⁻¹, and r_cap is ≈25 nm for an average pore diameter in the cathode catalyst layer of ≈50 nm.³¹ With these values, the capillary pressure p_cap would range from 0–25 bar for contact angles ranging from 90 to 0° (at contact angles >90°, no H₂ pressure buildup would be required for the expulsion of water from the cathode electrode). Unfortunately, the exact value for θ is difficult to measure, as the contact angle of membranes and ionomer bonded electrodes depends on the hydration state and the history of the electrode. Yu et al. investigated the contact angle of membranes and of catalyst layers composed of Pt/C and ionomer by the sessile drop method and by environmental scanning electron microscopy.⁴⁸ For a Nafion membrane they observed a decrease of the contact angle from initially 93.9° to 87.8° after extended contact with liquid water. For the catalyst layer of new MEAs, they found contact angles of ≈145°, while it decreased to 33–98° for aged catalyst layers, most likely due to the gradual oxidation of the carbon support surface. We believe that the actual contact angle of conditioned and water soaked cathode electrodes is likely equal to the contact angle of the ionomer (i.e., of the Nafion membrane), which would suggest a contact angle of ≈87.8°, equating to a capillary pressure of p_cap ≈ 2 bar_a. Further insights on this can be gained by the following analysis of the electrolyzer performance versus hydrogen pressure under differential pressure conditions (i.e., the anode compartment remaining at ambient pressure).

Zoom In Zoom Out Reset image size

Since only the H₂ side of the cell is pressurized, the activities of H₂O and O₂ are not affected by the high pressure operation on the H₂ side. Consequently, the shift of the reversible cell voltage ΔE_rev compared to ambient pressure operation () can be calculated by:

For the maximum cathode pressure p_cath = 30 bar_a (), the cell voltage should be shifted by ΔE_rev = 60.9 mV at 80°C. In Fig. 11, the iR-free cell voltages are plotted for the measurements at a cathode pressure between 1–30 bar_a. The curves were corrected by ΔE_rev to exclude the pressure induced shift of the reversible cell voltage from the analysis. Assuming that ΔE_rev is the only factor influencing the performance, the curves should lie on top of each other. For current densities of ≈200 mA cm⁻² this is in fact the case. At smaller current densities, the pressure fluctuated due to the low gas flow, which can explain some fluctuations of the cell voltages. In general, however, in this current range the cell voltage at high pressure seems to be slightly lower than expected. This can be explained by the H₂ crossover through the membrane. For a H₂ permeability of 3 · 10⁻⁹ ,⁴⁹ the H₂ crossover current at a cathode pressure of 30 bar_a is ≈30 mA cm⁻². Consequently, for small current densities, the H₂ flux to the anode is in the range of the O₂ production. Since H₂ does not react with O₂ at the anode but accumulates,⁵⁰ its presence effectively lowers the O₂ partial pressure at the anode, which leads to a reduction of E_rev according to Eq. 4. This, of course, is an undesireable operating condition, as significant concentrations of H₂ build up in the O₂ compartment.

Zoom In Zoom Out Reset image size

While the dilution effect of crossover H₂ becomes negligible at current densities of >300 mA cm⁻², the cell voltage for high pressure operation gets increasingly lower than expected from ΔE_rev, i.e., the electrolyzer performance at high pressure actually improves when corrected for ΔE_rev (s. black squares at 30 bar_a vs. red diamonds at 1 bar_a in Fig. 11). This can be seen more clearly in the inset of Fig. 11, which shows the difference between the ΔE_rev–corrected voltage at any given pressure referenced to that at 30 bar_a, clearly indicating a maximum improvement of ≈20 mV at 3 A cm⁻² when raising the H₂ pressure from 1 to 30 bar_a. Therefore, high pressure operation of the H₂ cathode must result in a reduction of one of the voltage loss terms. A similar phenomenon has been observed in other studies and was ascribed to a reduction of the size of produced gas bubbles at high pressure^5,6 or to improved OER kinetics, inferred from a lower apparent Tafel slope at high pressure.³ However, in these studies the O₂ side of the cell was also pressurized (balanced pressure), in contrast to the experiments in our study which were carried out at differential pressure conditions. Consequently, in our measurements, the OER kinetics and the O₂ mass transport should not change, suggesting that high H₂ pressure must affect some so far not addressed voltage loss on the H₂ side of the electrolysis cell. We believe that its most likely origin is a mass transport overpotential on the cathode caused by an increased partial pressure of H₂ in the cathode catalyst layer, . This is illustrated by assuming a difference in H₂ partial pressure between the cathode catalyst layer and the flow-field channel of p_H2,cat - p_H2,_channel ≈ 2 bar_a, similar to the capillary pressure p_cap ≈ 2 bar_a calculated by Eq. 16 with a contact angle of 87.8° and an average pore radius of 25 nm. The corresponding mass transport overpotential at pressures of either 1 or 30 bar_a in the H₂ flow-field would be:

From this exemplary calculation it becomes clear that the overpotential due to H₂ mass transport would be much larger at ambient pressure than at high pressure if the contact angle in the cathode electrode is <90°. The above estimated decrease of η_mt with increasing pressure would be expected to be most pronounced at high current densities, where a possible bypass of hydrogen through partially flooded pores would become more unlikely, thus leading to a gradual increase of the difference between the 1 and 30 bara voltage shown in the inset of Fig. 11. The overall difference of ≈20 mV at 3 A cm⁻² (s. inset of Fig. 11), which is clearly caused by the hydrogen cathode performance corresponds, we believe, to the unassigned mass transport losses in Fig. 9a and to a large extent to those in Fig. 9b.

A summary of the contributions of the various voltage losses to the overall cell voltage is given in Fig. 12. for the MEA with the optimum ionomer content of 11.6 wt% in the anode. The highest overpotentials are due to the OER kinetic losses determined by a Tafel analysis, which account for ≈350 mV at 3 A cm⁻² (s. purple area in Fig. 12) and the ohmic losses calculated from the HFR, which account for ≈155 mV at 3 A cm⁻² (s. orange area in Fig. 12). The losses due to proton conduction resistance in the cathode and anode electrodes add up to ≈20 mV at 3 A cm⁻² (s. green and blue area in Fig. 12). For the calculation of the proton conduction resistance in the anode electrode the lower limit of is used (s. Fig. 9). The overpotential for H₂ mass transport is calculated from the difference between the data at 1 bar_a and the data at 30 bar_a as shown in the inset in Fig. 11, which results in ≈20 mV at 3 A cm⁻². The remaining losses account for less than 10 mV at 3 A cm⁻² and can likely be attributed to O₂ mass transport.

Zoom In Zoom Out Reset image size