Degradation Mechanisms of Carbon Supports under Hydrogen Passivation Startup and Shutdown Process for PEFCs

.

Polymer electrolyte fuel cells (PEFCs) convert chemical energy directly to electrical energy with low emissions and high energy efficiency and have shown promise to be an eco-friendly power source for fuel cell vehicles (FCVs) and residential co-generation systems. [1][2][3] Nevertheless, PEFCs still have several problems to be solved, such as limited lifetime and reliability and high cost, before large-scale commercialization can be realized. [4][5][6] The minimization of the PEFC cost can be achieved by improving the specific mass activity (MA) of the catalyst for the oxygen reduction reaction (ORR) at the cathode. The conventional cathode catalysts have consisted of Pt nanoparticles dispersed on high surface area carbon black supports (Pt/CB) to maximize the electrochemically active surface area (ECSA) for the ORR. 4 However, as is widely known, Pt/CB cathode catalysts are degraded under PEFC operating conditions such as load change cycles and startup/shutdown (SU/SD) cycles due to a combination of processes, which include ECSA loss due to the agglomeration or dissolution of Pt nanoparticles, [6][7][8][9][10][11][12] and the corrosion of the CB support material. 7,[11][12][13][14][15] During the SU/SD cycles, air and H 2 coexist transiently in the anode until the replacement of the former with the latter (or vice versa) is completed. Reiser et al. showed that this situation causes the cathode potential to climb to more than 1.5 V due to the so-called "reverse current mechanism", which significantly accelerates the degradation of the catalyst due to the oxidation of the CB and the agglomeration or dissolution of the Pt nanoparticles. 12 The latest papers on this topic have been reviewed. 13 Several approaches have been taken to both understand and mitigate the decrease of PEFC performance during operation. 13 One approach to mitigate the decrease of the cell performance is to use transition metal oxide support materials, for example, titanium-based oxides [16][17][18][19][20] and tin-based oxides. [21][22][23][24][25][26] Shintani et al. proposed that that the reverse current during SU/SD can be decreased by decreasing the ORR current generation on the Ta-doped TiO 2 -supported Pt anode due to its high resistivity in air. 27 In order to clarify the degradation mechanism of the cathode catalysts during SU/SD, Ishigami et al. visualized the distribution of the oxygen partial pressure at the anode in real time and space during the SU/SD. 28 Durst et al. used a segmented cell to understand the local degradation and performance of the cathode. 29 Park et al. investigated the degradation of Pt/CB cathode catalysts by comparing square wave and triangular wave potential cycling. 30 In our previous research, we have investigated the durability and degradation mechanisms of the Pt/CB catalyst layer (CL) both under accelerated SU conditions for FCVs, which included gas exchange (air-SU/SD), and under the standard protocol of cathode potential cycling, which simulates SU/SD conditions. 31.32 It was found that two essential factors for the maintenance of cell performance under SU/SD conditions were the high corrosion resistance of the support material and the uniform dispersion of the Pt nanoparticles on the support. We were able to improve not only the cell performance but also the durability under air-SU/SD conditions by use of a graphitized CB (GCB)-supported Pt catalyst prepared by the "nanocapsule method" (n-Pt/GCB) as the cathode material, compared with those for a cathode prepared with a commercial Pt/GCB. 32 However the ECSA for n-Pt/GCB still decreased by more than 50% after 1000 cycles of air-SU/SD, compared with that before cycling.
In order to improve the SU/SD durability of PEFCs in FCVs, the hydrogen passivation process (H 2 -SU/SD) has been proposed. 33 In the SD portion of this process, the O 2 in the cathode is consumed by the H 2 permeated from the anode after the closing of the cathode gas valves at both the inlet and outlet, and then the supply of H 2 is also stopped. In the SU process, the valves are re-opened, and then the air and H 2 are re-supplied to the cathode and anode, respectively. The H 2 -SU/SD process prevents the construction of local cells, which cause the reverse current processes in the anode and cathode, and it also suppresses the severe carbon oxidation reaction (COR) and Pt aggregation that are observed during the air-SU/SD process. However, it was found that carbon corrosion still occurred during the H 2 -SU/SD process, even though it was less severe than that during the air-SU/SD process. In this research, we investigated the mechanism of the COR during the H 2 -SU/SD process as follows: (1) the effect of the presence of Pt catalyst on the CO 2 generation, (2) the timing of the CO 2 generation and (3) the effect of Pt oxidation state on the CO 2 generation. We propose a solution to the problem of the carbon corrosion in the H 2 -SU/SD process, which is based on the mechanisms elucidated here. were designated as "Pt/CB//Pt/CB" and "Pt/CB//CB", respectively. A commercial Nafion membrane (NRE 211, DuPont, 25 μm thickness) was used as the polymer electrolyte membrane (PEM). The CCMs were prepared in the same manner as reported in our previous work. Briefly, the catalyst pastes were prepared from each of the Pt-loaded carbon catalysts with Nafion binder (ion exchange capacity 0.9 meq g −1 , DE521, E. I. Du Pont de Nemours & Co., Inc.), and the mass ratio of Nafion binder (dry basis) to CB (Nafion binder/CB) was adjusted to 0.7. The catalyst paste was directly sprayed onto the PEM to prepare the CCM by use of a pulse-swirl-spray apparatus (PSS, Nordson Co.), and then dried at 60 • C in an electric oven. The Pt loading amount was 0.50 ± 0.05 mg -Pt cm −2 . For two experiments, Pt-free CB CLs were used, and the loading amount of the CB layer was 0.56 ± 0.05 mg -CB cm −2 , which was the same as that of the CB in the Pt/CB CL. The active geometric area of the electrode was 29.2 cm 2 . The CCM was pressed at 140 • C and 1.0 MPa for 3 min. The CCM was sandwiched between two gas diffusion layers (GDLs, 25BCH, SGL Carbon Group Co., Ltd.) and was then assembled into a single serpentine pattern cell (Japan Automotive Research Institute (JARI) standard cell) containing two carbon separator plates. Fig. 1 shows the schematic of the cell with gas lines. Two solenoid valves were included, on the inlet and outlet of the cathode. The H 2 -SU/SD process was performed under the gas conditions shown in Table I at 50 • C. These gases were humidified at 30% relative humidity (RH) by use of heated water bubblers, controlled by a measurement system (FCE-1, Panasonic Production Technology). In step 1, H 2 and CO 2free artificial air (N 2 80%; O 2 20%), were supplied to the anode and cathode, respectively. In step 2, both solenoid valves were closed in order to seal off the cathode as the SD step. In step 3, both solenoid valves were opened in order to supply the air to the cathode as the SU step. These three steps were defined as one "air-air" cycle. During the early air-air cycles, amorphous portions of the CB particles can easily be oxidized, and the CO 2 concentration in the cathode exhaust, which was generated during the air-air cycles, decreased with increasing number of cycles. We confirmed that the CO 2 concentration did not change greatly after 100 cycles. Therefore, all data for the H 2 -SU/SD durability evaluations were obtained after 99 air-air cycles. The CO 2 concentration in the cathode exhaust was measured by use of a non-dispersive infrared (NDIR, GMP343, Vaisala) detector on the downstream side of the solenoid valve on the cathode outlet. Thus, the total amount of CO 2 generated in steps 2 and 3 was measured after both solenoid valves were opened. In order to investigate the effects of O 2 on the COR in step 2, N 2 was supplied to the cathode during the first step. These three steps were defined as one N 2 -air cycle, as shown in Table I. From a comparison of the results between the air-air cycles and the N 2 -air cycles, we can separate the values of CO 2 generated in steps 1 and 2.

Procedure of H 2 -SU/SD and experimental setup.-
Holding of the cathode potential in step 2 of the air-air cycles.-The cathode potential was decreased at a scan rate of 50 mV s −1 and then held at various potentials from 0.6 V to 1.0 V in steps of 0.1 V during step 2 of the air-air cycles (E hold ) in order to investigate the influence of the Pt oxidation-state upon the oxidation amount of the CB. The cathode potential was controlled with a potentiostat (HZ-5000 Automatic Polarization System, Hokuto Denko Co.). The anode was used as both reference electrode and counter electrode. The cathode potential increased after the cell was switched to open circuit at the end of step 2.

Procedure of CCM evaluation for durability and electrochemical properties.-
The ECSA of the Pt catalyst in the cathode was evaluated both before and after 100 H 2 -SU/SD cycles by use of cyclic voltammetry (CV). The CV was measured between 0.075 V and 1.0 V at a scan rate of 20 mV s −1 with a potentiostat both at 40 • C (100% RH) and 50 • C (30% RH), with H 2 in the anode and N 2 in the cathode. The anode was used as both reference electrode and counter electrode. The value of ECSA was determined from the hydrogen adsorption charge (Q Pt-H ), referred to Q H • = 0.21 mC cm −2 , the value used conventionally for polycrystalline platinum. 34 The electric charge for Pt oxidation (Q Pt-oxidation) was obtained by integration of the anodic peak area (ca. 0.6 V to 1.0 V) and was corrected for the anodic baseline of the electric double layer capacitive (EDLC) current of the voltammogram. The electric charge for the Pt oxide reduction (Q PtO-reduction ) was obtained by the integration of the cathodic peak area (1.0 V to each final potential, to 0.6 V in steps of 0.1 V), with correction for the cathodic baseline of the EDLC current. The current-potential (I-E) curves were measured both before and after the 100 H 2 -SU/SD cycles galvanostatically supplied to the anode and the cathode by use of an electronic load (PLZ-664WA, Kikusui Electronics Co.) operated in the constant current mode (5 min acquisition at each current), controlled by a measurement system (FCE-1, Panasonic Production Engineering Co.) at 65 • C with 100% RH H 2 and air under ambient pressure (1 atm) with gas utilizations of 70% and 40%, respectively.

Comparison of the extent of carbon oxidation during air-SU/SD and H 2 -SU/SD.-Concentrations
of CO 2 generated in the cathode of a cell using Pt/CB for both the anode and cathode were measured during the air-SU/SD and H 2 -SU/SD cycling. The steps of the air-SU/SD process, in which air and H 2 are alternately supplied to the anode and artificial air is continuously supplied to the cathode, are shown in Table II. 32 Fig. 2a shows the changes of both the cell voltage (E cell ) and CO 2 concentration at the 100th cycle during the air-SU/SD and H 2 -SU/SD processes (air-air cycle in Table I). In step 1 of the  air-SU/SD process, the E cell decreased immediately after the air feed was initiated and approached 0 V, and the CO 2 concentration temporarily increased to about 30 ppm. In step 2, the E cell increased to over 1.1 V after the H 2 was initiated, and the CO 2 concentration increased again nearly simultaneously, reaching 133 ppm and then decreasing to 0 ppm. These phenomena were caused by the COR in the cathode outlet region as a result of the depletion of H + due to their consumption in the anode outlet region for the ORR, which is the well-known reverse current phenomenon. In the H 2 -SU/SD process, the E cell gradually decreased from 1.0 V to 0.15 V after both valves were closed in step 2. This voltage decrease was caused by O 2 consumption in the cathode, due to reaction with the H 2 permeating from the anode to cathode through the membrane. The permeation rate of H 2 obtained from linear sweep voltammetry was 0.86 mA cm −2 at 50 • C and 30% RH under ambient pressure with H 2 -supplied anode and N 2 -supplied cathode. The total volume of remaining O 2 was 22.3 mL, which was calculated as the sum of the volumes of the channel of the cell (1.5 mL), of the gas lines between the cell and valves (20.3 mL), of the GDL (0.5 mL), and of the CL (26.2 μL). Thus, the estimates for the time required to react the remaining O 2 were 51.3 min. for the total volume and 1.2 min. for the sum of the volumes of of the GDL and CL. The actual time during the H 2 -SU/SD cycles (12.0 min.) was shorter than that estimated for the consumption of O 2 in the total volume, and longer than that estimated for the total of the CL and GDL volumes in Fig. 2a. We consider that this difference was caused both by the decrease of the cathode pressure as a result of the ORR and the influence of the O 2 supply into the CL from the channel. In step 3 of the H 2 -SU/SD process (artificial air resupplied), the E cell rose to 1.0 V after the valves were opened and air was introduced to the cathode. The gas volumes of the channel of the cell, the CL and the GDL were 1.5 mL, 26.2 μL and 0.5 mL, respectively. The air flow rate was 100 ml/min. Thus, the air-front passage time was estimated to be 1.2 s. The CO 2 concentration nearly simultaneously increased to 28 ppm and then decreased to 0 ppm. Fig. 2b shows the weight ratios of the oxidized CB to the total CB in the CL at the 100th cycle during both the air-SU/SD and H 2 -SU/SD processes, which were calculated from the CO 2 peak areas in Fig. 2a. We found that the carbon oxidation in the H 2 -SU/SD process was only one-eighth of that in the air-SU/SD process, and thus that the H 2 -SU/SD process is more attractive for the durability of FCVs. Fig. 3a shows cathode CVs at 40 • C obtained before and after 100 H 2 -SU/SD cycles. The cathode ECSA decreased about 13% to 60 m 2 g −1 from the initial value 69 m 2 g −1 (Fig. 3b). Fig. 4a shows the I-E curves at 65 • C obtained before and after the H 2 -SU/SD durability evaluation. The MA at 0.85 V decreased about 26% to 102 A g −1 from the initial value 137 A g −1 (Fig. 4b).
We consider that the degree of the carbon corrosion occurring during the H 2 -SU/SD cycling cannot be ignored for the long-term operation of FCVs, because the extent of oxidative degradation (55 ppm after 100 cycles) during the H 2 -SU/SD process would be expected to lead to a 5.5% degradation after 1,000 cycles, and the degradations of the ECSA and MA even at 100 cycles were approximately 13% and 26%, respectively.

Effect of presence of Pt catalyst on CO 2 generation during H 2 -SU/SD cycling.-
We investigated the effect of the presence of the Pt catalyst on CO 2 generation during the H 2 -SU/SD process using an asymmetric cell, i.e., Pt/CB//CB. Fig. 5 shows the changes of the E cell and CO 2 concentration during the H 2 -SU/SD process (air-air cycle in Table I) with CB anode//Pt/CB cathode and Pt/CB anode//CB cathode cells. The E cell values of the CB anode//PtCB cathode cells and Pt/CB anode//CB cathode in step 1 were 1.0 V and 0.6 V, respectively, and gradually decreased to 0 V after both valves were closed in step 2. We consider that, even though the hydrogen oxidation reaction (HOR) and hydrogen evolution reaction activity of the CB is significantly small compared with that for the Pt catalyst, it is sufficient to act as an reversible hydrogen electrode (RHE) electrode under open circuit voltage (OCV) conditions. A significant amount of CO 2 was detected from the cathode in the cell in which Pt catalyst was only present in the cathode; however, a negligible amount was detected in the cell in which the Pt catalyst was present in the anode and only CB in the cathode. These results indicate that the COR hardly occurred in the absence of Pt catalyst in the cathode and was greatly accelerated by the presence of Pt catalyst. 2 can be generated in both step 2 (O 2 was consumed by H 2 permeating into the cathode) and step 3 (artificial air was introduced into the cathode) of the H 2 -SU/SD process, because H 2 and O 2 coexisted in the cathode and could form a local cell in the cathode CL. However, we were not able to detect separately the CO 2 concentration from each step, because the CO 2 detector was located downstream from the outlet of the cathode. In order to investigate the timing of the COR, N 2 was fed to the cathode in step 1 (N 2 -air Figure 5. Changes of the E cell and CO 2 concentration with CB and Pt/CB both in the anode and cathode during the H 2 -SU/SD process (air-air cycle in Table I).  Table I. cycle) for suppression of the COR in step 2. In step 3 of the N 2 -air cycle, artificial air was fed to the cathode. The CO 2 generation in that case can considered to occur as a result of the COR from step 3 only. The difference in the amounts of oxidized CB comparing the air-air and N 2 -air cycles shows the amount of CB oxidized in step 2 of the H 2 -SU/SD process. Fig. 6a shows the changes of both voltage and CO 2 concentration for a cell using Pt/CB for both anode and cathode after 100 H 2 -SU/SD cycles and then subjected to an air-air cycle and an N 2 -air cycle. In step 2 of the air-air cycle (valves closed), the E cell gradually decreased from 1.0 V to 0.15 V. In step 3 of the air-air cycle (artificial air resupplied), the E cell rose to about 1.0 V, and the CO 2 concentration nearly simultaneously increased to 28 ppm and then decreased to 0 ppm. In step 1 of the N 2 -air cycle (180 s, N 2 purge), the E cell was approximately 0.1 V, and then in step 2 (valves closed), the voltage decreased to nearly 0 V. This result indicates that the COR was suppressed due to the absence of O 2 in the cathode. In step 3 of the N 2 -air cycle (artificial air supplied), the E cell rose to about 1.0 V, and the CO 2 concentration increased to 18 ppm, and then decreased to 0 ppm. Fig. 6b shows the ratios of oxidized CB to total CB in the CL after the 100 H 2 -SU/SD cycles during the air-air and N 2 -air cycles, which were calculated from the peak areas of CO 2 concentration in Fig. 6a. The amount of CB oxidized during the N 2 -air cycle was less by 23% compared with that generated during the air-air cycle. These results indicate that the amount of CO 2 generated in step 3 was significantly larger than that in step 2 and that 23% of the COR occurred in step 2 (valves closed), and the remaining 77% of the COR occurred in step 3 (artificial air resupplied). The COR occurred both during step 2, when O 2 was present in the cathode, and step 3, in spite of the absence of O 2 . This result indicates that the reduction and re-oxidation of the Pt oxide were significantly involved with the COR during both steps 2 and 3. In addition, the result suggests that the Pt oxidation state is related to the degree of COR in the H 2 -SU/SD process. (a) Changes of the E cell and CO 2 concentration during air-air cycles with and without the E hold , (b) CVs at 50 • C with 30% RH N 2 in the cathode and 30% RH H 2 in the anode and (c) relationships both of the oxidized CB ratio to the total CB in the CL during the H 2 -SU/SD process and Q PtO-reduction with the lower limit of the cathode potential.

Effect of Pt oxidation state on CO 2 generation in the H 2 -SU/SD
process.-In order to investigate the effect of the Pt oxidation state, the cathode potential was held at various potentials from 0.6 V to 1.0 V in steps of 0.1 V after the gas valves were closed during the H 2 -SU/SD process after the N 2 -air cycle (E hold 0.6 to 1.0 V in Fig. 7a). Fig. 7a shows the changes both of the E cell and the CO 2 concentration under the air-air cycle with and without the E hold in the step 2. It was found that CO 2 concentration decreased with increasing cathode holding potential.
In order to relate the amount of oxidized CB to the Pt oxidation state, the values of the electric charges for both oxidation of Pt and reduction of the Pt oxide (Q Pt-oxidation and Q PtO-reduction ) at the various holding potentials of the cathode were obtained from the CVs of the cell using Pt/CB//Pt/CB at 50 • C under 30% RH H 2 and N 2 . The Q PtO-reduction values with various lower limits of cathode potential and Q Pt-H were calculated from the areas drawn in the CVs of the type shown in Fig. 7b. Fig. 7c shows the relationship between the cathode holding potential and values of both the ratio of oxidized CB to total CB and the Q PtO-reduction obtained during the air-air cycles, with and without the E hold in step 2 of the H 2 -SU/SD process. Both the ratios of oxidized to total CB and Q PtO-reduction values decreased with increasing cathode holding potential. Takei et al. also indicated that high potential operation (0.73 V−0.93 V) enhanced Pt oxidation compared with low potential operation (0.63 V−0.93 V). 35 The result of Fig. 7c clearly indicates that the COR was affected by the degree of Pt oxidation. Linse et al. have also indicated that higher cathode potential decreased the extent of the COR, with increasing degree of Pt oxidation in the cathode potential range from 0.4 V to 1.0 V. 36 In addition, we can propose that the holding of the cathode potential at higher values during the H 2 -SU/SD process is able to suppress the COR, which is a practical strategy that could be implemented in FCV systems.

Mechanisms of the COR during the H 2 -SU/SD process.-Based
on the results described above, we propose three different types of mechanism for the COR in the cathode CL in both steps 2 and 3 of the H 2 -SU/SD process (Fig. 8). In step 1 of the H 2 -SU/SD process, the cathode CL was filled with air, and oxygen-containing species, which we refer to simply as "oxide," were adsorbed on the Pt particles.
In step 2, H 2 permeates from the anode to cathode, and the Pt oxide is reduced when the H 2 reaches the Pt particles (Reaction 1), and the HOR (Reaction 2) also occurs in the cathode CL.
Some Pt oxides in the cathode are electrochemically reduced by the H + and electrons thorough the ionomer and carbon support, respectively, when the cathode potential decreases with the increase of H 2 concentration (Reaction 3).
However, we note several specific situations that can arise in the cathode CL as follows, which would be caused by non-uniform distributions of either ionomer or Pt particles or both. Such inhomogeneities can give rise to local electrochemical cells involving the COR. Inhomogeneities in either composition or environment are well known to induce local cell corrosion in metals. 37 In the case of mechanism (i) in Fig. 8, the H 2 supply to some regions of PtO x is limited due to the presence of thicker ionomer over layers, which exist when there is a non-uniform distribution of the ionomer in the cathode CL. In H 2 -rich regions, the HOR will occur, and the complementary reduction reaction would be Reaction 3. This can lead to a shortage of H + with which to reduce the Pt oxide, which would then be supplied by the COR (Reaction 4). In the case of mechanism (ii), the H + supply to some regions of PtO x can also be limited due to the presence of thin ionomer layers on the carbon support. In order to supply H + and electrons to the sites of Pt oxide reduction, the COR can occur at the carbon adjacent to the Pt oxide.
In step 3, the gas valves are opened, and the air is re-introduced to the cathode. Then the HOR commences in the H 2 -abundant regions, and the ORR (Reaction 5) commences in the air-abundant regions in the cathode CL.
In the case of mechanism (iii), however, if the ORR sites are located at a distance from the HOR sites, the supply of H + can also be limited due to the presence of thin ionomer layers when there is a nonuniform distribution of the ionomer. Therefore, the COR can occur at the carbon surface adjacent to the Pt particles in order to supply both H + and electrons to the ORR sites. Based on mechanism (iii), the carbon corrosion during step 3 should be severer both nearby the membrane and in the outlet region, if the supply of O 2 to the catalyst surface were impeded. However, it was found that the extents of carbon corrosion in the inlet and outlet regions were nearly the same, by using Raman spectroscopy, as reported in our previous work. 31,32 It is considered that the difference of carbon corrosion between the inlet and outlet regions was not so large under the present H 2 -SU/SD process, because the time for the gas exchange was short (1.2 s).
We can explain the fact that a N 2 purge in step 1 (Fig. 6) can prevent the COR during step 2, because the Pt oxide that was necessary for both Oxidation mechanisms of the carbon support due to local cells during step 2 (H 2 permeating to the cathode from the anode) both with (i) a single Pt particle and (ii) two or more Pt particles, and (iii) during the step 3 (artificial air resupplied) during the H 2 -SU/SD process. mechanisms (i) and (ii) was absent. We also note that the COR caused by mechanism (iii) during step 3 was accelerated by the enhanced ORR activity of the metallic Pt that was formed by reduction during step 2.
The effect of the E hold of the cathode in step 2 (Fig. 7) on the suppression of the COR can also be explained by the mechanisms outlined above. The degree of Pt oxidation increased with increasing holding potential. The preservation of the Pt oxide state is able to prevent the COR in both mechanisms i and ii, because the progress of the reduction reaction of the Pt oxide was inhibited by the E hold . The Pt oxide preservation also inhibits the ORR in step 3 and is able to suppress the COR in mechanism (iii). Nevertheless, the COR was still observed even in the case with E hold at 1.0 V. We consider that the metallic Pt that remained at the cathode holding potential of 1.0 V could have caused the COR, as described below. We calculated that the oxide coverages of Pt, which were ratios both of the Q Pt-oxidation and Q PtO-reduction versus the Q Pt-H as a hydroxyl (OH ad ) of one electron unit, were about 0.4 at 50 • C and 30% RH in Fig. 7, and were about 1.0 at 40 • C and 100% RH in Fig. 3, respectively. Wakisaka et al. indicated that not only the OH ad but also atomic oxygen (O ad ) of two electron units is adsorbed at 1.0 V in O 2 -saturated 0.1 M HF solution. 38 The calculated Pt oxide coverage of 0.4 was obtained at low humidity of the CL with the ionomer, and that condition seemed to decrease the amount of exposed metallic Pt due to adsorption of the anionic groups of the ionomer, i.e., sulfonate. 39 However, we consider that the adsorbed anions can be desorbed by the effect of generated H 2 O by the ORR (Equation 4) during step 3. Moreover, in fact, the oxide coverage is considered to be lower than 0.4 and thus the coverage of bare metal Pt to be greater in the cathode CL. The metallic Pt functions as an ORR site, and then the generated water tends to desorb the ionomer and thus to accelerate the COR.
In order to improve the durability of PEFCs in FCVs, the use of higher stability support materials, such as graphitized CB 32 or metal oxides [16][17][18][19][20][21][22][23][24][25][26] is an effective approach to inhibit the COR during the H 2 -SU/SD process. Based on the COR mechanisms (i), (ii) and (iii), the uniform distribution of the ionomer in the cathode CL prevents both the inhibition of H 2 supply and the shortage of H + which cause the COR associated with the reduction of the Pt oxide and the ORR. The increase of the exchange speed of air in the cathode also mitigates the COR caused by mechanism (iii). The effects of these approaches on the durability versus the SU/SD process will be able to be enhanced by the preservation of the Pt oxidation state.

Conclusions
We investigated the COR mechanisms during the H 2 -SU/SD process of PEFCs, in which the O 2 in the cathode is consumed by the H 2 permeating from the anode after the closing of the gas valves at both the inlet and outlet of the cathode. It was found that the use of the H 2 -SU/SD process decreased the amount of oxidized CB to approximately one-eighth compared with that of the air-SU/SD process, because the reverse current, which caused the severe COR, was suppressed. However, the degree of carbon corrosion that occurs during the H 2 -SU/SD process cannot be ignored for the long-term operation of FCVs. In order to clarify the mechanisms of the COR during the H 2 -SU/SD process, we investigated the following: (1) the effect of the presence of Pt catalyst on the COR by comparison of CBs both with and without Pt on each side of the membrane; (2) the timing of the COR during the H 2 -SU/SD process by comparison of the amounts of oxidized CB between N 2 and air supply to the cathode in the first step; and (3) the correlation between the COR and the Pt oxidation state by comparison of the amounts of oxidized CB during the H 2 -SU/SD process both with and without holding the cathode potential. The results indicated the following: (1) the COR occurred to a negligible extent in the absence of Pt catalyst in the cathode and was greatly accelerated by the presence of Pt catalyst; (2) the COR percentages that occurred during the H 2 permeation and air re-introduction were 23% and 77%, respectively; and (3) the COR was suppressed by the increase of the Pt oxide amount. Based on these results, we proposed the degradation mechanisms of the carbon support due to the non-uniform distributions of both ionomer ant Pt particles. The COR was caused by local cells that arose due to (i) a limited access of H 2 or limited access of protons associated with (ii) the reduction of the Pt oxide during the H 2 permeation from the anode to cathode and (iii) the ORR at metallic Pt sites during the air re-introduction.