Microstructured Membrane Electrode Assembly for Direct Methanol ...

Journal of The Electrochemical Society, 154 共10兲 B1034-B1040 共2007兲

B1034

0013-4651/2007/154共10兲/B1034/7/$20.00 © The Electrochemical Society

Microstructured Membrane Electrode Assembly for Direct Methanol Fuel Cell Hee-Tak Kim,z Tatyana V. Reshentenko, and Ho-Jin Kweon Samsung SDI Company, Limited, Yeongtong-gu, Suwon-si, Gyeonggi-do 443-391, Korea The concept of microstructured membrane electrode assembly for direct methanol fuel cell featuring a double-catalyst layer and micropatterned interface between the membrane and the catalyst layer was suggested and realized in this work. The doublecatalyst layer, which consists of a dense first layer and a porous second layer, was designed to achieve effective mass transfer in the catalyst layer. The micropatterned interface was generated by forming a micropattern on the membrane surface prior to catalyst coating on the membrane. The introduction of the double-catalyst layer resulted in a power density increase of 18%, and the generation of the micropatterned interface caused an additional power density increase of 23% at 50°C and 0.4 V. The I-V polarization and impedance analysis indicates that mass transfer in the catalyst layer was significantly improved owing to the large pores in the second catalyst layer. The micropatterned interface did not influence the cathode reaction rate; however, it caused a profound increase of the anode reaction rate. © 2007 The Electrochemical Society. 关DOI: 10.1149/1.2767406兴 All rights reserved. Manuscript submitted May 11, 2007; revised manuscript received June 8, 2007. Available electronically August 15, 2007.

Direct methanol fuel cell 共DMFC兲 has recently received special attention as a portable power source for mobile electronic devices such as notebook personal computer, portable media player 共PMP兲, and personal digital assistant 共PDA兲 owing to its high energy density and easy handling of methanol.1-5 In these applications, volume reduction of the DMFC system with satisfying power and energy requirement is of primary importance; however, fuel cell system volume is still larger than master devices. Therefore, there is a strong incentive to reduce the volume of each component of the DMFC system, which includes a power-generating unit called stack, a fuel delivery unit, control unit, and fuel tank. Membrane electrode assembly 共MEA兲 is one of the important units that affects the DMFC system volume. MEA generates electricity as a result of electrochemical reduction of oxygen in ambient air at the cathode catalyst layer and of electrochemical oxidation of methanol supplied from the fuel tank at the anode catalyst layer. The size of the stack, fuel tank, and fuel delivering unit at a specific power and energy is determined from MEA characteristics. For that reason, MEA performance enhancement strongly contributes to miniaturization of DMFC system. For facile electrochemical reaction, protons should rapidly move from anode to cathode passing through the membrane. Air and methanol aqueous solution, which are cathode and anode feed, respectively, should be rapidly delivered to the catalyst layer through the diffusion layer. The protonic transfer, electronic transfer, and mass transfer have to be balanced because the slowest processes limit power generation.1,2,6-8 The area of active catalyst surface where the protonic, electronic, and fuel transfer are all satisfied should be large enough to gain high power density because the electrochemical oxidation and reduction only occur at the active surface. In that sense, the morphological design of the catalyst layer is generally directed to enlarge active catalyst surface, not scarifying any diffusion processes.9-12 The difficulty in realizing a large, active catalyst area and fast diffusion simultaneously can be shown with the change of catalyst loading level. At low loading level, diffusion processes are fast because the catalyst layer is thin; however, the amount of active catalyst surface is often not enough to gain high power density.13-17 By contrast, in the high loading level regime, the increased catalyst layer thickness with the increase of catalyst loading level slows diffusion processes, making the catalyst inactive.13-17 In this work, we suggest a microstructured MEA which provides a large, active catalyst area and fast fuel diffusion with an aim to improve MEA power density. The microstructured MEA is characterized by the double-catalyst layer and the micropatterned interface.

z

E-mail: [email protected]

The double-catalyst layer consists of the dense first layer and the porous second layer. The concept of the double-catalyst layer is schematically illustrated in Fig. 1a and b. The large pores in the second catalyst layer 共see Fig. 1b兲 allow facile transfer of reactant to the first catalyst layer and rapid removal of by-product from the first catalyst layer. The second catalyst layer not only helps fuel transfer to the dense first layer but also contributes to power generation. The micropatterned interface was conceptually intended to enlarge the active catalyst surface as shown in Fig. 1c. Because a catalyst placed close to the membrane is most active due to facile proton transfer,6-8 the increase of the membrane/catalyst layer interface would effectively increase the amount of the electrochemically active catalyst. It was reported that membrane surface etching,18,19 sand-blasting,19 and Nafion spraying20,21 were utilized to increase the membrane roughness and then enlarge the reaction interface between membrane and electrode catalyst layer. These approaches generate irregular surface morphology. In order to generate a more regular and reproducible micropatterned interface, we tried to transfer a predetermined regular pattern to the membrane, followed by coating of catalyst ink on the surface-patterned membrane. The patterned interface dictates the surface patterns of the membrane. We combined the micropatterned interface and the double-catalyst layer to generate the MEA structure described in Fig. 1d. We investigated the morphological and electrochemical properties of these microstructured MEAs. The influence of the microstructure on power performance is discussed based on polarization and impedance results. Experimental Preparation of micropatterned membrane.— The micropatterned membrane was prepared by transferring the pattern of metal mesh to Nafion 115 membrane. Two different shapes of stainless steel mesh, M250 and M400, dimensions of which are drawn in Fig. 2, were used for membrane surface patterning. The meshes are placed on both sides of Nafion 115 and the stack was pressed under 250 kgf/cm2 at a temperature in the range of 125–145°C. After cooling the laminate down to room temperature, the metal meshes were peeled off from the membrane, leaving their patterns on both sides of the membrane. When the mesh was removed at a temperature higher than 100°C, mechanical damage was found on the membrane surface. Catalyst layer formation.— Pt black 共Johnson Matthey, Hispec 1000兲 and Pt-Ru black 共Johnson Matthey, HISPEC 6000兲 were used as catalysts for cathode and anode, respectively. Catalyst ink, consisting of appropriate amounts of the unsupported catalyst, Nafion solution, and isopropyl alcohol, were homogenized to disperse the catalyst. The ratio of the catalyst to the ionomer was maintained to be 1:0.12 共dry weight兲 for both anode and cathode. To generate the first catalyst layer, the ink was sprayed onto


B1035

Figure 1. Design of microstructured MEA: 共a兲 conventional single-catalyst layer without micropattern. 共b兲 doublecatalyst layer, 共c兲 micropatterned interface, and 共d兲 double-catalyst layer with micropatterned interface.

Nafion 115. In the course of the spray coating, the Nafion 115 membrane was immobilized on a vacuum table, the temperature of which was controlled at 60°C. After depositing the catalyst ink, the resulting coats were further dried at 60°C for 2 h under vacuum to completely remove residual solvent. For brevity, the first cathode catalyst layer-membrane-first anode catalyst layer is denoted as catalystcoated membrane 共CCM兲. The spray coating was also applied for the formation of the second catalyst layer on the diffusion layer. The diffusion layer used in this work was 10AA 共SGL, Germany兲 and 10DA 共SGL Germany兲 for anode and cathode, respectively. We coated the microporous layer that consists of Vulcan XC-72共Cabot, USA兲 and PTFE 关XC72:PTFE⫽1/1共wt/wt兲兴 on 10AA 共carbon loading level: 0.2 mg/cm2兲 and 10 DA 共carbon loading level: 1.3 mg/cm2兲 prior to the catalyst coating. The second catalyst layer-coated diffusion layer is called the catalyst-coated substrate 共CCS兲. The Pt loading for the first cathode catalyst layer and the second cathode catalyst layer was 1 and 6 mg/cm2, respectively. The Pt-Ru loading for the first anode catalyst layer and the second anode catalyst layer was also 1 and 6 mg/cm2, respectively. The area of the

catalyst layer was 10 cm2. The loading level was calculated based on the geometric area of the catalyst layer 共10 cm2兲. MEA preparation.— The resulting CCM and CCS were adhered together by pressing at enhanced temperature. The CCS cathode and CCS anode were placed on each side of the CCM, and they were pressed at 500 kgf/cm2 at 125°C for 3 min, resulting in bonding of the first and the second catalyst layer. Electrochemical characterization.— The MEAs were sandwiched between two plates with triple serpentine flow channel and mechanically pressed, ensuring a tight seal. An electrical heater and thermocouple were embedded in the plates for temperature control; 1 M aqueous MeOH solution was pumped to the anode side at a stoichiometry of 3. Ambient air was fed by pump into the cathode side at a stoichiometry of 3 without back pressure. The cell temperatures were varied in the range of 50–70°C. The MEAs were conditioned by operating at 0.4 V at 50°C for 2 h. After the conditioning, I-V polarization was obtained. This process was repeated for 10 days. The power performance was generally increased and stabilized within the first 3 days. Further test for 7 more days did not cause any difference in performance. The I-V polarizations measured at the third day were typically presented in this work. Electrochemical impedances of the MEAs were recorded by using IM6 共Zahner, Germany兲 at 50°C in the frequency range from 100 kHz to 1 mHz. Impedances were measured under galvanostatic control. The amplitude of the sinusoidal voltage signal was set to 5 mV. In order to separate anode and cathode impedance from total cell impedance, we measured anode impedance by injecting hydrogen to the cathode compartment.22,23 By subtracting anode impedance from total cell impedance, cathode impedance can be extracted.22,23 Results and Discussion

Figure 2. Structure of metal mesh for micropattern generation.

Morphological characteristics of double-catalyst layer.— Figure 3 compares the morphology of the first catalyst layer formed on the membrane surface and the second catalyst layer formed on the diffusion layer. A profound difference in morphology was found.

B1036


Figure 3. SEM image of catalyst layer 共a兲 first cathode layer, 共b兲 second cathode layer, 共c兲 first anode layer, 共d兲 second anode layer.

The second catalyst layer contains ⬃100 ␮m-sized large pores which originally exist in the diffusion layer; however, the first catalyst layer does not include large pores. From the scanning electron microscopy 共SEM兲 images, we are convinced the first and the second layers were prepared as we originally intended. Because the spray coating of the catalyst ink on the diffusion layer did not block the large pores in the diffusion layer, the morphology of the second catalyst layer resembles that of the diffusion layer. Meanwhile, the dense and flat nature of the membrane surface resulted in the morphological characteristic of the first catalyst layer. Morphological characteristics of micropatterned interface.— The SEM images of the patterned membrane surfaces prepared with M250 and M400 metal mesh are given in Fig. 4a and b, respectively. The patterns of the metal mesh were clearly dictated on the membrane surface. Even though we successfully generated the surfacepatterned membrane, the stability of the pattern had to be checked. Because the membrane is made of thermally deformable polymer, the surface pattern can be disrupted during lamination at enhanced temperature. Therefore, whether the micropattern is preserved at the lamination temperature of 125°C or not is a matter of concern. In order to check this issue, we measured the thickness of the patterned membrane after aging at 125°C, which is the lamination temperature for bonding CCM and CCS. As shown in Table I, the thickness of the patterned membrane was significantly decreased from 135 to 118 ␮m after aging at 125°C for 5 min, when the patterning temperature was 125°C. It is because the shape of the pattern became blunt during the aging, as shown in Fig. 5a and b. This behavior could be explained as such that the polymer chain is in an extended conformation after the pattern forming at 125°C and the recoiling of the polymer chain to relieve internal stress occurs when it is heated again, leading to the disruption of the pattern. Based on this consideration, we increased the patterning temperature, expecting stable and random chain conformation due to higher degree of thermal motion. The patterns formed at 135 and 145°C were found to be more thermally stable compared to that formed at 125°C, as given in Table I and Fig. 5c and d. The pattern forming at the higher temperatures renders the polymer chain lower in internal stress and reduces driving force for deformation during the aging.

Figure 4. SEM image of the micropatterned membrane: 共a兲 Surface of patterned membrane with M250, 共b兲 surface of the patterned membrane with M400. Patterning temperature: 125°C.


B1037

Table I. Thickness of micropatterned membranes with various pattern-forming temperatures before and after aging at 125°C for 5 min. After pattern After aging at Pattern forming 125°C Pristine forming, before Pattern temperature 共°C兲 membrane aging 共␮m兲共␮m兲 M250 M250 M250 M400

125 135 145 135

102 102 102 102

135 136 134 127

118 126 128 123

The temperature for preparing patterned membrane was selected to be 135°C, because at higher temperature such as 145°C, thermal degradation could occur. It is a generally stated problem that, during the direct deposition of the catalyst ink on membrane, the solvent in the ink readily swells the membrane, causing significant expansion of the membrane. In order to minimize dimensional change, the solvent in the deposited ink has to be rapidly removed before it swells the membrane. Because the excessive swelling of the membrane during the first catalyst layer coating could lead to the disruption of the membrane pattern, we carefully chose the spraying condition to minimize dimensional changes. The preservation of the pattern after coating the first catalyst layer was evidenced by optical microscopic 共OM兲 investigation. To prevent any deformation during specimen preparation, an epoxy resin was filled in the catalyst layers and cured at 60°C. Figure 6a shows the OM of the cross section of CCM. It is clear that the patterned interface is well maintained even after depositing the first catalyst layer on the membrane. The patterned interface was also observed after bonding the CCM and the CCS. The CCS was mechanically peeled off from the MEA and the remaining catalyst on the patterned membrane was also mechanically rubbed out to check whether the micro pattern is preserved after thermal lamination of CCM and CCS. The OM of the catalyst-removed membrane 共see Fig. 6b兲 showed superimposed pattern of the originally formed micropattern and the fibrous patterns of the CCS. The darker parts correspond to the site where the catalyst-coated fiber in CCS was bonded to the first layer of CCM. Even though high pressure was applied at 125°C during the lamination of CCM and CCS, the micropatterned interface was maintained at the pressurized sites. The two layers revealed acceptable interfacial bonding after the thermal lamination. When CCS was peeled off from the laminate of CCM and CCS, some part of the second catalyst layer remained on

Figure 6. 共Color online兲 Optical microscopic images of 共a兲 cross section of the CCM with the patterned interface and 共b兲 surface of the patterned interface after removing the CCS and the catalyst of CCM.

CCM, indicating the adhesion between the two catalyst layers is comparable to the cohesion of the second catalyst layer. Performance of the microstructured MEA.— The characteristics of the MEAs studied in this work are listed in Table II. The conventional MEA 共MS-1兲 was prepared by spraying the cathode and anode catalyst layers on each side of Nafion 115. In order to investigate the effect of the double-catalyst layer, we combined a CCM with unpatterned membrane and CCS 共MS-2兲. Two kinds of the micropatterned CCM were also combined with CCS to investigate the effect of the micropatterned interface 共MS-3 and MS-4兲. Figure 7 and Table III represent I-V characteristics on the third day at 50, 60, and 70°C, and power densities at 0.4 V for the first 3 days of measurements, respectively. For all the temperatures investigated, the double-catalyst layer showed significantly improved power density compared to the single-layered, conventional catalyst layer. For MS-1 共single layer兲, the voltage starts to drop more rapidly at current density of 0.16 A/cm2, which indicates that significant mass transfer limitation starts at this current density.7 However, for the double-catalyst layered MEA 共MS-2兲, significant voltage drop did not occur, even in the high current density regime. The

Table II. Characteristics of MEAs studied in this work.

Sample name

Figure 5. 共Color online兲 Comparison of optical micrographs of the patterned membranes 共N115 patterned with M250兲.

MS-1 MS-2 MS-3 MS-4

Electrode structure

Membrane/electrode interface

Loading level 共mg/cm2兲 anode/cathode

Single layer Double layer Double layer Double layer

Unpatterned Unpatterned Patterned with M250 Patterned with M400

7/7 7/7 7/7 7/7

B1038

Journal of The Electrochemical Society, 154 共10兲 B1034-B1040 共2007兲 Table III. Power densities of MEAs at 0.4 V (stoichiometry: 3 both for anode and cathode) „mW/cm2…. Third day Sample name MS-1 MS-2 MS-3 MS-4

First day 50°C

Second day 50°C

50°C

60°C

70°C

14.2 32.1 28.6 26.0

58.9 72.6 52.6 52.0

57.2 67.9 83.5 79.6

73.2 87.9 111.2 104.6

87.3 103.0 134.9 127.0

tical. MeOH crossover was measured by linear sweep voltammetry technique. The crossover current at open-circuit voltage 共OCV兲 at 50°C with 1 M MeOH feed was found to be 102 mA/cm2 for MS-2 and 121 mA/cm2 for MS-3. The higher crossover for MS-3 is due to the enlarged interface between anode and membrane. The fact that the micropatterned MEA revealed higher power density in the low current density regime 共⬍100 mA/cm2兲 in spite of the increased MeOH crossover strongly suggests that the electrochemical reaction rate is faster for the micropatterned MEA as we originally intended. In the high current density regime 共⬎100 mA/cm2兲, the merit of the micropatterned interface still holds. The power density of MS-3 at 0.4 V and 50°C was 23% higher than that of MS-2. MS-3 and MS-4 MEAs did not show significant mass transfer limitation at high current density. This is because these MEAs have the double-catalyst layer structure that provides facile mass transfer within the catalyst layer. The shape of the micropattern also seems to affect the cell performance. M250 共MS-3兲 revealed better power characteristics than M400 共MS-4兲. The difference can be explained in terms of the interfacial area of the micropattern. The membrane thickness of M250 and M400 patterned membrane was 136 and 127 ␮m, respectively. Because M250 generates deep indent on the membrane, it is believed that the membrane patterned with M250 has slightly larger interfacial area than that with M400. The design of the micropattern would be another important subject in the design of the microstructured MEA.

Figure 7. 共Color online兲 I-V polarizations of the MS-1, MS-2, MS-3, and MS-4 at various temperatures.

large difference in voltage at high current density between MS-1 and MS-2 shows that the introduction of the porous second layer is quiet effective for mass transfer. The power density at 50°C and 0.4 V was enhanced by 18% with the introduction of the double-catalyst layer. The comparison of MS-2, MS-3, and MS-4 shows the effect of the micropatterned interface. MS-3 and MS-4 revealed higher power density than MS-2 in all current density range investigated. The power density at lower current density generally depends on methanol crossover, and electrochemical reaction rate. Because the same catalyst was used, the catalytic activity can be assumed to be iden-

Impedance analysis of the microstructured MEA.— Even though significant improvement of power density by introducing the double-layered structure and the micropatterned interface was obtained, the mechanism of the performance enhancement is still ambiguous. In order to elucidate these effects, impedance analysis was performed for MS-1, MS-2, and MS-3. As described in the Experimental section, the anode and the cathode impedance were separated. The impedance was measured at 50°C at current density of 90 mA/cm2. Figure 8 compares the impedances of MS-1 and MS-2 to analyze the effect of the doublecatalyst layer. The total impedance shown in a Nyquist plot was larger for MS-1, which is consistent with I-V characteristics. As shown in Fig. 8b and c, MS-2 revealed smaller cathode and anode impedance than MS-1, implying the double-catalyst layer structure could enhance the charge transfer reaction rate for both cathode and anode. The increased reactant concentration is responsible for the decrease of the charge transfer resistance. Even though the double layer enhanced both the anode and the cathode performance, improvement of the cathode performance seems to be more significant. It suggests that mass transfer in the MS-1 cathode was slow. The water generated in the cathode also would be responsible for the low air diffusion in the single-cathode catalyst layer. Because there is a macropore for the double-layered catalyst layer 共MS-2兲, air can be more easily supplied even in the existence of flooded water. Figure 9 compares total, anode, and cathode impedances of MS-2 and MS-3 to investigate the role of the micropatterned interface on the power enhancement. The total impedance was significantly reduced by introducing the micropattern, which is in good


Figure 8. 共Color online兲 Nyquist plot of 共a兲 total impedance, 共b兲 anode impedance, 共c兲 cathode impedance of MS-1 and MS-2 measured at 50°C and 90 mA/cm2.

agreement with the I-V polarization result. Interestingly, cathode impedance was nearly unchanged, but anode impedance was reduced profoundly with the micropattern. The invariance of the cathode impedance indicates that the cathode reaction is mainly limited not by electrode kinetics but by dif-

B1039

Figure 9. 共Color online兲 Nyquist plot of 共a兲 total impedance, 共b兲 anode impedance, 共c兲 cathode impedance of MS-2 and MS-3 measured at 50°C and 90 mA/cm2.

fusion process, as predicted from the effect of the double-layered structure. By contrast, the anode reaction seems to be quite sensitive to the interfacial structure. For MS-2 and MS-3, degree of mass transfer in the anode catalyst layer is the same because they have the same electrode structure. Also, the current density of 90 mA/cm2 at

B1040


which impedance was detected was low enough to exclude any disturbance of mass transfer limitation. Therefore, catalyst activity, reaction order, and electrochemical surface area that determine the kinetics of methanol oxidation reaction would be mainly responsible for the anode impedance difference. Because the same catalyst was used, the catalyst activity and the reaction order have to be identical. The enlarged interface could provide efficient proton transfer through the catalyst layer and increase electrochemical surface area, consequently leading to the increased methanol oxidation reaction rate. Conclusion The concept of the microstructured MEA that employees doublecatalyst layer and micropatterned interface has been realized by combination of CCM and CCS and by direct coating of the catalyst ink on the patterned membrane surface. The micropatterned interface was maintained even after catalyst coating and hot lamination process. The introduction of the double-catalyst layer structure leads to 18% power density increase at 50°C and 0.4 V, and the generation of the micropatterned interface provides 23% additional power density increase. The I-V characteristics and the impedance analysis indicate that the porous layer in the double-layered catalyst layer enhances mass transfer in the anode and, more significantly, in the cathode. Whereas the micropatterned interface reduces only anode impedance, which implies the cathode oxygen reaction is relatively insensitive to active catalyst area, the facile anode reaction requires large active catalyst surface. Samsung SDI Company assisted in meeting the publication costs of this article.

References 1. A. K. Shukla, A. S. Arico, and V. Antonucci, Renewable Sustainable Energy Rev., 5, 137 共2001兲. 2. A. S. Arico, S. Srinivasan, and V. Antonucci, Fuel Cells, 1, 133 共2001兲. 3. H. Dohle, H. Schmitz, T. Bewer, J. Mergel, and D. Stolten, J. Power Sources, 106, 313 共2002兲. 4. A. Blum, T. Duvdevani, M. Philosoph, N. Rudoy, and E. Peled, J. Power Sources, 117, 22 共2003兲. 5. C. Y. Chen, P. Yang, Y. S. Lee, and U. F. Lin, J. Power Sources, 141, 24 共2005兲. 6. A. Hamnett, Catal. Today, 38, 445 共1997兲. 7. M. P. Hogarth and T. R. Ralph, Platinum Met. Rev., 46, 146 共2002兲. 8. T. Schultz, S. Zhou, and K. Sundmacher, Chem. Eng. Technol., 24, 12 共2001兲. 9. Handbook of Fuel Cells, W. Vielstich, A. Lamm, and H. A. Gasteiger, Editors, Vol. 4, John Wiley & Sons, Chichester 共2002兲. 10. Z. Wei, S. Wang, B. Yi, J. Liu, L. Chen, W. Zhou, W. Li, and Q. Xin, J. Power Sources, 106, 364 共2002兲. 11. A. Lindermeir, G. Rosenthal, U. Kunz, and U. Hoffman, J. Power Sources, 129, 180 共2004兲. 12. S. Q. Song, Z. X. Liang, W. J. Zhou, G. Q. Sun, Q. Xin, V. Stergiopoulos, and P. Tsiakaras, J. Power Sources, 145, 495 共2005兲. 13. L. Liu, C. Pu, R. Viswanathan, Q. Fan, R. Liu, and E. S. Smotkin, Electrochim. Acta, 43, 3657 共1998兲. 14. C. Y. Chen, P. Yang, Y. S. Lee, and K. F. Lin, J. Power Sources, 141, 24 共2005兲. 15. N. Nakagawa and Y. Xiu, J. Power Sources, 118, 248 共2003兲. 16. X. Zhao, X. Fan, S. Wang, S. Yang, B. Yi, Q. Xin, and G. Sun, Int. J. Hydrogen Energy, 30, 1003 共2005兲. 17. S. Surampudi, S. R. Narayanan, E. Vamos, H. Frank, G. Halpert, A. La Conti, J. Kosek, G. K. Surya Prakash, and G. A. Olah, J. Power Sources, 47, 377 共1994兲. 18. T. Sakai, Y. Kawami, H. Takenaka, and E. Torikioi, J. Electrochem. Soc., 137, 3777 共1990兲. 19. P. Millet, T. Alleau, and R. Durand, J. Appl. Electrochem., 23, 322 共1993兲. 20. Y. Zhang, C. Wang, N. Wan, Z. Liu, and Z. Mao, Electrochem. Commun., 9, 667 共2007兲. 21. S. Wang, G. Sun, G. Wang, Z. Zhou, X. Zhao, H. Sun, X. Fan, B. Yi, and Q. Xin, Electrochem. Commun., 7, 1007 共2005兲. 22. J. T. Mueller and P. M. Urban, J. Power Sources, 75, 139 共1998兲. 23. J. T. Müller, P. M. Urban, and W. F. Hölderich, J. Power Sources, 84, 364 共1999兲.