On the impact of water activity on reversal tolerant fuel cell anode performance and durability

Journal of Power Sources 328 (2016) 280e288

Contents lists available at ScienceDirect

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

On the impact of water activity on reversal tolerant fuel cell anode performance and durability Bo Ki Hong a, b, 1, Pratiti Mandal a, 1, Jong-Gil Oh b, Shawn Litster a, * a b

Department of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, PA, 15213, USA Fuel Cell Engineering Design Team, Hyundai Motor Company, Yongin-Si, Gyeonggi-Do, 446-716, South Korea

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


Water is essential to both carbon corrosion and oxygen evolution reaction (OER).
Addition of OER catalyst in anode does not guarantee prolonged reversal durability.
Performance of OER catalyst surprisingly drops at water-excess condition.
OER catalyst's performance exhibits volcano type dependence on water activity.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 17 March 2016 Received in revised form 29 June 2016 Accepted 1 July 2016 Available online 12 August 2016

Durability of polymer electrolyte fuel cells in automotive applications can be severely affected by hydrogen starvation arising due to transients during the drive-cycle. It causes individual cell voltage reversal, yielding water electrolysis and carbon corrosion reactions at the anode, ultimately leading to catastrophic cell failure. A popular material-based mitigation strategy is to employ a reversal tolerant anode (RTA) that includes oxygen evolution reaction (OER) catalyst (e.g., IrO2) to promote water electrolysis over carbon corrosion. Here we report that RTA performance surprisingly drops under not only water-deficient but also water-excess conditions. This presents a significant technical challenge since the most common triggers for cell reversal involve excess liquid water. Our findings from detailed electrochemical diagnostics and nano-scale X-ray computed tomography provide insight into how automotive fuel cells can overcome critical vulnerabilities using material-based solutions. Our work also highlights the need for improved materials, electrode designs, and operation strategies for robust RTAs. © 2016 Elsevier B.V. All rights reserved.

Keywords: Fuel starvation Cell reversal Reversal tolerant anode Oxygen evolution reaction Water electrolysis Carbon corrosion

1. Introduction Polymer electrolyte fuel cells (PEFCs) have gained significant

* Corresponding author. Scott Hall, Room no. 5107, Mechanical Engineering Department, Carnegie Mellon University. 5000 Forbes Avenue, Pittsburgh, PA, 15213, USA. E-mail address: [email protected] (S. Litster). 1 These authors contributed equally to this work. http://dx.doi.org/10.1016/j.jpowsour.2016.07.002 0378-7753/© 2016 Elsevier B.V. All rights reserved.

attention as highly efficient and clean power source for fuel cell electric vehicles (FCEVs) due to their high power density, high efficiency, and zero-emission features [1e6]. For the automotive fuel cells to be commercially viable, three major challenges of cost, performance, and durability must be resolved. Automotive fuel cells have several major components, i.e., membrane-electrode assembly (MEA), gas diffusion layer (GDL), and bipolar plate, in order to produce electricity. Among the critical components of fuel cells, the electrodes in the MEA, mostly based on carbon-supported

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

platinum (Pt/C) catalysts, are of paramount importance to generate electricity for vehicles efficiently. As Fig. 1 shows, under normal fuel cell operating conditions, hydrogen and oxygen (in air) gases are supplied to the anode and cathode, respectively, where the hydrogen oxidation reaction (HOR) and oxygen reduction reaction (ORR) respectively occur as follows [3,4]: H2 $ 2Hþ þ 2e, Eo ¼ 0.000 V (vs. SHE)

(1)

1/2O2 þ 2Hþ þ 2e $ H2O, Eo ¼ 1.229 V (vs. SHE)

(2)

where, Eo is the standard electrode potential and SHE is the standard hydrogen electrode. Although often overlooked, hydrogen starvation at the anode arising from either a hydrogen supply malfunction or hydrogen channel blockage by liquid water or ice is a critical issue for FCEVs that can be exacerbated when FCEVs are operated under transient conditions such as start-up and rapid load change [5,6]. When an anode in PEFC stack is starved of hydrogen, the anode lacks a facile Faradaic source of protons and electrons at low potentials and it thus capacitively charges due to the external current from the high cumulative voltage of the adjacent cells stacked in series. With charging, the anode half-cell potential (Ean) increases relative to the cathode's (Eca) and the individual cell voltage (Ecell) reverses (that is, Ecell ¼ Eca e Ean < 0 V). With continued charging, the anode halfcell potential becomes sufficiently high for the oxygen evolution reaction (OER), i.e., water electrolysis, to provide protons and electrons. If the water electrolysis cannot support the full current, the anode half-cell potential further increases leading to Faradaic carbon oxidation (i.e., the carbon corrosion shown in Fig. 1) [5e9]. Although carbon corrosion is thermodynamically favorable, water electrolysis proceeds at a higher rate because it seems to be kinetically favorable [10]. However, the sluggish carbon corrosion reaction becomes sufficiently fast at high voltages (i.e., >1 V) to cause severe carbon corrosion [4e7]. The presence of platinum also aggravates carbon corrosion at high anode-half-cell potentials [11,12]. Carbon corrosion typically occurs through oxidation of carbon into carbon dioxide (CO2) or carbon monoxide (CO) as follows [7e10]: C þ 2H2O $ CO2 þ 4Hþ þ 4e, Eo ¼ 0.207 V (vs. SHE)

(3)

C þ H2O $ CO þ 2Hþ þ 2e, Eo ¼ 0.518 V (vs. SHE)

281

(4)

Thus, if unabated, these cell voltage reversal events typically cause the MEA to be electrically shorted due to a significant amount of heat generated in the membrane as the anode potential increases to high values, eventually resulting in catastrophic cell failure [4e6]. Unlike the inherently transient start-up/shut-down carbon corrosion degradation [13,14], cell reversal degradation is a sustained process that must be survivable for several hours over the lifetime of the fuel cell stack. The anode's voltage rise and cell reversal must be intentionally stopped or delayed through system control strategies. Many control approaches to reversal tolerance have been developed over the past decade, such as cell voltage and exhaust gas monitoring [10,15] and flushing of the anode compartment to eliminate accumulated nitrogen and/or water [16]. However, these control strategies can limit the robustness of operation, hinder performance, and leave the cell susceptible to reversal damage. Instead of complex active control system, a material-based solution that robustly prevents degradation without active intervention is desired. One such approach is a reversal tolerant anode (RTA) [4,6]. The key to an RTA is inclusion of highly active OER catalyst into the anode that promotes oxygen evolution over carbon corrosion. A variety of water electrolysis or OER catalysts, e.g., IrO2, RuO2, TiO2, IrxSn1-xO2, PtIr, IrRu, etc., have been added into anodes to suppress the drop of Ecell (i.e., the increase in Ean) such that facile carbon corrosion potentials are avoided [4,6,10,17]. One of the constraints of the OER catalyst for RTAs is their stability in acidic media and, to a lesser degree, their robustness to a limited number of potential swings between roughly 0 and 2 V vs. SHE. Another challenge with RTAs is the associated material cost. Most FCEVs use a large-sized stack (e.g., 100 kW) to propel the vehicle which consists of several hundreds of unit cells in series, with MEAs having active areas on the order of several hundred square centimeters [18]. This entails significant amounts of OER catalyst and requires a careful design and operation of PEFCs to maximize the RTA OER catalyst's effectiveness. A key aspect of a successful RTA strategy is that, it is effective under all operating conditions, including high temperature (i.e., >80  C) and extremely high and low relative humidity (RH). Operation of fuel cells at high temperature is normally desirable

Fig. 1. A schematic illustration of fuel cell operation under normal and hydrogen starvation conditions. On the left, electrochemical reactions at the anode and cathode under the normal hydrogen supply condition are shown. On the right, electrochemical reactions under the hydrogen-starved condition for the MEA with and without an RTA are shown.

282

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

because it reduces the FCEV's cooling radiator size due to an enhanced heat rejection capability between fuel cell temperature and environment, improving the vehicle's package layout [19,20]. To this end, however, it is essential to overcome critical vulnerability of RTA at high temperature through a fundamental understanding of its dependence on overall water activity (i.e., RH) range. Herein, we unveil a surprising, repeatable vulnerability of RTAs in response to variations in water activity. In particular, at high water activity we observed the opposite to our expectation of higher water activity enabling robust water electrolysis. Instead, we found dramatically diminished RTA performance at high water activity. We present an in-situ investigation of this finding and report the fundamental behavior of RTAs and MEA degradation. Our methods include electrochemical diagnostics, i.e., polarization curves, electrochemical impedance spectroscopy (EIS), cyclic voltammetry (CV), and nano-scale X-ray computed tomography (nano-CT). In this study, among a variety of RTAs, the IrO2 was selected as a representative model catalyst because it is a common OER catalyst with well-characterized properties. 2. Experimental section 2.1. Sample preparation Water electrolysis catalyst, IrO2, was synthesized by the Adams fusion method according to the literature [21,22]. The amount of IrO2 in the anode was fixed to be 50 wt% with respect to the amount of Pt catalyst. The loading of IrO2 used is high with respect to practical implementation in commercial systems. Nevertheless, the higher loading provided us with a model system to study the degradation trend. A standard MEA consisting of a perfluorinated sulfonic acid ionomer membrane (Nafion® NRE211CS (Hþ) grade, DuPont, USA), carbon-supported platinum catalyst (Pt/C: HISPEC4000® grade, Johnson Matthey, UK), and ionomer binder (Nafion® D2021 grade, DuPont, USA) was used throughout the study. The ionomer content in the electrode was fixed to be 30 wt% with respect to the total solid content of dried electrode. The Pt loadings of the anode and cathode were 0.1 and 0.4 mg cm2, respectively. An in-house catalyst-coated membrane (CCM)-type MEA was produced through decal transfer process using a polyimide (Kapton® HN, DuPont, USA) film. 2.2. Electrochemical diagnostics A single cell with an active area of 4 cm2 was used for all the tests. The CCM was stacked in between gas diffusion layers with micro-porous layer (25BC grade, SGL Technologies, GmbH, Germany) directly in the cell hardware without hot-pressing. Machined-graphite flow field plates with single serpentine gas channels were used for both the anode and the cathode. Under hydrogen (anode) and air (cathode) conditions, the single fuel cell was fully activated and then the beginning-of-life (BOL) voltagecurrent density polarization curve was measured using a fuel cell test station (FCT-150S model, Bio-Logic, France). All polarization curves were measured at least twice and the average voltage values were reported. For all electrochemical measurements including polarization curves, EIS, and CV, the cell temperature and RH were fixed at 65  C and 100%, respectively. We used high stoichiometry ratio at both anode and cathode. The gas flow rates were fixed at 0.13 and 0.43 slpm for anode and cathode respectively, for all the tests including the reversal tests. This minimizes the spatial variance of the electrode degradation because we used a differential cell configuration. For CV measurements, we used 0.13 slpm for both gas streams. EIS was performed three times for each MEA sample using a frequency range of 0.1 to 100,000 Hz at a current

density of 0.1 A cm2 using a potentiostat (VSP model, Bio-Logic, France). A current density perturbation with an amplitude of 0.005 A cm2 was used. Before the CV measurements, the cell was purged with the same gases for each CV test for at least 1 h. Cathode and anode CVs were measured by supplying H2/N2 and N2/H2 for anode/cathode, respectively. The CVs were scanned from 0.04 to 1.2 V at a rate of 50 mV s1 and measured at least 30 times per MEA sample. Electrochemically-active surface area (ECSA) of Pt catalyst in the cathode was estimated from each cathode CV. The reversal tests were performed with the cell at 90  C. The inlet RH of the reactant gases was varied by controlling the humidifier temperature. The RH conditions tested were 14%, 28%, 36%, 55%, 68%, and 82% to achieve a wide spectrum of humidity conditions within the cell. The 28% and 82% RH cases were tested three and two times, respectively, to account for the variability observed in their reversal behavior. The electrochemical test results reported for each of these two cases are an average for all the tests and the standard deviation has been shown using an error bar. Prior to a reversal test, the cell was conditioned under the specific reversal condition for at least 1 h to equilibrate the cell to the new operating condition. For the reversal test, a constant current density of 0.2 A cm2 was drawn from the cell, which was supplied with hydrogen and air at the anode and cathode, respectively, at the high flow rates as mentioned earlier. After 0.5 min, the hydrogen gas was switched to nitrogen, thus simulating fuel starvation at the anode. The cell voltage response was recorded and the lower voltage cutoff was set at 2.5 V to prevent permanent failure of the cell. When the cell voltage reached the cut-off value, the cell would return to the open circuit condition and the cell was shut down and left overnight. Cell polarization curves, EIS, cathode and anode CVs were measured the following day. The reversal testing was repeated until the voltage at 0.4 A cm2 (representing the cell performance) fell below roughly 30% of its BOL value and was then considered to be at the end-of-life (EOL) state. 2.3. Nano-CT imaging Specimens were prepared for nano-CT imaging in the pristine and EOL states. All the specimens were extracted from the central region of the 2 cm  2 cm active area of the MEAs. The cell was operated with high reactant flow rates and stoichiometric ratios such that the spatial variation of the degradation is expected to be minimal. For the pristine MEA, a small 0.5 mm  0.5 mm specimen was extracted from the MEA and then mounted onto a polyimide film (Kapton®, DuPont, USA), fixed on a flattened pin head using silver-epoxy composite adhesive, such that the anode was the topmost layer. Here the pristine MEA represents an as-received fresh MEA without experiencing any cell assembly or electrochemical measurements. The anode side of the MEA was ablated using a high-precision (~1 mm resolution) laser mill (QuickLaze 50ST2, ESI®) with a low energy laser beam of 355 nm ultraviolet wavelength, leaving only the pillar-shaped anode specimen with diameter of approximately 65 mm on the membrane surface. The pillar-shaped anode specimen was made such that most of the volume of the specimen remains in the field-of-view (FOV) for all nano-CT radiographs, which ensures higher signal-to-noise ratio in the reconstructed images. The EOL MEA sample was taken out of the cell for post-mortem analysis. In case of the EOL MEA, the GDL on the cathode side was detached readily after the cell was disassembled and the MEA/GDL samples were left intact for at least two days under ambient conditions, while the GDL on anode side remained attached to the MEA firmly until the end of nano-CT imaging. The three-dimensional images were quantitatively analyzed through a combination of commercial software (Avizo Fire, FEI) and custom codes (Matlab, Mathworks).

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

3. Results and discussion 3.1. Electrochemical diagnostics and analysis In order to investigate the voltage reversal behavior of the RTA MEA, cell voltage reversal tests under various RHs were performed. Fig. 2a shows time series of cell potentials during reversal tests. It is clear from the inset in Fig. 2a that an MEA without RTA survives for less than a minute, while the MEA with RTA sustains the load current for about an hour under complete hydrogen starvation. The reversal time shown in Fig. 2a inset was measured for test performed at 90  C and 36% RH. Initially, the fuel cells were operated at a constant current density of 0.2 A cm2 under hydrogen (anode) and air (cathode) conditions for 0.5 min, followed immediately by operation under nitrogen (anode) and air (cathode) conditions until the cell voltage reached 2.5 V. The time for the cell voltage to

Fig. 2. Voltage reversal behavior of fuel cells: a, Cell voltages at 0.2 A cm2 as a function of voltage reversal time during the first reversal test (FRT) at four representative RHs. The inset shows cell voltages at 0.2 A cm2 during first reversal test on a conventional MEA (without RTA) and on an MEA with RTA (RTA 50 wt%) tested at 90  C and 36% RH. b, The FRTs of fuel cells as a function of RH. The dotted line indicates a guideline for the eye. c, Polarization curves for the fuel cell at BOL, i.e., with accumulated reversal time (ART) of 0 min and after each reversal tests at 82% RH (with ARTs progressing from 12 min to 14 min, which was considered to be EOL). d, The first degradation rates of fuel cells at 0.4 A cm2 as a function of RH. The data points denote the average value for multiple measurements and the error bars denote the standard deviation of the measured parameters.

283

decrease from 0.0 to 2.5 V is defined as the voltage reversal time e the time at which catastrophic failure is imminent. After the first reversal test, the PEFC performance was evaluated under regular operation with hydrogen and the electrochemical properties were characterized. The most critical priority for developing RTAs is to understand the properties that dictate the first voltage reversal time (FRT). In addition to the FRT, another key figure of merit is the degradation rate before and after the first reversal test. These two parameters correspond to how long the RTA will survive under harsh reversal conditions, without any external aid from fuel cell control system side, and the resulting deterioration of the PEFC performance. The next priority is to evaluate the performance decay and material degradation as a function of accumulated voltage reversal time (ART) since it is related to effectiveness of RTAs with repetitive reversal events. The ART was estimated by summing all the individual voltage reversal times. Fig. 2a shows the first reversal behavior of RTA MEAs when operated at different relative humidity conditions. When operated under a typical automotive fuel cell condition with an intermediate gas inlet RH of 68%, the initial cell voltage with hydrogen supplied to the anode was approximately 0.7 V. Upon switching the hydrogen supply to nitrogen after 0.5 min, the cell voltage decreased gradually as the hydrogen was purged and then suddenly dropped to 0.8 V followed by the first long plateau to approximately 1.3 V. In the past work, this plateau has been attributed to OER and water electrolysis [6,10]. The FRT of the cell at 68% RH was measured to be about 76 min. If the HOR voltage under normal hydrogen supply is roughly 0.0 V, the anode half-cell voltage during the first reversal test corresponds to the difference between cathode half-cell voltage and the cell voltage. If we assume that galvanostatic cell reversals do not significantly affect the ORR at the cathode, we assume the cathode half-cell potential remains very close to the cell voltage during the initial hydrogen operation and during the reversal tests. This has been verified in past work with a reference electrode [5]. In this case, the first plateau in the reversal corresponds to an anode half-cell voltage of 1.5e2.0 V, which is sufficient for the OER at the anode as well as minor carbon oxidation [6]. In our work and the prior work in the literature, after certain reversal time, the OER catalyst in the anode eventually gets deactivated [6,7]. As the RTA becomes deactivated for water electrolysis, the cell voltage decreases (i.e., higher anode half-cell voltages), and briefly plateaus in the potential range of carbon corrosion, approximately 1.5 to 2.0 V corresponding to an anode potential of 2.2e2.7 V. These ranges are consistent with the results in the literature [6]. When using an RTA, the carbon corrosion plateau is very short (~60e90 s) in duration relative to the water electrolysis due to the finite amount of carbon in the electrochemically active layer. For CO2 evolution, the current density of 0.2 A cm2 yields a bulk carbon thickness removal rate of nearly 2 mm min1 (see Appendix A for details of the calculation). Due to the low Pt loading needed for HOR, typical anodes are quite thin (<5 mm) and are quickly depleted of carbon under these conditions. It is also noteworthy that the voltage reversal time of the second reversal test becomes about 5 min, which is much shorter than that of the first one (Appendix A, Fig. S1). It implies that most of the anode carbon in the MEA was oxidized or corroded during the first voltage reversal test and only a small amount of electrochemically active carbon was left in the anode for further carbon corrosion reactions during the successive reversal tests, causing much shorter reversal times. When the gas RH was set to a lower value of 28%, the FRT of the cell was lower with periodic voltage oscillations, whose reproducibility was verified by three independent cell tests in total. All three cell tests exhibit similar periodic oscillation behavior to one another, but show a broad range of the FRT values ranging from

284

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

about 25 to 79 min. The average and standard deviation of the FRT at 28% RH was 53 ± 27 min, implying highly dynamic, unstable, and transient nature of this operating condition, which is not desirable in order to maintain uniformity and predictability of automotive operation. A similar oscillatory behavior was also observed for the cell at 36% RH (not shown here for simplicity). Such oscillations have been reported for periodic build-up and removal of liquid water, poisoning of the catalysts, and complex water management related water absorption, desorption and diffusion in the Nafion® ionomer [23,24]. To verify that the phenomena were not due to oscillation in the gas stream water content, additional independent fuel cell tests were done with hydrogen delivered at 0.2 A cm2 under 28% and 36% RHs. We found that there were no significant periodic oscillations or fluctuations of the cell voltages under the normal hydrogen supply condition (Appendix A, Fig. S2), indicating that the observed periodic voltage oscillations under nitrogen supply can be attributed to complex interactions between water availability, transport, and electrolysis in the anode. For example, Nafion® ionomer binder in the anode may be oscillating between different hydration states, causing preferential switch between carbon corrosion and OER, which in turn yields an oscillating voltage at a fixed current. When the fuel cell was operated under an extremely dry condition of 14% RH, the cell exhibited the shortest FRT of about 1 min due to a severe deficiency of water reactant for water electrolysis as well as carbon corrosion, implying a highly uncontrollable and unfavorable state of the fuel cell. In order to understand the dependence of the RTA MEA on the overall RH values at a high cell temperature of 90  C, the cell was also tested under a well humidified condition of 82% RH which seems to be the highest verifiable RH value attainable at 90  C with the testing apparatus used. Strikingly contrary to our expectations, however, the cell voltage decreases to 2.5 V quickly under this condition. Another independent cell was tested to verify the reproducibility and it exhibited very similar reversal behavior. The average FRT value of the two independent tests was 9 min. This abrupt drop of the reversal tolerance under the high humidity condition is notably opposite to our expectation of a favorable role of increased water activity for stabilizing and facilitating the water electrolysis, because water is an essential reactant for water electrolysis. A more detailed analysis for better understanding of this reversal behavior will be discussed later through EIS, CV, and nanoCT analysis. In Fig. 2b, a quantitative comparison among the FRTs as a function of RH clearly exhibits that the FRT of the RTA MEA increased significantly up to intermediate RHs of 55e68%, followed

by a drastic drop at the highest RH of 82%. Therefore, it is possible to classify the fuel cell operation into four distinct regimes: Regime I (i.e., 14% RH) which is highly dry (i.e., H2O limited), uncontrollable and unfavorable to operate PEFCs; Regime II (i.e., 28e36% RH) which features highly unstable, transient, and dynamic characteristics; Regime III (i.e., 55e68% RH) which is highly robust and stable to operate fuel cells, thus regarded as a ‘robust’ zone; Regime IV (i.e., 82% RH) which is well humidified, deactivated, and unfavorable. In order to investigate the effects of repetitive reversal conditions on fuel cell performance and durability, polarization curves were measured after each reversal test. The reversal tests were terminated after the cell reached its EOL state (i.e., when the cell voltage at 0.4 A cm2 fell to approximately 30% of its BOL value, the cell was considered to be at its EOL state). One exception is that the cell operated at 14% RH never reached this EOL criterion due to the water starved conditions, causing rapid drops of the cell voltage. For simplicity, only the polarization curves for 82% RH before and after successive reversals are shown in Fig. 2c, since the overall decay pattern of the polarization curves under all RH conditions appears similar to one another. The cell performance decreases significantly as the voltage reversal tests are repeated, which can be attributed to carbon corrosion at the anode as we discuss later in more detail. Fig. 2d shows the degradation rate of the first reversal in terms of cell voltage loss percentage at 0.4 A cm2 before and after the first reversal test divided by the FRT for the different RH values tested. It is observed that the fuel cells operated under the driest (14% RH) and the wettest (82% RH) conditions degraded faster than the others, showing very high degradation rates. In order to investigate the long-term reversal behavior under repetitive reversal tests at various RHs, the normalized polarization curve cell voltages at 0.4 A cm2 as a function of ART at various RHs are also presented in Appendix A, Fig. S3. It is noteworthy that the cell voltage under the 82% RH condition also decreased very fast during the repetitive reversal tests. In order to understand the fundamental origin of the cell performance decay due to reversal, EIS at 0.1 A cm2 and CV of both the anode and the cathode were performed after each reversal test. Here we focus on elucidating the unexpected behavior of Regimes II, III, and IV as shown in Fig. 3. Even though the EIS spectra were recorded after every reversal, only the BOL, the first reversal, and EOL data are presented for simplicity. In Fig. 3a, the EIS at 28% RH exhibits that the Ohmic resistance (ROhm), measured from the

Fig. 3. Electrochemical diagnostics of fuel cells before and after the reversal tests. a, b, and c are EIS data at 0.1 A cm2 under 28%, 68%, and 82% RH conditions, respectively. d, e, and f are anode CVs under 28%, 68%, and 82% RH conditions, respectively.

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

intercept of the real-axis (Zre) at high frequency, increased slightly under the repetitive reversal tests, whereas the anode’s charge transfer resistance (Rct,an), the first high frequency semicircle on the left, increased significantly. At an EOL ART of 100 min, the size of the Rct,an appears to be almost comparable to that of the cathode’s charge transfer resistance (Rct,ca), indicated by the lower frequency semicircle on the right, which is a very rare occurrence in PEFCs. This could be attributed to combined degradation from severe carbon corrosion at the anode and Pt dissolution caused by periodic voltage oscillations that is similar to the degradation under anode potential cycling. On the other hand, with a higher RH of 68%, it is clearly seen in Fig. 3b that the ROhm increased significantly after the reversal tests, indicating an increased Ohmic loss due to an increase in either ionic resistances or electrical contact resistances [25,26]. We speculate that the carbon loss occurred primarily at the membrane-electrode interface due to an Ohmic limited reaction. This fact is evident from the nano-CT images of the degraded anode (Fig. 5 and Fig. S5), where the anode-GDL interface remains smooth, like in the pristine MEA (corresponding to the smooth Kapton®anode interface before decal transfer onto the Nafion® membrane) while there are asperities, voids, and inhomogeneity at the EOL anode-membrane interface. The loss of carbon near the membrane will result in electrically isolated Pt and increased high-frequency resistance (i.e., Ohmic resistance) as the portion of the anode near the polymer electrolyte membrane (PEM) behaves as an inactive, high resistance electrolyte. The degree of increase in ROhm was higher for 68% and 82% RH cases, which also corresponds to the higher loss in anode thickness for those two cases as compared to the 28% RH case. In addition, not only did Rct,an increase significantly, but Rct,ca also increased, indicating that the cathode degrades under reversal conditions, although to a lesser degree. In Fig. 3c, the overall EIS patterns at 82% RH are similar to those at 68% RH, with the voltage reversal time being the major difference between the two regimes. In the case of the anode CV data, the pattern at 28% RH changed slightly after the first voltage reversal test (Fig. 3d). A minor new peak appeared at approximately 0.6 V, which might have resulted from carbon corrosion at the anode [27] because of the exceptionally high anode potentials during voltage reversal tests. As the reversal tests were repeated further, the anode CVs seemed to lose most of the characteristic peaks of Pt/C catalysts, finally leaving little sign of any characteristic platinum hydrogen adsorption/ desorption peaks in the 0.1e0.3 V range and only a very narrow electrical double-layer capacitance remained. This change implies a large decrease in ECSAs in the anode after reversal tests. Unfortunately, it is difficult to quantify the ECSA values at the anode in these cases due to the ambiguous electrical double layer currents. The degradation of the anode at 28% RH appears to be an accelerated and combined one: presumably composed of carbon corrosion due to high anode potential and Pt dissolution due to potential cycling, which is in good agreement with the results of EIS. At 68% RH, the anode CVs exhibit a pronounced evolution of the carbon corrosion peaks at around 0.6 V after the first and EOL reversal tests. The carbon corrosion peaks could be attributed to the surface oxide formation due to the hydroquinone-quinone (HQ-Q) redox reaction on the carbon support surface as follows [28,29]: C  OH $ C ¼ O þ e þ Hþ

(5)

The overall patterns of the anode CVs at 82% RH were again similar to those at 68% RH, with the voltage reversal times being the major difference. In these cases, the hydrogen adsorption/desorption current peaks were not as notably lost as in the case of 28% RH, consistent with our hypothesis that the oscillating voltage at low RH accelerates Pt dissolution. The role of water on carbon corrosion

285

has been contradictory depending on various experimental conditions and materials used: carbon corrosion can increase with increasing RH [30,31] or decreasing RH [32]. Even though the role of high water activity for carbon corrosion and water electrolysis reactions appears complicated, requiring further future study, it is reasonable to infer at this point that a severe deactivation of the IrO2 catalyst may occur under the wettest condition, eventually disrupting the water electrolysis. In order to examine the contribution of cathode degradation during reversal tests to the performance loss, the cathode CVs were also measured after all the reversal tests. The cathode CVs at 82% RH before and after the voltage reversals are presented as a representative case in Appendix A, Fig. S4a. The cathode CV patterns at all the different RH values were similar. The BOL cathode CV showed the typical peak characteristics of Pt/C catalyst: Pt-H desorption/adsorption peaks at around 0.1e0.3 V and Pt-OH desorption/adsorption peaks at about 0.7e1.1 V [3]. The cathode CV patterns after reversal tests are similar to that of BOL, but the cathode ECSAs of all the fuel cells decreased after the repetitive reversal tests (Appendix A, Fig. S4b), indicating that the cathode degradation during voltage reversals also contributes to the overall cell performance loss. 3.2. Three-dimensional morphological analysis A section of the MEA was taken and the anode surface was imaged using a scanning electron microscope (SEM) (Quanta 200 FEG, FEI) with its back-scattered electron detector. The GDL was carefully separated from the degraded anode using tweezers to expose the anode’s surface. Fig. 4a and b show the anode surface on the pristine MEA and Fig. 4c and d show the anode surface of the 68% RH EOL MEA. The bright zones in the SEM images shown in Fig. 4 are the agglomerates of IrO2 in the anode. A very heterogeneous distribution of IrO2 is observed. Comparing the pristine anode surface (Fig. 4a and b) and the EOL anode surface (Fig. 4c and d), we find no apparent change in the content or distribution of IrO2. There was also no significant change observed in the PEM or the cathode of the MEAs (Appendix A, Fig. S5). The SEM images gave us only limited 2D information for the anodes. In order to obtain volumetric data to characterize and analyze possible alteration in the morphology of the anode in three-dimensions (3D) we used nano-CT. Nano-CT also allows us to inspect the internal interfaces between components, such as void formation at the membrane-anode interface. It is well known that carbon corrosion can yield significant changes to PEFC electrode morphology due to the removal of the catalyst’s carbon support and the collapse of the electrode [7e9,27]. Nano-CT provides ability to image within samples in a non-invasive and non-destructive nature and has been previously used to investigate the 3D structure of fuel cell electrodes [33,34]. In order to gain further insight into the nature of the reversal degradation, nano-CT analysis was performed on pristine anode samples and EOL anode samples after being tested at different RH conditions. Fig. 5 shows the 3D images of the pristine anode and the EOL anode for the 68% RH case. Other cases are shown in Appendix A, Fig. S6. The pristine anode has a significant amount of IrO2 particles (shown in orange in Fig. 5a) distributed within the anode catalyst layer, which apparently remains similar in the EOL anode (Fig. 5b), as also seen in Fig. 4. The tomographic slice images on the right side (in greyscale) of the 3D renderings show virtual cross-sections of the anodes. The brightest phase represents IrO2 agglomerates, while the grey region is the Pt/C porous network and dark phases indicate mainly ionomer binder and pores that have low X-ray absorption contrast. The nano-CT images were obtained in the Xray absorption mode with an FOV of 65 mm and a voxel resolution of

286

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

Fig. 4. Electron micrographs of anode surface: a and b are SEM images (back-scattered electron) of the anode surface of the pristine MEA imaged with magnifications of 3000 and 12000, respectively. c and d are SEM images (back-scattered electron) of the anode surface for the 68% RH EOL MEA, imaged with magnification of 3000 and 12000, respectively. The higher intensity particles are the IrO2 agglomerates, while the Pt/C/ionomer matrix is in dark grey. The darkest zones are pores and cracks on the surface of the anode.

64 nm. After the reversal tests, it is observed that the thickness of the EOL anode decreased significantly (Fig. 5b). In particular, the cross-sectional tomograph of the EOL anode exhibits a thinner and brighter anode as compared to that of the pristine one, which is attributed to higher concentration of Pt catalysts in the anode, due to the loss of carbon support via carbon corrosion during the reversal tests. The morphological changes in the other EOL anode samples at 28% and 82% RHs (Appendix A, Fig. S6) are similar to that at 68% RH. Quantitative analysis of the 3D images reveals that the anode thicknesses, IrO2 agglomerate volumes, and surface areas for all the EOL anodes at 28%, 68%, and 82% RHs generally decreased (Fig. 5c). The reduction of anode thickness mostly resulted from severe carbon corrosion [13,35e37], while the decrease in IrO2 volume and surface area might be attributed to the degradation, redistribution, or densification of the initially nanoporous IrO2 catalysts during the reversal tests. It is noteworthy that the EOL anodes at all RHs retain a significant amount of IrO2 catalysts. Thus, the eventual end of the water electrolysis plateau cannot be attributed to the loss of IrO2. Rather, we postulate that there is some electrochemical deactivation mechanism, which is the focus of

ongoing and future work. We speculate that the volcano-type behavior of the reversal time with respect to the water-activity (proportional to RH) might have resulted from a competition between the increase of IrO2 OER activity with higher RH and the opposing increase in the rate of deactivation with higher RH (which could be related to the carbon oxidation reaction's water activity dependence). Videos with cross-sections of the imaged 3D volume of the pristine and the EOL anode samples at 28%, 68%, and 82% RHs are provided as Supplementary data, Videos S1eS4. Supplementary videos related to this article can be found at http://dx.doi.org/10.1016/j.jpowsour.2016.07.002.

4. Conclusion The key finding from this study is that, there is a significant and unexpected vulnerability in the use of RTAs for preventing cell reversal degradation when operating in a high humidity condition. This vulnerability is in addition to the expected challenges of low relative humidity RTA operation. When the RH was increased from 68% to 82%, there was roughly an order of magnitude shorter

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

287

of the most common scenarios for hydrogen starvation are based on start up from cold temperatures or liquid water flooding that will yield high RH in the gas streams. In addition, these results give new insight on the electrode degradation mechanism and deactivation of the OER catalyst during cell reversal. Specifically, at medium to high RH levels we observed that the fuel cell performance loss following carbon corrosion due to reversal has a primarily Ohmic origin. We also conclude that the eventual loss of electrolysis activity is not solely due to a disappearance of the IrO2 catalyst. Author contributions B.K.H. and S.L. conceived the project and designed the experiments. B.K.H. designed the electrode formulation and J.-G.O synthesized the IrO2 catalyst and prepared the MEA samples. B.K.H. and P.M. made fuel cells and performed electrochemical experiments. P.M. executed nano-CT analysis. B.K.H., P.M. and S.L. wrote the paper and all authors discussed the results and commented on the manuscript. B.K.H. and P.M. contributed equally to this work. Conflict of interests The authors declare no competing financial interests. Acknowledgements This work was supported by the collaborative research project (R-141594.0001) of Hyundai Motor Company and Carnegie Mellon University. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jpowsour.2016.07.002. References

Fig. 5. Morphological analysis of the pristine and EOL anodes using nano-CT. a and b are representative 3D images (volume renderings) of the pristine and EOL (reversal tested at 68% RH) anodes in the MEAs, respectively. The agglomerated IrO2 catalysts (surface rendering after segmentation) are represented in orange and tomographic slices (virtual cross-section through the imaged volume) are shown on the right side of each volume rendering. c shows the relative anode thickness (the height of the bar shows the average thickness value and the error bar shows the standard deviation across the MEA area), IrO2 volume and surface area of EOL anodes with regard to those of the pristine one at 28%, 68%, and 82% RHs (footprint area for analysis ¼ 36 mm  36 mm). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

reversal time the RTA could sustain and it significantly underperforms. This repeatable finding has not been previously disclosed and it requires careful consideration in the design of RTAs for future FCEVs. For example, to account for typical relative humidity variation in a cell and stack, one can implement spatial patterning of OER catalyst in the RTAs such that the most vulnerable regions have the highest loading of the OER catalyst [38,39]. The reduced RTA performance at high RH is particularly concerning because several

[1] M.K. Debe, Electrocatalyst approaches and challenges for automotive fuel cells, Nature 486 (2012) 43e51. [2] B.C.H. Steele, A. Heinzel, Materials for fuel-cell technologies, Nature 414 (2001) 345e352. [3] J.H. Ohs, U.S. Sauter, S. Maass, Degradation caused by dynamic operation and starvation conditions, in: D. Stolten, B. Emonts (Eds.), Fuel Cell Science and Engineering: Materials, Processes, Systems and Technology, Wiley-VCH Verlag GmbH & Co, Weinheim, 2012, pp. 543e570. [4] R.T. Atanasoski, L.L. Atanasoska, D.A. Cullen, Efficient oxygen evolution reaction catalysts for cell reversal and start/stop tolerance, in: M. Shao (Ed.), Electrocatalysis in Fuel Cells: A Non- and Low-Platinum Approach, SpringerVerlag, London, 2013, pp. 637e663. [5] A. Taniguchi, T. Akita, K. Yasuda, Y. Miyazaki, Analysis of electrocatalyst degradation in PEMFC caused by cell reversal during fuel starvation, J. Power Sources 130 (2004) 42e49. [6] T.R. Ralph, M.P. Hogarth, Catalysis for low temperature fuel cells: Part II. The anode challenges, Platin. Met. Rev. 46 (2002) 117e135. [7] P. Mandal, B.K. Hong, J.-G. Oh, S. Litster, 3D imaging of fuel cell electrode structure degraded under cell voltage reversal conditions using nanoscale Xray computed tomography, ECS Trans. 69 (2015) 443e453. [8] K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, John Wiley & Sons, New York, 1988, pp. 316e352. [9] F. Zhou, S.J. Andreasen, S.K. Kær, D. Yu, Analysis of accelerated degradation of a HT-PEM fuel cell caused by cell reversal in fuel starvation condition, Int. J. Hydrogen Energy 40 (2015) 2833e2839. [10] S. Ye, Reversal-tolerant catalyst layers, in: J. Zhang (Ed.), PEM Fuel Cell Electrocatalysts and Catalyst Layers: Fundamentals and Applications, SpringerVerlag, London, 2008, pp. 835e860. [11] F. Maillard, A. Bonnefont, F. Micoud, An EC-FTIR study on the catalytic role of Pt in carbon corrosion, Electrochem. Comm. 13 (2011) 1109e1111. , A. Crisci, F. Maillard, Carbon [12] Z. Zhao, L. Castanheira, L. Dubau, G. Berthome corrosion and platinum nanoparticles ripening under open circuit potential conditions, J. Power Sources 230 (2013) 236e243. [13] J. Durst, et al., Degradation heterogeneities induced by repetitive start/stop events in proton exchange membrane fuel cell: inlet vs. outlet and channel vs. land, Appl. Catal. B: Environ. 138e139 (2013) 416e426.

288

B.K. Hong et al. / Journal of Power Sources 328 (2016) 280e288

[14] J. Dillet, et al., Impact of flow rates and electrode specifications on degradations during repeated startups and shutdowns in polymer-electrolyte membrane fuel cells, J. Power Sources 250 (2014) 68e79. [15] R.H. Barton, Cell voltage monitor for a fuel cell stack, US Patent 6,724,194 B1, Apr. 20, 2004. [16] I.C. Halalay, et al., Anode materials for mitigating hydrogen starvation effects in PEM fuel cells, J. Electrochem. Soc. 158 (2011) B313eB321. [17] R.T. Atanasoski, D.A. Cullen, G.D. Vernstrom, G.M. Haugen, L.L. Atanasoska, A materials-based mitigation strategy for SU/SD in PEM fuel cells: properties and performance-specific testing of IrRu OER catalysts, ECS Electrochem. Lett. 2 (2013) F25eF28. [18] H.A. Gasteiger, J.E. Panels, S.G. Yan, Dependence of PEM fuel cell performance on catalyst loading, J. Power Sources 127 (2004) 162e171. [19] S.D. Knights, K.M. Colbow, J. St-Pierre, D.P. Wilkinson, Aging mechanisms and lifetime of PEFC and DMFC, J. Power Sources 127 (2004) 127e134. [20] Y. Shao, G. Yin, Z. Wang, Y. Gao, Proton exchange membrane fuel cell from low temperature to high temperature: material challenges, J. Power Sources 167 (2007) 235e242. [21] R. Adams, R.L. Shriner, Platinum oxide as a catalyst in the reduction of organic compounds. III. Preparation and properties of the oxide of platinum obtained by the fusion of chloroplatinic acid with sodium nitrate, J. Am. Chem. Soc. 45 (1923) 2171e2179. [22] J.-G. Oh, W.H. Lee, H. Kim, The inhibition of electrochemical carbon corrosion in polymer electrolyte membrane fuel cells using iridium nanodendrites, Int. J. Hydrogen Energy 37 (2012) 2455e2461. [23] D.G. Sanchez, D.G. Diaz, R. Hiesgen, I. Wehl, K.A. Friedrich, Oscillations of PEM fuel cells at low cathode humidification, J. Electroanal. Chem. 649 (2010) 219e231. [24] R. Lin, C. Cao, J. Ma, E. Gülzow, K.A. Friedrich, Optimizing the relative humidity to improve the stability of a proton exchange membrane by segmented fuel cell technology, Int. J. Hydrogen Energy 37 (2012) 3373e3381. [25] A.P. Young, J. Stumper, S. Knights, E. Gyenge, Ionomer degradation in polymer electrolyte membrane fuel cells, J. Electrochem. Soc. 157 (2010) B425eB436. [26] X. Zhang, L. Guo, H. Liu, Recovery mechanisms in proton exchange membrane fuel cells after accelerated stress tests, J. Power Sources 296 (2015) 327e334.

[27] L.M. Roen, C.H. Paik, T.D. Jarvi, Electrocatalytic corrosion of carbon support in PEMFC cathodes, Electrochem. Solid-State Lett. 7 (2004) A19eA22. [28] K.H. Kangasniemi, D.A. Condit, T.D. Jarvi, Characterization of Vulcan electrochemically oxidized under simulated PEM fuel cell conditions, J. Electrochem. Soc. 151 (2004) E125eE132. [29] J. Wang, G. Yin, Y. Shao, S. Zhang, Z. Wang, Y. Gao, Effect of carbon black support corrosion on the durability of Pt/C catalyst, J. Power Sources 171 (2007) 331e339. [30] J.H. Kim, et al., Effects of cathode inlet relative humidity on PEMFC durability during startup-shutdown cycling, J. Electrochem. Soc. 157 (2010) B633eB642. [31] K.H. Lim, H.-S. Oh, S.-E. Jang, Y.-J. Ko, H.-J. Kim, H. Kim, Effect of operating conditions on carbon corrosion in polymer electrolyte membrane fuel cells, J. Power Sources 193 (2009) 575e579. [32] R.L. Borup, J.R. Davey, F.H. Garzon, D.L. Wood, M.A. Inbody, PEM fuel cell electrocatalyst durability measurements, J. Power Sources 163 (2006) 76e81. [33] 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. [34] E.A. Wargo, T. Kotaka, Y. Tabuchi, E.C. Kumbur, Comparison of focused ion beam versus nano-scale X-ray computed tomography for resolving 3-D microstructures of porous fuel cell materials, J. Power Sources 241 (2013) 608e618. , E. Rossinot, Heteroge[35] L. Dubau, J. Durst, F. Maillard, M. Chatenet, J. Andre neities of aging within a PEMFC MEA, Fuel Cells 12 (2012) 188e198. [36] L. Dubau, et al., Carbon corrosion induced by membrane failure: the weak link of PEMFC long-term performance, Int. J. Hydrogen Energy 39 (2014) 21902e21914. [37] L. Dubau, et al., A review of PEM fuel cell durability: materials degradation, local heterogeneities of aging and possible mitigation strategies, WIREs Energy Environ. 3 (2014) 540e560. [38] H. Zhang, H. Haas, A. Leow, M. Davis, R. Fellows, Reversal tolerant membrane electrode assembly for a fuel cell, US Patent 8,968,960 B2, Mar. 3, 2015. [39] J.-P. Suchsland, A. Ghielmi, P. Braun, Barrier layer for carbon corrosion protection in electrochemical devices, US Patent 2015/0354072 A1, Dec. 10, 2015.