Communication—Patterned Electrodes to Increase Water Back-Diffusion in Hydroxide Exchange Membrane Fuel Cells

Hydroxide exchange membrane fuel cells (HEMFCs) support non-precious-metal catalysts and have demonstrated high performance, but suffer from flooding on the anode and possible drying on the cathode. We present a novel electrode patterning approach for redistributing water within the cell. Bare membrane channels provide additional pathways for water back-diffusion from anode to cathode, yielding modest performance improvements in the mass transport polarization regime. © The Author(s) 2016. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0071607jes] All rights reserved.

Hydroxide exchange membrane fuel cells (HEMFCs) are an emerging low-cost alternative to proton exchange membrane fuel cells (PEMFCs). 1 HEMFCs intrinsically support inexpensive, earthabundant catalysts like Ni and Ag 2,3 and show increasingly competitive performance. 4 Due to HEMFCs' distinctive water stoichiometries (+2 for hydrogen oxidation on the anode where water is produced, and -1 for oxygen reduction on the cathode where it is consumed), water management is a potentially serious issue. 5 The anode is quick to flood and the extent to which the cathode dries out is unknown. Substantial back-diffusion of water from anode to cathode is needed for optimal performance. 6 Membrane thickness plays a key role in the distribution of water inside an HEMFC. Thicker membranes present higher resistance against water transport; accordingly, high-performance membranes, such as Tokuyama A901, tend to be very thin (∼10 μm). But the catalyst layers (CLs) where water is produced or consumed may themselves act as barriers to water transport. To address this issue, we propose patterned electrodes that alternate between normal active electrode area and inactive water diffusion channels consisting of bare membrane. While the channels do not directly generate electricity, they alleviate flooding and/or drying by equalizing any water concentration gradients between cathode and anode, boosting the performance of nearby electrode area (Fig. 1).
We quantify the degree of patterning as the ratio of the total electrode perimeter that borders water channels to the total active area of the electrode. In principle, the greater the interfacial perimeter per active area, the better the water transport. Note that this ratio is appropriate only when the water channels are wide, as in this work. For a highly patterned electrode the ratio of total channel area to total active electrode area might be a better descriptor.
Electrode patterning is especially well-suited to HEMFCs. PEM-FCs do not appear to suffer from water imbalance as severely as HEM-FCs. In addition, proton exchange membranes tend to be thicker than hydroxide exchange membranes for two reasons: their ionic conductivity is higher, 1,7 and they are typically prepared from hydrophobic fluorinated polymers, which have high gas permeability. [8][9][10] Clearing off part of the CL in a PEMFC might have little effect on the water distribution if a thick membrane already acts as a significant barrier to transport. By contrast, in an HEMFC the relatively low ionic conductivity and low gas permeability of non-fluorinated hydroxide exchange electrolytes drive the membrane down to nearly the same thickness as the CLs.
Here we study membrane-electrode assemblies (MEAs) based on five types of electrodes (see Fig. 1) and two membranes (Tokuyama A201 and A901). The electrodes are identically composed and differ only by the mask with which they are fabricated. We hypothesize that * Electrochemical Society Member. z E-mail: yanys@udel.edu the performance per active area should correlate with the amount of electrode perimeter that borders water channels. Further, the performance enhancement should be more pronounced in MEAs based on the thinner A901 membrane, for which the CLs account for a greater proportion of the total resistance to water transport.

Experimental
High-performance MEAs were fabricated with a robotic sprayer (Sono-Tek ExactaCoat) by a catalyst-coated membrane (CCM) method as described elsewhere. 5 Electrolytes (A201/A901 membranes and AS-4 ionomer solution) were supplied by Tokuyama Corp.; 50% Pt on high-surface-area C catalyst, by Tanaka Kikinzoku Kogyo (TKK); unwetproofed TGP-H-060 carbon paper gas diffusion layers, by Toray; and the fuel cell enclosure, by ElectroChem.
MEAs were tested with a fuel cell test system (Scribner 850e). Regardless of electrode type, the cell was always assembled with the same gaskets (window size: 5 cm 2 ). The flow rate of the H 2 and O 2 feeds was 0.2 mL/min, the cell temperature was 60 • C, and the humidifier temperature for both feeds was 70 • C. Performance is reported on the basis of the active electrode area, excluding any channels of bare membrane.
Safety note: all tests were conducted at standard pressure. To avoid explosive crossover, exercise extreme caution when testing patterned MEAs at elevated pressure.

Results and Discussion
MEAs were fabricated with masks based on the designs in Figure 1 and tested under identical conditions. To ensure the best possible comparison, four of the five masks were applied together and sprayed simultaneously with the same batch of ink.
The results suggest a modest relationship between electrode perimeter and peak power density (PPD), a measure of performance. With a thick membrane (A201, 28 μm) no relationship is visible (Fig. 2, black circles). However, as expected, with the thinner A901 membrane (10 μm) a significant trend emerges (red triangles). Representative polarization curves are shown in Figure 3. Whereas all of the MEAs behave similarly at low current, at high current in the mass transport regime the patterned electrodes generally perform better, indicating that the water diffusion channels are effective in controlling flooding and/or drying.
Interestingly, the trend is not monotonic: increasing the degree of patterning beyond 2 × 2 hurts performance. While this trend should be verified with additional experiments, the error bars are sufficiently tight that some further speculation is warranted. The decrease in performance for the 4 × 4 patterned cell could represent gas crossover losses, an issue specific to thin membranes like A901. Just as the CL acts as a barrier to water transport, it also supplements the membrane in separating H 2 on the anode from O 2 on the cathode. Removing part Figure 1. Illustration of enhanced water diffusion in patterned MEAs. The face view shows mask designs used to prepare a series of MEAs. A photograph of an MEA fabricated with the 4 × 4 pattern from (e) is also shown. The edge view shows the anode side of the fuel cell sandwich. The gasket and gas diffusion layer (GDL) are not to scale; in reality each is about 20 times thicker than the membrane or CL. Dashed blue arrows indicate transport routes whereby water on the anode must pass through both the CL and membrane to reach the cathode side. Solid arrows indicate preferred routes that utilize water channels (white), macroscopic gaps in the CL. In (a), there are no such gaps because the electrode is too large for the gasket opening. In (c-e), channels are introduced deliberately. The total electrode area of 3.5 cm 2 excludes these channels. Under the hypothesis that the water channels improve performance, after normalizing to the active electrode area (i.e., the total black area contained within the dashed line in the face view), (a) should perform worst and (e) best.
of the CL is expected to increase crossover. The trouble with this explanation is that crossover losses are normally visible at open circuit, yet for these data the patterned and unpatterned MEAs behave identically until a relatively high load is applied. Further, it is unclear why Figure 2. Peak power density of conventional and patterned MEAs made with thin A901 or thick A201 membranes. Error bars indicate ±1 standard error (n = 3). The electrode perimeter-to-area ratio is defined as the amount of perimeter length bordering the active electrode patches divided by the total active electrode area within the 5 cm 2 window. The five specific x-values correspond to five different patterns as illustrated at the top of the plot (see also Fig. 1). For the large (>5 cm 2 ) electrode, the perimeter-to-area ratio is zero because there is no exposed membrane in the active 5 cm 2 gasket window. redistributing a given amount of channel area should affect crossover losses; intuitively, many small channels should result in the same gas crossover as few large ones. Another possible explanation is that due to crossover losses the feed flow rate is inadequate to fully fuel the electrode area at the end of the flow path, but this seems unlikely with the relatively high flow rate employed (200 mL/min for 5 cm 2 electrode area).
Despite these concerns, as additional evidence of greater gas crossover for 4 × 4 vs. 2 × 2 patterning, we measured the "initial" open-circuit voltage (IOCV) immediately after connecting the  cell to the H 2 and O 2 feeds. This IOCV measurement differs from the normal OCV point at the beginning of a polarization scan in that the cell has not yet been conditioned. No water has been generated inside the cell and too little time has passed for the humidified feeds to wet the electrolyte. Without the swelling that results from wetting, the volume of hydrophilic regions in the membrane and CLs is lower, so hydrophobic regions occupy a larger fraction of the total volume. Thus, both because the total volume is lower and because transport of nonpolar gases is faster through hydrophobic media, 9 conductance of H 2 and O 2 through the electrolyte should be highest just after the feeds are connected. Therefore, IOCV can be thought of as an OCV measurement that is especially sensitive to gas crossover. Indeed, under these conditions the IOCV is much lower for highly patterned A901 MEAs (Fig. 4), consistent with the hypothesis that crossover is more significant at high levels of patterning. As expected, when A201 is used instead, patterning makes no difference because the thick membrane already blocks virtually all gas crossover.

Summary
Overall, electrode patterning does appear to slightly increase cell performance as long as the membrane is thin (or has a high water diffusivity) compared to the CLs. However, at high levels of patterning the tradeoff between enhanced gas crossover and enhanced water diffusion leads to a performance optimum. At best, the electrode patterns proposed here could increase cell performance by several percent. For most applications, such a small improvement is unlikely to be practical when the water channels could instead be filled in with additional electrode area. Speculatively, electrode patterning could offer an advantage in scenarios where the gas permeability of the membrane is low, fuel efficiency matters more than capital cost, and/or the catalyst is especially expensive compared to other device components.