Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Characterization of structural degradation in a polymer electrolyte membrane fuel cell cathode catalyst… Young, Alan P. 2010

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2010_fall_young_alan.pdf [ 2.9MB ]
Metadata
JSON: 24-1.0058873.json
JSON-LD: 24-1.0058873-ld.json
RDF/XML (Pretty): 24-1.0058873-rdf.xml
RDF/JSON: 24-1.0058873-rdf.json
Turtle: 24-1.0058873-turtle.txt
N-Triples: 24-1.0058873-rdf-ntriples.txt
Original Record: 24-1.0058873-source.json
Full Text
24-1.0058873-fulltext.txt
Citation
24-1.0058873.ris

Full Text

CHARACTERIZATION OF STRUCTURAL DEGRADATION IN A POLYMER ELECTROLYTE MEMBRANE FUEL CELL CATHODE CATALYST LAYER by  Alan P. Young  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCE  in  The Faculty of Graduate Studies (Chemical & Biological Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  October 2010  © Alan P. Young, 2010  ABSTRACT  This study investigated the structural degradation of a polymer electrolyte membrane fuel cell (PEMFC) due to carbon corrosion and ionomer degradation. Cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS), scanning electron microscopy (SEM), and polarization analyses were completed to characterize and correlate the structural degradation to the performance. Accelerated stress tests (AST) were used to produce the different known degradation mechanisms. Both failure mechanisms had unique fingerprints on the performance degradation.  The carbon corrosion results showed a clear  thinning of the cathode catalyst layer (CCL) and gas diffusion carbon sub-layer, and a reduction in the effective platinum surface area caused by the carbon support oxidation. The degradation of the CCL and carbon sub-layer altered the water management, as evidenced by an increase of the voltage losses associated with oxygen mass transport and CL ohmic resistance. The ionomer degradation AST showed that greater ionomer in the CCL resulted in greater platinum content in the membrane and a higher fluoride washout rate, suggesting the higher ionomer content facilitated the mass transfer of contaminants (such as dissolved platinum and iron) into the membrane. It is proposed that H2O2 was produced at the anode, diffused into the membrane, and decomposed at the platinum and/or iron sites bound in the membrane structure. The decomposition products attacked the ionomer both in the bulk phase and CCL causing: i) membrane thinning which exacerbated H2 crossover, ii) lower membrane  ii  conductivity, and iii) CL structure degradation, resulting in increased reaction penetration into the CL and decreased effective oxygen diffusivity due to changes in CL water content. A method using an electrochemical quartz microbalance (EQCM) was investigated to further evaluate ionomer degradation. Mimicking the ionomer films in the CCL on the EQCM would enable a quantitative method to further evaluate the degradation reactions and overall mechanism. While this technique was not fully developed, background on the EQCM and the work to date is presented as a starting point for future development.  iii  TABLE OF CONTENTS Abstract ..…………………………………………………………………………………ii Table of Contents ………………………………………………………………………iv List of Tables ………………………………………………………………………...…vii List of Figures ………………………………………………………………………….viii List of Acronyms ……………………………………………………………………...…x List of Symbols …………………………………………………...……………………xii Acknowledgements …………………………………………………………………...xv Dedication ……………………………………………………………………………..xvi Co-authorship Statement ……………………………………………………………xvii  1.0  Introduction ................................................................................................ 1  1.1  PEMFC Background ............................................................................... 2  1.1.1  Gas diffusion layer ........................................................................... 5  1.1.2  Catalyst layer ................................................................................... 7  1.1.3  Membrane electrolyte .................................................................... 10  1.2  Experimental ......................................................................................... 12  1.2.1  MEA components .......................................................................... 12  1.2.2  MEA structure ................................................................................ 13  1.2.3  Operating conditions...................................................................... 13  1.2.4  Cyclic voltammetry ........................................................................ 15  1.2.5  Electrochemical impedance spectroscopy..................................... 15  1.2.6  Polarization analysis ...................................................................... 18  1.2.7  Limiting current diffusivity .............................................................. 23  1.2.8  Scanning electron microscopy ....................................................... 24  1.3 2.0  References ........................................................................................... 25 Characterizing the Structural Degradation in a PEMFC Cathode Catalyst  Layer: Carbon Corrosion ..................................................................................... 28 2.1  Introduction ........................................................................................... 29  iv  2.2  Experimental ......................................................................................... 33  2.3  Results and Discussion ........................................................................ 34  2.3.1  Carbon loss ................................................................................... 34  2.3.2  SEM analysis ................................................................................. 35  2.3.3  Cyclic voltammetry ........................................................................ 36  2.3.4  Electrochemical impedance spectroscopy..................................... 41  2.3.5  Polarization analysis ...................................................................... 49  2.4  Conclusions .......................................................................................... 52  2.5  References ........................................................................................... 55  3.0  Ionomer Degradation in Polymer Electrolyte Membrane Fuel Cells ......... 58  3.1  Introduction ........................................................................................... 59  3.2  Experimental ......................................................................................... 69  3.3  Results and Discussion ........................................................................ 71  3.3.1  Condensate analysis ..................................................................... 71  3.3.2  SEM analysis ................................................................................. 75  3.3.3  Cyclic voltammetry ........................................................................ 78  3.3.4  Electrochemical impedance spectroscopy..................................... 80  3.3.5  Polarization analysis ...................................................................... 84  3.3.6  Diffusivity analysis ......................................................................... 88  3.4  Conclusions .......................................................................................... 90  3.5  References ........................................................................................... 92  4.0  Application of Electrochemical Quartz Crystal Microbalance to Study  Degradation of Ionomer Films ............................................................................. 96 4.1  Introduction ........................................................................................... 97  4.1.1  EQCM background ........................................................................ 97  4.1.2  EQCM operation .......................................................................... 102  4.1.3  EQCM applications ...................................................................... 104  4.2  Experimental ....................................................................................... 109  4.3  Results and Discussion ...................................................................... 109  4.3.1  EQCM characterization ............................................................... 109  v  4.3.2  Contamination ............................................................................. 111  4.3.3  Film behaviour ............................................................................. 114  4.3.4  Bubble formation ......................................................................... 116  4.4  Conclusions ........................................................................................ 117  4.5  References ......................................................................................... 119  5.0 5.1  Concluding Chapter ............................................................................... 121 References ......................................................................................... 125  vi  LIST OF TABLES Table 1 – Test conditions for the standard fuel cell operation, CV/EIS experiments, and the accelerated stress tests. ................................................... 14  vii  LIST OF FIGURES Figure 1.1 – PEMFC components and operation. ................................................. 3 Figure 1.2 – Unit structure for a DuPont Nafion® electrolyte. .............................. 10 Figure 1.3 – Masked Membrane Electrode Assembly (MEA). ............................ 13 Figure 1.4 – Transmission line circuit and EIS response. ................................... 16 Figure 1.5 – Voltage loss breakdown. ................................................................. 19 Figure 2.1 – Carbon oxidation degradation mechanism...................................... 31 Figure 2.2 – Cumulative carbon loss versus corrosion cycle time....................... 35 Figure 2.3 – SEM picture at 0 and 30 hours of corrosion cycling. ....................... 36 Figure 2.4 – CVs over the corrosion period. ....................................................... 37 Figure 2.5 – Double layer capacitance over the corrosion period. ...................... 38 Figure 2.6 – EPSA over the corrosion period. ..................................................... 39 Figure 2.7 – EIS spectra as a function of gas RH. .............................................. 41 Figure 2.8 – CCL ionomers conductivity for 23 and 33 wt% Nafion content. ...... 43 Figure 2.9 – Comparison of EIS resistance performed during the AST. ............. 45 Figure 2.10 – Nyquist plot after 0 and 30 h of corrosion cycling. ........................ 46 Figure 2.11 – HF cell resistance versus RH over the corrosion period. .............. 47 Figure 2.12 – Changes in CL resistance and CL conductivity over the corrosion period as a function of gas RH. ........................................................................... 48 Figure 2.13 – Oxygen and air polarization curves under 100% and 60% gas RH for the 23 wt% MEA. ........................................................................................... 50 Figure 3.1 – Potential-pH (Pourbaix) diagram for H2O2 at 25°C.......................... 63 Figure 3.2 – Total MEA cumulative F- release versus time for different membrane thickness and electrode potentials. ..................................................................... 70 Figure 3.3 – Cumulative fluoride release during the AST. ................................... 72 Figure 3.4 – Analysis of product water conductivity during AST testing. ............. 73 Figure 3.5 – SEM Analysis at 0 and 440 h of the Pt dissolution AST for the 33wt% Nafion MEA. ............................................................................................ 76 Figure 3.6 – Membrane thinning during the Pt dissolution AST. ......................... 76 viii  Figure 3.7 – RH cycling effect on degradation of 33 wt% Nafion MEA. .............. 78 Figure 3.8 – EPSA and hydrogen crossover current during the Pt dissolution AST. .................................................................................................................... 79 Figure 3.9 – HF cell resistance for different membrane thickness. ..................... 80 Figure 3.10 – CL ionic resistance for different ionomers contents. ..................... 81 Figure 3.12 – Polarization loss analysis for the 23wt%Nafion MEA. ................... 85 Figure 3.13 – Polarization loss analysis for the 33 wt% Nafion MEA. ................. 85 Figure 3.14 – Average reaction penetration under oxygen and air. .................... 87 Figure 3.15 – Limiting current density versus oxygen concentration................... 89 Figure 4.1 – 9 MHz AT-cut quartz resonator with Pt electrodes. ......................... 97 Figure 4.2 – Quartz crystal cut at 45 degree. ..................................................... 98 Figure 4.3 – Correlates physics of quartz crystal to mass on a spring model. .... 99 Figure 4.4 – Simulated Bode plot for the motional segment of a QCM resonator. .......................................................................................................................... 101 Figure 4.5 – Equivalent circuit representing quartz, additional mass, and liquid on the QCM resonator. .......................................................................................... 104 Figure 4.6 – Equivalent circuit representing a film deposited on an EQCM resonator. .......................................................................................................... 106 Figure 4.7 – Measured Bode plot from impedance spectroscopy of a QCM resonator. .......................................................................................................... 110 Figure 4.8 – Calculating the mass constant from CV and EQCM. .................... 111 Figure 4.9 – Identifying region of contaminate adsorption on the EQCM. ......... 113 Figure 4.10 – Mass change with addition of Nafion film. ................................... 115 Figure 4.11 – Effect of gas bubbles on EQCM surface. .................................... 117  ix  LIST OF ACRONYMS AST – accelerated stress test Cdl – double layer capacitance CCL (CL) – cathode catalyst layer CCM – catalyst coated membrane CV – cyclic voltammetry DI – dionized water EDX – energy dispersive x-ray spectroscopy EIS – electrochemical impedance spectroscopy EPSA – effective platinum surface area EQCM (QCM) – electrochemical quartz crystal microbalance ESR – electron spin resonance spectroscopy EW – equivalent weight FTIR – fourier transform infrared spectroscopy GDE – gas diffusion electrode GDL – gas diffusion layer HRTEM – high resolution transmission electron microscopy ICPMS – inductively coupled plasma mass spectroscopy MEA – membrane electrode assembly MS – mass spectroscopy NMR – nuclear magnetic resonance spectroscopy OCV – open circuit voltage PEMFC – polymer electrolyte membrane fuel cell  x  RH – relative humidity; over-saturated or 120%RH implies a gas with 100% water vapour with additional liquid water. SEM – scanning electron microscope SS – steady state XRD – x-ray diffraction spectroscopy  xi  LIST OF SYMBOLS A  area  cm2  a  Tafel constant  V  b  Tafel slope  V decade-1  Cch  channel oxygen concentration  mol m-3  Cdl  double layer capacitance  F  CF  QCM film elasticity capacitance  F  Cf  QCM mass sensitivity factor  Hz cm2 ng-1  Cq  QCM energy storage capacitance  F  Cs  QCM static capacitance  F  c  quartz elastic constant  kg m-1 s-2  Dq  quartz dielectric constant  Deff  effective oxygen diffusion coefficient  cm2 s-1  E  cell voltage  V  Eo  standard equilibrium half cell potential  V  Ee  equilibrium cell voltage  V  F  Faradays constant  C mol e-1  ∆f  QCM frequency change  Hz  fo  QCM initial resonant frequency  Hz  i  current density  A cm-2  io  exchange current density  A cm-2  iH2X  hydrogen crossover current density  A cm-2  iL  limiting current density  A cm-2  k  EPSA loss rate constant  cycle-1 or h-1  L  EIS inductance  H  LF  QCM film motional inductance  H  LL  QCM inductance due to solution  H  LM  QCM inductance due to mass  H  Lq  QCM motional inductance  H  ∆m  QCM mass change  ng  n  electron number xii  ncyc  cycle number  R  gas constant  J mol-1 K-1  RF  QCM film motional resistance  Ω  RD  QCM film dissipation resistance  Ω  RL  QCM resistance due to solution  Ω  Rc  high frequency cell resistance  Ω  Ri*  operational CCL resistance  Ω  Ritotal  total CCL resistance  Ω  Rq  QCM motional resistance  Ω  Rpol  polarization ohmic resistance  Ω cm2  Rsh  shorting resistance  Ω  r  QCM energy dissipation factor  kg m-3 s-1  S  normalized EPSA  Smin  minimum normalized EPSA  T  temperature  K  Tw  Warburg time constant  s-1  tGDL  gas diffusion layer thickness  µm  tGDE  gas diffusion electrode thickness  µm  tcat  cathode catalyst layer thickness  µm  tq  quartz thickness  mm  treaction  average reaction penetration into CCL  µm  Zw  EIS Warburg impedance  Ω  α  transfer coefficient  ε  ionomer volume fraction  κ  piezoelectric stress constant  C m-2  εo  permittivity of free space  F m-1  ηL  QCM solution viscosity  g cm-1 s-1  ηMT  mass transport polarization loss  mV  ρq  QCM quartz density  g cm-3  ρL  QCM solution density  g cm-3  xiii  µq  QCM quartz shear modulus  g cm-3  σBULK  bulk ionomer conductivity  S cm-1  σCL  cathode catalyst layer ionomer conductivity  S cm-1  τ  cathode catalyst layer ionomers tortuosity factor  ω  angular frequency  rad s-1  xiv  ACKNOWLEDGEMENTS I would like to thank my many colleagues and mentors at Ballard Power Systems and UBC who supported my work and established the groundwork for this study. I would like to recognize and thank Dr. Elod Gyenge, Dr. Jürgen Stumper, and Shanna Knights for their guidance and mentorship. I would also like to thank Ballard Power Systems Inc. and the National Science and Engineering Research Council of Canada (NSERC) for funding this work. Special thanks to my very supportive wife Joslyn.  xv  DEDICATION  To my wife & daughter  xvi  CO-AUTHORSHIP STATEMENT This work was planned by Alan Young with guidance from Dr. Jürgen Stumper and Dr. Elod Gyenge. Experimentation and data analysis was completed by Alan Young. Papers and publications written by Alan Young with guidance and editing by Shanna Knights, Dr. Jürgen Stumper, and Dr. Elod Gyenge.  xvii  1.0  Introduction  1  There has been significant research conducted to evaluate different degradation mechanisms in the air-hydrogen polymer electrolyte membrane fuel cell (PEMFC) including carbon corrosion, platinum dissolution, and membrane degradation1-3. However, there has been little work to correlate the component structural damage to performance.  This thesis attempts to correlate the  structural damage in the cathode catalyst layer (CCL) and membrane to the overall MEA performance, and establishes a ‘fingerprint’ for carbon corrosion and ionomer degradation mechanisms, within specified accelerated stress tests (AST). Several ASTs have been developed in the industry4 to mimic degradation experienced in the field. Similar ASTs have been used in this study, and in some cases further developed, as defined further in subsequent chapters. Development of a technique using an electrochemical quartz crystal microbalance (EQCM) was investigated to further understand the ionomer degradation mechanism in the CCL. The goal of the technique was to measure ionomer degradation via mass changes with the EQCM, as a function of ionomer film thickness and peroxide concentration.  1.1  PEMFC Background  As shown in Fig. 1.1, a PEMFC is comprised of a two-electrode system separated by a solid polymer electrolyte. The role of the electrodes is to perform oxidation (anode electrode) and reduction (cathode electrode) reactions,  2  whereby a negative Gibbs free energy is available for conversion to high value electrical energy. An electrochemical potential for the oxidation and reduction reactions can be calculated from the change in Gibbs free energy.  When  combined in a cell these half-cell potentials provide the voltage driving force for the combined cell reaction.  Figure 1.1 – PEMFC components and operation. (Reference: Ballard Power Systems)  The electrons produced in the oxidation reaction flow through the anode electrode, through an external circuit, thus providing electrical power, and back to  3  the cathode electrode where they reduce the reactant. In order to complete the circuit the ions produced in the oxidation/reduction reaction must flow through the membrane electrolyte; thus, the role of the membrane electrolyte is to block electrons and reactants, but allow ionic current to pass with low resistance. In the air-hydrogen PEMFC, hydrogen is oxidized to form protons and electrons as shown in Eq. 1.1  2 H 2 → 4 H + + 4e −  [1.1]  The oxygen in air adsorbs onto the cathode catalyst and is reduced by reacting with protons and electrons to form water and heat, as shown in Eq. 1.2.  O2 + 4 H + + 4e − → 2 H 2 O + Heat  [1.2]  The complexity of the system increases with the need to deliver the reactant gases to the electrodes while still maintaining a conductive path to the current collectors. An electrically conductive gas flow field is used to deliver the gas to the cell. This is typically formed in a metal or carbon composite collector plate, providing the multiple functions of reactant access and removal, electron and heat transport, and structural support to the membrane electrode assembly. However, to promote gas access to the catalyst under the landings between the flow fields and provide support to the thin membrane electrolyte across the flow field channel span, a gas diffusion layer (GDL) is added to the structure for both the cathode and anode electrodes. The entire PEMFC is under compression to provide good contact between all of the component interfaces.  4  In order to properly design a PEMFC it is important to understand the materials used for the components in the PEMFC, the impact of these materials on the overall performance, and the degradation of these materials in the PEMFC operating environment.  The following brief discussion outlines the  materials commonly used for the GDL, catalyst layers (CL), and membrane electrolyte. Components that are not discussed here are the collector plates and seal materials; however, a more thorough review including these components is presented by Rama et al.3  1.1.1 Gas diffusion layer The GDL serves three purposes: 1) to provide a pathway for mass transfer of reactants (oxygen and hydrogen) and products (water) to and from the CL, 2) to ensure structural stiffness to the MEA, and 3) to provide an electrically and thermally conductive pathway between the CL and the current collectors. The GDL (100-300 µm in thickness) is often made of a carbon fibre material with a carbon microporous layer or sub-layer consisting of carbon particles and a hydrophobic binder. The GDL design can be significantly different for different applications and operating regimes.  For example, for a PEMFC operating at high current  densities where oxygen is consumed and a significant amount of water is produced, it is important to have a high porosity to allow for greater mass transfer to and from the CL. Alternatively, at low current densities a lower porosity may be required to keep water in the MEA to hydrate the membrane electrolyte, 5  especially if operating with dry gases5. The GDL thickness can also be levered in a similar manner depending on the operating conditions; however, there are other constraints that provide a minimum thickness required depending on the channel and landing spacing of the gas flow field. For example, a wide landing will require a thicker GDL to provide gas access to the catalyst under the landing. This will also be a function of the current density and the oxygen diffusion gradient through the GDL. Of course, this is a trade-off with the electrical contact area of the GDL and landing.  In order to reduce ohmic losses and uneven  current distribution between channel and landings, while maintaining gas access to the CL, an optimized contact area is required. Models have been produced to couple these variables and provide trade-off analysis6. The carbon sub-layer is added to the carbon fibre paper to make the GDLCL interface more uniform and smooth.  Without the sub-layer the relatively  rough carbon fibre paper could puncture the membrane, especially when the PEMFC is under compression. To smooth the GDL while still allowing electrical current and gas access, a conductive material with adequate porosity and fine particle size is required. The most common material used is fine carbon with a hydrophobic binder.  The hydrophobic binder provides the required adhesion  between particles and the carbon fibre paper, while also keeping liquid water out of the pore space of the sub-layer. This binder material is often added to the carbon fibre paper itself for improved water management7, 8.  6  1.1.2 Catalyst layer The porous CL (5-25 µm in thickness) is comprised of a mixture of platinum catalyst supported on electrically and thermally conductive carbon particles, and proton conducting ionomer. which has three roles:  The CCL typically contains 20-40 wt% ionomer,  1) to act as a binder between the platinum/carbon  particles, 2) to provide a proton conductive link from the membrane to the catalyst for protonic current flow, and 3) to make the platinum catalyst electrochemically active by forming the electronic-ionic charge transfer layer, also known as the Helmholtz double layer. The platinum loading, platinum to carbon ratio, carbon type, and ionomer loading and type are all structural design levers used to improve the CL for performance and durability. Perhaps the most obvious, the platinum loading has large implications on both performance and durability. The high price of platinum is a significant driver to reduce the platinum loading in the PEMFC; however, doing so generally reduces performance and durability. Under certain operating conditions the platinum catalyst dissolves, and either agglomerates, or diffuses through the membrane electrode assembly (MEA). This results in a reduction in the platinum surface area available for the half-cell reactions, resulting in greater polarization losses and lower efficiencies. Designing the PEMFC with a lower platinum loading can have this effect on performance, while catalyst dissolution and agglomeration results in lower overall durability. Development of a more stable catalyst that is not subject to a reduction in surface area over its lifetime will facilitate the reduction of platinum loading and PEMFC cost.  7  The carbon type used for the catalyst support has a significant impact on CCL durability.  Yu and Ye9,  10  outline different catalyst preparation, carbon  support types, thermal treatments, and Pt-C interactions as they pertain to performance and durability. Acidic oxygen surface groups on the carbon (vs. basic groups) increase hydrophilicity making the surface more accessible to aqueous solution of metal precursor, resulting in greater platinum dispersion. Greater dispersion results in smaller catalyst particle sizes, greater surface area, and improved kinetic performance. Alternatively, less acidic surface groups are more stable and provide better platinum anchorage; thereby, enhancing the thermal stability of platinum. Thermal treatment has also been shown to increase basic π sites on the carbon surface, which have been shown to strengthen the Pt-C interaction, making them more stable. As will be discussed in detail in chapter two, carbon oxidation, or corrosion, can occur under several operating scenarios in the PEMFC. Graphitized carbons have fewer surface groups susceptible to corrosion attack and have been shown to have greater corrosion resistance11-13.  Unfortunately, in most carbons the  graphitization level and carbon surface area are closely linked (increasing the graphitization results in reduced surface area). With a small surface area and fewer surface groups, the Pt-C interfacial area is smaller; therefore, the Pt distribution and particle size is limited, resulting in lower Pt surface area, and reduced performance compared to catalysts with less graphitic carbon supports. The balance between performance (high surface area and catalyst dispersion)  8  and durability (graphitization) for various applications will dictate the optimized carbon type, until a corrosion resistant high surface area carbon is developed. The presence and weight percent of platinum in the catalyst has been shown to significantly impact the corrosion rate of carbon in both ex-situ thermal13 and in-situ PEMFC potential tests14. The presence of platinum lowered the activation energy for carbon corrosion compared to carbon only electrodes. Greater platinum percent increased the amount of carbon being corroded, thus increasing the corrosion rate, most likely due to increased contact area between platinum and carbon. Studies have been conducted to evaluate the optimal Pt/C ratio regarding performance15. Generally, a greater Pt/C ratio results in a thinner CL, which changes the reaction distribution through the CL. The shorter paths of reactants, oxygen and protons, potentially improve performance; however; higher Pt/C ratios also increase the starting platinum particle size, which reduces the platinum surface area and therefore, the kinetic performance. Alternatively, a lower Pt/C creates a thicker layer with a lower average Pt particle size, greater platinum surface area, and improved kinetics; however, a longer oxygen and proton path length may increase mass transport and ohmic voltage losses. Therefore, a trade-off exists between the kinetic and non-kinetic losses. While an increased starting particle size in the higher Pt/C catalysts will result in lower performance, they are less prone to further dissolution16 and will reach a more stable particle size sooner. In this case, performance is sacrificed for durability.  9  The ionomer loading in the CCL has a large impact on performance due to the direct impact on ionic conductivity. This is discussed in detail in chapter three as is the effects of catalyst ionomer degradation on performance. Since it is the same material, it is speculated that the ionomer in the CL behaves similar to the membrane electrolyte as will be discussed next.  1.1.3 Membrane electrolyte The role of the membrane electrolyte is to block electrons and reactants, but allow ionic current to pass with low resistance from the anode to cathode electrode. The most widely used proton conducting ionomer in the PEMFC industry is Nafion® made by DuPont.  The Nafion chemical structure is comprised of a  fluoropolymer backbone with sulphonic acid site side chains as shown in Fig. 1.2.  Figure 1.2 – Unit structure for a DuPont Nafion® electrolyte.  The fluoropolymer backbone provides the structure to the ionomer and is hydrophobic. The sulphonic acid side chains form clusters that are hydrophilic and provide the proton conductivity functionality. The proton conduction depends  10  on both the ionomer structure and the level of hydration within the ionomer. The equivalent weight (EW) of the electrolyte is the inverse of the ion exchange capacity, and is defined by the gram of material per mol of sulphonic acid sites. Increasing the number of sulphonic acid side chains makes lower EW electrolytes. Lower EW (high ion exchange capacity) electrolytes generally have higher ionic conductivity, are more hydrophilic and can adsorb more water, resulting in greater swelling. Two main design levers for membrane performance are the proton conductivity and membrane thickness.  As discussed the main method to  improve conductivity is by decreasing the EW of the polymer; however, this comes at a sacrifice of ionomer stability17. The membrane thickness directly impacts the ohmic resistance of the MEA; however, as the thickness is decreased the potential for punctures and gas leaks increases, as does the diffusion of reactant gases through the membrane. Gas crossover results in parasitic reactions that will decrease overall PEMFC efficiency and can produce radicals that degrade the membrane electrolyte, impacting durability.  This particular degradation mechanism is  discussed in greater detail in chapter three.  11  1.2  Experimental  Due to the similarity in test apparatus and methods, the experimental section of chapters two and three have been reduced from their published format, and where appropriate, combined here. 1.2.1 MEA components MEAs for this work were prepared by bonding cathode and anode gas diffusion electrodes (GDEs) to a DuPont Nafion membrane at 150 °C, 15 bar, for 2.5 minutes in a compaction press. The carbon corrosion AST used two layers of DuPont NR211 to form a 60 µm membrane; however, the ionomer degradation AST used a single 30 µm DuPont NR211 membrane for most of the testing. The anode GDE consisted of a Toray TGP-60 carbon fibre paper with a carbon and Teflon® sub-layer and a graphitic carbon supported platinum (50 wt% carbon/50 wt% platinum) CL with a platinum loading of 0.3 mg cm-2.  The  cathode GDE was similar to the anode GDE, only with greater nominal platinum loading of 0.6 mg cm-2. The ionomer content was varied between 23 and 33 wt% of the CCL, to show differences in the CL ionomer conductivity and any influence on degradation. To facilitate better bonding to the membrane, both the anode and cathode were sprayed with an additional 0.2 mg cm-2 Nafion solution. It was assumed that this spray coat stayed near the surface of the CL and did not penetrate and contribute significantly to the amount of ionomer in the CL.  12  1.2.2 MEA structure The MEAs had a 50 cm2 anode geometric area and a masked cathode with a 4 cm2 geometric area as shown in Fig. 1.3. The remainder of the 50 cm2 cathode GDE was stripped of catalyst with tape and covered with a 12 µm thick polyester mask. Since the CL was only slightly thicker (16 µm) than the polyester mask, even compression was achieved. The smaller masked area was used to help provide a one-dimensional operating condition, where temperature, pressure, and concentration gradients were reduced to the point of being negligible.  Figure 1.3 – Masked Membrane Electrode Assembly (MEA).  1.2.3 Operating conditions The cell was conditioned at the standard steady state (SS) operating conditions listed in Table 1.  The gas flow rates correlate to a reactant 13  stoichiometry of greater than 50, which approximated a one-dimensional environment, where temperature and concentration gradients were minimized across the active area of the MEA. Table 1 – Test conditions for the standard fuel cell operation, CV/EIS experiments, and the accelerated stress tests.  Nominal Operating Conditions  Temp (°C)  Pressure (psig)  Oxidant Flow Rate (slpm)  Fuel Flow Rate (slpm)  Relative Humidity %  Standard (SS)  70  30  25 (Air)  15 (H2)  100  CV & EIS  70  30  25 (N2)  15 (H2)  50 -120  Corrosion AST  70  30  0.3 (N2)  0.3 (H2)  120  Ionomer  90  30  0.3 (Air)  0.3 (H2)  120  Degradation AST  The PEMFC hardware used a pressurized bladder to provide the necessary compression to maximize contact between the MEA and the conductive flow field plates. The performance varied with bladder pressure due to the ohmic contact resistance between the plate and MEA, until the contact resistance was minimized given the GDL and channel geometry. The bladder pressure was maintained at 80 psig to reduce contact resistance and prevent over compression, which could crush the GDL material and decrease performance and durability.  14  1.2.4 Cyclic voltammetry Cyclic voltammetry (CV) was conducted using a Solartron SI1287 potentiostat in order to calculate the effective platinum surface area (EPSA), the H2 crossover current, and the double layer capacitance (Cdl). Hydrogen present at the anode acted as the reference electrode as the cathode potential was cycled between 0.1 and 1.0 VRHE. The EPSA was determined by CO stripping, assuming a charge density of 420 µC cm-2 to break the linear Pt-CO bond15. Hydrogen crossing through the membrane was oxidized at the cathode catalyst.  The corresponding H2 crossover current was determined from the  anodic shift of the voltammograms on the current axis. The H2 crossover current density (iH2X) was used as a general indicator of gas crossover throughout this study. The double layer capacitance Cdl was determined from the CVs at 0.45 VRHE due to the absence of faradaic reactions.  1.2.5 Electrochemical impedance spectroscopy Electrochemical impedance spectroscopy (EIS) measurements were taken by applying a 10mV AC perturbation signal with a 0.45 VRHE DC bias potential. A SI1287 Solartron potentiostat and a 1250 Solartron high frequency (HF) response analyzer were used in a 4-electrode configuration to sweep the frequency between 50 kHz-0.05Hz. The bias potential was applied to eliminate any pseudo-capacitive effects that result from hydrogen and oxygen adsorption-  15  desorption18. Z-plot and Z-view software were used to conduct and analyse the EIS spectra to determine the cell and CL ionomer resistance. Measurements were taken in a nitrogen and hydrogen atmosphere on the cathode and anode, respectively. Gas relative humidity (RH) was varied between 50 and 120% to obtain a relationship between the CL ionomer resistance and RH over the degradation period. The EIS spectra were fit to an equivalent circuit representing a transmission line circuit as shown in Fig. 1.4 a).  Figure 1.4 – Transmission line circuit and EIS response.  16  The transmission line circuit consists of two rails of resistive components, ionic and electronic, connected by capacitive elements that represent the Cdl. At high frequencies the Cdl offered no resistance and only the membrane and cell electrical resistance were measured (HF cell resistance - Rc). As the frequency decreased, the Cdl at the membrane/catalyst interface charged and created impedance. Further decreasing the frequency, current penetrated further into the pore measuring both the impedance of the Cdl of the pore wall and the CL ionomer resistance (Ri). Once the entire pore was measured the impedance increased toward infinity as the capacitance became fully charged and blocked all current. This is shown in Fig. 1.4 b) with the Nyquist plot for the experimental EIS spectra and the equivalent circuit fit.  The open circuit Warburg circuit  element represents De Levie porous electrode behaviour19 and is described by a cotanh function. Expanding this function into a series expression, the resistance reduces to Ri/3 as the frequency approaches zero. The TW term represents the time constant (resistance x capacitance). By dividing TW by Ri the Cdl can be calculated. In nitrogen atmosphere at 0.45 VRHE the limited H2 crossover caused the only possible electrochemical reaction, which had an excessively high charge transfer resistance due to the low H2 concentration.  The charge transfer  resistance was ignored in the transmission line circuit. The equivalent circuit also showed an inductance (L) and a shorting resistance (Rsh). The inductance is a system artifact and is often subtracted from the EIS spectra; however, it was presented here for completion. The shorting resistance represents membrane shorting behaviour that may occur in the MEA. Both the inductance and shorting  17  resistance terms were included in the equivalent circuit to improve the fitting accuracy. Pickup et al.18,  20-22  and Makharia et al.23 have used this EIS method to  describe the ionomer resistance in the porous structure for several PEMFC CL structures with and without load applied.  This method was used here to  understand the structural change in the CL due to oxidation of the carbon support and degradation of ionomer.  1.2.6 Polarization analysis Voltage-current polarization curves were measured at the standard operating conditions in Table 1 with pure oxygen and air.  Traditionally the  polarization curve is separated into different mechanistic phenomena: kinetic, ohmic, and mass transport. This study combined the impedance analysis to further breakdown this mechanistic phenomena into the main MEA components as shown in Fig. 1.5 a). The breakdown of components was similar to Neyerlin et al.24; however, a more empirical curve fitting approach was used to obtain the kinetic and CL ohmic losses, as opposed to a model based approach. The oxygen polarization data was fitted to Eq. 1.3 in order to determine the kinetic and ohmic losses:  E = Ee − b log(i + iH 2 X ) − a − i ⋅ R pol  [1.3]  where Ee is the equilibrium cell potential, i is the current density, iH2X is the H2 crossover current density, Rpol is the area-specific ohmic resistance, b is the 18  Tafel slope, and a is the Tafel constant. It was assumed that no mass transport losses occurred with pure oxygen. The effect of the iH2X was included to obtain more accurate Tafel parameters25. a)1.3 Theoretical Maximum Voltage  1.2  Cell voltage (V)  1.1 Kinetic: Oxygen - (line 1)  1.0 0.9 0.8  + Kinetic: Air - (line 2)  0.7  + Ohmic: Membrane & cell electrical (Rc) - (line 3)  0.6  + O2 Gas mass transport - (line 4)  0.5  + Ohmic: Catalyst layer ionic (Ri*) - (line 5)  0.4 0.0  0.5  1.0  1.5  2.0  2.5  3.0  3.5  4.0  4.5  5.0  -2 Current density (A cm )  b)  Figure 1.5 – Voltage loss breakdown.  19  The fitted Tafel parameters were used to calculate the exchange current and the transfer coefficient given in Eq. 1.4 and 1.5  b=  2.303 ⋅ R ⋅ T α ⋅n⋅F  a = −b ⋅ log(io )  [1.4]  [1.5]  where R is the gas constant, T is the temperature (K), n is the number of electrons exchanged, F is Faraday’s constant, α is the transfer coefficient, and io is the exchange current. The kinetic air polarization curve was then calculated from the iR-free oxygen curve, using the concentration corrected exchange current density according to Eq. 1.626,  io (air ) = io (oxygen) ⋅ 0.21(1−α ) = 10 ( − a / b ) ⋅ 0.21(1−α )  [1.6]  The difference between the oxygen and air polarization curves due solely to the oxygen concentration difference was typically around 50mV. The membrane ionic and cell electrical resistance (Rc), including the GDL, flow field plates, and interfacial contact resistances, was measured by the EIS analysis at HF (Fig. 1.5 a). Traditionally, mass transport losses are fit according to Eq. 1.727  η MT =  2.303 ⋅ R ⋅ T i log(1 − ) α ⋅n⋅F iL  [1.7]  20  and exhibit a well-defined abrupt (knee shaped) drop in potential reflecting the limiting current density (iL).  This loss was calculated by measuring the  approximate iL in an air environment.  It was added to the iR corrected air  polarization curve as shown in Fig. 1.5 a). This well-defined limited abrupt voltage drop rarely occurs in the operating fuel cell and more often a gradual drop is observed.  It is believed that this  gradual drop is caused by greater CL ionomer ohmic losses before the mass transfer limiting current is attained. The performance drop between the calculated air polarization (including ohmic and mass transport losses) and the measured air polarization can be attributed to the ohmic loss associated with the ionic current distribution in the CL.  This ohmic loss is influenced by the CL ionic conductivity, and it is  intrinsically linked also with the distribution of kinetic activity and mass transport losses across the CL. For example, low CL ionic conductivity will create a low oxygen reduction reaction penetration depth into the CL starting from the membrane, forcing oxygen molecules to travel farther to reach the reaction zone. The ionic conductivity can be used as an indicator of the proton availability in the CL, which is obviously a reactant in the ORR.  Furthermore, a high charge  transfer resistance would increase the reaction penetration into the CL toward the GDL to access more platinum surface area, and mass transport limitations would also force a greater reaction penetration due to higher oxygen concentration at the CL-GDL interface. Some of these concepts are depicted in Fig. 1.5 b).  21  The concept of balancing kinetics, mass transport limitations, and CL reaction penetration has been previously discussed in literature28-30. In the mass transport free region, Baker et al. defined the reaction penetration depth by the ratio of the charge transfer resistance and the CL ionic resistance31. Neyerlin et al. further calculated the effective or operational catalyst utilization based on the reaction penetration into the CL from the membrane24. Although the protonic current through the CL is distributed, equation 1.8 estimated the average reaction penetration by calculating the ratio of the CL ionic resistance during operation, Ri*, and the total CL ionic resistance measured by EIS, Ritotal. This ratio was equated to the ratio of the average reaction penetration length, treaction, into the CL, to the CL thickness, tcat. average ⋅ reaction ⋅ penetration =  t Ri * = reaction Ritotal t cat  [1.8]  The analysis done in this study was a simplified empirical approach and contains some inaccuracies, such as neglecting the effect of the reaction penetration on the operational platinum surface area and kinetic performance. Despite these deviations, separating the polarization losses by component provided understanding of the operating characteristics of each MEA component and gave a quick method to evaluate degradation and compare different MEA designs.  22  1.2.7 Limiting current diffusivity The limiting current effective diffusivity measurement used by Baker et al.31, to show differences between GDLs under varying water contents, was used as a supplementary test to verify any change in the mass transport limitations over the degradation period. The effective diffusivity of oxygen through the MEA was measured based on Fick’s Law as described by Eq. 1.9  iL =  Deff ⋅ (Cch) ⋅ n ⋅ F t GDE ⋅10 2  [1.9]  where iL is the limiting current density, Deff is the effective diffusivity, Cch is the channel oxygen concentration, n is the mole electrons per mole oxygen, F is Faraday’s constant, and tGDE is the diffusion path length. Deff includes oxygen diffusivity through the pore space as well as through the Nafion electrolyte. By plotting the limiting current density against the channel oxygen concentration the effective diffusion coefficient was calculated from the slope of the curve. In this test the diffusion path length was approximated by the gas diffusion electrode (GDE) thickness. Any tortuosity effects were considered part of the effective diffusivity coefficient. The effective diffusivity measurements were conducted under the standard operating conditions listed in Table 1. With the same flow rates the oxygen concentration was varied between 0.5 and 6.5% O2 balanced with N2, while the limiting current was measured.  At each  concentration the current was allowed to stabilize at a cathode potential of approximately 0.15 VRHE, which was assumed to be sufficiently low to assure oxygen mass transport limiting conditions.  23  1.2.8 Scanning electron microscopy After operation the MEAs were analyzed with a Philips XL30 scanning electron microscope (SEM) to detect changes in the MEA structure.  MEA  samples were cast into epoxy pucks, polished using a Struers TegraPol-11 polisher with 120-1200 grit silicon carbide paper, and given a carbon topcoat using an Edwards Scancoat6 coater. Pictures were taken using a backscatter detector at 400x magnification and 15 kV. The membrane and CL thickness were measured and used in conjunction with the component loadings to calculate the CL porosity, and ionomer volume fraction. The CL ionomer conductivities were also calculated by normalizing the EIS resistance measurements by the CL thickness.  24  1.3  References  1.  R. Borup et al., Chem. Rev., 107, 3904 (2007)  2.  S. Zhang, X. Yuan, J. Hin, H. Wang, K. Friedrich, M. Schulze, J Power Sources (2009)  3.  P. Rama, R. Chen, and J. Andrews, Proc. ImechE Part A: J. Power Energy 222, p.421 (2008)  4.  S. Zhang, X. Yuan, H. Wang, W. Merida, H. Zhu, J. Shen, S. Wu, J. Zhang, Int. J. of Hydrogen Energy, 34, p.388 (2009)  5.  E.R. Gallagher, U.S. Pat. 7,374,838 (2008).  6.  W. Sun, B. Peppley, and K. Karan, J. Power Sources 144, 42 (2005).  7.  U. Pasaogullari, and C. Wang, J. Electrochemical Society 151, A399 (2004).  8.  G. Park, J. Sohn, T. Yang, and Y. Yoon, J. Power Sources, 131, 182 (2004).  9.  X. Yu, and S. Ye, J. Power Sources, 172, 133 (2007)  10. X. Yu, and S. Ye, J. Power Sources, 172, 145 (2007) 11. P.T.Yu, W. Gu, R. Makharia, F.T.Wagner, and H.A. Gasteiger, ECS Transactions 3, 797 (2006). 12. J.E. Owejan, P.T.Yu, R. Makharia, ECS Transactions, 11, 1049 (2007). 13. D.A. Stevens, M.T. Hicks, G.M. Haugen, and J.R. Dahn, J. Electrochem. Soc., 152, A2309 (2005).  25  14. L.M. Roen, C.H. Paik, and T.D. Jarvi, Electrochemical and Solid State Letters 7, A19 (2004). 15. T.R. Ralph, G.A. Hards, J.E. Keating, S.A. Campbell, D.P. Wilkinson, M. Davis, J. St-Pierre, and M.C. Johnson, J. Electrochem. Soc., 144, 3845 (1997). 16. R. Darling, and J. Meyers, J. Electrochem. Soc. 150, A1523 (2003). 17. J.Qiao, M.Saito, K. Hayamizu, and T. Okada, J. Electrochem. Soc., 153, A967 (2006). 18. E. B. Easton, and P. G. Pickup, Electrochimica Acta, 50, (2005). 19. R. De Levie, Electrochimica Acta, 9, 1231 (1964). 20. M.C. Lefebvre, R.B. Martin, and P.G. Pickup, Electrochemical Solid State Letters, 2, 259 (1999). 21. X. Ren, and P.G. Pickup, Electrochemica Acta, 46, 4177 (2001). 22. G. Li, and P.G. Pickup, J. Electrochem. Soc., 150, C745 (2003). 23. R. Makharia, M.F. Mathias, and D.R. Baker, J. Electrochem. Soc., 152, A970 (2005). 24. K.C. Neyerlin, W. Gu, J. Jorne, A. Clark, and H.A. Gasteiger, J. Electrochem. Soc., 154, B279 (2007). 25. H.A. Gasteiger, S.S. Kocha, B. Sompalli, and F.T. Wagner, Applied Catalysis B: Environmental, 56, 9 (2005).  26  26. A.J Bard, and L.R. Faulkner, Electrochemical Methods: Fundamentals and Applications, p.101, John Wiley & Sons, New York (1980). 27. A.J Bard, and L.R. Faulkner, Electrochemical Methods: Fundamentals and Applications, p.110, John Wiley & Sons, New York (1980). 28. T. E. Springer, M.S. Wilson, and S. Gottesfeld, J. Electrochem. Soc. 140, 3513 (1993). 29. E. Gyenge, J. Power Sources, 152, 105 (2005). 30. M. Eikerling, and A.A. Kornyshev, J. Electroanal. Chem., 475, 107 (1999). 31. D. Baker, C. Wieser, K.C. Neyerlin, and M.W. Murphy, ECS Transactions, 3, 989 (2006).  27  2.0 Characterizing the Structural Degradation in a PEMFC Cathode Catalyst Layer: Carbon Corrosion  A version of this chapter has been published: A. P. Young, J. Stumper, and E. Gyenge, “Characterizing the Structural Degradation in a PEMFC Cathode Catalyst Layer: Carbon Corrosion”, J. Electrochem. Soc., 156, B913 (2009).  28  2.1  Introduction  It is evident that there are several mechanisms that can affect the performance of the PEMFC including hydrogen and oxygen mass transfer to the reaction sites, proton conductivity through the ionomer, hydrogen and oxygen electrode reaction rates, and water management. All of these mechanisms exist within the electrode structure, which consists of a GDL and a CL. The GDL (100300 µm) is often made of a carbon fibre material with a carbon sub-layer consisting of carbon particles and a hydrophobic binder. The GDL serves three purposes: 1) to provide a pathway for mass transfer of reactants (oxygen and hydrogen) and products (water) to and from the CL, 2) to ensure structural stiffness to the MEA, and 3) to provide an electrically and thermally conductive pathway between the CL and the current collectors. The porous CL (5-25 µm) is comprised of a platinum catalyst supported on electrically and thermally conductive carbon particles, and proton conducting ionomer. The CCL typically contains 20-40 wt% ionomer, which has three roles:  1) to act as a binder  between the platinum/carbon particles, 2) to provide a proton conductive link to the membrane for protonic current flow, and 3) to make the platinum catalyst electrochemically active by transferring protons to and from the catalyst. Unfortunately, the structure of the CL is difficult to control during manufacture; therefore, it is difficult to control the factors that affect its performance such as gas mass transfer, and proton conductivity. The current method of manufacturing the CL mixes the supported catalyst powder with ionomer solution by various mixing techniques, such as ball milling, sonication,  29  and impeller mixing. The resulting catalyst ink is then applied either onto the GDL, to produce a GDE, or onto the membrane, to produce a catalyst coated membrane (CCM). The goal is to strengthen the interfaces between layers to increase performance, while using the most cost effective manufacturing process. While most MEA developers are currently developing CCM technology, this study used a GDE structure. Early PEMFC research used platinum black as the catalyst at the cathode and anode electrodes. These electrodes had very high platinum loadings (>>1.0 mg cm-2), hence were very costly. One of the ways of reducing the platinum loading was switching from platinum black to platinum supported on carbon. The smaller platinum particles on carbon enabled a reduced loading to less than 0.5 mg cm-2, while maintaining the required platinum surface area. Although cost was reduced the durability of the fuel cell was negatively impacted. At potentials greater than 0.2 VRHE (reversible hydrogen electrode), the carbon support is thermodynamically able to oxidize to carbon dioxide (CO2) and/or carbon monoxide (CO) Eq. 2.1 and 2.21, leaving the platinum unsupported and inactive. Furthermore, due to the loss of support the platinum particles have been shown to agglomerate into larger particles, dissolve into the ionomer, or get washed out of the system2,3. C(s) + 2H2O  CO2 (g) + 4H+ + 4e-  Eo = 0.207 VRHE  25°C [2.1]  C(s) + H2O  CO (g) + 2H+ + 2e-  Eo = 0.518 VRHE  25°C [2.2]  30  Even in the presence of platinum, the kinetics of carbon oxidation/corrosion is slow; therefore, carbon is quite stable under normal PEMFC operating conditions. In practice, elevated cathode potentials of greater than 1.2 VRHE are required to oxidize carbon at reaction rates high enough to cause significant structural degradation. Normal PEMFC operation occurs between 0-1.0 VRHE; however, upon fuel starvation or gas switching conditions (start-up & shutdown protocols) the cathode potential can exceed 1.2 VRHE as shown in Fig. 2.1.4  Figure 2.1 – Carbon oxidation degradation mechanism.  When hydrogen in the anode compartment is replaced by air, or vice versa, a temporary potential gradient is formed along the anode electrode. Hydrogen oxidation occurs in the hydrogen rich portion of the electrode, and proton and/or  31  oxygen reduction occurs in the transient portion of the cell. The voltage across the MEA in the hydrogen rich region of the anode drives current in the air transient portion of the cell, elevating the half-cell potential on the cathode to values greater than 1.2 VRHE.  At these potentials both water and carbon  oxidation takes place as shown in Eq. 2.1 to 2.3. H2O  ½ O2 + 2H+ + 2e-  Eo = 1.230 VRHE  25°C  [2.3]  Meyers et al.5 modelled this process and suggested possible mitigation strategies to reduce the amount of carbon oxidation, including a nitrogen purge step before introducing a gas change at the anode electrode, or controlling the cathode potential by shorting the cell. They calculated a reduction in the carbon oxidation through both methods, with the shorting mitigation having the greatest impact. This phenomenon is similar to connecting a power supply to a fuel cell with air at both electrodes. The power supply voltage drives current through the fuel cell, causing carbon and water oxidation at the original fuel cell cathode which now becomes in effect the anode and proton and oxygen reduction at the original anode which now in effect becomes the cathode. Water electrolysis technology operates by the same principle, only the system is designed to maximize water oxidation and eliminate carbon oxidation.  The AST in this study used this  technique by connecting a potentiostat to the fuel cell with nitrogen at the cathode electrode and hydrogen at the anode electrode. Both carbon and water oxidation reactions were forced by applying the voltage across the MEA to 1.5 VRHE, where the anode was treated as both counter and reference electrode.  32  The carbon lost at electrode potentials of greater than 1.2 VRHE reduces the effective platinum surface area and kinetic performance, and also weakens the CL structure. Once a significant amount of carbon is lost as CO2 and/or CO, the CL structure collapses and becomes thinner leaving only the unsupported platinum and Nafion ionomer. Using EIS, CV, and polarization analysis, the CL structure and structural degradation caused by carbon corrosion were investigated and the sensitivity of the EIS conductivity measurement to gas RH and MEA structure was evaluated. This study provides the basis for using these techniques to further characterize structural damage to the CL.  2.2  Experimental  Stevens et al.6, conducted high temperature thermal stability tests on carbon supports for platinum catalysts under both dry and humidified gas conditions.  They showed an increased rate of carbon oxidation under a  humidified air atmosphere.  Therefore, the cathode corrosion AST was  completed in an over-saturated nitrogen (cathode) and hydrogen (anode) atmosphere. The over-saturated atmosphere is designated by 120% RH, which represents 100% gas RH with additional liquid water. The voltage was cycled stepwise from the open circuit voltage of approximately 0.15 VRHE, to 1.5 VRHE for 30 and 150 s, respectively. The corrosion voltage of 1.5 VRHE had a greater duration to accelerate the corrosion degradation. The duration at the recovery voltage (0.15 VRHE) was reduced to minimize the test time. It was assumed that most carbon was oxidized to CO2 as opposed to CO; therefore, the exhaust  33  gases were monitored for CO2 to calculate the amount of carbon lost. Since too much flow diluted the gas stream below the limits of the Fuji Electric ZRH Infrared CO2 gas analyzer, the gas flow rates were reduced from 25 slpm to 0.3 slpm during the AST. During the AST period, CV and EIS were done every hour or 20 cycles. The full diagnostic test protocol, which included CV, EIS, air and oxygen polarizations, was completed at 0, 5, 10, 20, and 30 h. SEM analysis was conducted with new and degraded MEAs. The assumptions made in the voltage loss breakdown and limiting current diffusivity diagnostics are invalid after severe structural degradation occurs; therefore, these diagnostics were not analyzed for the carbon corrosion AST; however, the polarization curves are presented and discussed. The structural degradation due to carbon corrosion changed the water management characteristics; therefore, polarization curves were measured under both 100 and 60 % gas RH.  2.3  Results and Discussion 2.3.1 Carbon loss  The carbon dioxide in the exhaust gases was measured and the amount of carbon lost due to corrosion was calculated as shown in Fig. 2.2. The rate of carbon loss was not dependent on the catalyst support loading or ionomer loading, suggesting a secondary carbon source. The CLs contained between 2-3 mg of carbon depending on the specific MEA tested. More carbon oxidized (> 4.5 mg) than catalyst support, confirming a secondary carbon source susceptible  34  to corrosion. This carbon source was the carbon sub-layer between the CL and the GDL as shown in the SEM analysis. 5.0 4.5 4.0  Carbon Loss (mg)  3.5 3.0 2.5 2.0  MEA#1 - 23wt% Nafion  1.5  MEA#2 - 33wt% Nafion  1.0 0.5  C + 2 H 2O → CO2 + 4 H + + 4e −  0.0 0  10 20 30 Corrosion Cycle Time (hr)  40  Figure 2.2 – Cumulative carbon loss versus corrosion cycle time.  2.3.2 SEM analysis Fig. 2.3 shows the SEM pictures of a new MEA (left image) compared to an MEA that was corroded for 30 cycle h (right image). After 30 h of corrosion cycling the CCL was a third of the thickness compared to the new MEA and was much brighter suggesting a higher density of platinum. The lighter grey patches show the Teflon®, which roughly mark the edges of the GDL and the interface with the carbon sub-layer. The new MEA shows the carbon sub-layer between the GDL and the CL, whereas after 30 h of corrosion cycling the sub-layer was no longer present. This suggests that the carbon sub-layer was oxidized over  35  the degradation period.  There was no significant change in the anode or  membrane thickness over the corrosion period.  Figure 2.3 – SEM picture at 0 and 30 hours of corrosion cycling.  2.3.3 Cyclic voltammetry CV was conducted as part of the diagnostic protocol at 0, 5, 10, 20, and 30 h of corrosion cycling time, as shown in Fig. 2.4. Cdl increased within the first 10 h of corrosion followed by a decrease over the last 20 h. The initial increase in capacitance is likely due an increased concentration of carbon corrosion products with oxygen functionalities at the surface of the carbon support, which are hydrophilic and offer a greater Cdl7-9. The competing effect of the loss of carbon and platinum surface area due to the loss of the carbon support decreased the Cdl.  At 8 h these opposing effects became balanced and a  maximum capacitance occurred.  Loss of carbon became the dominant  36  mechanism after 8 h of corrosion cycling and Cdl decreased over the remaining corrosion period.  0.05 0.04 0.03 Current (A)  0.02 0.01 0.00 -0.01  0h  -0.02  5h 10 h  -0.03  30 h 20 h  -0.04 -0.05 0.0  0.2  0.4 0.6 0.8 Voltage (V)  1.0  1.2  Figure 2.4 – CVs over the corrosion period.  This effect is shown clearly in Fig. 2.5 and was also demonstrated by EIS. Evidently there was an offset between the CV and EIS measurements, suggesting an inaccuracy in either measurement. Since the heterogeneous and non-ideal EIS results were modelled to a homogeneous ideal structure this offset was not unexpected. The data from the CV response is believed to be more accurate; however, since the capacitance values obtained from EIS were within the same order of magnitude and sensitive to the effects of degradation, the EIS results were validated.  37  0.50 Cdl - CV - 23wt% Nafion Cdl - CV - 33wt% Nafion Cdl - EIS - 23wt% Nafion Cdl - EIS - 33wt% Nafion  0.45 0.40  Capacitance (F)  0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0  5  10  15  20  25  30  35  40  Corrosion Cycle Time (hr)  Figure 2.5 – Double layer capacitance over the corrosion period.  The CV’s in Fig. 2.4 show decreased H2 adsorption/desorption and Pt oxide peaks over the degradation period, signifying a reduction in the platinum surface area. The first observable H2 adsorption/desorption peak decreased more than the second; suggesting a preferential degradation of the Pt (110) crystal face compared to Pt (100) and Pt (111). Grgur et al.10, showed the Pt (110) crystal face to be the most kinetically active with the greatest exchange current density, and therefore, to have the greatest impact on the kinetic performance. Carbon monoxide stripping quantified this reduction in the EPSA as shown in Fig. 2.6.  38  225  23wt% Nafion  200  33wt% Nafion  175 150 125 100  2  2  EPSA (cm Pt / cm geometric area)  250  75 50 25 0 0  5  10  15  20  25  30  35  40  Corrosion Cycle Time (hr)  Figure 2.6 – EPSA over the corrosion period.  The EPSA decreased by 56% over the corrosion period for both the 23 and 33 wt% ionomer CLs. Debe et al,11 found similar results when cycling between 0.6-1.2 VRHE under a N2 cathode/H2 anode conditions. At 70 °C they calculated a reaction rate constant of 5.5*10-4 cycle-1, using a first order kinetic rate model described in Eq. 2.4 and 2.5.  dS dncyc = −k * S  [2.4]  ln[( S − S min ) (1 − S min )] = −k * ncyc  [2.5]  where S = normalized EPSA, Smin = minimum normalized EPSA, k = rate constant, and ncyc = number of cycles.  39  By plotting the left hand term in Eq. 2.5 against the number of cycles and taking the slope, the reaction rate constant was calculated for the data presented in Fig. 2.6. The reaction rate constant, 5.3 - 6.4*10-3 cycle-1, was higher than Debe et al.’s results, which was expected due to the greater voltage limits of 1.5 VRHE used for this AST. The decrease in platinum surface area was due to varying degrees of platinum agglomeration, washout, dissolution into the membrane, and possible catalyst islanding.  Catalyst islanding is defined as a lack of ionic and/or  electronic connectivity to the electrode, rendering the catalyst inactive. Significant platinum migration into the membrane has been shown in SEM pictures5 as a bright white band in the membrane, which was not evident in this study. Some studies have shown the location of this platinum band to be a function of the hydrogen and oxygen gas crossover and therefore, can be manipulated by the gas partial pressures12,13. The absence of platinum in the membrane in this study suggests the AST conditions were not conducive to significant platinum migration into the membrane. Considering the catalyst was rarely exposed to voltages where Pt becomes thermodynamically unstable14, this result was expected. The reduction of the CL thickness shows that oxidation of the carbon support caused the CL to collapse into a much denser network of platinum and ionomer with limited porosity (see Fig. 2.3).  Therefore, the  dominant mechanisms for the decreased EPSA are likely platinum agglomeration and possibly catalyst islanding. Debe et al. showed supporting X-ray diffraction (XRD) data, which resulted in similar platinum peak intensities before and after a  40  90% EPSA loss. This signified very little change in the amount of platinum in the electrode.  2.3.4 Electrochemical impedance spectroscopy It is well known that polymer electrolyte membranes require water to achieve their high proton conductivity, which is directly related to the inlet gas RH15. Fig. 2.7 shows experimental EIS spectra measured as a function of the nitrogen and hydrogen gas RH. The shapes of the curves were very similar; only the HF intercept to the real axis (HF cell resistance) increased, signifying increased membrane resistance.  Also, the approximate 45° linear regions  extended with lower RH, indicating increased CL ionomer resistance.  -1.0  100%RH 90%RH 80%RH 70%RH 60%RH 50%RH  -0.6  -0.4  -0.10  Z" - Reactance (Ohm)  Z" - Reactance (Ohm)  -0.8  -0.2  -0.05  0.00  0.0 0.05 0.00 0.2 0.0  0.2  0.4  0.6  0.8  1.0  0.05 0.10 Z' - Resistance (Ohm)  0.15  1.2  Z' - Resistance (Ohm)  Figure 2.7 – EIS spectra as a function of gas RH.  41  The data from Fig. 2.7 was modelled with an equivalent circuit consisting of an inductor, resistors, and Warburg element described in Fig. 1.4. The difference between the modelled fit in Fig. 1.4 and the data in Fig. 2.7 was noted. The experimental data showed a slight slope in the high impedance region signifying a non-ideal fit to the given equivalent circuit. The equivalent circuit described in Fig. 1.4 assumes a network of cylindrical pores with homogeneous diameter, as defined by the cotanh function originally described by De Levie. However, the CL is a heterogeneous porous network with varying pore shapes and sizes. Some research has shown how different pore sizes and pore shapes can alter the shape of the impedance response16-19. The geometry can change the ratio between the Cdl and the CL proton resistance, and therefore, change the shape of the Nyquist plot.  This can explain deviations that occur between the  experimental data and the model fit. Several published models20-22 have used Bruggeman’s relationship to describe the effective proton conductivity in the porous CL, as defined by Eq. 2.6. σ CL = σ BULK * ετ  [2.6]  where σ = ionic conductivity (S cm-1) , ε = ionomer volume fraction, τ = tortuosity factor = 1.5 Based on this relationship decreasing the ionomer volume fraction decreases the CL ionomer conductivity.  This is demonstrated in Fig. 2.8  comparing the CL ionomer conductivity between the 23 and 33 wt% ionomer MEAs.  42  CL Ionomer Conductivity (S/cm)  0.016 23wt%Nafion - Exp 33wt%Nafion - Exp 23wt%Nafion - Model 33wt%Nafion - Model  0.014 0.012 0.010 0.008 0.006 0.004 0.002 0.000 20  40  60  80  100  120  140  RH  Figure 2.8 – CCL ionomers conductivity for 23 and 33 wt% Nafion content.  The ionomer volume fraction was calculated based on the known amounts and densities of carbon (2.1 g cm-3), platinum (21.5 g cm-3), and ionomer (1.98 g cm-3) in the CL and the CL thickness measured from the SEM analysis. The ionomer volume fractions for the 23 and 33 wt% CLs were 0.10 and 0.17, respectively. The experimental values follow Bruggeman’s relationship very well using a tortuosity factor equal to 1.5, except in the over-saturated gas humidity (water vapour and liquid) condition where the Bruggeman model deviates from the experimental data.  In over-saturated conditions several factors may  contribute to the deviations shown in Fig. 2.8. Ionomer swelling due to increased water uptake can change the CL ionomer volume fraction, tortuosity, and may change the pore shape and size distribution sufficiently to introduce greater deviation from the ideal equivalent circuit data fit. Liquid water residing in the pore space may also contribute to the proton conductivity in the CL.  43  Makharia et al23 used similar EIS methods under current operation to measure the resistances in the cell. The HF cell resistance (0.06 – 0.07 Ω cm-2) was similar; however, since Makharia used different ionomer contents the CL ionomer resistance could not be directly compared. Makharia had 7 and 15 vol% ionomer resulting in 0.24 and 0.10 Ω cm-2, respectively. This study had 10 and 17 vol% ionomer resulting in 0.26 and 0.15 Ω cm-2, respectively. Given the greater ionomer contents in this study, these values are higher than Makharia, which could be caused by the different CL structure (GDE versus CCM), the lower temperature (70 versus 80°C), or the test method (without and with current operation). The HF cell resistance and CL ionomer resistance were measured by EIS every hour (20 cycles) during the AST and every 10 h as part of the full diagnostic testing (see Fig. 2.9). Under the AST conditions both the 23 and 33 wt% ionomer MEAs exhibited the same trends, where the catalyst ionomer resistance and HF cell resistance increased. This increase was likely due to loss of water from carbon and water oxidation (Eq. 2.1 – 2.3). After the SS diagnostic testing the MEA re-hydrated and the resistance decreased, as shown by the discontinuities at 10 and 20 h in Fig. 2.9. The rate of resistance increase during the AST increased with greater degradation. This suggests either 1) an increase in the reaction rate of carbon and/or water oxidation or 2) less water in the catalyst and membrane layers.  44  0.120 2  HF Cell Resistance (Ohm.cm )  2  CL Ionomer Resistance (Ohm.cm )  5.0  0.100 0.080 0.060 0.040 0.020  4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0  0.000 0  10  20  30  40  0  Corrosion Cycle Time (hrs) 23wt% Nafion - AST  33wt% Nafion - AST  23wt% Nafion - SS  33wt% Nafion - SS  10  20  30  40  Corrosion Cycle Time (hr)  Figure 2.9 – Comparison of EIS resistance performed during the AST.  After 20 h of the corrosion-AST the HF cell resistance measured during the AST, started to decrease. The reason for this was not verified; however, it is speculated that this decrease was caused by the compaction of the MEA due to the degradation of the CCL and carbon sub-layers. Fig. 2.10 shows the impedance spectra at 0 and 30 h of corrosion cycling. It is noted that unlike Fig. 2.7, the shapes of the curves are different. This may signify a different pore shape and/or size distribution as discussed previously.  45  2  Z" - Reactance (Ohm.cm )  -1.2 -1.0  0hrs  -0.8  30hrs  -0.6 -0.4 -0.2 0.0 0.2 0.00  0.05  0.10  0.15  0.20  0.25  0.30  Z' - Resistance (Ohm.cm2)  Figure 2.10 – Nyquist plot after 0 and 30 h of corrosion cycling.  The HF cell resistance measured under SS conditions increased over the degradation period and was most evident at sub-saturated gas RH as shown in Fig. 2.11. This occurred with both 23 and 33 wt% ionomer catalyst structures. The HF cell resistance can increase by the following mechanisms: 1) loss of carbon decreasing electronic connectivity between carbon particles, making the CL more electrically tortuous 2) decreasing ionic connectivity between catalyst particles and membrane, 3) increased layer interface resistance due to layer separation/delamination caused by carbon loss, and 4) contaminants released from the carbon support oxidation could bind to the membrane sulphonic acid sites via a cation exchange process. This last mechanism would decrease the ion exchange capacity and lower the water content in the membrane24. The hardware compression sensitivity showed the cell performance became more sensitive to compression over the corrosion period. SEM pictures showed the 46  disappearance of the carbon sub-layer over the 30 h corrosion period. Both results support mechanism 3). Cathode - 23wt% Nafion  Cathode - 33wt% Nafion 0.30 0 hrs 5 hrs 10 hrs 20 hrs 30 hrs  0.20  2  0.25  0 hrs  HF Cell Resistance (Ohm.cm )  2  HF Cell Resistance (Ohm.cm )  0.30  0.15 0.10 0.05 0.00  0.25  10 hrs 20 hrs  0.20  30 hrs 0.15 0.10 0.05 0.00  20  40  60  80  100  120  140  Relative Humidity %  20  40  60  80  100  120  Relative Humidity %  Figure 2.11 – HF cell resistance versus RH over the corrosion period.  Under SS conditions Fig. 2.12 (left images) shows both the 23 and 33 wt% ionomer CL’s decrease in ionomer resistance over the AST period. The effect was greater in the 23 wt% ionomer CL since this MEA started out with much higher resistance due to its lower ionomer volume fraction. However, both MEAs reached very similar proton resistance after 30 h of corrosion cycling. As the carbon support corroded, the CL lost its structural integrity and collapsed. This reduced the porosity of the CL and created a very thin compact layer of ionomer and platinum. The ionomer volume fraction increased and resistance decreased, changing the CL ionomer conductivity.  47  140  Cathode - 23wt% Nafion  Cathode - 23%Nafion  0.014 0h 5h 10 h 20 h 30 h  7.0 6.0  -1  8.0  CL Ionomer Conductivity (S cm )  2  CL Ionomer Resistance (ohm cm )  9.0  5.0 4.0 3.0 2.0 1.0  20  40  60  80  100  120  0.008 0.006 0.004 0.002  20  140  40  60  80  100  120  RH  RH  Cathode - 33wt% Nafion  Cathode - 33wt% Nafion  140  0.014  CL Ionomer Conductivity (S cm )  3.5 3.0  -1  2  0.010  0.000  0.0  CL Ionomer Resistance (ohm cm )  0h 30 h  0.012  0h 10 h  2.5  20 h 2.0  30 h  1.5 1.0 0.5 0.0  0.012  0h  0.010  30 h  0.008 0.006 0.004 0.002 0.000  20  40  60  80 RH  100  120  140  20  40  60  80  100  120  140  RH  Figure 2.12 – Changes in CL resistance and CL conductivity over the corrosion period as a function of gas RH.  Fig. 2.12 (right images) shows the conductivity of the 23 and 33 wt% ionomer CLs at 0 and 30 h of corrosion cycling. At sub-saturated humidification the conductivity increased over the corrosion period; however, at saturated humidification the conductivity decreased. The increased conductivity under sub-  48  saturated conditions can be explained by the compaction of the CL and increased ionomer volume fraction.  The decreased conductivity at saturated  conditions suggests a significant change in the water management. Assuming the equivalent circuit and model fitting was valid, it was hypothesized that the decreased conductivity was a result of the inability of liquid water to penetrate and humidify the CL due to the collapsed platinum/ionomer CL structure. Corrosion of the carbon in the sub-layer would leave a porous Teflon® network.  Similar to GoreTex® material, the degraded structure blocks liquid  water but is porous enough to allow water vapour to diffuse and humidify the layer.  GoreTex® is made of a hydrophobic porous Teflon® material25, which  repels liquid water but allows smaller water vapour molecules to pass through and remain breathable.  This hypothesis is supported by the polarization  analysis.  2.3.5 Polarization analysis Air and oxygen polarization curves were completed with saturated (100%) gas RH, at periodic intervals over the corrosion-AST for both 23 and 33 wt% ionomer structures. Despite the increased CL ionomer conductivity in the 33 wt% ionomer CL, both CL structures demonstrated similar performance and performance degradation.  Fig. 2.13 shows the oxygen and air polarization  curves for the 23 wt% CL after 0, 10, and 30 h of corrosion.  49  1.0 a) 0 hours  0.9 Cell voltage (V)  0.8 0.7 0.6 0.5 0.4 0.3 0.2 Oxygen - 100%RH Oxygen - 60%RH  0.1 0.0 0.0  0.5  1.0  1.5  Air - 100%RH Air - 60%RH 2.0  2.5  3.0  3.5  2  Current density (A/cm ) 1.0 0.9  b) 10 hours  Cell voltage (V)  0.8 0.7 0.6 0.5 0.4 0.3 0.2 Oxygen - 100%RH Oxygen - 60%RH  0.1 0.0 0.0  0.5  1.0  1.5  Air - 100%RH Air - 60%RH 2.0  2.5  3.0  3.5  2  Current density (A/cm ) 1.0 0.9  c) 30 hours  Cell voltage (V)  0.8 0.7 0.6 0.5 0.4 0.3 0.2  Oxygen - 100%RH Oxygen - 60%RH  0.1 0.0 0.0  0.5  1.0  1.5  Air - 100%RH Air - 60%RH 2.0  2.5  3.0  3.5  2  Current density (A/cm )  Figure 2.13 – Oxygen and air polarization curves under 100% and 60% gas RH for the 23 wt% MEA.  50  An operating ohmic resistance of 0.12 Ω cm-2 was obtained from the saturated oxygen voltage performance. Without any mass transport limitations the operating ohmic resistance is expected to be the summation of the membrane resistance, cell electrical resistance, and 1/3 of the total CL ionomer resistance13. This was verified by summing the EIS HF cell resistance (0.63 Ω cm-2) and 1/3 of the CL ionomer resistance (0.73 Ω cm-2) for a total of 0.146 Ω cm-2. Although not shown, after 5 h of corrosion the ohmic performance improved with an approximate 0.008 Ω cm-2 decrease in the total ohmic resistance, coinciding with the initial drop in the CL ionomer resistance shown in Fig. 2.12. On the other hand, mass transport losses increased drastically at higher current densities (>2.0 A cm-2).  As the corrosion cycling continued the performance  degraded significantly. There was an obvious kinetic contribution to the total voltage loss due to the decreased platinum surface area as previously shown in Fig. 2.6.  The increased HF cell resistance and decreased CL ionomer  conductivity shown in Fig. 2.11 and 2.12, respectively, supports the decrease in the MEA’s ohmic performance. To further understand the performance loss, air and oxygen polarization curves were compared under sub-saturated (60 %RH) and saturated (100 %RH) gas conditions.  At 0 h of corrosion the sub-saturated voltage performance  curves (Fig. 2.13 a) were all lower than the saturated performance, as expected, due to the lower membrane and CL conductivity. In the sub-saturated curves the ohmic performance improved with greater current density, as the water produced  51  humidified the cell and increased the membrane and catalyst ionomer conductivity. At both 10 (Fig. 2.13 b) and 30 h (Fig. 2.13 c) of corrosion cycling the sub-saturated voltage performance was greater than the saturated performance over most of the current density range. The current hypothesis suggests that in saturated conditions liquid water at the GDL/CL interface was unable to penetrate into the CL. Liquid water at this interface would impede oxygen diffusion to the catalyst, causing an additional mass transport and CL ionomer ohmic loss.  With poor oxygen diffusion in the CL, a shift in the  electronic and protonic current distribution through the CL would occur. The protonic current or proton penetration into the CL from the membrane would shift toward the GDL, effectively increasing the CL ionomer ohmic loss. This hypothesis was demonstrated further by comparing the 100 and 60 %RH oxygen curves after 10 h of the corrosion-AST. The dry oxygen curve appeared to have a lower operating ohmic resistance at current densities less than 1.5 A cm-2 correlating to the sub-saturated zone in the conductivity profile in Fig. 2.12.  Under this condition both oxygen and water vapour were able to  diffuse into and out of the CL. At greater than 1.5 A cm-2 the ohmic resistance increased and appeared to match the saturated feed gas ohmic resistance, where oxygen may have been blocked by liquid product water.  2.4  Conclusions  CCLs containing 23 and 33 wt% Nafion ionomer were subjected to a corrosion-AST to determine how the CL structure was altered and how these 52  changes affected performance.  Both CL structures demonstrated similar  performance and performance degradation. Monitoring the carbon dioxide in the gas exhaust revealed the rate of carbon loss was independent of catalyst or ionomer content.  More carbon left the cell than there was catalyst support,  supporting the SEM analysis, which showed both carbon catalyst support and carbon sub-layer thinning. The CCL thinned from approximately 15 to 5 µm after 30 h of corrosion cycling. No change was observed in the membrane or anode CL thickness. CV revealed a 55% decrease in EPSA over the 30 h corrosion period mostly due to platinum agglomeration as the result of carbon loss. The rate of EPSA loss followed a first order kinetic model and compared to literature results. The Cdl increased 200% over the first 8 h and then slowly decreased over the remaining 22 h of corrosion cycling. Previous studies have shown the initial increase to be caused by production of oxygen containing carbon surface groups.  Eventually this increase reached a maximum where the decreased  surface area of the catalyst started to dominate, at which point the capacitance started to decrease. EIS was used to measure the HF cell resistance and the CCL ionomer resistance as an indicator of the structural and performance degradation. In a new MEA the CL ionomer resistance was shown to obey Bruggeman’s relationship with a tortuosity factor of close to 1.5. Over the 30 h corrosion period the HF cell resistance increased between 15 and 30% in saturated conditions and greater than 100% in sub-saturated conditions. This resistance increase  53  also affected the hardware compression sensitivity where increased interlayer contact resistance was hypothesized. Over the 30 h corrosion period the CL ionomer conductivity increased in sub-saturated conditions and decreased in saturated conditions.  The collapse of the CL due to carbon oxidation was  hypothesized to block liquid water from the CL but allow water vapour to diffuse into its structure. Air and oxygen polarization curves were conducted and showed similar performance and performance degradation between the two CL structures tested. After 10 h of corrosion cycling the MEA performed better in sub-saturated conditions compared to saturated conditions.  It was hypothesized that liquid  water impeded oxygen diffusion and caused a shift in the current distribution or proton penetration through the CL, resulting in greater ohmic losses. The above methods were successful in explaining and quantifying the structural degradation effects of carbon oxidation on the performance of the PEMFC.  54  2.5 1.  References  K. Kinoshita, Carbon: Electrochemical and Physiochemical Properties, p.316, John Wiley & Sons, New York (1998).  2.  W. Bi, G.E. Gray, and T.F. Fuller, Electrochemical and Solid-State Letters, 10, B101 (2007).  3.  X. Yu, and S. Ye, J. Power Sources, 172, 145 (2007).  4.  M.V. Lauritzen, P. He, A.P. Young, S. Knights, V. Colbow, and P. Beattie, J. New Mat. Electrochem. Systems 10 (2007).  5.  J.P. Meyers, and R.M. Darling, J. Electrochem. Soc., 153, A1432 (2006).  6.  D.A. Stevens, M.T. Hicks, G.M. Haugen, and J.R. Dahn, J. Electrochem. Soc., 152, A2309 (2005).  7.  K.H. Kangasniemi, D.A. Condit, and T.D. Jarvi, J. Electrochem. Soc., 151, E125 (2004).  8.  M.J. Bleda-Martinez, J.A. Macia-Agullo, D. Lozano-Castello, E. Morallon, D. Cazorla-Amoros, and A. Linares-Solano, Carbon, 43, 2677 (2005).  9.  C.H. Paik, G.S. Saloka, and G.W. Graham, Electrochemical and Solid-State Letters, 10, B39 (2007).  10. B.N. Grgur, N.M. Markovic, and P.N. Ross, Can. J. Chem., 75, 1465 (1997). 11. M.K. Debe, A.K. Schmoeckel, G.D. Vernstrom, and R. Atanasoski, J. Power Sources, 161, 1002 (2006).  55  12. E. Guilminot, A. Corcella, F. Charlot, F. Maillard, and M. Chatenet, J. Electrochem. Soc., 154, B96 (2007). 13. J. Zhang, B.A. Litteer, W. Gu, H. Liu, and H.A. Gasteiger, J. Electrochem. Soc., 154, B1006 (2007). 14. R. Borup et al., Chem. Rev., 107, 3904 (2007). 15. X. Zhou, Z. Chen, F. Delgado, D. Brenner, R. Srivastava, J. Electrochem. Soc., 154 (1) B82 (2007). 16. H. Keiser, K. Beccu, and M. Gutjahr, Electrochimica Acta, 21, (1976). 17. H. Song, J. Sung, Y. Jung, K. Lee, L. Dao, M. Kim, and H. Kim, J. Electrochem. Soc., 151, (2004). 18. K. Eloot, F. Debuyck, M. Moors, and A.P. Van Peteghem, J. Applied Electrochemistry, 25, 326 (1995). 19. K. Eloot, F. Debuyck, M. Moors, and A.P. Van Peteghem, J. Applied Electrochemistry, 25, 334 (1995). 20. D. Bernardi and M. Verbrugge, J. Electrochem. Soc., 139, 2477 (1992). 21. F. Gloaguen and R. Durand, J. Applied Electrochemistry, 27, 1029 (1997). 22. Havranek, and K. Wippermann, J. Electroanalytical Chem., 567, 305 (2004). 23. R. Makharia, M.F. Mathias, and D.R. Baker, J. Electrochem. Soc., 152, A970 (2005). 24. S. Nagasaka, M. Harada, and T. Okada, Analytical Chimica Acta, 525, 115 (2004).  56  25. R.W. Gore, and S.B. Allen Jr., U.S. Pat., 4,194,041 (1980).  57  3.0 Ionomer Degradation in Polymer Electrolyte Membrane Fuel Cells  A version of this chapter has been published: A.P. Young, J. Stumper, S. Knights, and E. Gyenge, “Ionomer Degradation in Polymer Electrolyte Membrane Fuel Cells”, J. Electrochem. Soc., 157, (3) (2010)  58  3.1  Introduction  Several mechanisms can affect the performance of the PEMFC, such as the hydrogen and oxygen mass transfer to the reaction sites, proton conductivity through the Nafion ionomer, hydrogen and oxygen reaction rates, and water management in the cell.  All of these mechanisms exist within the electrode  structure, which consists of a GDL and a CL. The GDL (100-300 µm) is often made of a carbon fibre mat with an additional layer consisting of carbon particles and a hydrophobic binder (carbon sub layer). The GDL serves three purposes: i) to provide a pathway for mass transfer of reactants (oxygen and hydrogen) and products (water) to and from the CL, ii) to ensure structural stiffness to the MEA, and iii) to provide an electrically and thermally conductive pathway between the CL and the current collectors. The porous CL (5-25 µm) is comprised typically of carbon supported platinum catalyst interacting with the proton conducting ionomer. The CCL typically contains 20-40 wt% ionomer, which has three roles: i) to act as a binder between the platinum/carbon particles, ii) to provide a proton conductive link between the membrane and Pt catalyst sites for protonic current flow, and iii) to make the platinum catalyst electrochemically active by transferring protons to and from the catalyst. The most widely used proton conducting ionomer in the PEMFC industry is Nafion made by DuPont.  The Nafion chemical structure is comprised of a  fluoropolymer backbone with sulphonic acid site side chains. The fluoropolymer backbone provides the structure to the ionomer and is hydrophobic.  The  59  sulphonic acid side chains form clusters that are hydrophilic and provide the proton conductivity functionality. The mechanism for proton conduction through bulk ionomer has three potential pathways commonly described in literature1. i) Grotthus Transport and bulk ionic conductivity: Proton mobility is dependent on hydronium ions being solvated with water. The Grotthus transport mechanism occurs when protons are passed from one water molecule to the next. Studies and models suggest the Grotthus mechanism dominates proton mobility through Nafion ionomer2. ii) Surface Hopping and surface ionic conductivity: In Nafion, the sulfonic acid groups are considered too far apart for direct proton hopping; therefore protons ‘hop’ between acid groups and water molecules along the surface of the pore. This mechanism is not as fast as the bulk Grotthus transport due to the electrostatic attraction and binding of protons with the negatively charged sulfonate groups (i.e. undissociated acid). iii) Vehicular Transport: Refers to the convective transport of protons by the water flux in the membrane. It is typically a minor effect compared to Grotthus transport. All mechanisms described require water to make the ionomer conductive and the relative contribution of the various mechanisms depends, to some extent, on the physico-chemical properties of the membrane. Assuming the proton transport mechanisms are the same in the CL ionomer compared to the bulk ionomer in membranes, the effective proton conductivity of the CL is expected to follow Bruggeman’s relationship as shown in Eq. 3.1 σ CL = σ BULK  *  ετ  [3.1]  60  where σ is the conductivity (S cm-1) , ε is the ionomer volume fraction, and τ is the tortuosity factor, which varies from 1-2 depending on the CL structure3-5. There have been different attempts at measuring the effective conductivity of the CL.  Boyer et al.3 sandwiched a CL between two membranes and  measured the total ionic resistance through the sandwich. By subtracting the known resistance of the two membranes, the ionic resistance/conductivity of the CL was measured. In their work they found that the tortuosity factor equalled unity. Extensive work has been done using EIS to measure the CL ionic resistance based on De Levie transmission line equivalent circuits6-9. The work presented in chapter two verified this method against Bruggeman’s relationship and used the measurement to help characterize CL structural degradation due to carbon corrosion10. This method was used in this study to measure ionomer degradation in the CCL. Ionomer degradation occurs due to in-situ formation of small amounts of hydrogen peroxide (H2O2) in the PEMFC.  It is generally accepted that the  degradation mechanism involves the decomposition of H2O2, which produces radicals (OH•, OOH•) that attack the chemical bonds in Nafion membrane. Hydrogen peroxide can be produced either by oxidation of water or by the reduction of oxygen, as shown in the Pourbaix diagram in Fig. 3.111. During normal operation of a PEMFC the oxygen reduction reaction can proceed by a direct four-electron reaction or a two-step peroxide pathway as shown in Eq. 3.2 and 3.3 a,b, respectively12.  61  Pt  O2 + 4e − + 4 H + → 2 H 2 O Pt / C  O 2 + 2e − + 2 H + ⇔ H 2 O 2 Pt / C  H 2 O2 + 2e − + 2 H + ⇔ 2 H 2 O  Eo = 1.230 VRHE  25°C  [3.2]  Eo = 0.672 VRHE  25°C  [3.3a]  Eo = 1.770 VRHE  25°C  [3.3b]  While the direct four-electron reaction occurs only on platinum, the two-step pathway occurs on both carbon13 and platinum depending on the platinum crystal face14-16. Rotating ring-disc electrode studies have shown H2O2 formation on Pt (111) and Pt (100), but much less on Pt (110). Liu et al.17 have verified the presence of H2O2 in the fuel cell in a CV study. This was accomplished by comparing the CV of platinum in known concentrations of H2O2 solutions to the CV of platinum electrodes placed in an operating fuel cell. Teranishi et al.18 measured H2O2 using mass spectroscopy of the cathode gas exhaust in an operating fuel cell. In the PEMFC H2O2 formation occurs primarily on the anode where the operating potential (e.g. 0.05 to 0.2 VRHE) is favourable for creating a high negative overpotential for oxygen reduction to peroxide on either Pt or the exposed carbon support surface sites. Therefore, the H2O2 formation is heavily dependent on the rate of oxygen crossover through the membrane. The gas crossover depends on the membrane material characteristics, such as chemical composition, thickness, and hydration, as well as the anode and cathode gas partial pressures. All of these factors interact in a complex manner to alter the rate of H2O2 production.  62  Understanding the mechanism for H2O2 decomposition is equally, if not more important, than H2O2 production. The Pourbaix diagram in Fig. 3.1 shows that in an acidic environment, such as Nafion, H2O2 will electrochemically decompose between 0.67 and 1.77 VRHE depending on the concentration and pH.  Figure 3.1 – Potential-pH (Pourbaix) diagram for H2O2 at 25°C. (Reprinted from Ref. [11] © NACE International (1974)).  Eq. 3.3a shows the peroxide oxidation to oxygen at E > 0.67 VRHE, while Eq. 3.3b presents the peroxide reduction to water at E < 1.77 VRHE. If both reactions take place at the same rate on the same, or electronically connected, surface sites, the net result is decomposition of peroxide to water and oxygen via electrochemical pathways (Eq. 3.4).  63  H 2 O2 → H 2 O + 1 O2 2 In  addition  to  [3.4] electrochemical  decomposition,  decomposes through thermo-catalytic reaction.  hydrogen  peroxide  In the presence of metal  contaminants such as iron, copper, chromium, or cobalt, the Fenton’s reaction19,20 produces intermediate radicals, hydroxyl (OH•) and hydroperoxyl (HOO•), as shown in Eq. 3.5 and 3.6. H 2 O2 + M x → M x +1 (OH ) + OH •  [3.5]  H 2 O2 + OH • → HOO • + H 2 O  [3.6]  Polymer fragments containing fluoride and sulphur have been found in the condensate collected after degradation, employing either in-situ fuel cell operation or ex-situ Fenton’s tests, suggesting that Fenton’s reaction occurs in the MEA. Kadirov et al.21 accelerated ionomer degradation with UV irradiation and used electron spin resonance (ESR) to measure radical formation. Others have analysed data from nuclear magnetic resonance (NMR), Fourier transform infrared (FTIR) spectroscopy, and mass spectroscopy (MS) on degraded MEAs to show degradation products that are consistent with radical attack on both the sulfonic acid side chains and at unstable end groups of the fluorocarbon backbone18,22,23.  Membrane manufacturers, DuPont and 3M have produced  membrane materials with stabilized end groups24,25, which have reduced the amount of degradation but did not eliminate it. The current work used DuPont NR211 ionomer material with stabilized end groups in both the membrane and CL ionomer.  64  Most literature suggests that metal contaminants in the membrane or CL are responsible for the decomposition of hydrogen peroxide, and most ex-situ tests attempt to replicate membrane degradation by this mechanism. Aoki et al.26 showed that while the presence of Fenton’s metal contaminants increased the concentrations of hydroxyl radicals, these radicals were still present in the absence of a Fenton’s catalyst.  Mittal et al.  27,28  experienced membrane  degradation with no anode catalyst suggesting a mechanism that directly produces radical formation at the cathode catalyst surface, as opposed to the H2O2 intermediate. The rate of membrane degradation has been linked to the presence of platinum in the membrane due to platinum catalyst dissolution. Some studies29,30 showed an increase in the membrane degradation and suggest that the presence of platinum in the membrane increased the H2O2 production and/or decomposition by heterogeneous reaction of crossover gases H2 and O2. However, others31 showed a decrease in the degradation rate by intentionally placing platinum throughout the membrane. These differences may be resolved by further understanding the platinum dissolution mechanism and the incorporation of the platinum in the membrane. The fuel cell open circuit voltage (OCV) of approximately 1.0VRHE is favourable for either Pt electro-dissolution or oxide formation as described by Eq. 3.7 and 3.8  Pt → Pt 2+ + 2e −  Eo = 0.963 (1.188) VRHE  25°C  [3.7]  65  Pt + H 2 O → PtO + 2 H + + 2e −  Eo = 0.98 VRHE  25°C  [3.8]  The equilibrium potentials in Eq. 3.7 are from different potential-pH diagrams, by Lee and Pourbaix (in brackets) respectively, as outlined by Borup et al32. The platinum oxide could chemically dissolve forming Pt2+ according to Eq. 3.9  PtO + 2 H + → Pt 2+ + H 2 O  [3.9]  which explains why the platinum dissolution rate is 104 times lower for constant voltage tests compared to voltage cycling conditions where platinum oxide forms and dissolves repeatedly33. The dissolved platinum may agglomerate by redepositing on existing platinum catalyst particles (Ostwald ripening), wash out of the system, and/or diffuse into the ionomer of the CL and the membrane. The platinum cations that travel into the membrane are reduced by the hydrogen crossing over from the anode, creating a layer of platinum. Changing the partial pressures of oxygen and hydrogen can alter the location of the platinum layer in the membrane34,35. High-resolution transmission electron micrographs (HRTEM) have revealed different shapes and platinum surfaces in the reduced platinum36,37. Ferreira and Shao-Horn suggest that the type of platinum that precipitates depends on the concentration of dissolved platinum and hydrogen37. At the cathode there is a low concentration of hydrogen and the platinum ions precipitate to form dendritic structures that are made up of Pt (111) surfaces. In the presence of hydrogen both Pt (111) and Pt (100) surfaces form truncated octahedrons, truncated  66  square cuboids, and truncated tetrahedrons.  Pt (111) and Pt (100) surfaces  produce H2O2; however, Pt (110) surfaces produce relatively little H2O2. Therefore, it is plausible that the platinum in the membrane may increase or decrease the membrane degradation rate depending on the location and/or platinum surfaces present. Yu et al.38 gave evidence to suggest that the degradation of the Nafion membrane occurs near the cathode/membrane interface.  By measuring the  sulphur distribution across the membrane cross section before and after degradation, they showed a decrease in sulphur at the cathode/membrane interface. This can be explained by i) increased H2O2 production at the platinum in the membrane, which is often close to the cathode/membrane interface, and ii) diffusion of H2O2 from the anode to the cathode where the conditions are more suitable for H2O2 decomposition; or iii) H2O2 generation and decomposition at the cathode. Teranishi et al.18 reported that operating at OCV increased the ionomer degradation rate compared to operating under load. This implied that greater degradation occurred at higher electrode potentials due to the greater decomposition of H2O2, and formation of peroxide radicals. Liu et al.17 showed an increased production of H2O2 with thinner membrane due to greater H2 or O2 crossover. However, greater H2 crossover also creates a lower cathode mixed potential, reducing the OCV. Therefore, the membrane thickness will impact both the H2O2 production and decomposition.  67  Mittal et al.28 demonstrated this effect of membrane thickness on ionomer degradation. An increase in membrane degradation was shown for membrane thickness up to 175 µm, followed by a gradual decrease in degradation for thickness up to 500 µm.  Increasing the membrane thickness up to 175 µm  reduced the O2 and H2 crossover and increased the OCV, resulting in greater H2O2 decomposition and ionomer degradation.  Therefore, for membrane  thickness up to 175 µm the cathode potential was the dominating stressor for ionomer degradation. For membranes thicker than 175 µm the gas crossover was decreased to the level of reducing H2O2 formation, and therefore, reducing ionomer degradation as indicated by fluoride release. While a significant amount of work has focused on the degradation of Nafion membrane, little has been done regarding the degradation of ionomer in the CCL. It has been suggested that the ionomer in the CL does not degrade as fast as the membrane due to the proximity of the platinum catalyst27. Aoki et al.31 have developed an ex-situ test to measure the degradation of Nafion ionomer in the CL. They replicate the gas mixtures that are produced by gas crossover in a normal operating PEMFC anode and cathode compartments. The gas mixture was passed across a GDE, which consists of the CL and GDL.  The gas  condensate was collected and measured for fluoride ions. The ionomer in the CL showed signs of degradation, in both a cathode and anode gas environment; however, greater degradation occurred under anode conditions. Neither rate of fluoride loss was as high as when the membrane was present. This supports the  68  theory that the anode is the major producer of H2O2; however, this also suggests greater degradation closer to the anode/membrane interface. The present work investigates the Nafion degradation in both the membrane and CCL and assesses its impact on fuel cell performance. Properties indicative of ionomer degradation, specifically the membrane and CL ionic conductivity and fluoride release, were analyzed. EIS, CV, and polarization analysis, were used in conjunction with imaging techniques to evaluate the changes in the CL structure caused by ionomer degradation.  3.2  Experimental  As listed in Table 1, the cell was conditioned for approximately 16 hours at the standard operating conditions before conducting the AST testing.  The  ionomer degradation protocols were based on previous work18, 19, and exposed the MEA to either the OCV or a constant cathode potential of 1.0 VRHE with air at the cathode and hydrogen at the anode. Both anode and cathode gas exhaust condensates were collected and monitored for conductivity, pH, and fluoride ion concentration using an Accumet Research duel channel pH/Ion/Conductivity meter with Orion 9609BN ion plus sure-flow fluoride and Orion pH probes. CV and EIS measurements were done approximately every 24 h during the 440 h AST. The full diagnostic test protocol was completed at 0, 150, 300, 440 h. The initial structure in this study was made with two NR211 membranes (60 µm) to maintain a low gas crossover (as indicated by iH2X = 3.0 mA cm-2) and  69  high OCV (0.96 VRHE). Subsequently the membrane thickness was decreased to 30 µm with a single NR211 membrane, increasing the gas crossover (iH2X = 5.9 mA cm-2). This resulted in a lower OCV (0.90 VRHE) and a lower degradation rate as shown by the cumulative fluoride release in Fig. 3.2.  Air/H2  N2/H2  Air/H2  N2/H2  35 60 um NR211 at OCV  30  30 um NR211 at 1.0 V  25 20  -  -2  Cumulative F (µmol cm )  30 um NR211 at OCV  15 10 5 0 0  50  100  150  200 250 Time (h)  300  350  400  450  Figure 3.2 – Total MEA cumulative F- release versus time for different membrane thickness and electrode potentials.  A third test was done using a single 30 µm NR211 membrane (iH2X = 5.7 mA cm-2) and holding the cathode potential at 1.0 VRHE using a Xantrex XHR 60-18 DC power supply. In this test the cell operated in reverse with hydrogen (from gas crossover) being oxidized at the original positive cathode electrode, which in effect became the anode, while protons and possibly oxygen (from gas  70  crossover) being reduced at the original negative anode electrode, which in effect became the cathode. This resulted in greater fluoride release compared to the OCV condition with a single NR211 membrane, demonstrating that the cathode potential had a strong impact on the overall degradation rate for membrane thickness less than 60 µm. These final conditions were used as the standard ionomer degradation AST throughout the rest of the study. The effect of inlet gas RH was also investigated to help understand the ionomer degradation mechanism. The standard AST used 100%RH in both inlet gas streams. The conditions were changed to 100/3%RH cathode/anode and vice versa, to determine if the water flux would change the degradation rate by acting as a carrier of H2O2 or a Fenton’s reagent.  3.3  Results and Discussion 3.3.1 Condensate analysis  The exhaust gas condensates were collected and monitored for fluoride ions, conductivity, and pH. The fluoride ions were measured in both the anode and cathode gas condensates and were roughly equal, failing to indicate any preferential degradation close to the cathode or anode electrode. Fig. 3.3 shows the total cumulative fluoride (anode and cathode) for both the 23 and 33 wt% Nafion CCL ionomer contents. The 23 wt% Nafion MEAs degraded slower than the 33 wt% Nafion MEAs. The second 23 wt% Nafion MEA experienced an accidental 6 h - drying period at 300 h, which accounted for the brief deviation in 71  the degradation rate.  The total MEA fluoride content was calculated to be  approximately 230 µmol cm-2, based on the density (1.98 g cm-3) and equivalent weight (1100 g mol-1 SO3-) of Nafion. The membrane contained approximately 97% of the fluoride including the Nafion spray coats, and the CCL contained 3%. The difference in fluoride release between the 23 and 33 wt% Nafion MEAs cannot simply be explained by the difference in CL Nafion content. The cause of the slower degradation of the 23 wt% Nafion MEA is explored in later discussion.  Air/H2  70  Air/H2  N2/H2  Air/H2  N2/H2  23wt% Nafion Run#1 23wt% Nafion Run#2 33wt% Nafion Run#1 33wt% Nafion Run#2  60 Cumulative F- (mmol.cm-2)  N2/H2  50  40  30  20  10  0 0  100  200  300 Time (h)  400  500  600  Figure 3.3 – Cumulative fluoride release during the AST.  The imposed water flux investigation showed greater degradation when water was forced from anode to cathode. This was not conclusive since the membrane had previously been degraded, which could have impacted the degradation rate when altering the water flux. However, this is consistent with  72  the theory that H2O2 is produced at the anode and diffuses through the membrane to where it decomposes resulting in radical attack of the ionomer. The conductivity and pH of the gas condensates were measured to support the trends given by the fluoride release, as there is an expected correlation between the sample conductivity and fluoride content. However, this correlation was unclear; therefore, the condensates were also measured for dissolved metals by inductively coupled plasma mass spectrometry (ICPMS) to determine if any contaminants washed out of the cell that would contribute to increased conductivity, such as dissolved platinum or any Fenton’s reagents. The conductivity-pH plot in Fig. 3.4 was broken down by the major ions found in solution (H+, OH-, Al3+, F-). 1000 DI H2O DI H2O w Al ion DI H2O w Al & F ions Sample  -1  Conductivity ( µ S cm )  100  10  1  0.1  0.01 2  4  6  8  10  12  pH  Figure 3.4 – Analysis of product water conductivity during AST testing.  73  The conductivity of each species was calculated from their measured concentrations and respective specific molar conductivity. The conductivity of DI water (H+ and OH-) was calculated based on the measured pH of the condensate samples.  The conductivity due to Al3+ washout was calculated from the  concentration in the condensates (measured by ICPMS in one sample) and added to the DI water conductivity. The concentration of fluoride ions in the condensates increased the solution conductivity further, which compared to the measured sample conductivity. Condensates were also collected without the MEA to establish a baseline and to determine if the source of the dissolved ions was from the MEA. The Al3+ ions were only present with the MEA and decreased to 0.14 mg L-1 in a nitrogen atmosphere compared to 0.5-1.5 mg L-1 in the air AST. Therefore, the Al3+ ions are believed to be a side product of the degradation. The aluminum was likely an impurity in the GDL that gradually leached out over time.  The ionomer  degradation products (fluoride) may have affected the leaching out of the metal contaminants. The cathode GDE’s were analysed for metals by acid extraction (standard method EPA 3050) and contained small amounts of aluminum (1-2 µg g-1) and an even greater amount of iron (10-40 µg g-1). The amount of aluminium in the condensates could not fully be explained by the amount of aluminium in the cathode GDE’s; therefore, another source must have been present. Because aluminium is inert regarding H2O2 decomposition and is often used as a H2O2 vessel material, its source was not investigated further.  However, iron is a  74  known Fenton’s catalyst, which likely contributed to the H2O2 decomposition and ionomer degradation that occurred here.  Iron was only found in the gas  condensates at 0.22 mg L-1 when the water flux was forced through the membrane by a large differential in gas RH. Previous studies showed metal ions are exchanged with protons at the sulfonic acid sites of the membrane by a cation exchange process39,40, where ion selectivity determines the extent of the cation exchange.  Nagasaka et al. showed iron bipyridine had very high  selectivity and resulted in almost complete saturation of the ionomer. The Fe2+ exchange reduced the saturated water content from 23% to 8%. Therefore, iron impregnated membrane would have fewer sulfonic acid sites available for proton transfer and a much lower water content, which increases the ionic resistance. In this test, given only the amount of iron found in the cathode GDEs through acid extraction, the effect on the availability of membrane sulfonic groups would be negligible; therefore, another contamination source would be needed to explain changes in the proton conductivity. This is discussed further in the EIS results.  3.3.2 SEM analysis Fig. 3.5 compares SEM micrographs at 0 h (left image) and 440 h (right image), revealing no significant thinning of the cathode or anode electrodes. Alternatively, the Nafion membrane thinned 40-50% over the 440-h degradation period.  This was not a linear decrease as shown in Fig. 3.6, which plots  membrane thickness as a function of cumulative fluoride release.  75  Figure 3.5 – SEM Analysis at 0 and 440 h of the Pt dissolution AST for the 33wt% Nafion MEA. 40  Membrane Thickness ( µm)  35 30 25 20 15 10 5 0 0  10  20  30  40  50  60  70  Cumulative F- Release (µmol.cm-2)  Figure 3.6 – Membrane thinning during the Pt dissolution AST.  76  There was greater increase in H2 crossover through the membrane in the 30-60 µmol cm-2 fluoride release range where little membrane thinning occurred, suggesting that both uniform and localized membrane degradation occurred. Tang et al.23 showed the formation of voids and pinholes in SEM micrographs after soaking the membrane in an H2O2-metal cation solution. It should be noted that there was no apparent degradation under the cathode mask, where there was only membrane and anode catalyst. The only other apparent difference between the new and degraded MEAs was the platinum found in the membrane. The 33 wt% Nafion MEA visibly had greater platinum in the membrane and it was much closer to the CCL. More ionomer in the CL may have provided an improved pathway for dissolved ions to reach the membrane. This precipitated platinum may have played an essential role in the membrane degradation via greater H2O2 production and/or H2O2 decomposition, depending on the potential at the Pt sites. Employing SEM-energy dispersive x-ray spectroscopy (EDX) no Fe was found in the membrane. However, based on the amount of Fe measured from the GDEs, the Fe content in the membrane might have been below the sensitivity of the instrument. Greater platinum and/or iron in the membrane would explain why the 33 wt% Nafion CL had a greater fluoride release rate during the AST. Fig. 3.7 shows a severely degraded MEA subjected to successive changes in cathode and anode RH (120%/120%RH, 120%RH/3%RH, 3%RH/120%RH). This MEA experienced delamination between the membrane and anode CL, tears in the membrane close to the platinum layer in the membrane, and a 77  whitening of the anode catalyst at the membrane/catalyst interface signifying a greater platinum density.  Catalyst interlayer cracking and delamination was  observed in both the anode and cathode.  Cycling the gas humidity causes  ionomer swelling and contraction. The resulting mechanical stresses have been shown to cause ionomer degradation41, and likely contributed to the damage seen in Fig. 3.7.  Cathode Pt band and membrane tear  Anode cracks, brighter catalyst at membrane/anode interface, and delamination from membrane  Anode  Figure 3.7 – RH cycling effect on degradation of 33 wt% Nafion MEA.  3.3.3 Cyclic voltammetry A decrease in the EPSA and an increase in the H2 crossover current were observed in CVs over the duration of the AST as shown in Fig. 3.8.  SEM  analysis already showed that Pt dissolves and precipitates into the membrane, decreasing the amount of Pt in the CCL. The drop in EPSA caused an expected decrease in the kinetic performance as will be shown in the polarization loss analysis.  The increase in H2 crossover current was a direct result of the 78  membrane thinning due to ionomer degradation.  Unfortunately, high H2  crossover current distorts the CV and increases the measurement error in both the H2 crossover current and EPSA; therefore, no EPSA could be measured after 290 h of the AST in the 33 wt% Nafion MEA.  160  1.0 0.9  120 EPSA / EPSAo  0.7 100  0.6  80  0.5 23wt% Nafion EPSA 0.4  33wt% Nafion EPSA  0.3  23wt% Nafion i H2X  60 40  33wt% Nafion i H2X  0.2  -2  0.8  H2 Crossover Current (mA.cm )  140  20  0.1 0.0  0 0  10  20  30  40  50  60  -2  Cumulative Fluoride Release (µmol.cm )  Figure 3.8 – EPSA and hydrogen crossover current during the Pt dissolution AST.  Using the first order kinetic treatment developed by Debe et al.42, and discussed in detail in chapter two, the rate constant for EPSA loss was calculated as 0.004-0.015 h-1. Comparatively, in the carbon corrosion AST, the EPSA loss rate constant was much higher, i.e., 0.105-0.128 h-1. In the latter case the rate of EPSA loss was increased by the significant loss of carbon in addition to platinum agglomeration. Carbon loss did not occur in the present case.  79  3.3.4 Electrochemical impedance spectroscopy Variation in the HF cell resistance and CCL ionic resistance for new MEAs as a function of RH was obtained using both 23 and 33 wt% Nafion contents in the CCL (Fig. 3.9 and 3.10). As defined earlier the HF cell resistance represents the sum of the membrane protonic and cell electronic resistances. 200 30um  Rc - HF Cell Resistance (mΩ cm2)  180  60um  160 140 120 100 80 60 40 20 0 40  60  80  100  120  140  %RH  Figure 3.9 – HF cell resistance for different membrane thickness.  Fig. 3.9 shows the HF cell resistance for two different membrane thicknesses. Taking the difference between the two sets of data and normalizing for thickness, the membrane conductivity was calculated to be 0.12 S cm-1 at 100% RH and 70°C.  This is consistent with literature43, which showed the  membrane conductivity to be between 0.09-0.24 S cm-1 at saturation depending  80  on the water content.  Approximately 4 mΩ was due to the cell electrical  resistance.  1000  9000  2  Ri - CL Ionic Resistance (mΩ cm )  10000 800  8000 600  7000 400  6000  243  200  146  5000 0  4000  60  80  3000  100 120 % RH  140  23%Nafion  2000  33%Nafion  1000 0 40  60  80  100  120  140  % RH  Figure 3.10 – CL ionic resistance for different ionomers contents. (The error bars indicate the standard deviation based on 4 and 7 repeats for the 23 and 33 wt% Nafion MEAs, respectively).  The CL ionic resistance for the MEAs shown in Fig. 3.10 was normalized for area and thickness giving CL ionic conductivities at saturation of 0.008 and 0.010 S cm-1 for 23 and 33 wt% Nafion contents respectively. These values are in agreement with Bruggeman’s relationship as discussed in chapter two. The HF cell resistance and CL ionic resistance measured during the AST were plotted as a function of cumulative fluoride loss in Fig. 3.11. Compared to the sample variability in Fig. 3.9, the HF cell resistance increased significantly at both 60% RH and 120% RH for the 33 wt% Nafion MEA.  81  a)  180  2  HF Rc - Cell Resistance (mΩ.cm )  160 140 120 100 80 60 40 20  23%N - 120%RH  33%N - 120%RH  23%N - 60%RH  33%N - 60%RH  0 0  10  20  30  40  50  60  70  -2  Cumulative Fluoride Release (µmol.cm )  4000  33%N - 120%RH  23%N - 60%RH  33%N - 60%RH  700  3500  600  2  Ri - CL Ionic Resistance (mΩ.cm ) - 120%RH  800 23%N - 120%RH  2  Ri - CL Ionic Resistance (mΩ.cm ) - 60%RH  b) 4500  3000 500 2500 400 2000 300 1500 200  1000  100  500 0  0 0  10  20  30  40  50  60  70  Cumulative Fluoride Release (µmol.cm-2)  Figure 3.11 – MEA resistance during the Pt dissolution AST.  For equivalent fluoride loss, the 23 wt% Nafion MEA had a smaller increase in the HF cell resistance, revealing the degradation had a larger impact on the 33  82  wt% Nafion MEA.  Subtracting the 4 mΩ cell electronic resistance, and  accounting for the decreased membrane thickness, the membrane conductivity was calculated to be 0.067 and 0.035 S cm-1 at 120%RH for the 23 and 33 wt% Nafion MEAs after the 440 h AST. These values are significantly lower than the baseline conductivity. This result is supported by Fig. 3.3, which showed greater fluoride loss in the 33 wt% Nafion MEAs. During the AST it is believed the equivalent weight of the membrane increased either due to the loss of sulfonic acid sites, or the contamination of metal ions such as Fe2+, causing the increased resistance. A negligible amount of iron was found in the GDE material; therefore, the loss of sulfonic acid sites was likely the cause of the decreased conductivity. Both MEAs subjected to the AST protocol showed an initial decrease in the CCL ionic resistance, followed by an increase. It is speculated that the initial CCL ionomer fluoride loss increased the hydrophilicity and water content of the CCL, which lowered the ionic resistance. An example of this phenomena was shown when Duca et al.44 increased the hydrophilicity of polyvinylidene fluoride (PVDF) by decreasing the F/C ratio and increasing the O/C ratio through an Ar plasma treatment. The eventual increase in the CCL ionic resistance was likely caused by the loss or contamination of the sulfonic acid sites, which was also suspected to occur in the membrane.  An overall CCL resistance increase of 38% was  observed in the 23 wt% Nafion MEA at 120%RH. The EIS spectra for the MEA with 33 wt% Nafion in the CCL became convoluted due to the high hydrogen crossover current; therefore, the CCL ionomer resistance could not be measured  83  accurately beyond the 40 µmol cm-2 cumulative fluoride release level. Based on the observed trends in Fig. 3.11b, the authors believe the 33 wt% Nafion CCL had an even greater resistance increase than the 23 wt% after 440 h of the AST.  3.3.5 Polarization analysis Fig. 3.12 and 3.13 shows the polarization curves (a) and the corresponding voltage loss breakdown analysis (b) for the 23 and 33 wt% CCL Nafion content MEAs, after being subjected to the AST protocol.  Comparing the fuel cell  polarization curves in Fig. 3.12a and 3.13a, the CCL with 23 wt% Nafion performed significantly better, showing little performance loss during 440 h of AST; reaching a superficial current density of 2.5 A cm-2 at almost 0.6 V. In contrast, the 33 wt% Nafion MEA after 440 h of degradation could generate only about 0.3 V at 2.1 A cm-2. Clearly the latter MEA suffered significant damage, which was ultimately reflected in the polarization performance. The polarization breakdown analysis in Fig. 3.12b indicates that the 23 wt% Nafion MEA showed a slight decrease in the kinetic losses after the first 150 h of degradation. This may have been a result of the improved ionomer conductivity that the CL experienced at the same time. The increased conductivity would enable easier access to more platinum catalyst sites. The CCL ohmic loss in the 23 wt% Nafion MEA decreased by 18 mV over the first 300 h of the AST, also supporting the initial decrease in the CCL ionic resistance measured by EIS. The following increase in the 23 wt% Nafion CCL kinetic loss would be caused by not only the decreased EPSA, but possibly the increased CCL ionomer resistance 84  (Fig. 3.12b). Greater resistance would restrict the access to platinum catalyst sites. None of the other polarization losses increased in the 23 wt% Nafion MEA over the degradation period. 1.0  600  a) Performance Loss (mV) at 2.0 A cm  -2  0.9  Cell voltage (V)  0.8 0.7 0.6 0.5 0.4 0h 148 h 290 h 437 h  0.3 0.2 0.1  500  b)  448 420 437 409  400  300  200 89 85 83  100  0.0  87 19 17  25273129  0 0.0  0.5  1.0  1.5  2.0  2.5  3.0  3.5  Air Kinetics  Current density (A cm-2)  Cell Resistance Ohmic  Catalyst Layer Ohmic  Mass Transport  Figure 3.12 – Polarization loss analysis for the 23wt%Nafion MEA. 1.0  600 Performance Loss (mV) at 2.0 A cm  Cell voltage (V)  0.8 0.7 0.6 0.5 0.4 0h  0.3  148 h 0.2  290 h  0.1  437 h  0.0  b)  513  -2  a)  0.9  500 403  428 409  400  300  200 90  100  126 114 98  168 21 39 40  41 53 25 31  0 0.0  0.5  1.0  1.5  2.0  2.5 2  Current density (A cm )  3.0  3.5  Air Kinetics  Cell Resistance Ohmic  Catalyst Layer Ohmic  Mass Transport  Figure 3.13 – Polarization loss analysis for the 33 wt% Nafion MEA.  85  Significant losses occurred in the 33 wt% Nafion MEA (Fig. 3.13b). The kinetic and CCL ohmic losses increased due to the increased CCL ionomer resistance, H2 crossover, and decreased EPSA.  The 33 wt% Nafion CCL  ionomer losses increased over the degradation period, supporting the 40 mΩ cm2 increase in the CCL ionic resistance shown in Fig. 3.11. The HF cell ohmic resistance, calculated from the EIS data, translated to a 45 mV drop at 2.5 A cm2  .  As discussed, the cause of this was likely a decrease in the number of  available sulfonic acid sites or ion exchange capacity of the membrane available for proton transport. This shows that the polymer backbone of the membrane was not the only location of radical attack, and the side chains can be cleaved and lost. Tang et al.23 showed side chain fragments by FTIR spectroscopy after soaking the membrane in an H2O2/metal cation solution. The average reaction penetration depth in the CCL from the membrane toward the diffusion layer, as defined by Eq. 1.8, was determined for both oxygen and air, in the case of the 33%wt Nafion CCL (Fig. 3.14). The error in this estimation was large at low current densities; therefore, the reaction penetration is shown as a function of current density between 0.5 and 3.0 A cm-2. With oxygen for the new MEA (i.e. before degradation tests were performed) the average reaction penetration was approximately 20% and decreased to 5% over the degradation period, demonstrating the effect of the CCL ionic resistance increase (Fig. 3.14a).  86  100  100 a) air  90  0h 150 h 300 h 430 h  80 70 60  Reaction Penetration %  Reaction Penetration %  90  50 40 30 20  b) oxygen 0h 150 h 300 h 430 h  80 70 60 50 40 30 20  10  10  0  0 0.0  0.5  1.0  1.5  2.0  2.5  3.0  3.5  -2  Current Density (A cm )  0.0  0.5  1.0  1.5  2.0  2.5  3.0  -2  Current Density (A cm )  Figure 3.14 – Average reaction penetration under oxygen and air. The average reaction penetration with air was quite different and typically increased with current density (Fig. 3.14b).  For the new MEA, the reaction  penetration was approximately 30% until 1.75 A cm-2, when mass transport limitations due to the lower oxygen content starting forcing a deeper reaction penetration into the CCL. After 150 h of the AST the mass transport limitations had a larger effect forcing the average reaction penetration even further into the CCL.  After 300 h AST the penetration decreased similar to the oxygen  performance, likely due to the increased CCL ionic resistance. After 440 h of the AST the reaction penetration increased dramatically starting at 0.5 A cm-2, showing the impact of the increased H2 crossover. Because the H2 that crossed over reacts with the available oxygen, the oxygen concentration at the membrane/CCL interface is reduced. It is evident from this analysis that the platinum surface area taking part in the oxygen reduction reaction depends on the reaction penetration. Because the reaction penetration in an air environment 87  3.5  is a function of current density, the active platinum surface area and kinetic performance, are also a function of current density. Although not shown here, under oxygen the 23 wt% Nafion CCL started with a lower reaction penetration of 8% due to its higher initial CCL ionic resistance compared to the 20% reaction penetration of the 33 wt% Nafion CCL.  The  reaction penetration first increased slightly to 10% due to the decreased CCL ionic resistance and then decreased to 3.5% over the remainder of the AST as the CCL ionic resistance increased.  3.3.6 Diffusivity analysis The limiting current versus oxygen concentration curves are presented in Fig. 3.15 for both the 23 and 33 wt% Nafion CCLs. This was done to show how the effective oxygen diffusivity and total H2 crossover changed over the AST. The slopes of the trend lines are proportional to the effective oxygen diffusion coefficient. Linear trend lines were fit to the first three points to measure the effective diffusivity of the CCL structure with negligible water production.  At  greater oxygen concentrations and limiting current density, the curves bend, signifying a decreased effective diffusion coefficient, most likely due to increased water content in the electrode.  88  1.6  2.0  a) 23wt% Nafion  1.8  0h  1.6  290 h  -2  1.2  J lim (A.cm )  -2  0h  1.2  148 h  1.4  J lim (A.cm )  b) 33wt% Nafion 1.4  437 h  1.0 0.8  148 h 290 h  1.0  437 h  0.8 0.6  0.6  0.4 0.4  Slope = Deff * n * F / LGDE  0.2  0.2 Slope = Deff * n * F / LGDE  0.0  0.0 0.0  1.0  2.0  3.0  4.0  5.0  -3 Channel Concentration (mol.m )  6.0  7.0  0.0  1.0  2.0  3.0  4.0  5.0  6.0  Channel Concentration (mol.m-3)  Figure 3.15 – Limiting current density versus oxygen concentration.  The 23 wt% Nafion CCL in Fig. 3.15a, showed similar slopes over the AST period with an effective diffusivity of 1.56 x 10-2 cm2 s-1, supporting the polarization results that showed virtually no change in the mass transport region, thus, it can be inferred that the CL structure and morphology were unaffected by the AST. However, after 440 h of AST the curve shifted along the x-axis due to the increased H2 crossover reacting with the oxygen at the cathode. Fig. 3.15b shows for the new 33 wt% Nafion CL the effective oxygen diffusivity was similar to the 23 wt% case (1.65 x 10-2 cm2 s-1), which was expected because of fairly similar CCL porosity (63 and 71% porosity for the 33 and 23 wt% Nafion CCLs, respectively) and pore structure morphology. However, the 33 wt% Nafion CCL demonstrated greater mass transport limitations over the AST as indicated by the diminishing slopes in Fig. 3.15b (final effective diffusivity of 7.40 x 10-3 cm2 s-1), and shown before in the polarization results. Hence, the diffusivity analysis confirms the serious degradation of the 33 89  7.0  wt% Nafion CCL after 440 h of the AST, which created a more hydrophilic structure and possibly also changed the porosity and pore structure morphology.  3.4  Conclusions  The effect on PEMFC performance of holding the cathode potential at 1.0 VRHE in air for up to 440 h was investigated using 23 and 33 wt% Nafion CCL designs. The EPSA experienced a 40-50% reduction due to platinum dissolution and/or agglomeration. Significant degradation of the membrane was also found (general thinning by approximately 44% coupled with indications of pinhole formation) corresponding to a 40 and 70% decrease in the membrane conductivity for the 23 and 33 wt% Nafion MEAs, respectively. The CCL ionomer resistance showed an initial decrease during the early phases of the AST, followed by an increase, with an overall 38% resistance increase in the 23 wt% Nafion MEA case. The overall CCL ionomer resistance increase could not be measured accurately for the 33 wt% Nafion MEA due to excessive hydrogen crossover. The increased ionomer content in the CCL (23 versus 33 wt%) was found to exacerbate the performance degradation.  The 33 wt% Nafion MEA had  greater platinum content in the membrane and a higher fluoride washout rate, suggesting that the higher ionomer content in the CCL provided mass transfer pathways for contaminants (such as dissolved platinum and iron) to diffuse into the membrane. Hence, the hypothesis is supported that H2O2 was produced at the anode, diffused into the membrane, and decomposed at the platinum and/or 90  iron sites bound in the membrane structure.  The decomposition products  attacked both the bulk phase and CL ionomer causing membrane thinning, decreased membrane conductivity, and CL ionomer structure degradation. The amount of fluoride released indicated that the majority of fluoride loss originated from the membrane. The average penetration depth of the ORR into the CCL relative to the catalyst/membrane interface was determined using a novel method relying on EIS and steady state polarization analysis. The results showed for the fuel cell operated with an air cathode, a drastic increase in the average penetration depth versus current density for the CCL with 33 wt% ionomer content. This was most prominent after 440 h of AST, suggesting an oxygen mass transport limitation in the CCL. This finding was corroborated by oxygen effective diffusivity analysis. It is proposed that this extended ORR penetration depth from the membrane into the CCL was caused by a combination of factors: H2 crossover consuming the O2 at the membrane-CCL interface, increased hydrophilicity of the CCL due to degradation of the organic backbone of the ionomer (especially at sites close to the membrane interface) increasing the water content of the CCL, loss of platinum active sites and changes in the CCL ionic conductivity.  91  3.5  References  1.  P. Choi, N. Jalani, and R. Datta, J. Electrochem. Soc., 152, E123 (2005).  2.  E.L. Thompson, T.W. Capehart, T.J. Fuller, and J. Jorne, J. Electrochem. Soc., 153, A2351 (2006).  3.  C. Boyer, S. Gamburzev, O. Velev, S. Srinivasan, and A.J. Appleby, Electrochimica Acta, 43, 3703 (1998).  4.  M. Nakamura, J. Appl. Phys., 57, 1449 (1985).  5.  R. Makharia, M. Mathias, and D. Baker, J. Electrochem. Soc., 152, A970 (2005).  6.  M. Lefebvre, R. Martin, and P. Pickup, Electrochemical and Solid State Letters, 2, 259 (1999).  7.  X. Ren, and P. Pickup, Electrochimica Acta, 46, 4177 (2001).  8.  G. Li, and P. Pickup, J. Electrochem. Soc., 150, C745 (2003).  9.  E. Easton, and P. Pickup, Electrochimica Acta, 50, 2469 (2005).  10. A. Young, J. Stumper, and E. Gyenge, J. Electrochem. Soc., 156, B913 (2009). 11. M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, p. 106-111, National Assoc. of Corrosion Engineers, Houston, TX (1974). 12. K. Kinoshita, Carbon: Electrochemical and Physiochemical Properties, p.361, John Wiley & Sons, New York (1998).  92  13. O. Antoine, and R. Durand, J. Appl. Electrochem., 30, 839 (2000). 14. B.N. Grgur, N.M. Markovic, and P.N. Ross, Can. J. Chem., 75, 1465 (1997). 15. U.A. Paulus, T.J. Schmidt, H.A. Gasteiger, and R.J. Behm, J. Electroanal. Chem., 495, 134 (2001). 16. M. Inaba, H. Yamada, J. Tokunaga, and A. Tasaka, Electrochemical and Solid State Letters, 7, A474 (2004). 17. W. Liu, and D. Zuckerbrod, J. Electrochem. Soc., 152, A1165 (2005). 18. K. Teranishi, K. Kawata, S. Tsushima, and S. Hirai, Electrochemical and Solid State Letters, 9, A475 (2006). 19. E. Endoh, S. Terazono, H. Widjaja, and Y. Takimoto, Electrochemical and Solid State Letters, 7, A209 (2004). 20. U. Jung, S. Jeong, K. Chun, K. Park, H. Lee, D. Choi, and S. Kim, J. Power Sources, 170, 281 (2007). 21. M.K. Kadirov, A. Bosnjakovic, and S. Schlick, J. Phys. Chem. B, 109, 7664 (2005). 22. J. Healy, C. Hayden, T. Xie, K. Olson, R. Waldo, M. Brundage, H. Gasteiger, and J. Abbott, Fuel Cells, 5, 302 (2004). 23. H. Tang, S. Peikang, S.P. Jiang, F. Wang, and M. Pan, J. Power Sources, 170, 85 (2007). 24. J.W. Bauman, DOE Hydrogen, Fuel Cells and Infrastructure Technologies Program Review, DuPont Fuel Cells (2004).  93  25. M. Hicks, DOE Hydrogen, Fuel Cells and Infrastructure Technologies Program Review, 3M Company: Fuel Cell Components Program, No. DEFC36-03GO13098 (2004). 26. M. Aoki, H. Uchida, and M. Watanabe, Electrochemistry Comm., 8, 1509 (2006). 27. V. Mittal, H.R. Kunz, and J.M. Fenton, Electrochemical and Solid State Letters, 9, A299 (2006). 28. V. Mittal, H.R. Kunz, and J.M. Fenton, J. Electrochem. Soc., 154, B652 (2007). 29. A. Ohma, S. Suga, S. Yamamoto, and K. Shinohara, J. Electrochem. Soc., 154, B757 (2007). 30. J. Zhang, B.A. Litteer, W. Gu, H. Liu, and H.A. Gasteiger, J. Electrochem. Soc., 154, B1006 (2007). 31. M. Aoki, H. Uchida, and M. Watanabe, Electrochemistry Comm., 7, 1434 (2005). 32. R. Borup, J. Meyers, B. Pivovar, Y.S. Kim, R. Mukundan, N. Garland, D. Myers, M.Wilson, F. Garzon, D. Wood, et al., Chem. Rev. (Washington, D.C.), 107, 3904 (2007). 33. X. Wang, R. Kumar, and D.J. Myers, Electrochemical and Solid State Letters, 9, A225 (2006). 34. W. Bi, G.E. Gray, and T.F. Fuller, Electrochemical and Solid State Letters, 10, B101 (2007). 94  35. L. Kim, C.G. Chung, Y.W. Sung, and J.S. Chung, J Power Sources, 183, 524, (2008). 36. T. Akita, A. Taniguchi, J. Maekawa, Z. Siroma, K. Tanaka, M. Kohyama, and K. Yasuda, J Power Sources, 159, 461 (2006). 37. P.J. Ferreira, and Y. Shao-Horn, Electrochemical and Solid State Letters, 10, B60 (2007). 38. J. Yu, B. Yi, D. Xing, F. Liu, Z. Shao, Y. Fu, and H. Zhang, Phys. Chem. Chem. Phys. 5 (2003). 39. A. Steck, and H.L. Yeager, Anal. Chem., 52, 1215 (1980). 40. S. Nagasaka, M. Harada, and T. Okada, Analytica Chimica Acta, 525, 115 (2004). 41. X. Huang, R. Solasi, Y. Zou, M. Feshler, K. Reifsnider, D. Condit, S. Burlatsky, and T. Madden, J Polymer Science: Part B, 44, 2346 (2006). 42. M.K. Debe, A.K. Schmoeckel, G.D. Vernstrom, and R. Atanasoski, J. Power Sources, 161, 1002 (2006). 43. X. Zhou, Z. Chen, F. Delgado, D. Brenner, and R. Srivastava, J. Electrochem. Soc., 154, B82 (2007). 44. M. Duca, C. Plosceanu, and T. Pop, Polymer Degradation and Stability, 61, 65 (1998)  95  4.0 Application of Electrochemical Quartz Crystal Microbalance to Study Degradation of Ionomer Films  96  4.1  Introduction 4.1.1 EQCM background  The quartz crystal microbalance (QCM) utilizes piezoelectric phenomena to measure nanogram changes in mass and viscosity.  Piezoelectric materials  produce a voltage upon mechanical deformation. This mechanism has been used in a wide range of applications including transducers for sonar, speakers, and microphones. Ultrasonic sound waves deform the piezoelectric material, which produces a measurable voltage signal. Alternatively, applying a voltage produces mechanical deformation of the material. Quartz is an acentric crystalline silicon dioxide structure with a density of 2.649 g cm-3 and a melting point of 1650°C. The orientation of atoms in an acentric crystal creates a polar structure. When a voltage is applied across the crystal, the dipoles shift creating a mechanical strain in the material as shown in Fig. 4.1. a)  b) Pt AT-cut quartz crystal  Electrode leads Figure 4.1 – 9 MHz AT-cut quartz resonator with Pt electrodes. (Adapted from Buttry & Ward.6)  97  When quartz is heat-treated above 573°C the crystalline structure changes from an alpha-quartz to a beta-quartz structure. Alpha-quartz displays the most stable piezoelectric behaviour and is most often used for QCM resonators. The quartz crystal material is cut at specific angles for different applications. The quartz most often used for resonator applications is an AT-cut crystal, which is cut at an angle of 35° from the optical plane as shown in Fig. 4.2. BT, IT, SC, and FC cut crystals are cut at varying angles for varying applications1.  Figure 4.2 – Quartz crystal cut at 45 degree. (Reprinted from Ref. [1]).  As shown in Fig. 4.1, the AT-cut quartz is sandwiched between two electrodes that can apply an alternating voltage creating an oscillating motion in the crystal. This is analogous to the oscillating behaviour of a pendulum or a mass on a spring. Fig. 4.3 correlates this mechanical behaviour to an electrical equivalent circuit.  98  a)  b) CS  r  static  Cm motional Rq  Cq  Lq  Force Figure 4.3 – Correlates physics of quartz crystal to mass on a spring model. (Adapted from Buttry & Ward6).  The mass on a spring analogy is comprised of a spring compliance term (Cm), which represents the energy stored in the spring, the energy dissipation factor (r), which is the resistance to motion, and the force of the mass on the spring (M). This correlates to the motional segment of the QCM equivalent circuit model, where Rq is equivalent to the resistance to motion, Cq is the energy storage capacitance, and Lq is the inductance associated with the inertia of motion. CS is the static flat plate capacitance that arises due to the dielectric of quartz between the two electrodes. This can be calculated as a simple flat plate capacitor as shown in Eq. 4.1, where Dq is the quartz dielectric constant (~4.5), εo is the permittivity of free space (8.85 x 10-12 F m-1), A is the piezoelectric active area (~0.196 cm-2), and tq is the quartz thickness (~0.185 mm). This results in an approximate static capacitance of 4.22 x 10-12 F.  99  CS =  Dq ⋅ ε o ⋅ A tq ⋅ 10  [4.1]  As discussed the motional segment of the equivalent circuit is composed of a resistor-capacitor-inductor in series. The resistance to the oscillating motion of the resonator is a function of the mechanical energy dissipation factor (r) and piezoelectric stress constant (κ), as shown in Eq. 4.2. This resistance can be measured at the resonant frequency of the quartz crystal where the magnitude of the impedance of the capacitor and inductor are equal, and sum to zero. For the quartz crystal shown in Fig. 4.1, the resistance in air was measured to be 15+/- 5 Ω. tq ⋅ r 3  Rq =  8 ⋅ 105 ⋅ A ⋅ κ 2  [4.2]  The motional capacitance represents the ability of the quartz to store energy and is equated to the piezoelectric stress constant and quartz elastic constant (c) as shown in Eq. 4.3. 8 ⋅ A ⋅κ 2 Cq = 10 ⋅ tq ⋅ π 2 ⋅ c  [4.3]  The motional inductance represents the inertial component due to the mass of the crystal, and is equated to the quartz density (2.649 g cm-3) divided by the piezoelectric stress constant as shown in Eq. 4.4. tq ⋅ ρ q 3  Lq =  8 ⋅ 102 ⋅ A ⋅ κ 2  [4.4]  100  The impedance of the motional segment was simulated in a Bode plot shown in Fig. 4.4. At the resonant frequency of approximately 9 MHz the total impedance reaches a minimum and the phase angle changes from capacitive (90°) to inductive (90°).  Based on this curve, if either of the capacitance or  inductance is known, the other can be determined by equating the two at the QCM resonators known natural frequency, as will be shown later. Bode Plot Simulation  |Z| Impedance (Ohm)  1.0E+08 1.0E+06  Capacitive Impedance (-1/wC)  Resonant Frequency (9MHz)  Inductive Impedance (wL)  1.0E+04 1.0E+02  Phase Angle  1.0E+00 1.0E+04  -100.0 -80.0 -60.0 -40.0 -20.0 0.0 20.0 40.0 60.0 1.0E+04 80.0 100.0  1.0E+05  1.0E+06  1.0E+07 Frequency (Hz)  1.0E+08  1.0E+09  1.0E+10  Capacitive Phase Angle = -90  Inductive Phase Angle = 90 1.0E+05  1.0E+06  1.0E+07  1.0E+08  1.0E+09  1.0E+10  Frequency (Hz)  Figure 4.4 – Simulated Bode plot for the motional segment of a QCM resonator.  An electrochemical quartz crystal microbalance (EQCM) makes use of the electrodes on either side of the quartz to study the electrochemistry of different electrodes. For example, platinum electrodes can be used to elucidate reaction steps in a catalytic reaction such as adsorption-desorption, by measuring the 101  mass change as a function of voltage. In this technique the EQCM becomes the working electrode with a counter and reference electrode in a typical electrochemical wet cell.  4.1.2 EQCM operation When an external mass is attached to the QCM (or EQCM) resonator the natural frequency changes. In 1959 Sauerbrey correlated the mass change to the change in the resonant frequency as described by Eq. 4.5. − 2 ⋅ f  ⋅ ∆m ⋅ 10−9 ( µ q ⋅ ρ q )1 / 2 ⋅ A 2  ∆f =  [4.5]  where fo is the initial resonant frequency of the crystal, Δm (ng) is the external mass, A is the active area of the crystal, ρq = 2.648 g cm-3 is the density of quartz, and μq = 2.947×1011 g cm-1 s-2 is the shear modulus of quartz. The constants in Eq. 4.5 are often grouped into the mass sensitivity factor Cf, simplifying to Eq. 4.6. ∆f = −C f ⋅  ∆m ⋅ 10−9 A  [4.6]  For a quartz crystal with a resonant frequency of 9 MHz, the mass sensitivity factor in ambient air is 0.183 Hz cm2 ng-1. The EQCM can be utilized to verify the mass sensitivity constant by adding or stripping a known mass onto or from the working electrode and measuring the resulting frequency change. Combining the EQCM with CV provides a way to quantify the amount of mass on  102  the electrode. Faraday’s law is used to calculate the mass from the total charge required to strip copper from the surface of the platinum through the amount of charge passed, as will be shown by copper deposition data discussed later. Other environments that have a higher viscosity such as liquid water also increase the natural frequency of the EQCM resonator. Kanazawa and Gordon2 showed that in an absence of a mass change the addition of a liquid to one side of the resonator increased the natural frequency according to Eq. 4.7, where ρL and ηL are the density and viscosity of the liquid, respectively. − f  ⋅ ( ρ L ⋅ η L )1 / 2 ( µ q ⋅ ρ q )1 / 2 ⋅ π 1 / 2 3/ 2  ∆f =  [4.7]  For example, adding water with a density of 1.0 g cm-3 and a viscosity of 8.9 x 10-3 Poise (g cm-1 s-1) to one side of a 9 MHz resonator resulted in a frequency change of approximately –1700 Hz. Summing the frequency changes for both a mass change and liquid effect (Eq. 4.5 to 4.7) leads to Eq. 4.8, which describes an EQCM resonator in solution. ∆f = −C f ⋅ [  ρ ⋅η ∆m ⋅ 10−9 + ( L L )1 / 2 ] A π ⋅ 4 ⋅ fo  [4.8]  The equivalent circuit that represents Eq. 4.8 is very similar to that in Fig. 4.3b and is shown in Fig. 4.5.  103  Quartz  CS  Mass  Liquid static  motional Rq  Cq  Lq  LM  RL  LL  Figure 4.5 – Equivalent circuit representing quartz, additional mass, and liquid on the QCM resonator. (Adapted from Buttry & Ward6)  The motional segment of the circuit is comprised of the resistor-capacitorinductor associated with the quartz with the addition of an inductor associated with any mass change, and a resistor-inductor in series that is associated with the liquid.  The resistor RL represents the resistance to motion due to the  viscosity of the liquid.  For water this resistance has been measured at  approximately 320 Ω. The inductor LL represents the inertial component based on the density of the liquid and is approximately 5.4 x 10-9 H.  4.1.3 EQCM applications The QCM has been used to study film deposition onto the QCM electrodes and film characterization3-5. polypyrrole/disulfide  electrode,  Ye and Blanger3 studied the formation of a Skomspka and Hillman4  investigated  the  electrodeposition and crosslinking of poly(N-vinylcarbazole) films, and Chen et al. 5  used a CNT/Nafion film to monitor humidity.  104  Several factors must be considered when applying films such as the examples given above, such as the mass loading and uniformity, surface roughness, film adhesion, and any viscoelastic characteristics of the film. One can intuitively understand how these factors could interfere with the Sauerbrey frequency-mass correlation of the QCM by understanding the interaction between the film and the piezoelectric quartz crystal. For example, if the film loading is too thick or not uniform the film would change the mass sensitivity constant defined by Eq. 4.5 and 4.6 by deviating too far from the properties of the piezoelectric quartz. Surface roughness can create non-uniformities across the active area, which again could change the mass sensitivity constant. Poor film adhesion or slippage would create a situation where the QCM is resonating and the film is stationary or only partially moving, which would increase the resistance to motion of the QCM and could act as a very viscous fluid as described by Eq. 4.7 and 4.8, as opposed to an increase in mass due to the film itself. An equivalent circuit proposed by Buttry and Ward6 representing a film deposited on the QCM resonator is shown in Fig. 4.6. The quartz, mass, and liquid segments are the same as described in Fig. 4.5.  The film segment  consists of the resistance to motion of the film (RF), the inertia of motion (LF), an energy dissipation term (RD), and an elasticity component (CF).  105  Figure 4.6 – Equivalent circuit representing a film deposited on an EQCM resonator. (Adapted from Buttry & Ward6)  The resistance to motion and inertia of motion are typical and can be added to the Sauerbrey relationship. A film that is not rigid and oscillates at different rates through its thickness (ie. top and bottom of layer are no longer coupled) is described as viscoelastic. A film’s viscoelasticity is represented by the energy dissipation and elasticity terms in Fig. 4.6. Typically quartz has relatively low resistance to motion (RQ) and high inertia of motion (LQ), which creates a stable oscillation and makes this an excellent material for timepiece applications. When the total resistance to motion of the system, including the dissipation resistance due to a viscoelastic film, becomes greater than the impedance due to the inertia of motion the oscillations become unstable and the Sauerbrey mass-viscosity relationship breaks down. In this case one must consider and correct for the energy dissipation. The energy dissipation effect can be quantified by monitoring the decay constant of the resonator motion upon stopping the applied voltage. This measurement has been developed in a modified EQCM tool7.  This is  discussed further when applying a Naifon film to the EQCM.  106  As previously discussed combining CV with the EQCM can be a powerful tool to characterize the surface chemistry as a function of voltage. Adsorption of water and oxygen on platinum occurring in PEMFCs, can be studied with the EQCM by measuring the change in mass while cycling the platinum electrode potential. Santos et al.8 showed this relationship by plotting the mass change versus the charge transfer. These plots gave linear trends with inflection points to distinguish different absorbing/desorbing species. The molar mass of the different species adsorbing/desorbing on the platinum surface was calculated by correcting the slopes by Faraday’s constant and assuming a single electron transfer, as shown in Eq. 4.9. g g C mole − ) = Slope( ) ⋅ F ( ) ⋅ n( ) MW ( mol C mole − moladsorbate  [4.9]  Santos et al. showed the adsorption of water on a Pt EQCM with a molar mass close to 18 g mol-1, which supports the mechanism proposed by Gloagen et al.9 in Eq. 4.10. Pt ( H ) + xH 2 O → Pt ( H 2 O) x + H + + e −  [4.10]  Hydrogen is replaced by water as it is stripped from the platinum surface between 0 to 0.3 VRHE, where x is less than or equal to 1. Secondary slopes were believed to be hydrated anion adsorption ClO4- + 2H2O (135.56 g mol-1) and HSO4- + 2H2O (133 g mol-1) for the HClO4 and H2SO4 solutions respectively.  107  The degradation of electrode materials has been studied through the current and mass change as a function of voltage. Dam and de Bruijn10 showed the mass change due to dissolution of the platinum electrode as a function of voltage. At both 0.85 and 1.15 VRHE the mass of the EQCM initially increased due to the formation of the platinum oxide layer described in Eq. 4.11 and 4.12. Pt + H 2 O ⇔ PtO + 2 H + + 2e −  E o = 0.98 − 0.059 pH  25°C [4.11]  PtO + H 2 O ⇔ PtO2 + 2 H + + 2e −  E o = 1.05 − 0.059 pH  25°C [4.12]  Once the mass reached approximately 2µg the mass started to decrease, suggesting that the platinum oxide was required before significant platinum dissolution occurred as shown in Eq. 4.13 and 4.14. At high potentials direct platinum dissolution can occur as shown in Eq. 4.15. +  PtO + 2 H + ⇔ Pt 2 + H 2 O  log[ Pt 2+ ] = −7.06 − 2 pH  PtO2 + 4 H + + 2e − → Pt 2+ + 2 H 2 O  E o = 0.84 − 0.12 pH − 0.03 log[ Pt 2+ ] [4.14]  Pt ⇔ Pt 2+ + 2e −  E o = 1.12 + 0.029 log[ Pt 2+ ] 25°C [4.15]  25°C [4.13]  Increasing the potential from 0.85 to 1.15 VRHE resulted in a higher dissolution rate due to the greater driving voltage as well as both indirect and direct platinum dissolution. Wang et al.11 showed how cycling the potential across the platinum oxide formation and dissolution potentials (approximately 0.8 to 1.2 VRHE) resulted in greater platinum dissolution compared to potentiostatic tests.  108  Although not discussed further here, Dam and de Bruijn also used the EQCM to study corrosion of the catalyst carbon support and the stability of different catalysts. Buttry and Ward6 gave examples of where others have used the EQCM as a corrosion sensor to study corrosion rates of various metals. The present work investigated the use of the EQCM to study degradation of Nafion ionomer.  4.2  Experimental  Except were noted the EQCM working electrode was submersed in 0.5 M H2SO4 solution with a HgSO4 reference electrode (MSE #XR200-740-14-006) and a platinum mesh counter electrode at room temperature.  4.3  Results and Discussion 4.3.1 EQCM characterization  The effect of the electrolyte was measured according to Eq. 4.7.  The  frequency change from adding water to the system was measured at –2300Hz; similar to the –1700Hz calculated by Eq. 4.7. Impedance spectroscopy was used to determine all or part of the equivalent circuit shown in Fig. 4.3 and simulated in Fig. 4.4.  For example, the total  capacitive impedance was measured by impedance spectroscopy between 100 Hz to 1 MHz as shown in Fig. 4.7. The motional capacitance (2.56 x 10-11 F) was determined by subtracting the static capacitance from the total measured  109  capacitive impedance. The motional inductance was then calculated (1.22 x 10-5 H) assuming a resonant frequency of 9 MHz, resulting in a piezoelectric stress constant of 2.96 C m-2, an elastic constant of 2939 kg m-1 s-2, and an energy dissipation factor of 4.34 x 109 kg m-3 s-1 calculated from Eq. 4.2 - 4.4. Bode Plots: QCM922 - 100mV amplitude, 1 Hz - 10 MHz  |Z| Impedance (Ohm)  1.0E+10 1.0E+08 1.0E+06 1.0E+04  SI1287 is unreliable >1MHz Run#1 Run#2 Run#3  1.0E+02 1.0E+00 1.0E+00 1.0E-02  1.0E+01  1.0E+02  1.0E+03  1.0E+04  1.0E+05  1.0E+06  1.0E+07  1.0E+08  Frequency (Hz)  Figure 4.7 – Measured Bode plot from impedance spectroscopy of a QCM resonator.  As discussed previously, copper adsorption/desorption is commonly used to quantify the mass constant in Eq. 4.6. This was done here by adding copper sulphate solution to the EQCM electrolyte to facilitate copper adsorption onto the platinum electrode. The charge required to cycle the potential from 0 to 1.1 VRHE was measured and correlated to the frequency change according to Eq. 4.6 resulting in a sensitivity constant of ~0.193 Hz cm2 ng-1 as shown in Fig. 4.8. This is close to the theoretical mass constant of 0.183 Hz cm-2 ng-1 discussed earlier based on the density and shear modulus of the quartz alone.  110  Cyclic Voltammetry & EQCM: Cu Deposition 0.003  4000 3000  0.002  Charge = 0.0104 C 0.001  Current (A)  1000 0.000  0 0  0.2  0.4  0.6  0.8  1  1.2 -1000  -0.001  ∆ frequency = -3350 Hz  Frequency Change (Hz)  2000  -2000  -0.002 -3000 -0.003  -4000 Voltage vs RHE  Figure 4.8 – Calculating the mass constant from CV and EQCM.  4.3.2 Contamination Upon attempting to measure the platinum electrode CV it was apparent that contamination was present in the system.  In an effort to eliminate the  contamination, the system was cleaned with hydrogen peroxide and DI water. The MSE reference electrode was replaced with a standard hydrogen electrode (SHE) in case of a possible leaky reference electrode.  Unfortunately, the  contamination was still present. By using the change in mass-charge curves discussed above and using Eq. 4.9, the EQCM was used to help describe the unknown contamination. The potential of the working electrode was held at -0.58 VMSE (0.1 VRHE) for forty  111  minutes to promote contaminate adsorption.  The voltage was cycled while  measuring the current and the mass change. As shown in Fig. 4.9 a) the current peaks and mass loss decreased with each successive cycle signifying a reduction in the amount of contaminates adsorbed on the platinum surface. It should also be noted that there was no hydrogen desorption peaks, due to the contaminate species covering the platinum surface.  After 5 cycles the  contamination on the platinum electrode reached an equilibrium, with a mass loss of approximately 200 ng cm-2. To confirm this finding the potential was held at 0.8 VMSE (1.5 VRHE) for ninety minutes to promote desorption of the contaminate species. The potential was again cycled revealing an increase in the current peaks above 0 VMSE (0.68 VRHE) as shown in Fig. 4.9 b). The hydrogen desorption peaks started high, as hydrogen was able to adsorb on the ‘clean’ platinum surface, and decreased as the contaminate species began to cover the platinum. The mass loss increased steadily until reaching equilibrium of approximately 200 ng cm-2, similar to the equilibrium reached in Fig. 4.9 a). The mass increased between -0.48 VMSE (0.2 VRHE) and 0.18 VMSE (0.5 VRHE), signifying the adsorption of contaminates.  112  a)  Cyclic Voltammetry 6.0E-05  Mass Change 400  Current (A)  -2  Mass Change (ng cm )  200  4.0E-05  2.0E-05  0.0E+00 -0.8 -0.6 -0.4 -0.2 0  0.2  0.4  0.6  0.8  -2.0E-05  0 -0.8 -0.6 -0.4 -0.2 0 -200  0.2  0.4  0.6  0.8  -400 -600 -800  -4.0E-05 Voltage (V MSE)  Voltage (V MSE)  Cyclic Voltammetry 6.0E-05  Mass Change  Current (A)  -2  4.0E-05  2.0E-05  0.0E+00 -0.8 -0.6 -0.4 -0.2 0  200  Mass Change (ng cm )  b)  -1000  0.2 0.4 0.6 0.8  -2.0E-05  -4.0E-05 Voltage (V MSE)  100 0 -0.8 -0.6 -0.4 -0.2  0  0.2 0.4 0.6 0.8  -100 -200 -300 Voltage (V MSE)  Figure 4.9 – Identifying region of contaminate adsorption on the EQCM.  The molar mass of the contaminate was estimated by calculating the moles of contaminate from the charge of adsorption or desorption and normalizing for the mass change from the EQCM, similar to the molar mass calculated in Eq.  113  4.14. For the contamination in Fig. 4.9 this calculation gave 88 +/- 15 g mol-1 electron.  Based on this information alone it would be difficult to identify the  contaminate, but coupled with other spectroscopic methods the contaminate could be identified.  4.3.3 Film behaviour As shown in Fig. 4.10, Nafion solution was added to the EQCM resonator in approximately 1-µL steps with a spin coat treatment. Each step increased the overall weight by approximately 15-25 µg with the exception of the fourth addition. This linearity supports rigid film behaviour; therefore, the Sauerbrey relationship was accurate for the first three films. The fourth addition became distinctly non-linear, and may have introduced viscoelastic behaviour by becoming too thick. Any future work would require staying below this threshold thickness (<50 µg – 1.2 µm). Submersing the Nafion film in water increased the mass due to the hydration of the Nafion film. This mass increase should correlate to the water content of the Nafion film, which has been characterized in literature as 14-22 mol H2O per mol SO3- site in the Nafion structure12. As noted in Fig. 4.10, the first film showed this correlation; however, subsequent films deviated, with the fourth film actually showing a mass loss. Therefore, thicker hydrated Nafion films can be considered viscoelastic films and would require additional dissipation  114  monitoring to account for the viscoelasticity. Only very thin hydrated Nafion films (<25ug) can be used with the current system. Mass Change vs Nafion Film Addition 80 70  Nafion Film Nafion Film w H2O  Mass Change (µg)  60 50 40 30 20 14.2 10  30.0 2.9 mol H2O/ mol SO3-  -14.4  0 ~1  ~2  ~3  ~4  Nafion Film Additon (µL)  Figure 4.10 – Mass change with addition of Nafion film.  New technology13 by Q-Sense® couples the EQCM with dissipation monitoring and takes the viscoelastic effects into account (QCM-D model) and has enabled the study of mass changes in viscoelastic materials7,14-16. Similar QCM studies have been done using hydration of various films17 as a sensor to monitor gas humidity5,18. patented19  a  coupled  QCM-heat  Masscal Scientific Instruments® have sensor  (QCM/HCC  –  quartz  crystal  microbalance/heat conduction calorimeter) to measure both mass and thermal changes at a gas-solid interface. One such application of their system quantified thermal effects on vapour adsorption in polymers20. 115  4.3.4 Bubble formation As discussed in chapter 3, the degradation of ionomer requires the production and decomposition of H2O2. Using the thinnest film shown in Fig. 4.10, 0.1 wt% H2O2 was added to the electrolyte in an attempt to degrade the ionomer. As shown in Fig. 4.11, it initially appeared as though there was an immediate mass loss; however, the frequency change was not stable. It was discovered that upon adding H2O2, bubbles immediately formed on the surface of the film.  When these bubbles were purged from the surface the frequency  change decreased and current was measured as more H2O2 reacted on the platinum electrode.  Based on these results it was discovered that the initial  frequency increase was not due to ionomer loss, but rather the elimination of the liquid mass effect described by Eq. 4.7 and the equivalent circuit in Fig. 4.5, due to the formation of gas bubbles. This frequency change due to the liquid density and viscosity creates an added complexity if the liquid-electrode interface is not uniform, such as when gas bubbles are present on the electrode surface. The fractional coverage of gas bubbles was estimated by taking the ratio of frequency change with bubbles (approximately 700Hz in Fig. 4.11) versus the frequency change that occurred when adding liquid initially (-2300Hz Eq. 4.7). In this case approximately 30% of the surface was covered.  116  Frequency Change vs Time: Addition of H2O2 -600  9.8E-03 1mM H2O2 addition  8.8E-03 7.8E-03  -200 0  400  800  1200  1600  2000  2400  2800  3200  36006.8E-03  0  5.8E-03  200  4.8E-03  400  3.8E-03 2.8E-03  600 Frequency Change Current  800 1000  1.8E-03 8.0E-04 -2.0E-04  Time (seconds)  Figure 4.11 – Effect of gas bubbles on EQCM surface.  The concentration of H2O2 was reduced to 0.01 wt% in an effort to eliminate the gas bubble effect; however, bubbles still formed on the surface.  The  formation of gas bubbles would make it difficult to quantify any mass loss due to ionomer degradation.  4.4  Conclusions  The EQCM is an invaluable tool for analyzing mass and viscosity changes on various electrodes and in various solutions. Using piezoelectric phenomena the EQCM correlates changes in the natural frequency of the quartz crystal to mass and/or viscosity changes as described by the Sauerbrey relationship. The Sauerbrey relationship is valid for rigid films only and when the resistance to motion becomes greater than the inertia of motion, the film exhibits  117  Current (A)  Frequency Change (Hz)  -400  Bubbles purged from electrode surface  viscoelastic behaviour and is no longer considered rigid.  In this case the  dissipation energy must be considered to predict mass changes, using additional models. The EQCM is a powerful technique and has been used by many different industries to study different applications.  Examples such as reaction  mechanisms, adsorption/desorption behaviour, contamination, and degradation are outlined here, and are just a few of the possible applications for this tool. Although using an EQCM to quantify ionomer degradation was unsuccessful in this study, problems were highlighted and potential solutions given for future study. The obstacles that can be solved require a combination of using more sophisticated equipment that can incorporate energy dissipation monitoring to interpret viscoelastic behaviour, and removal of any contamination sources. Problems that require greater attention involve either eliminating bubble formation on the electrode surface during H2O2 decomposition or finding an alternate method of degrading the ionomer film. Improved control of ionomer film deposition on the EQCM electrode would increase film adhesion resulting in more repeatable measurements.  118  4.5 1.  References  H. Jie Thesis; “Technical background, applications and implementation of quartz crystal microbalance systems”, 2006.  2.  K.K. Kanazawa, and J.G. Gordon, Anal. Chem., 57, 1770 (1985).  3.  S. Ye, and D. Blanger, J. Phys. Chem., 100, (1996).  4.  M. Skompska, and AR. Hillman, J. Electroanal. Chem., 433, 127 (1997).  5.  H.W. Chen, R.J. Wu, K.H. Chan, Y.L. Sun, and P.G. Su, Sensors and Actuators B, 104, 80 (2005).  6.  D.A. Buttry, and M.D. Ward, Chem. Rev., 92, 1355 (1992).  7.  www.q-sense.com/application_notes-28.asp (07 Voigt viscoelastic model vs Sauerbrey – QS405-07-2).  8.  M.C. Santos, D.W. Miwa, and S.A.S. Machado, Electrochem. Comm., 2 692 (2000).  9.  F. Gloaguen, J.M. Leger, and C. Lamy, J. Electroanal. Chem., 467, 186 (1999).  10. V.A.T. Dam, and F.A. de Bruijn, J. Electrochem. Soc., 154, B494 (2007). 11. X. Wang, R. Kumar, and D. Myers, Electrochem. Solid State Letters, 9, A225 (2006). 12. J.T. Hinatsu, M. Mizuhata, and H. Takenaka, J. Electrochem. Soc., 141, 6, 1493 (1994). 13. M. Rodahl, F. Hook, A. Krozer, and B. Kasemo, US Patent 6,006,589 (1999).  119  14. M.V. Voinova, M. Rodahl, M. Jonson, and B. Kasemo, Physica Scripta, 59, 391 (1999). 15. MV Voinova, M Jonson and B Kasemo, Biosensors & Bioelectronics 17, 835 (2002). 16. Y. Zhang, B. Du, X. Chen, and H. Ma, Analytical Chemistry, 81 (2), 642 (2008). 17. B.D. Vogt, E.K. Lin, W. Wu, and C.C. White, J. Phys. Chem. B., 108, 12685 (2004). 18. www.q-sense.com/application_notes-28.asp (18 Analyzing humidity effects using QCM-D – QS405-18-1). 19. A. Smith, U.S. Pat. 6,190,035 (2001). 20. A. Smith, R.B. Mulligan, and H.M. Shirazi; http://www.masscal.com/library/VaporSorptionInPolymers.pdf  120  5.0 Concluding Chapter  121  Correlating the structural and performance degradation of the PEMFC helped identify the fingerprints for both carbon corrosion and ionomer degradation failure mechanisms. Carbon corrosion resulted in CCL and carbon sub-layer thinning, which caused platinum agglomeration and altered water management. The reduced platinum surface area reduced the kinetic performance, while the altered water management increased CCL ohmic losses due to oxygen mass transport limitations. Mitigating the voltage degradation due to carbon corrosion can be done by changing the operating conditions or by changing the MEA design to either reduce the corrosion or the corrosion effects. This study showed that operating with sub-saturated reactant gases after significant corrosion had occurred reduced the oxygen mass transport limitations and improved CCL ohmic performance. Owejan et al.1 showed reduced voltage losses by using a more graphitic carbon in the sub-layer. This supports the hypotheses discussed in chapter two, which suggested the corrosion of the sub-layer had a significant role in changing the water management due to the Teflon structure that remained.  It is  recommended to further investigate the impact of the sub-layer structure on the performance loss due to carbon corrosion.  122  Ionomer degradation resulted in membrane thinning, lower membrane conductivity, and CCL structure degradation resulting in reduced CCL ionic conductivity, reduced EPSA, and decreased effective oxygen diffusivity due to changes in CCL water content. Mitigating these effects can also be done by changing the operating conditions or the MEA design. As discussed in the AST development of chapter three, the degradation was dependent on the gas crossover and more importantly, the cell potential. Reducing either of these stressors will reduce either the generation or decomposition of H2O2 and thus reduce the ionomer degradation rate. Reducing the cell potential had a much greater impact unless the gas crossover is decreased significantly to reduce H2O2 production. Hydrocarbon membranes have comparatively little gas crossover, produce much less H2O2, and experience little chemical attack; however, their mechanical stability is quite poor during RH cycle testing2. It has been shown in the present study that reducing the ionomer content in the CCL reduces the performance degradation. While the mechanism remains unclear it has been hypothesized that lower ionomer content reduces the pathway for contaminants diffusing into the membrane. Adding an antioxidant to the membrane has been shown in literature to reduce the membrane degradation by acting as a radical scavenger3,4 and is commonly used in the industry to prolong membrane durability.  123  Although  using  an  EQCM  to  quantify  ionomer  degradation  was  unsuccessful in this study, problems were highlighted and potential solutions given for future study. The obstacles that can be solved require a combination of using more sophisticated equipment that can incorporate energy dissipation monitoring to interpret viscoelastic behaviour, and removal of any contamination sources. Problems that require greater attention involve either eliminating bubble formation on the electrode surface during H2O2 decomposition or finding an alternate method of degrading the ionomer film. Improved control of ionomer film deposition on the EQCM electrode would increase film adhesion resulting in more repeatable measurements. As summarized, the EQCM has been used by others in industry to study other degradation mechanisms, such as platinum dissolution, and is a very powerful tool for studying electrochemical systems.  124  5.1 1.  References  J.E. Owejan, P.T.Yu, and R. Makharia, ECS Transactions,11(1), 1049 (2007).  2.  V.A. Sethuraman, J.W. Weidner, A.T. Haug, and L.V. Protsailo, J. Electrochem. Soc., 155 (2), B119 (2008).  3.  D. Zhao, B.L. Yi, H.M. Zhang, and H.M. Yu, J. Membrane Science, 346, 143 (2010).  4.  L.M. Bonorand, U.S. Pat. App. 218,334 A1 (2007).  125  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0058873/manifest

Comment

Related Items