Microscale measurements of oxygen concentration across the thickness of diffusion media in operating polymer electrolyte fuel cells

Journal of Power Sources 306 (2016) 674e684

Contents lists available at ScienceDirect

Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour

Microscale measurements of oxygen concentration across the thickness of diffusion media in operating polymer electrolyte fuel cells William K. Epting, Shawn Litster* Department of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, United States

h i g h l i g h t s

g r a p h i c a l a b s t r a c t


Oxygen concentration measurements in polymer electrolyte fuel cell diffusion media.
Measured in pores during fuel cell operation using 25 mm microsensor probe.
Spatial resolution of 20 mm through the thickness of diffusion media.
Spatially-correlated to structure imaged by micro X-ray computed tomography.
Local O2 transport resistance and liquid water onset locations in diffusion media.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 12 August 2015 Received in revised form 3 November 2015 Accepted 3 December 2015 Available online xxx

Although polymer electrolyte fuel cells (PEFCs) offer promise as efficient, low emission power sources, the large amount of platinum catalyst used for the cathode's oxygen reduction (ORR) results in high costs. One approach to using less Pt is to increase the oxygen concentration at the catalyst by reducing the oxygen transport resistances. An important resistance is that of the diffusion media (DM). The DM are highly heterogeneous porous carbon fiber substrates with a graded composition of additives across their thickness. In this work we use an oxygen microsensor with a micro-positioning system to measure the oxygen concentration and presence of liquid water in the pores at discrete points across the thickness of a commercial carbon felt DM in operating PEFCs. Under conditions with no liquid water, the DM accounts for 60% of the oxygen depletion, with 60e70% of that depletion being due to the thin microporous layer (MPL) on the catalyst layer (CL) side. Using concentration gradient data, we quantify the non-uniform local transport resistance across the DM and relate it to high resolution 3D X-ray computed tomography of the same DM. © 2015 Elsevier B.V. All rights reserved.

Keywords: Polymer electrolyte fuel cell Diffusion media Gas diffusion layer Oxygen sensor X-ray computed tomography In-situ measurements

1. Introduction Polymer electrolyte fuel cell (PEFC) systems have to the

* Corresponding author. Scaife Hall 323, 5000 Forbes Avenue, Pittsburgh, PA 15213, United States. E-mail address: [email protected] (S. Litster). http://dx.doi.org/10.1016/j.jpowsour.2015.12.016 0378-7753/© 2015 Elsevier B.V. All rights reserved.

potential to displace combustion engines and batteries in a wide range of transportation, portable, and stationary power applications because of their high efficiency, power and energy density, and zero exhaust carbon emissions. Despite significant reductions in estimated costs over the past decade, projected mass produced automotive PEFC costs are still twice the desired goal [1,2]. A key contributor to that cost is the raw material cost of the

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

Pt catalyst in the cathode. Despite the notable promise of new low and zero Pt-group metal catalysts [3,4], engineering to enhance the oxygen transport efficacy and increase oxygen activity at the catalyst remains a valuable pathway to retaining cathode performance while reducing Pt loading. The O2 transport resistances include those through the flow field [5,6], DM [7e9], the CL [10e12], catalyst agglomerate [13,14], polymer electrolyte films [12,15e17], and component interfaces [18,19]. Unfortunately, due to the enclosed design of PEFCs and the associated micro/nano length scales of these layers and interfaces, determining a direct breakdown of how each resistance and each feature contributes to the overall transport loss remains a significant challenge. Our focus herein is the oxygen transport resistances specific to the DM and its spatially distinct internal features. The DM is also often referred to as the gas diffusion layer (GDL) or porous transport layer (PTL) [20]. It is typically a porous carbon layer placed between the CLs and the solid gas distribution plates (i.e., bipolar plates in a PEFC stack). It serves to distribute gases and electrical current over the entire area of the CL as well as provide mechanical support to the delicate polymer membrane. Since PEFCs are low temperature fuel cells, they can accumulate liquid product water that can additionally impede oxygen transport. To repel liquid water and maintain gas pathways, the fiber substrate is often coated with hydrophobic polytetrafluoroethylene (PTFE). A microporous layer (MPL) of carbon black and PTFE is commonly applied to the fiber substrate to improve contact with the CL [19,21]. There are a number of ways to characterize oxygen transport losses using global, cell-level measurements, largely focusing on the effect of different operating conditions and electrode compositions on PEFC performance. In-depth analysis of a PEFC's IeV curve, coupled with testing the effects of different gas mixtures, can yield insight into which mechanisms of voltage loss are dominant [22e25]. This distinction includes relative weighting of oxygen transport losses due to Knudsen diffusion in the CL pores or dissolved oxygen diffusion in the ionomer binder versus that of molecular diffusion in the DM [23,24]. In a related but more specific method, researchers also study the effect of different operating conditions on the PEFC's limiting current in order to extract cathode transport parameters and identify the dominant loss mechanisms [17,23,24,26e32]. Most recently, limiting current studies in PEFCs with carefully varied platinum loadings and feed gas conditions have yielded insight into the oxygen transport resistance local to the catalyst [28e31]. A number of modeling [33e37] and experimental [7,38e41] works have studied the interplay between water and oxygen transport in the DM. Imaging studies using synchrotron x-ray imaging [38,39] and neutron radiography [41] have demonstrated that there is significant liquid water throughout the DM in a variety of conditions. Experiments [7,39,40] and pore-network DM modeling [36,37] suggest that water forms breakthrough pathways or “capillary fingers.” As more water is generated, it tends to travel through these established finger-like pathways e rather than forming a new path. Owejan et al. [42] showed that in the absence of MPL cracks, liquid water cannot penetrate the hydrophobic MPL and water is restricted to vapor phase transport. Similarly, Zenyuk et al. [43] showed that introducing large openings in the MPL can enhance liquid water removal from the CL. Outside of actual operating PEFCs, several studies have made exsitu and computational determinations of the oxygen diffusivity of DMs [44e47] providing valuable input to DM and design and parameters for cell-scale models (computational, analytical, and theoretical). Unfortunately, the prior ex-situ characterizations only yield the effective properties of the complete layer and do not

675

resolve the local impact of morphological heterogeneity [45]. While computational characterizations of pore-scale transport in reconstructed volume offer significant spatial detail, they often lack important physics such as Knudsen diffusion and liquid water saturation [48]. In addition, spatially distributed experimental data to validate such models has been lacking. A range of studies have been carried out on PEFC electrodes and hardware concerning in-situ measurements of in-plane distributions of various quantities [49e54]. In-situ measurements within the PEFC's membrane-electrode assembly have been limited. Prior work includes using Pt-wire probes embedded in the membrane to measure the presence of hydrogen peroxide [55,56], oxygen and hydrogen concentrations [52,53,57], and water content [58]. Some of these studies featured multiple probes embedded through the thickness of the membrane to obtain through-plane measurements [53,55,58]. Perhaps the most relevant of these studies to this work are those that amperometrically sensed oxygen concentrations at the cathode CL/PEM interface [52,56,57], one of which did so with multiple probes across the plane of the PEM to obtain in-plane distributions [52]. Other researchers have applied oxygen sensitive paints on transparent plates to measure oxygen concentration along the gas channels and the DM surface [59]. Unfortunately, optical techniques are not amenable to microscale measurements within the porous components. To our knowledge, only one prior existing study reports distributed oxygen concentration measurements through the thickness of DM [60]. However, this intrusive approach first involved milling a large 100 mm hole through the entire DM and then inserting a 45 mm diameter fiber optic sensor through that hole. Thus, the effects of sub-100 mm features and, more critically, the restricted diffusion of the small pores in the MPL are not resolved. In this paper, we present the in-situ characterization of oxygen transport through the diffusion medium (DM) of a working polymer electrolyte fuel cell (PEFC) using an oxygen concentration microsensor. The slender microsensor had a 25 mm tip diameter, made from a pulled glass capillary, which allowed us to measure concentrations within the DM pores. The 25 mm tip diameter is significantly smaller than the characteristic fiber spacing in the DM's carbon fiber felt and should not significantly affect the DM's microstructure or transport, yielding minimally intrusive measurements. Wargo et al.'s [47] pore size characterization showed inplane chord length distribution peaks that were slightly larger than 50 mm in the fiber domain, which is twice the sensor diameter. The sensor tip diameter is also only twice that of the DM's carbon fibers, so its presence should not significantly impact oxygen transport relative to the existing materials, which include dense bundles of fibers. The sensor can also pass through the MPL, as first tested exsitu, only affecting transport locally after the experiment is fully complete as we will discuss. Using a precision stage with a micromanipulator, we measured the spatial distribution of O2 concentration at 20 mm resolution through the thickness of unmodified commercial DMs in operating PEFCs. We performed these measurements at different current densities and different values of relative humidity (RH), on multiple fuel cell builds. We then analyzed those concentration measurements to yield position-dependent transport properties through the DM thickness, measured here inoperando for the first time, and to find where liquid water appears in the DM under different conditions, measured here inoperando for the first time using a non-imaging-based method. Finally, since the transport properties of a porous medium arise from its microstructure, we employed microscale X-ray computed tomography (micro-CT) to characterize the morphology of the DM and give further insight to the results of the oxygen concentration analysis.

676

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

2. Experimental 2.1. Fuel cell hardware The PEFC hardware used in this study was a slightly modified 5 cm2 fuel cell test fixture from Scribner Associates Inc. (Southern Pines, NC), featuring a single serpentine channel 800 mm in width and depth, in both the cathode and anode flow plates; the flow plates themselves are 12.7 mm thick. The membrane electrode assembly (MEA) was a catalyst-coated Nafion NR-211 membrane, with a cathode Pt loading of 0.3 mg cm2 (Ion Power, New Castle, DE). For this study, the original 25 cm2 active area was reduced to 5 cm2 using adhesive, 25 mm thick Kapton masks (DuPont, Wilmington, DE). On both the anode and cathode sides, we used Sigracet 10BC DM (SGL, Wiesbaden, Germany), which is made from carbon felt with a nominal 80% porosity, coated with 5 wt% polytetrafluoroethylene (PTFE), and coated with an MPL on the side that faces the CL. The specified nominal thickness of the 10BC DM is 420 mm. However, the thickness can vary significantly from that value. As we will describe later, we used micro-CT to determine these DMs to be about 300 mm thick. Surrounding the DMs, we employed 230 mm thick PTFE/fiberglass gaskets on both sides during compression, leading to a roughly 30% compression under the lands (ribs) of the flow plate. Note that despite the compression under the lands, it is typical for a DM to remain at its full, uncompressed thickness in the center of a channel of this width [61]. 2.1.1. Microsensor insertion The oxygen microsensor used in this study (OX-25, Unisense A/ S, Århus, Denmark) was made from a pulled glass capillary with a 25 mm tip. Internal to the sensor was a small electrochemical cell for amperometric measurements based on limiting oxygen reduction reaction (ORR) current and included a “guard cathode” set back from the main working electrode designed to consume extraneous oxygen entering from behind. Fig. 1c shows a photograph of the sensor, which was advanced forward in measured increments through the thickness of the cathode DM (the z axis) along a straight, through-thickness line using an experimental setup illustrated in Fig. 1aeb. The measured steps forward in space were accomplished using a micromanipulator (MM-33RT, Warner

Instruments, Hamden, CT) with a 10 mm resolution Vernier micrometer handle for movement in the through-thickness (z, axial to the oxygen sensor) direction. Controlled by the micromanipulator, the O2 microsensor entered the cathode via a 420 mm through-hole to allow sensor entry. The through-hole was placed in the center of the 800 mm inlet channel - the first pass before the serpentine channel turned. It was 15 mm from the channel inlet hole, or roughly 2/3 of the way along the channel before the first serpentine turn. The sensor through-hole was surrounded by a multi-step counter-sink, as illustrated in Fig. 1d, to allow space for the larger rear body of the sensor and to allow visual alignment of the sensor with the hole. The hole and counter-sinks shown in Fig. 1d, along with matching large holes in the hardware's endplate and current collector plate, were the only modifications to the hardware. As we will discuss below, the modification did not affect cell performance significantly. A custom mount ensured that the micromanipulator's z axis was parallel to the axis of the through-hole, and that the micromanipulator could not vibrate or move independently of the PEFC hardware, as lateral movements or vibrations can break the glass microsensor. In-plane alignment of the sensor with the through-hole was accomplished with the X and Y (non-axial) controls of the micromanipulator using visual feedback, aided by a digital microscope (Dino-Lite AM-311S, AnMo Electronics, Taipei, Taiwan). Once the sensor was aligned, the through-thickness (z) position was established by noting the micrometer gauge's reading when the tip of the sensor was microscopically observed to be flush with the outer surface of the counter-sink, indicated by z0 in Fig. 1d and e. Following this, the sensor's position was known relative to the channel and the DM as it was stepped forward in the z direction. Based on the known geometry and using the z coordinate (beginning from the PEM/cathode CL interface) marked in Fig. 1e, z0 has a value of 2.03 mm, channel entry occurs at zch ¼ 1.03 mm, the outer surface of the DM is encountered at zDM ¼ 0.32 mm, and the DM terminates (and the CL is reached) at zCL ¼ 0.02 mm. 2.2. Electrochemical PEFC characterization Gases were humidified and delivered to the PEFC using a

Fig. 1. a) Schematic of oxygen microsensor penetrating into PEFC cathode. b) Photograph of physical setup for sensor and PEFC experiments. c) Photograph of the OX-25 oxygen microsensor. d) Dimensioned drawing of modifications made to the cathode flow plate. e) Detail schematic of sensor, through-hole, flow channel, and DM (to scale).

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

677

Scribner Associates e850 fuel cell test stand, which also controlled the temperature, voltage, and current of the cell. The PEFC was maintained at 60  C through the entirety of this work, except where otherwise noted. In order to achieve differential cell behavior despite the presence of the through-hole, we performed all experiments at a cathode air flow rate of 2 standard liters per minute (SLPM); the anode's hydrogen gas was supplied at 0.5 SLPM and always provided at the same RH as the cathode gas. Each MEA was conditioned on its first day of operation by 8 h of voltage cycling at 0.8, 0.6, and 0.3 V, under 95% RH. Oxygen concentration experiments were performed under four different values of relative humidity: 62%, 72%, 83%, and 91%. For each of these RH cases, a new MEA was built, installed, conditioned, and characterized before the oxygen concentration measurements. For each separate cell build, we verified that the through-hole modification to the cathode flow plate did not significantly affect cell behavior. We did this by measuring two separate polarization curves at 2 SLPM: one with the through-hole sealed by adhesive Kapton, and the other with the hole open to the environment. Similarly, we ensured differential cell behavior by measuring the PEFC current at 0.3 V as the cathode flow rate was stepped from 1.5 to 1.75 to 2 SLPM; this was also done with the hole sealed and with the hole open for comparison. 2.3. Oxygen concentration measurement using oxygen flux interrupt After being positioned as described above, the oxygen microsensor was moved forward in the z direction (along its axis, deeper into the channel/DM) until it entered the channel (zch in Fig. 1b). At that point, we began taking oxygen measurements. The sensor's commercial software expresses its signal in voltage units of mV via the pico-ammeter that accompanies the sensor. Because of this, we will refer to the sensor signal as the voltage, Vsensor, despite the fact that the cO2 measurement inherently comes from a mass transport limited ORR current within the sensor. The oxygen sensor's signal was being recorded constantly, at a sampling rate of 10 Hz, during the entirety of the experiment. Oxygen measurements were taken every 150 mm through the channel. The z step between measurements was refined to 20 mm, beginning at the location ca. 60 mm away from the outer surface of the DM (z z 400 mm), and continuing through the rest of the experiment. At each position, without moving the sensor, oxygen concentration measurements were taken for 4 different PEFC current densities (iFC) e 0.5, 1, 1.5, and 1.8 A/cm2, galvanostatically controlled to ensure constant flux. These currents were selected to represent widely varying levels of oxygen flux and liquid water flooding, with all currents still being accessible in all PEFC builds and at all values of RH. After measuring cO2 at all four PEFC currents, the sensor was advanced further into the DM. Amperometric sensors generally require a calibration curve or correlation generated prior to measurements with known oxygen concentration (cO2) values. However, inside an operating fuel cell the sensor is subject to changing relative humidity and temperature, which changes the sensor's calibration points as the sensors diffusion barrier properties change. The sensor's signal also changes on a day to day basis. Fig. 2a shows the sensor's in-situ variation from RH and temperature without PEFC current. The curve with varying RH was taken at a constant 60  C, and the varying temperature curve was taken at a constant 72% RH. The fact that the two curves do not overlap at 60  C and 70% RH illustrates the day-to-day variation. As Fig. 2a shows, the signal (or apparent cO2) increases with both increased temperature and RH. Therefore, when the PEFC is operating, two competing effects confound precalibrated oxygen measurements: the increased RH and

Fig. 2. a) Sensor signals with the PEFC at open circuit, at different values of RH (at 60  C) and temperature (at 72% RH). b) A typical waveform of the sensor signal during O2 flux interrupt. The difference between Vmeas and VN2 represents the measured O2 concentration, and the difference between Vinterrupt and VN2 is the instantaneous calibration point, representing the O2 concentration of the channel gas.

temperature increase sensor signal, while the diminished concentration of oxygen reduces the signal. To address the continuously changing calibration points, we have a developed a method we term oxygen flux interrupt [62], where we interrupt the PEFC's current in an approach resembling the current interrupt method for measuring PEFCs' Ohmic resistances. When the PEFC current is interrupted, the oxygen concentration returns to the channel-level in a matter of milliseconds, while the changes in the sensor's behavior due to RH and temperature occur over many seconds. The result: an interruption of PEFC current provides an instantaneous calibration point Vinterrupt at cO2(channel), but under the hydration and temperature occurring during the prior measurement. Herein, the cO2(channel) is constant due to a very high stoichiometric ratio and a short air residence time in the channels. In these experiments, each PEFC current hold lasted 3 or 4 min, which we found to be more than enough time for the cell and the sensor to reach a steady state in high and low humidity cases, respectively. Each current interruption (where the PEFC was switched to open circuit) lasted only 5 s. To complete the linear, two-point calibration, we require a second oxygen concentration signal reference point. Here we use the sensor's signal when there is only humidified nitrogen in the cathode, and the PEFC is at open circuit. This value, VN2, was taken in a separate, prior experiment. Its value is typically on the order of 1/100th of the sensor's signal in oxygen, and we have found it to be very consistent over a wide range of RH and temperature. Thus, we take VN2 to be a constant value throughout all experiments. As illustrated in Fig. 2b, the difference between Vmeas (the pre-

678

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

interrupt sensor signal) and VN2 corresponds to the value of cO2 we desire to measure, while the difference between Vinterrupt and VN2 is a measure of cO2 in the channel and is used as an instantaneous calibration point. Given the high flow rate, this is also the inlet oxygen concentration of the humidified air. The expression used to calculate the dimensionless concentration as non-dimensionalized by the channel concentration, cO2 ¼ cO2 =cO2;channel , is:

cO2 ¼

Vmeas  VN2 Vinterrupt  VN2

(1)

2.4. Micro-CT of diffusion medium Since a porous medium's transport properties arise from its microstructure, interpreting the high resolution oxygen concentration measurements can be aided by knowledge of the local microstructural properties. Here we have used a laboratory microCT system (Versa 520, Carl Zeiss X-ray Microscopy, Inc., Pleasanton, CA) to characterize the 3D microstructure of the DM. A biopsy punch (Integra Miltex, Plainsboro, NJ) was used to extract a 1 mm diameter cylindrical sample of the whole thickness of the DM. This system generated a broadband “pink” X-ray beam with an accelerating voltage of 50 kV, and imaged the sample using a 10 objective. The voxel size was 1.5 mm. The sample was imaged across 360 of rotation in 1601 steps, at 1 s exposure time per angle step. The 1601 images were reconstructed using proprietary software and then processed and analyzed using Avizo Fire (FEI, Hillsboro, OR) and MATLAB (Mathworks, Natick, MA). 3. Results and discussion 3.1. Micro-CT of diffusion medium

which is from the same sample batch as those used in the fuel cell assembly. The 3D reconstruction of the cylindrical sample was cropped to a rectangular prism of 640  700  275 mm before phase segmentation. The actual DM extends slightly larger than this in the z direction, to about 300 mm thick, but the bottom-most 25 mm portion was cropped out as it included some artefacts of sample preparation and mounting. We then segmented the raw greyscale reconstruction into 3 phases e fibers, MPL/PTFE-binder, and pore. A virtual slice of the data before and after segmentation can be seen in Fig. 3(a) and (b), respectively. Note that the MPL phase has a porosity of roughly 40% due to pores in the 50e500 nm diameter range [47], which was not captured by the 1.5 mm voxels in the scan here. Instead, the MPL presented as one seemingly solid phase. Additional higher resolution scans, up to the minimum voxel size of the micro-CT, were not able to discern the MPL pores; those require nano-scale resolution CT systems [63] or FIB-SEM [47]. The PTFE and binder on the carbon fiber felt, on the other hand, has some sub-porosity that is observable by micro-CT. The porosity of the PTFE and the binder pores that are too small to resolve is minor enough that it does not significantly impact the overall porosity. The MPL and the PTFE and binder in the carbon fiber felt both had a similar intensity in the reconstruction, so they are both segmented as the same phase, though they are not the same material. The MPL material can be distinguished from the PTFE binder material qualitatively; e.g. in Fig. 3a, the MPL generally does not extend beyond about z ¼ 120 mm and has fewer voids. After segmenting the image, we plotted versus the z axis the phase fractions of the MPL/PTFE-binder phase, the fiber phase, and the pore phase. The results are in Fig. 3c. As expected, the MPL takes up almost 100% of the volume on the MPL-side of the DM e because of the MPL's sub-porosity, the actual porosity in this region would be much higher, comprised of small pores in which Knudsen diffusion is significant. We found that rather than a distinct interface, the MPL tapers off over a fairly wide (50e100 mm) range. Works by Fishman and Bazylak [64] and Wargo et al. [47] on the

Fig. 3 shows micro-CT imaging results for the SGL 10BC DM,

Fig. 3. Micro-CT characterization of the SGL 10BC DM sample, including a) a virtual tomography slice of the 3D reconstruction; b) a virtual slice of the segmented 3D reconstruction, showing the fiber phase (white), the MPL or PTFE-binder phase (grey), and the pore phase (black); c) the distribution of phase fractions as a function of the z coordinate through the DM; d) volume rendering of the 3D CT image.

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

10BC DM show a similar distribution. Following the tapering off of MPL content, there is a particularly high porosity, open region in the middle of the carbon fiber felt, followed by a particularly low porosity fiber/binder/PTFE region at the outer edge of the DM, similar to that observed by Wargo et al. [47]. To summarize, we identify four distinct regions in the DM: 1. 2. 3. 4.

Low porosity fiber region at the channel surface (z > 220 mm), High porosity fiber core region (120 mm < z < 220 mm), Fiber-MPL transition region (50 mm < z < 120 mm), and MPL core region (z < 50 mm).

3.2. Electrochemical characterization of PEFC Initial electrochemical characterization of each PEFC build successfully verified that the hole in the flow plate did not significantly affect cell behavior when using sufficiently high flow rates. Fig. 4 presents polarization curves for each of the 4 values of RH studied here, on the 4 different MEAs used. Some variation can be seen due to the particular MEA build e for example, the MEA used in the 72% RH case exhibits a higher Ohmic loss, likely due to assembly features outside the MEA. However, all of the MEAs at all of the RH conditions perform adequately for characterizing O2 transport in these experiments, considering that the cell's current density is controlled galvanostatically with high flow rates to ensure consistent oxygen flux. Moreover, all of the polarization curves show negligible difference between the sealed and open-hole case (solid vs. dashed lines), indicating little to no effect from the modification to the hardware at this flow rate. The inset in Fig. 4 shows the PEFC currents (normalized for each case by that its maximum value) at 0.3 V as the cathode gas flow rate is varied. All 4 RH cases with open and sealed holes are shown. All of the curves lie nearly on top of each other, with the highest flow rate of 2 SLPM giving only 5% more current density than the second-highest of 1.75 SLPM. This indicates that the PEFC is operating as a differential cell at flow rates of 2 SLPM, whether the through-hole is open or sealed.

679

the DM, the sensor's signal was logged constantly as oxygen flux interrupt procedures were performed. Fig. 5a shows an example of a normal sensor signal during all four current holds of an oxygen flux interrupt procedure at one z position. The sensor signal is steady as the PEFC holds current, and quickly jumps up to the calibration value corresponding to cO2(channel) when the PEFC is switched to open circuit for 5 s. During the course of the experiments, we identified that the sensor signal would become erratic upon contact with liquid water. Fig. 5b shows an example of erratic, anomalous sensor behavior indicating liquid water. This sensor behavior occurs only within the DM, and it occurs more often at higher values of RH and at higher PEFC currents e all consistent with liquid water flooding. The erratic behavior ceases within a few seconds when the PEFC current is stopped, which is consistent with the liquid water clearing away once it is no longer being produced. The anomalous sensor signal in liquid water makes the extraction of a cO2 value impossible e not only from a signal processing standpoint, but also because the liquid water significantly increases the timescale for cO2 to return to the channel calibration value when PEFC current is interrupted. Therefore, in the proceeding analysis, we consider the z-positions, currents, and RH values where the sensor exhibits erratic behavior to be flooded zones in the DM, and do not express a particular cO2 value for the flooded points.

3.3. Oxygen concentration measurements As the oxygen microsensor advanced through the channel and

Fig. 4. Polarization curves for all cell builds at all RH values studied, at a cathode flow rate of 2 SLPM, with the through-hole sealed (dashed line) and open (solid line). Inset is the PEFC current density (normalized by the max value in each case) at 0.3 V versus cathode gas flow rates for the sensor flow field through-hole sealed (dashed line) and open (solid line).

Fig. 5. Anecdotal time series of oxygen microsensor signal during oxygen flux interrupt procedure, in (a) a case with no liquid water (83% RH, near the outer surface of the DM) and (b) a case with significant liquid water flooding at all currents (91% RH, just within the DM).

680

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

Anecdotally, in the 72% RH experiment performed in this work, the sensor was advanced all the way through the DM's microporous layer (MPL). After the break-through and after all cO2 readings were complete, the sensor's readings during withdrawal from the cell indicated liquid water at all but the lowest current density, which was different from the previous behavior when the sensor was first advancing through the DM. We believe that the sensor had created a new 25 mm hole in the hydrophobic MPL that allowed facile outflow of water consistent with our group's prior MPL experiments and modeling [19]. To test this hole-creation hypothesis, during the same day but after the 72% RH experiment, we lowered the PEFC's RH to 62%. The erratic signal ceased. We measured cO2 at several z locations through the DM, including within the MPL, and all of the resulting cO2 measurements matched the non-flooded measurements from the 72% RH experiment. This demonstrated not only that the through-hole had created increased flooding in that region, but also that in the absence of liquid water, the hole in the DM created by the sensor did not affect cO2 through the DM. As further evidence for this, in other dry cases, the sensor could obtain repeatable readings when retreating back to an earlier z location, so long as the MPL had not been broken through. Considering that the existence of a hole in the DM ahead of the sensor exacerbates flooding, this method is particularly well-suited to determine the location of liquid water in the DM, particularly when compared to the oxygen probing method of Kobayashi et al. [60], which required a 100 mm hole drilled through the entire DM prior to the experiment. The oxygen flux interrupt procedure yielded measurements for four cases of RH (each with a separate MEA), and at four different PEFC currents, which are all plotted together in Fig. 6. Each concentration curve indicates where liquid water was encountered, if

at all, in that particular case. If flooding was observed, it persisted at every further measurement point deeper into the DM at that current density, but as noted previously, the flooding ceased when PEFC current was ceased. Also included in Fig. 6 is a virtual 2D slice of the 3D micro-CT reconstruction of the 10BC DM used in this study, shown to scale with the x-axis, to aid in interpreting the cO2 profiles. We also note here that the 72% RH case was the first experiment performed. In this case, the sensor was able to penetrate all the way through the MPL, as described above. That particular commercial oxygen sensor later broke in a separate experiment, and was replaced with another sensor of the same type. The sensors are made to order by the vendor. The second sensor was more prone to deflection that caused erroneous signal and could not penetrate through the MPL. For this reason, the 62% RH curves cease at roughly z ¼ 100 mm, where the sensor is first encountering the MPL and likely a rigid, solid fiber bundle embedded in the MPL (common in the micro-CT data). The 83% and 91% experiments encountered the same short circuit phenomenon, but it is not evident in Fig. 6 because liquid water flooding set in before the sensor reached the depth of the MPL. Other than the onset of liquid water, the different RH values' cO2 curves lie on top of each other in each of Fig. 6's sub-plots. Despite the sensor's behavior changing from day to day, and despite the DM and MEA being changed between each RH case, the cO2 measurements were repeatable across space (z) and PEFC current. Consistent transport properties between builds and RH values should be expected e the DM's microstructure does not change significantly based on RH, other than liquid water flooding. The results in Fig. 6 can now be used to extract the non-flooded oxygen transport properties as well as to evaluate where and under what conditions liquid water flooding occurs in the DM.

Fig. 6. Plots of dimensionless cO2 at four different values of RH, and at four different PEFC currents, indicated on each figure. A virtual tomography slice from micro-CT of the DM is included, to scale, to illustrate microstructural features affecting the concentration profiles.

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

3.4. Oxygen transport properties A detailed inspection of the cO2 profile at the interface between the channel gas and the DM surface can yield a mass transfer coefficient hm for the mass transfer boundary layer there. We performed this calculation based on the definition n_ O2 ¼ hm ðc∞  csurf Þ, using the data from the 62% RH and 83% RH cases, cases where smaller increments were applied at the channel surface. The oxygen flux n_ O2 is known from the cell current density iFC using Faraday's law of electrolysis. The concentrations c∞ and csurf are the values of cO2 in the channel (known) and at the surface of the DM (measured), respectively. The 62% RH and 83% RH cases agreed well across all current densities, and yielded an average mass transfer coefficient hm of 0.14 ± 0.02 m s1. The corresponding Sherwood number (Sh ¼ hL=Dhyd ) is 4.4 ± 0.6, which agrees well with the literature value 3.4 for fully-developed single phase flow and a varying increase to above 5 when liquid water droplets are present [65]. Moving deeper into the DM, the local slope dc=dz at any point (where there is no flooding) allows us to calculate the formation factor Kform by which the microstructure reduces the diffusivity, according to the following variation of Fick's law of diffusion:

n_ O2 ¼ Kform DO2=N2

dc dz

(2)

where, again, the oxygen flux n_ O2 is known from the cell current density iFC according to Faraday's law of electrolysis; and the binary diffusivity DO2/N2 is 2.52  105 m2 s1, a value corrected for temperature from the CRC Handbook [66]. In other words, Kform in a porous medium is the ratio of the effective diffusivity to the binary diffusivity, or Kform ¼ Deff =DO2=N2. As presented here, the Kform value also incorporates the effects of Knudsen diffusion since it is based on the molecular diffusion coefficient. Fig. 7 presents the values of Kform as a function of location in the DM for the four cases with sufficiently high O2 flux and with no liquid water flooding. All four curves agree well with each other, despite the changing RH and current density. Values of Kform that were calculated from unflooded iFC cases of 0.5 and 1 A cm2 agreed in value, but contained significant noise due to the lower signal with lower O2 flux and hence are not shown. Note that the 72% RH curves do not extend as far to the right in the figure (higher z values) because we advanced by a large of z-step when first entering the DM during that experiment. The 62% RH curves do not extend through the MPL (to the left) due to the previously described difficulty in penetrating the MPL with the second sensor in that case. The Kform curves in Fig. 7 have several notable features. Moving from right to left (channel to CL), the 62% RH case shows a significantly lower Kform upon initial entry into the DM, before moving to higher values. Lower values of Kform mean a higher transport resistance, due to lower porosity, more tortuous transport pathways, or other changes in diffusion behavior (e.g. Knudsen diffusion). The lower Kform at the DM entrance indicates some initial transport blockage close to the outer surface of the DM that is less severe deeper into the DM. In our micro-CT reconstruction of a 10BC sample (see Fig. 3), we did in fact observe a matching feature e a significantly lower porosity (0.55 minimum) closest to the outer surface relative to a much higher porosity in the middle (0.79 maximum). From the images, we observe a higher fraction of PTFE and binder near the outer surface of the DM that significantly reduces porosity. Based on as commonly assumed Bruggeman exponent of 1.5, the expected ratio of theoretical Kform (Kform ¼ ε1:5 ) for the low porosity fiber surface layer versus the high porosity fiber core is 0.58, and the experiments show a similar value of roughly 0.5 (i.e., a 50% reduction in the effective diffusion

681

coefficient in the low porosity fiber layer at the channel). The Kform results from the 72% RH case have similar values to the 62% case, ranging mostly from 0.5 to 0.7 in the high porosity fiber core region. The measurements in this case extend farther into the DM, into the MPL region. There is a gradual decline in Kform consistent with the fiber-MPL transition region seen in Fig. 3. In the MPL core, the formation factor drops to the 0.1e0.2 range. When we compare the formation factor with the porosity distribution in Fig. 3, it is clear that the formation factors follow a similar trend, but are consistently lower than the porosity. This is expected, as formation factor depends on a number of factors e porosity, tortuosity, and other transport effects like pore size dependent Knudsen diffusion [10]. For example, the Kform values in the carbon fiber phase, not including the low porosity fiber region near the channel, average to about 0.6, compared to a porosity of roughly 0.75 in that region. A common expression for the formation factor is the porosity divided by the tortuosity factor (Kform ¼ ε=t), where the tortuosity factor is the square of the average actual transport pathway length divided by the straight-line length through a porous medium [10]. Using the Kform expression, we estimate a tortuosity factor (t) value of 1.25. Wargo et al. [47] computed an overall Kform for the combined low porosity surface and high porosity core fiber regions in series and found a value of 0.45, which agrees well with the spatial-average of our spatially-resolved Kform. The relationship between Kform and porosity is somewhat different in the MPL, due to the smaller pore sizes. MPLs have pores on the order of 100 nm [47]. At this small pore size, gas diffusion occurs in the Knudsen transition regime, and the effective binary diffusivity is reduced to roughly half as can be estimated from the Bosanquet formulation [10]. Furthermore, the MPL's micropores are significantly more tortuous than the carbon fiber phase [47]. If we examine only the values of Kform in the MPL core, which is has an average of approximately 0.12, we can assume that this corresponds with the nearly 100% MPL region in Fig. 3. Using Wargo et al.’s [47] simulated MPL molecular diffusion formation factor of 0.23, we can estimate the Knudsen correction factor to be 0.52. As presented by Mu et al. [67], the Knudsen correction factor is the ratio of the effective Knudsen transition diffusivity to the diffusivity in the absence of Knudsen effects. In their work, Mu et al. found the factor range of 0.5e0.6 for a mean pore diameter range of 200e300 nm, respectively. This range of the Knudsen correction factor agrees well with our estimate of 0.52, given that the Wargo et al. results showed a mean pore diameter of 320 nm in the MPL. To summarize the oxygen transport properties, Table 1 lists our average values of Kform in the MPL region and the carbon fiber region, but not Kform in the entire DM e since oxygen transport resistance is dominated by the MPL, it does not make sense to simply average Kform through the entire DM, only in each sub-layer that has a consistent microstructure. Instead, we have also calculated the oxygen transport resistance RO2 for the entire DM, and for those same sub-regions, based on an electrical resistor analogy used with Fick's law of diffusion:

RO2 ¼

DcO2 n_ O2

(3)

Table 1 Summary of Kform and RO2 values. The MPL's Kform includes the Knudsen diffusion correction factor. Region (z range)

Average dry Kform

Average dry RO2 [s m1]

Entire DM (20e320 mm) MPL (20e100 mm) Carbon fiber (150e320 mm)

e 0.12 0.5

65 41 20

682

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

The Kform results in Table 1 are averages from the cases presented in Fig. 7, and RO2 was calculated from the concentrations in the 62% and 72% RH experiments in Fig. 6. Note that the values for RO2 in the sub-regions do not exactly add up to the total value because RO2 is not reported for the fiber-MPL transition region because of its ambiguity. The results show that the MPL core is responsible for 2/3 of the O2 transport resistance in dry conditions. The results tabulated here compare reasonably with several other works that have measured the overall oxygen transport resistance through the DM as a whole. For comparison, Chan et al. [45] measured Kform through an MPL-free SGL 10BA DM to be 0.36, which is reasonably close to the value of 0.5 for the carbon fiber region in Table 1 (the 10BA is identical to the 10BC without an MPL). Caulk and Baker [68] used limiting current methods to measure a value of roughly 50 s m1 for RO2, at atmospheric pressure, in an operating PEFC; however, this was for a slightly thinner, carbonfiber-only Toray 060 DM. Beuscher [27] used in-situ limiting current methods to measure an RO2 of 24 s m1 for a CARBEL DM, which is also MPL-free and roughly 230 mm thick e this would also be comparable to the carbon fiber region in the this work and agrees well with the corresponding value of 20 s m1 in Table 1. For unflooded conditions, we can also estimate the fraction of the total oxygen transport resistance that is due to the DM by evaluating the limiting current densities and the oxygen concentrations near the CL interface at a given current. Here we use the apparent limiting current density of 2.8 A cm2 for dry operation from the 62% RH and the 0.68 normalized oxygen concentration at the CL for 1.5 A cm2 that was obtained at 72% RH. Using the ratio of the current densities, we estimate the normalized oxygen concentration at the CL/MPL interface is 0.4. Thus, the DM is roughly responsible for 60% of the oxygen depletion and the total transport resistance. The oxygen concentration profiles in Fig. 6 also illustrate where liquid water flooding begins and under what conditions. We have compiled the z locations where the sensor first encountered flooding in each given case; Fig. 8 presents those locations for all values of PEFC current, and for all values of RH except for 62% RH, which had no flooding at all. Once flooding was encountered, the flooding behavior returned at each deeper point in z at that value of iFC for the remainder of that experiment. However, within the same experiment, a lower or even a higher current might still not show flooding until a deeper z location, or perhaps never at all (as in the

case of the lower currents in the 72% RH experiment). This shows that upon switching to lower currents, liquid water was either retreating towards the catalyst layer or evaporating. In other words, experimentally, the flooding was a space, RH, and PEFC current dependent phenomenon with no apparent memory for the previous value of iFC. Fig. 8 shows that at 91% RH, liquid water flooding was present at the outer DM surface and continued throughout the DM thickness. The 83% RH case shows liquid water flooding at all PEFC currents, with all of it beginning within the highly porous zone in the middle of the carbon fiber layer. Finally, only the highest current of 1.8 A cm2 exhibited any flooding in the 72% RH experiment, with a flooding onset within the MPL region. This is an unexpected finding, and could indicate that the sensor was passing through or near a small pore or a crack common to these MPLs and where liquid water is likely to be found [19]. As suggested previously [19], deliberate introduction of similar cracks or holes in the MPL could manage water removal in a way that prevents such widespread flooding. We can compare our liquid water onset measurements to the growing amount of work on high resolution, in-operando water imaging in PEFCs. Our measurement of the point of water onset at 91% RH and a lack thereof at 62% RH agrees with LaManna et al.'s [69] neutron radiography work measuring water distributions in PEFCs, which show liquid water throughout all of the DM other than the MPL in 95% RH, and nearly no liquid water at 50% RH. On the other hand, Eller et al. [38] observed significant liquid water throughout the DM even at 40% RH using synchrotron micro-CT; however, that group also found the synchrotron beam to have significant effects on the PEFC behavior and water content [70]. Another neutron radiography study by Fu et al. [41] found less liquid water within the DM's high porosity fiber core at 95% RH, but a significant uptick in liquid water content at the outermost surface of the DM, consistent with our observation of outer DM surface flooding at 91% RH. Yet another neutron radiography study by Kotaka et al. [71] found at 100% RH significant liquid water in the high porosity fiber core that reduced towards the outer surface, possibly due to physical occlusion of the water due to the lower porosity we see in Fig. 3 of the present work. All of these studies employ conditions and DM materials that are inconsistent with this study, and none of them study an intermediate humidity (e.g., 83%), where we observed no flooding until the high porosity fiber core

Fig. 7. Formation factor Kform for 62% and 72% RH cases at PEFC current densities of 1.5 and 1.8 A/cm2. A virtual tomography slice from micro-CT of the DM is included, to scale, to illustrate microstructural features affecting the values of Kform.

Fig. 8. Location of water onset at different values of RH, plotted versus PEFC current density. Location values of 0 indicate no flooding at all. A virtual tomography slice from micro-CT of the DM is included, to scale, to illustrate the location of flooding with respect to microstructural features.

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684

region was reached. However, it is worth noting that other in-situ studies of through-plane water distributions in PEFC DMs at high RH have found significant liquid water throughout at least the carbon fiber portion of the DM, a phenomenon we have mapped out in space, RH, and PEFC current in Fig. 8. We also note again the excessively high stoichiometry in this work, which has the effect of reducing liquid water accumulation. This means that at lower cathode stoichiometry ratios, flooding could more prevalent than what we measured here.

[7] [8]

[9]

[10]

4. Conclusions In this paper, we used a commercially available oxygen microsensor with a 25 mm tip to measure the oxygen concentration profile at 20 mm resolution through the thickness of a 10BC DM in an operating PEFC cathode. We analyzed the resulting concentration profiles alongside a microstructural characterization using micro-CT. From those measurements, we calculated a surface mass transfer coefficient and a distribution of the formation factor for diffusion through the DM that agrees with microstructural properties in this work and others, but until now had not been measured in-operando. This work provides the first data set from a single DM that separately identifies and spatially resolves the transport resistance in the four domains of the DM we studied, which include the low porosity fiber surface layer, the high porosity fiber core, the fiber-MPL transition, and the MPL core. The fiberMPL transition zone in particular has been previously imaged, but its effect on transport has not been measured until now. This data can be integrated into the optimization of those separate DM components in future DM design, and the technique itself can be used to further evaluate future DM designs. By evaluating liquid water onset data with a differential cell, we found liquid water flooding occurs throughout the DM when the cathode gas is 91% RH; it begins within the high porosity fiber core region at 83% RH; it only occurs at high currents at 72% RH; and no flooding occurs at all at 62% RH. Our microstructural characterization together with the flooding onset data revealed that the carbon fiber layer's highly porous core (compared to its denser surface) may be a common site for flooding to occur, suggesting that future DM designs may benefit from a more uniform porosity through the thickness of the fiber layer.

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

Acknowledgments

[23]

The authors gratefully acknowledge the assistance of Jeff Gelb and Arno Merkle of Carl Zeiss X-ray Microscopy, Inc. in providing CT imaging of the DM sample. This material is based upon work supported by the National Science Foundation under grant no. 1053752.

[24]

References [1] D. Papageorgopoulos, Fuel cells overview, in: US Dept. Energy Annu. Merit Rev. Proc., 2015. http://www.hydrogen.energy.gov/pdfs/review15/fc000_ papageorgopoulos_2015_o.pdf. [2] Fuel Cell Technologies Office Multi-Year Research, Development and Demonstration Plan, U.S. Dept. Energy, 2015. http://energy.gov/eere/fuelcells/ downloads/fuel-cell-technologies-office-multi-year-research-developmentand-22. [3] G. Wu, K.L. More, C.M. Johnston, P. Zelenay, High-performance electrocatalysts for oxygen reduction derived from polyaniline, iron, and cobalt, Science 332 (2011) 443e447, http://dx.doi.org/10.1126/science.1200832. [4] C. Chen, Y. Kang, Z. Huo, Z. Zhu, W. Huang, H.L. Xin, et al., Highly crystalline multimetallic nanoframes with three-dimensional electrocatalytic surfaces, Science 343 (2014) 1339e1343, http://dx.doi.org/10.1126/science.1249061. [5] N.S. Siefert, S. Litster, Voltage loss and fluctuation in proton exchange membrane fuel cells: the role of cathode channel plurality and air stoichiometric ratio, J. Power Sources 196 (2011) 1948e1954. [6] J.P. Owejan, T.A. Trabold, J. Gagliardo, D.L. Jacobson, R.N. Carter, D.S. Hussey, et

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

683

al., Voltage instability in a simulated fuel cell stack correlated to cathode water accumulation, J. Power Sources 171 (2007) 626e633. :// 000250066900049. S. Litster, D. Sinton, N. Djilali, Ex situ visualization of liquid water transport in PEM fuel cell gas diffusion layers, J. Power Sources 154 (2006) 95e105. P.K. Sinha, P.P. Mukherjee, C.Y. Wang, Impact of GDL structure and wettability on water management in polymer electrolyte fuel cells, J. Mater. Chem. 17 (2007) 3089e3103. ://000248335300021. F. Büchi, J. Eller, F. Marone, M. Stampanoni, Determination of local GDL saturation on the pore level by in situ synchrotron based X-ray tomographic microscopy, ECS Trans. 33 (2010) 1397e1405. http://link.aip.org/link/ abstract/ECSTF8/v33/i1/p1397/s1. S. Litster, W.K. Epting, E.A. Wargo, S.R. Kalidindi, E.C. Kumbur, Morphological analyses of polymer electrolyte fuel cell electrodes with nano-scale computed tomography imaging, Fuel Cells 13 (2013) 935e945, http://dx.doi.org/ 10.1002/fuce.201300008. G.Q. Wang, P.P. Mukherjee, C.Y. Wang, Optimization of polymer electrolyte fuel cell cathode catalyst layers via direct numerical simulation modeling, Electrochim. Acta 52 (2007) 6367e6377. ://000247834800009. J.P. Owejan, J.E. Owejan, W. Gu, Impact of platinum loading and catalyst layer structure on PEMFC performance, J. Electrochem. Soc. 160 (2013) F824eF833, http://dx.doi.org/10.1149/2.072308jes. W. Sun, B.A. Peppley, K. Karan, An improved two-dimensional agglomerate cathode model to study the influence of catalyst layer structural parameters, Electrochim. Acta 50 (2005) 3359e3374. W.K. Epting, S. Litster, Effects of an agglomerate size distribution on the PEFC agglomerate model, Int. J. Hydrogen Energy 37 (2012) 8505e8511, http:// dx.doi.org/10.1016/j.ijhydene.2012.02.099. N. Nonoyama, S. Okazaki, A.Z. Weber, Y. Ikogi, T. Yoshida, Analysis of oxygentransport diffusion resistance in proton-exchange-membrane fuel cells, J. Electrochem. Soc. 158 (2011) B416eB423, http://dx.doi.org/10.1149/ 1.3546038. H. Liu, W.K. Epting, S. Litster, Gas transport resistance in polymer electrolyte thin films on oxygen reduction reaction catalysts, Langmuir 31 (2015) 9853e9858, http://dx.doi.org/10.1021/acs.langmuir.5b02487. T. Suzuki, K. Kudo, Y. Morimoto, Model for investigation of oxygen transport limitation in a polymer electrolyte fuel cell, J. Power Sources 222 (2013) 379e389, http://dx.doi.org/10.1016/j.jpowsour.2012.08.068. T. Swamy, E.C. Kumbur, M.M. Mench, Characterization of interfacial structure in PEFCs: water storage and contact resistance model, J. Electrochem. Soc. 157 (2010) B77eB85, http://dx.doi.org/10.1149/1.3247585. I.V. Zenyuk, R. Taspinar, A.R. Kalidindi, E.C. Kumbur, S. Litster, Computational and experimental analysis of water transport at component interfaces in polymer electrolyte fuel cells, J. Electrochem. Soc. 161 (2014) F3091eF3103, http://dx.doi.org/10.1149/2.0161411jes. J.P. James, H.-W. Choi, J.G. Pharoah, X-ray computed tomography reconstruction and analysis of polymer electrolyte membrane fuel cell porous transport layers, Int. J. Hydrogen Energy 37 (2012) 18216e18230, http:// dx.doi.org/10.1016/j.ijhydene.2012.08.077. M.F. Mathias, J. Roth, J. Flemming, W. Lehnert, Diffusion media materials and characterization, in: W. Vielstich, A. Lamm, H.A. Gasteiger (Eds.), Handb. Fuel Cells, 2003, pp. 517e537. J. Benziger, E. Kimball, R. Mejia-Ariza, I. Kevrekidis, Oxygen mass transport limitations at the cathode of polymer electrolyte membrane fuel cells, AIChE J. 57 (2011) 2505e2517, http://dx.doi.org/10.1002/aic.12455. M.V. Williams, H.R. Kunz, J.M. Fenton, Analysis of polarization curves to evaluate polarization sources in hydrogen/air PEM fuel cells, J. Electrochem. Soc. 152 (2005) A635, http://dx.doi.org/10.1149/1.1860034. S. Sambandam, V. Ramani, Influence of binder properties on kinetic and transport processes in polymer electrolyte fuel cell electrodes, Phys. Chem. Chem. Phys. 12 (2010) 6140e6149, http://dx.doi.org/10.1039/b921916a. Y.W. Rho, S. Srinivasan, Y.T. Kho, Mass transport phenomena in proton exchange membrane fuel cells using O2/He, O2/Ar, and O2/N2 Mixtures: II. theoretical analysis, J. Electrochem. Soc. 141 (1994) 2089, http://dx.doi.org/ 10.1149/1.2055066. J. St-Pierre, B. Wetton, G.-S. Kim, K. Promislow, Limiting current operation of proton exchange membrane fuel cells, J. Electrochem. Soc. 154 (2007) B186, http://dx.doi.org/10.1149/1.2401045. U. Beuscher, Experimental method to determine the mass transport resistance of a polymer electrolyte fuel cell, J. Electrochem. Soc. 153 (2006) A1788, http://dx.doi.org/10.1149/1.2218760. Y. Ono, T. Mashio, S. Takaichi, A. Ohma, H. Kanesaka, K. Shinohara, The analysis of performance loss with low platinum loaded cathode catalyst layers, ECS Trans. 28 (2010) 69e78, http://dx.doi.org/10.1149/1.3496614. N. Nonoyama, S. Okazaki, A.Z. Weber, Y. Ikogi, T. Yoshida, Analysis of oxygentransport diffusion resistance in proton-exchange-membrane fuel cells, J. Electrochem. Soc. 158 (2011) B416eB423, http://dx.doi.org/10.1149/ 1.3546038. T.A. Greszler, D. Caulk, P. Sinha, The impact of platinum loading on oxygen transport resistance, J. Electrochem. Soc. 159 (2012) F831eF840, http:// dx.doi.org/10.1149/2.061212jes. J.P. Owejan, J.E. Owejan, W. Gu, Impact of platinum loading and catalyst layer structure on PEMFC performance, J. Electrochem. Soc. 160 (2013) F824eF833, http://dx.doi.org/10.1149/2.072308jes. D.R. Baker, D.A. Caulk, K.C. Neyerlin, M.W. Murphy, Measurement of oxygen

684

[33] [34] [35]

[36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

W.K. Epting, S. Litster / Journal of Power Sources 306 (2016) 674e684 transport resistance in PEM fuel cells by limiting current methods, J. Electrochem. Soc. 156 (2009) B991, http://dx.doi.org/10.1149/1.3152226. T. Berning, N. Djilali, A 3D, multiphase, multicomponent model of the cathode and anode of a PEM fuel cell, J. Electrochem. Soc. 150 (2003) A1589eA1598. A.Z. Weber, R.M. Darling, J. Newman, Modeling two-phase behavior in PEFCs, J. Electrochem. Soc. 151 (2004) A1715, http://dx.doi.org/10.1149/1.1792891. J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, Pore network modeling of fibrous gas diffusion layers for polymer electrolyte membrane fuel cells, J. Power Sources 173 (2007) 277e290, http://dx.doi.org/10.1016/ j.jpowsour.2007.04.059. P.K. Sinha, C.-Y. Wang, Pore-network modeling of liquid water transport in gas diffusion layer of a polymer electrolyte fuel cell, Electrochim. Acta 52 (2007) 7936e7945, http://dx.doi.org/10.1016/j.electacta.2007.06.061. E.F. Medici, J.S. Allen, The effects of morphological and wetting properties of porous transport layers on water movement in PEM fuel cells, J. Electrochem. Soc. 157 (2010) B1505, http://dx.doi.org/10.1149/1.3474958. J. Eller, T. Rosen, F. Marone, M. Stampanoni, A. Wokaun, F.N. Buechi, Progress in in situ X-Ray tomographic microscopy of liquid water in gas diffusion layers of PEFC, J. Electrochem. Soc. 158 (2011) B963eB970, http://dx.doi.org/ 10.1149/1.3596556. J. Lee, J. Hinebaugh, a. Bazylak, Synchrotron X-ray radiographic investigations of liquid water transport behavior in a PEMFC with MPL-coated GDLs, J. Power Sources 227 (2013) 123e130, http://dx.doi.org/10.1016/ j.jpowsour.2012.11.006. J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, On the role of the microporous layer in PEMFC operation, Electrochem. Commun. 11 (2009) 576e579, http://dx.doi.org/10.1016/j.elecom.2008.12.053. R.S. Fu, U. Pasaogullari, T. Shiomi, Y. Tabuchi, D.S. Hussey, D.L. Jacobson, Highresolution neutron radiography of through-plane liquid water distribution in polymer electrolyte membrane and gas diffusion layer, J. Electrochem. Soc. 159 (2012) F545eF553, http://dx.doi.org/10.1149/2.026209jes. J.P. Owejan, J.E. Owejan, W. Gu, T.A. Trabold, T.W. Tighe, M.F. Mathias, Water transport mechanisms in PEMFC gas diffusion layers, J. Electrochem. Soc. 157 (2010) B1456eB1464, http://dx.doi.org/10.1149/1.3468615. I.V. Zenyuk, E.C. Kumbur, S. Litster, Deterministic contact mechanics model applied to electrode interfaces in polymer electrolyte fuel cells and interfacial water accumulation, J. Power Sources 241 (2013) 379e387, http://dx.doi.org/ 10.1016/j.jpowsour.2013.03.165. R. Flückiger, S.A. Freunberger, D. Kramer, A. Wokaun, G.G. Scherer, F.N. Büchi, Anisotropic, effective diffusivity of porous gas diffusion layer materials for PEFC, Electrochim. Acta 54 (2008) 551e559, http://dx.doi.org/10.1016/ j.electacta.2008.07.034. C. Chan, N. Zamel, X. Li, J. Shen, Experimental measurement of effective diffusion coefficient of gas diffusion layer/microporous layer in PEM fuel cells, Electrochim. Acta 65 (2012) 13e21, http://dx.doi.org/10.1016/ j.electacta.2011.12.110. J. Becker, R. Flückiger, M. Reum, F.N. Büchi, F. Marone, M. Stampanoni, Determination of material properties of gas diffusion layers: experiments and simulations using phase contrast tomographic microscopy, J. Electrochem. Soc. 156 (2009) B1175, http://dx.doi.org/10.1149/1.3176876. E.A. Wargo, V.P. Schulz, A. Çeçen, S.R. Kalidindi, E.C. Kumbur, Resolving macroand micro-porous layer interaction in polymer electrolyte fuel cells using focused ion beam and X-ray computed tomography, Electrochim. Acta 87 (2013) 201e212, http://dx.doi.org/10.1016/j.electacta.2012.09.008. A. Cecen, E.A. Wargo, A.C. Hanna, D.M. Turner, S.R. Kalidindi, E.C. Kumbur, 3-D microstructure analysis of fuel cell materials: spatial distributions of tortuosity, void size and diffusivity, J. Electrochem. Soc. 159 (2012) B299eB307, http://dx.doi.org/10.1149/2.068203jes. D.J.L. Brett, S. Atkins, N.P. Brandon, V. Vesovic, N. Vasileiadis, A. Kucernak, Localized impedance measurements along a single channel of a solid polymer fuel cell, Electrochem. Solid-State Lett. 6 (2003) A63, http://dx.doi.org/ 10.1149/1.1557034. D.J.L. Brett, S. Atkins, N.P. Brandon, V. Vesovic, N. Vasileiadis, A.R. Kucernak, Measurement of the current distribution along a single flow channel of a solid polymer fuel cell, Electrochem. Commun. 3 (2001) 628e632, http:// dx.doi.org/10.1016/s1388-2481(01)00234-x. X.-G. Yang, N. Burke, C.-Y. Wang, K. Tajiri, K. Shinohara, Simultaneous measurements of species and current distributions in a PEFC under low-humidity

[52]

[53]

[54]

[55] [56] [57]

[58]

[59]

[60]

[61]

[62]

[63]

[64]

[65]

[66] [67]

[68]

[69]

[70]

[71]

operation, J. Electrochem. Soc. 152 (2005) A759, http://dx.doi.org/10.1149/ 1.1864492. S. Takaichi, H. Uchida, M. Watanabe, In situ analysis of oxygen partial pressure at the cathode catalyst layer/membrane interface during PEFC operation, Electrochim. Acta 53 (2008) 4699e4705, http://dx.doi.org/10.1016/ j.electacta.2008.01.094. S. Takaichi, H. Uchida, M. Watanabe, Distribution profile of hydrogen and oxygen permeating in polymer electrolyte membrane measured by mixed potential, Electrochem. Commun. 9 (2007) 1975e1979. R. Montanini, G. Squadrito, G. Giacoppo, Measurement of the clamping pressure distribution in polymer electrolyte fuel cells using piezoresistive sensor arrays and digital image correlation techniques, J. Power Sources 196 (2011) 8484e8493, http://dx.doi.org/10.1016/j.jpowsour.2011.06.017. W. Liu, D. Zuckerbrod, In situ detection of hydrogen peroxide in PEM fuel cells, J. Electrochem. Soc. 152 (2005) A1165, http://dx.doi.org/10.1149/1.1904988. S. Sambandam, V. Ramani, In situ measurement of oxygen permeability in polymer electrolyte membranes, ECS Trans. 25 (2009) 433e441. T. Mashio, A. Ohma, K. Shinohara, Advanced in-situ analysis of reactant gas partial pressure at catalyst layer/PEM interfaces, ECS Trans. 16 (2008) 1009e1018. F.N. Büchi, G.G. Scherer, Investigation of the transversal water profile in Nafion membranes in polymer electrolyte fuel cells, J. Electrochem. Soc. 148 (2001) A183, http://dx.doi.org/10.1149/1.1345868. Y. Ishigami, W. Waskitoaji, M. Yoneda, K. Takada, T. Hyakutake, T. Suga, et al., Oxygen partial pressures on gas-diffusion layer surface and gas-flow channel wall in polymer electrolyte fuel cell during power generation studied by visualization technique combined with numerical simulation, J. Power Sources 269 (2014) 556e564, http://dx.doi.org/10.1016/j.jpowsour.2014.07.017. T. Kobayashi, M. Uchida, J. Inukai, Y. Nakama, T. Ohno, Y. Nagumo, et al., Oxygen partial pressure inside gas diffusion layer of PEFC during power generation using micro probe, in: 226th Meet. Electrochem. Soc., 2014, p. 1158. I. Nitta, T. Hottinen, O. Himanen, M. Mikkola, Inhomogeneous compression of PEMFC gas diffusion layer, J. Power Sources 171 (2007) 26e36, http:// dx.doi.org/10.1016/j.jpowsour.2006.11.018. W.K. Epting, S. Litster, An oxygen flux interrupt method and in situ microsensor for characterizing oxygen transport in PEFCs, ECS Trans. 50 (2013) 269e277, http://dx.doi.org/10.1149/05002.0269ecst. W.K. Epting, J. Gelb, S. Litster, Resolving the three-dimensional microstructure of polymer electrolyte fuel cell electrodes using nanometer-scale X-ray computed tomography, Adv. Funct. Mater. 22 (2012) 555e560. Z. Fishman, A. Bazylak, Heterogeneous through-plane porosity distributions for treated PEMFC GDLs. II. Effect of MPL cracks, J. Electrochem. Soc. 158 (2011) B846, http://dx.doi.org/10.1149/1.3594636. S.G. Kandlikar, E.J. See, M. Koz, P. Gopalan, R. Banerjee, Two-phase flow in GDL and reactant channels of a proton exchange membrane fuel cell, Int. J. Hydrogen Energy 39 (2014) 6620e6636, http://dx.doi.org/10.1016/ j.ijhydene.2014.02.045. D.R. Lide, CRC Handbook of Chemistry and Physics, 94th ed., 2013, http:// dx.doi.org/10.1136/oem.53.7.504, 2013e2014. D. Mu, Z.S. Liu, C. Huang, N. Djilali, Determination of the effective diffusion coefficient in porous media including Knudsen effects, Microfluid. Nanofluid. 4 (2008) 257e260, http://dx.doi.org/10.1007/s10404-007-0182-3. D.A. Caulk, D.R. Baker, Heat and water transport in hydrophobic diffusion media of PEM fuel cells, J. Electrochem. Soc. 157 (2010) B1237, http:// dx.doi.org/10.1149/1.3454721. J.M. Lamanna, S. Chakraborty, J.J. Gagliardo, M.M. Mench, Isolation of transport mechanisms in PEFCs using high resolution neutron imaging, Int. J. Hydrogen Energy 39 (2014) 3387e3396, http://dx.doi.org/10.1016/ j.ijhydene.2013.12.021. J. Eller, J. Roth, F. Marone, M. Stampanoni, A. Wokaun, F.N. Büchi, Implications of polymer electrolyte fuel cell exposure to synchrotron radiation on gas diffusion layer water distribution, J. Power Sources 245 (2014) 796e800, http://dx.doi.org/10.1016/j.jpowsour.2013.07.025. T. Kotaka, Y. Tabuchi, U. Pasaogullari, C.-Y. Wang, Impact of interfacial water transport in PEMFCs on cell performance, Electrochim. Acta 146 (2014) 618e629, http://dx.doi.org/10.1016/j.electacta.2014.08.148.