The Pennsylvania State University. The Graduate School. College of Engineering STUDY OF A METHANOL REFORMING POLYMER ELECTROLYTE FUEL CELL SYSTEM

Transcription

1 The Pennsylvania State University The Graduate School College of Engineering STUDY OF A METHANOL REFORMING POLYMER ELECTROLYTE FUEL CELL SYSTEM A Thesis in Mechanical Engineering by Krishan Kumar Bhatia 2004 Krishan Kumar Bhatia Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2004

2 The thesis of Krishan Kumar Bhatia was reviewed and approved* by the following: Chao-Yang Wang Professor of Mechanical Engineering Thesis Advisor Chair of Committee Matthew M. Mench Assistant Professor of Mechanical Engineering Daniel Haworth Associate Professor of Mechanical Engineering André L. Boehman Associate Professor of Fuel Science Richard C. Benson Professor of Mechanical Engineering Head of the Department of Mechanical Engineering * Signatures are on file in the Graduate School ii

3 ABSTRACT As an alternative to on-board gaseous storage of hydrogen for fuel cell vehicles, a simple liquid hydrocarbon, such as methanol, could be stored and reformed into hydrogen as needed. However, carbon monoxide (CO), a by-product of both the on-board and offboard hydrocarbon reforming processes, is a poison to fuel cell catalysts. In addition, the size of either an on-board hydrogen storage system or hydrocarbon reforming system puts severe packaging constraints on vehicle architecture. This thesis is a comprehensive study of the effects of methanol reformate on the performance of a polymer electrolyte membrane (PEM) fuel cell. While investigating the problem of vehicle architecture constraints, it was found that humidification and pressurization of this fuel cell system can be optimized, and thus make room available on-board for either a methanol reforming/co treatment system or hydrogen storage system. In addition, it was found that methanol reformate, which contains dilute hydrogen and trace quantities of CO, is extremely detrimental to the performance of a PEM fuel cell. Furthermore, it was discovered, both experimentally and theoretically, that the transient process of poisoning is not only a function of CO concentration, but is also highly dependent on the level of hydrogen dilution. After studying the poisoning process, an actual methanol reforming fuel cell system was integrated and tested for overall efficiency. It was found that anode air injection was capable of greatly reducing the poisoning effect. This integrated methanol reforming system was compared to a direct methanol fuel cell system at various power levels. For automotive power applications, cost constraints proved the indirect system superior to the direct methanol system. However, with a well-to-wheel efficiency of 22%, the indirect methanol system was inferior to direct hydrogen fuel cell vehicles. iii

4 TABLE OF CONTENTS LIST OF TABLES.vii LIST OF FIGURES.. viii NOMENCLATURE.xiii ACKNOWLEDGMENTS... xvii CHAPTER 1: INTRODUCTION Background and Motivation The Fuel Cell Engine Hydrocarbon Reforming CO Clean-up...10 CHAPTER 2: PRESSURE AND HUMIDITY EFFECTS ON CELL PERFORMANCE Advantages of Lower Cell Pressure and Humidity Experimental Study of Low Cell Pressure and Humidity Analysis of Bulk Cell Exit Relative Humidity Conclusion..25 CHAPTER 3: TRANSIENT CARBON MONOXIDE POISONING Theoretical Background on CO Poisoning Theoretical Analysis Experimental Analysis Results and Discussion Conclusion.. 47 iv

5 CHAPTER 4: AIR-BLEED REMEDIATION OF CARBON MONOXIDE POISONING Advantages of Anode Air Injection Theoretical Background Experimental Study of Air Injection Results and Discussion Conclusion..69 CHAPTER 5: INTEGRATED METHANOL REFORMING FUEL CELL SYSTEM Integrated System Development Experimental Investigation Results and Discussion Conclusion..88 CHAPTER 6: COMPARISON OF INDIRECT AND DIRECT METHANOL FUEL CELL SYSTEMS Introduction Experimental Conditions and Methods of Comparison Portable Power (~20W) Auxiliary Power (~5kW) Automotive Power (~80kW) CHAPTER 7: CONCLUSION AND FUTURE WORK Conclusion Recommendations for Future Work.131 v

6 REFERENCES 133 APPENDIX A: Experimental Uncertainty. 141 APPENDIX B: Scanning Electron Photographs of Membrane Electrode Assembly APPENDIX C: Calculation Of Fuel Cell Exit Relative Humidity using MAPLE Software APPENDIX D: Transient CO Poisoning Kinetics Model Using MAPLE Software..160 APPENDIX E: Fuel Cell Vehicle Simulation using Matlab/Simulink Software..164 vi

7 LIST OF TABLES Table 2.1. Specifications for the Lynntech Membrane Electrode Assembly 27 Table 2.2. Specifications for the Lynntech Fuel Cell 28 Table 3.1. Constants used for solving transient CO poisoning kinetics model. 48 Table 4.1. Constants used for solving air injection kinetics model...71 Table 4.2. Specifications for the W.L. Gore Membrane Electrode Assembly.. 72 Table 5.1. Reformer Specifications and Operating Conditions. 90 Table 6.1. Operating Conditions for the DMFC and IDMFC.113 Table 6.2. Cost comparison for both systems at various power levels Table 6.3. System efficiency comparisons: Methanol to DC Electric Power Table 6.4. Vehicle specifications for IDMFC automobile Table A.1. Experimental conditions used for uncertainty analysis.144 vii

8 LIST OF FIGURES Figure 1.1. Fuel cell flow plate Figure 1.2. Disassembled fuel cell Figure 1.3. Fully assembled fuel cell Figure 1.4. Typical fuel cell polarization curve.. 17 Figure 2.1. Experimental Set-up..29 Figure 2.2. Cell performance map at 0.6 V versus cell pressure and cathode relative humidity, isometric view.. 30 Figure 2.3. Cell performance map at 0.6 V versus cell pressure and cathode relative humidity Figure 2.4. Cell performance map at 0.4 V versus cell pressure and cathode relative humidity, isometric view..32 Figure 2.5. Cell performance map at 0.4 V versus cell pressure and cathode relative humidity Figure 2.6. Diagram for analysis of cell exit relative humidity.. 34 Figure 2.7. Bulk cell exit relative humidity versus cathode inlet relative humidity...35 Figure 3.1. Steady state cell polarization for 100% H 2 and 40% H 2 anode fed gas...49 Figure 3.2. Cell polarization at various time steps (in minutes) throughout the poisoning process, with 100% H 2, 10ppm CO anode feed...50 Figure 3.3. Cell polarization at various time steps (in minutes) throughout the poisoning process, with 40% H 2, 10ppm CO anode feed Figure 3.4. Cell polarization at various time steps (in minutes) throughout the poisoning process, with 100% H 2, 100ppm CO anode feed viii

9 Figure 3.5. Cell polarization at various time steps (in minutes) throughout the poisoning process with, 40% H 2, 100ppm CO anode feed Figure 3.6. Current at 0.6V during the poisoning process vs. time for all four different anode feed gas compositions. The points represent actual experimental results, and the curves represent simulations based on the model developed.. 54 Figure 3.7. Current at 0.6V during the recovery process vs. time Figure 3.8. Normalized current density at 0.6V verses CO to hydrogen ratio multiplied by dimensionless time Figure 3.9. Computed normalized steady state current at 0.6V vs. CO concentration for various hydrogen dilution levels Figure Computed poisoning time constant vs. CO concentration for various hydrogen dilution levels...58 Figure 4.1. Cell polarization curves at atmospheric pressure with different anode feed gases and 10% anode air injection 73 Figure 4.2. Cell polarization curves at 15 psig with different anode feed gases and 10% anode air injection 74 Figure 4.3. Cell polarization curves at 30 psig with different anode feed gases and 10% anode air injection 75 Figure 4.4. Cell polarization curves at 30 psig with 100ppm CO simulated reformate gas feed and varying anode air injection levels.76 Figure 4.5. Cell power curves at 30 psig with 100ppm CO simulated reformate gas feed and varying anode air injection levels...77 ix

10 Figure 4.6. Cell polarization curves at 30 psig with 600ppm CO simulated reformate gas feed and varying anode air injection levels 78 Figure 4.7. Cell power curves at 30 psig with 600ppm CO simulated reformate gas feed and varying anode air injection levels...79 Figure 4.8. Percent available current (as compared to a pure hydrogen feed) verses air injection levels for two different simulated reformate gas feeds Figure 4.9. Normalized current density at 0.6 V verses CO to H 2 ratio for different air injection levels Figure Theoretically determined normalized current density at 0.6 V verses CO to H 2 ratio for different air injection levels Figure 5.1. Integrated methanol reformer and fuel cell system Figure 5.2. Integrated methanol reformer, fuel cell, and CO remediation system Figure 5.3. Cell polarization curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at atmospheric pressure...93 Figure 5.4. Cell power curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at atmospheric pressure Figure 5.5. Cell polarization curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 15 psig...95 Figure 5.6. Cell power curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 15 psig.. 96 Figure 5.7. Cell polarization curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 30 psig...97 x

11 Figure 5.8. Cell power curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 30 psig.. 98 Figure 5.9. Normalized peak power density verses pressure for pure hydrogen, the reformer feed alone, and reformer feed plus air injection Figure Cell polarization curves, for the reformer feed plus air injection as compared to a pure hydrogen feed, at atmospheric pressure with varying levels of cathode relative humidity Figure Peak cell power, as compared to a pure hydrogen feed, verses cathode relative humidity. This data was taken utilizing a reformer feed plus air injection Figure 6.1. Comparison of the cell polarization curves for the direct (DMFC) and indirect (IDMFC) fuel cell systems Figure 6.2. Comparison of the cell power density curves for the direct (DMFC) and indirect (IDMFC) fuel cell systems Figure 6.3. Historic cost of platinum Figure 6.4. Power density and efficiency (not including parasitic losses) for the indirect methanol fuel cell system (IDMFC) across the entire polarization curve..120 Figure 6.5. Power density and efficiency (not including parasitic losses) for the direct methanol fuel cell system (DMFC) across the entire polarization curve 121 Figure 6.6. FTP-75 Driving Cycle Figure 6.7. Vehicle power requirements during the FTP-75 driving cycle Figure 6.8. Cell voltage during drive cycle Figure 6.9. Cell current density during drive cycle xi

12 Figure Instantaneous well-to-wheel efficiency during drive cycle 126 Figure Cumulative well-to-wheel efficiency during drive cycle Figure Comparison of well-to-wheel efficiencies from various studies Figure A.1. Variation in cell polarization curve for 1000 runs..145 Figure A.2. Typical cell polarization curve showing the +/- 3σ uncertainty range Figure A.3. Variation of 3σ range, as a percentage of average value, with cell current density.147 Figure B.1. Fuel Cell Membrane Catalyst Layer Figure B.2. Fuel Cell Membrane Catalyst Layer, High Magnification Figure B.3. Gas Diffusion Layer 151 Figure B.4. Gas Diffusion Layer, High Magnification.152 Figure B.5. Micro-porous Structure. 153 Figure B.6. Micro-porous Structure, High Magnification Figure B.7. Cross Section of Entire Membrane Electrode Assembly (Gas Diffusion Layer, Micro-porous Structure, Catalyst Layer, and Membrane) Figure E.1. IDMFC Powered Automobile Simulation, Figure E.2. IDMFC Powered Automobile Simulation, Figure E.3. IDMFC Powered Automobile Simulation, xii

13 NOMENCLATURE A cell area [cm 2 ] b fc b fh b fo backwards forward ratio of CO adsorption [atm] backwards forward ratio of hydrogen adsorption [atm] backwards forward ratio of oxygen adsorption [atm] F Faraday s constant, [C mol -1 ] i current [A cm -2 ] i oc cathode exchange current density [A cm -2 ] k ec rate constant of CO electro-oxidation [A cm -2 ] k eh rate constant of hydrogen electro-oxidation [A cm -2 ] k fc rate constant of CO adsorption [A cm -2 atm -1 ] k fh rate constant of hydrogen adsorption [A cm -2 atm -1 ] k fo rate constant of oxygen adsorption [A cm -2 atm -1 ] k oc rate constant of CO chemical oxidation [A cm -2 ] k oh rate constant of hydrogen chemical oxidation [A cm -2 ] n ah2o inlet molar flow rate of water on the anode side [mol/s] n Air inlet molar flow rate of air [mol/s] n ch2o inlet molar flow rate of water on the cathode side [mol/s] n H2 n out inlet molar flow rate of hydrogen [mol/s] total exit molar flow rate [mol/s] n outair exit molar flow rate of air [mol/s] n outh2o exit molar flow rate of water [mol/s] xiii

14 n outh2 exit molar flow rate of hydrogen [mol/s] n prod P P sat moles of water produced in the cell [mol/s] total pressure [atm] saturation pressure of water at cell temperature [atm] R universal gas constant, [J mol -1 K -1 ] RH a RH c inlet relative humidity on the anode side inlet relative humidity on the cathode side RH out exit relative humidity R ohmic cell ohmic resistance [ohm cm 2 ] stoic a stoichiometry on the anode side stoic c stoichiometry on the cathode side t T V cell time [sec] temperature [K] cell voltage [V] V ideal ideal cell voltage [V] V o V max x c x h open circuit voltage [V] maximum reversible cell voltage [V] CO mole fraction hydrogen mole fraction y ah2o inlet mole fraction of water on the anode side y ch2o inlet mole fraction of water on the cathode side y out exit water mole fraction _ g f change in Gibbs free energy for a reaction [J/mol] xiv

15 _ h f enthalpy of reaction [J/mol] Greek Letters α charge transfer coefficient ρ molar area density of catalyst sites times Faraday s Constant [C cm -2 ] η η a η c η max efficiency anode surface overpotential [V] cathode surface overpotential [V] maximum reversible cell efficiency η ohmic ohmic voltage loss [V] θ c θ h θ o fractional surface coverage of CO fractional surface coverage of hydrogen fractional surface coverage of oxygen Acronyms AC ATR CO DC DMFC FCV FTP GC Alternating Current Autothermal Reforming Carbon Monoxide Direct Current Direct Methanol Fuel Cell Fuel Cell Vehicle Federal Test Procedure Gas Chromatograph xv

16 GDL HEV HHV ICE IDMFC MEA OSR PEM POX SR WGS Gas Diffusion Layer Hybrid Electric Vehicle Higher Heating Value Internal Combustion Engine Indirect Methanol Fuel Cell Membrane Electrode Assembly Oxidative Steam Reforming Polymer Electrolyte Membrane Partial Oxidation Steam Reforming Water Gas Shift xvi

17 ACKNOWLEDGMENTS First and foremost, I would like to thank Dr. Chao-Yang Wang for his continual patience and guidance during my graduate education. I would like to thank my family and friends for their constant support of my graduate education, and my committee members, Dr. Daniel Haworth, Dr. Matthew Mench, and Dr. André L. Boehman, for their direction and advice. I would also like to thank the United States Department of Energy, under cooperative agreement # DE-FC26-01NT41098, and the ConocoPhillips Company for sponsoring this work. I am extremely grateful to the National Science Foundation for supporting my graduate education. I am also greatly indebted to the late Dr. Donald Streit for introducing me to the world or automotive engineering, for teaching me to see the bigger picture, and for telling me to smile more often. xvii

18 Reliability is more important than horsepower. - Gary L. Neal xviii

19 CHAPTER 1 INTRODUCTION 1.1 Background and Motivation Several alternatives exist for the next generation of vehicles. The goals for such vehicles are to increase fuel economy, reduce both regulated and greenhouse gas emissions, and lower cost. Two of the most promising technologies are the Hybrid Electric Vehicle (HEV) and the Fuel Cell Vehicle (FCV). HEVs are already in commercial production while the FCV is still in the early prototype stage. HEVs rely on a conventional gasoline or diesel engine powertrain combined in series or parallel with an electric motor and energy storage system. Production examples of HEVs include the Toyota Prius and the Honda Insight. Prototype HEVs include the diesel electric Toyota ES 3 and the Daihatsu UFE, which has demonstrated a fuel economy of 55 km/l (130 mpg) on the Japanese 10/15 city cycle (Yamaguchi, 2001). Alternatively, the FCV relies on the use of hydrogen in an electrochemical fuel cell to generate electricity for an electric motor. The type of fuel cell most often associated with transportation applications is the low temperature Proton Exchange Membrane (PEM) fuel cell (although theoretically possible, the use of a direct methanol fuel cell as the main power source for an automobile has yet to be demonstrated). An example of a prototype FCV is the General Motors AUTOnomy and Hy-wire. While both technologies show great promise, the FCV shows the greatest potential for improved fuel economy and reduced emissions (Rousseau, 2003 and An, 2003). 1

20 However, the FCV faces many challenges, most notably cost and on-board hydrogen storage. As the fuel cell s precious metal loading continues to fall and improved bipolar plate manufacturing techniques evolve, the fuel cell engine s cost is expected to decline. With current technology under mass manufacturing scales, the cost of a fuel cell system is projected at $300/kW, which is six times higher than an equivalent conventional internal combustion engine (Ashley, 2001). However, Department of Energy (DOE) research goals set a target mass production cost of $35/kW by 2010 (U. S. DOE, 2002). The issue of on-board hydrogen storage for FCVs has also been actively pursued. Prototypes using direct hydrogen storage have incorporated high pressure storage tanks at minimum pressures of 25 MPa to contain sufficient fuel on-board. Examples of prototype fuel cell vehicles utilizing pressurized hydrogen storage include the Ford Focus FCV, the Nissan Xterra-FCV, and the Honda FCX-V3 (Ashley, 2001). Alternatively, low temperature cryogenic storage at or near 253 C can be used to contain enough hydrogen to meet a minimum vehicle range of 500 km on a single tank of fuel. Examples of prototype fuel cell vehicles utilizing liquid cryogenic hydrogen storage include the Daimler-Chrysler NECAR 4, the GM HydroGen1, and the Volkswagen Bora HyMotion (Ashley, 2001). While both such storage systems seem feasible, several issues associated with a hydrogen distribution infrastructure and on-board H 2 storage safety remain. The California Air Resource Board has estimated the cost to build a hydrogen infrastructure for 10% market penetration in the U.S. at $100 billion (Berlowitz et al., 2000). These issues have prompted the consideration of on-board hydrogen generation. Current research embraces the possibility of on-board reforming of hydrocarbons into hydrogen, as an alternative to direct hydrogen storage. Because hydrocarbons such as 2

21 gasoline and methanol are liquids at standard conditions, and have high specific energy densities, the problem of storing enough fuel to meet range is easily overcome. However, from a pure thermal well-to-wheel efficiency standpoint, only direct use of hydrogen onboard, from centralized methane reforming, can compete with the present diesel hybrid technology (Stodolsky et al., 1999). With the successful commercialization of HEVs, prototype FCVs must now try to hit a developed, well studied, and rapidly advancing and moving efficiency target (McNicol et al., 2001). Using slightly different predictions for possible improvements in fuel cell technology, Mizsey et al. found the percent increase in well-to-wheel efficiency over conventional ICE engines to be 66%, 40%, 23%, 39% respectively for the hybrid diesel, the on-board storage of hydrogen, for on-board methanol reforming, and for on-board gasoline reforming (Mizsey et al., 2001). Because a direct hydrogen fuel cell system can achieve better system wide efficiencies than the on-board hydrogen reforming FCV, and may achieve superior system wide efficiencies to the HEV, the direct hydrogen system is the consensus goal among industry and government leaders for the future of FCVs (Ashley, 2001). In spite of this, the reforming FCV still offers great promise. Recent breakthroughs in reforming technology could make on-board reforming competitive, on an overall efficiency basis, with the hybrid vehicle or the direct hydrogen vehicle. Secondly, vehicles which can reform gasoline or methanol on-board might provide an excellent transition from conventional gasoline vehicles to fully hydrogen FCVs while the infrastructure for such a vehicle fleet is being developed and built, and while allowing time for hydrogen storage technology to progress. It would also allow the industry to improve its fuel cell technology while simultaneously allowing the consumer time to adjust to the driving 3

22 characteristics of a FCV. For these reasons, on-board reforming technology of liquid hydrocarbons has been of great interest. Regardless of whether reforming is done on-board the vehicle or off-board at a central location, some form of reforming will be the fuel source for these advanced FCVs until carbon-free pathways for hydrogen generation, and effective hydrogen storage techniques, are developed. All hydrocarbon reforming methods, whether it be steam reforming, partial oxidation, or autothermal reforming, generate an effluent that is dilute in hydrogen and contains varying amounts of trace carbon monoxide (CO). Currently, the most widely used method of hydrogen generation is via steam reforming of natural gas (Kirk-Othmer, 1996). The effluent stream from reformers can be purified using a variety of methods (i.e. pressure swing adsorption, catalytic preferential oxidation, membrane separation, etc.), but in order to produce large quantities of economical, fuel cell grade hydrogen, the exact performance and durability effects of dilution and trace carbon monoxide must be known. This thesis is a comprehensive study of the effects of hydrocarbon reformate gas, specifically methanol reformate gas, on the performance of PEM fuel cells. The goals of this study are to determine the minimum purity requirements for fuel cell grade hydrogen, as well as evaluate the impacts that such an on-board reformer system, or offboard reformate gas fed fuel cell engine, will have on vehicle design. In addition to determining the effects of reformate gas on PEM fuel cells, this study was also undertaken to compare the efficiency of an indirect methanol reforming fuel cell (IDMFC) system to that of a direct methanol fuel cell (DMFC). However, before this, a 4

23 short introduction to the fuel cell engine, hydrocarbon reforming, and CO clean-up is essential. 1.2 The Fuel Cell Engine The polymer electrolyte fuel cell engine is a steady flow electrochemical device for converting chemical energy directly into electrical potential energy. The cell consists of anode and cathode plates, which act as flow fields for the fuel and oxidizer respectively (Figure 1.1). The cell flow plates must also act as terminals for connecting the positive and negative sides of the electric load to the cell. Current catalyst technology has limited the possible fuels for polymer electrolyte fuel cells to hydrogen and methanol. Due to the much higher cell efficiencies possible, hydrogen is typically the fuel considered for use in automotive fuel cells. Sandwiched between the anode and cathode flow plates is the membrane electrode assembly (MEA), which is where the fuel cell s core technology resides. Figure 1.2 shows a disassembled fuel cell, with the MEA between the flow plates, while Figure 1.3 shows a fully assembled cell. The MEA is comprised of a polymer electrolyte material for transporting hydrogen ions from the anode to cathode, with a platinum catalyst layer and gas diffusion layer (GDL) on either side. Equations 1.1 and 1.2 give the anode and cathode reactions respectively. H H + e [1.1] + 1 2H + 2e + O2 H 2O 2 [1.2] The hydrogen ions are transported through the polymer electrolyte from anode to cathode, while the electrons flow through the load circuit. As Equation 1.2 indicates, the 5

24 cathode exhaust from a hydrogen fed fuel cell is simply water. In practice, several individual fuel cells are stacked together in series to form a fuel cell engine. Similar to batteries in series, the voltage of the entire string of cells is the sum of the individual cell voltages. Also like batteries, for increased current requirements, the fuel cell s active area is increased. Thus, the number of cells and active area of any particular fuel cell engine varies based on the system s voltage and current requirements. The maximum efficiency of a fuel cell, given by Equation 1.3, is the ratio of the maximum reversible cell voltage to the ideal cell voltage. The actual efficiency of a fuel cell, given by Equation 1.4, is defined by the ratio of measured cell voltage to the ideal voltage. However, the value given by Equation 1.4 does not include other losses in the overall system such as fuel utilization and parasitic loses from ancillary equipment necessary for cell operation (i.e. compressors, fans, humidifiers, etc.). Equations 1.5 and 1.6 give the ideal voltage, V ideal, and the maximum reversible open circuit voltage, V max, respectively. These are defined in terms of the enthalpy of reaction, Gibbs free energy of reaction, and Faraday s constant (Larminie et al., 2000). For a hydrogen fuel cell, the ideal cell voltage, V ideal, is 1.48 V based on the higher heating value (HHV) of hydrogen. Thus, a fuel cell engine operating at 0.75 V per cell is roughly 50% efficient. V max max = V ideal η [1.3] c η = [1.4] V V ideal V ideal h f = [1.5] 2F 6

25 V max g f = [1.6] 2F Because the efficiency of a fuel cell is directly related to the cell voltage, the performance of a fuel cell is often characterized by its ability to supply current at a specific voltage. This current-voltage relationship is known as a polarization curve. A typical fuel cell polarization curve is shown in Figure 1.4. The polarization curve relates the voltage of an individual fuel cell to the available current draw per unit of active cell area. The initial drop in cell voltage is attributed to overcoming the activation potential of the anode and cathode reactions. This region of the polarization curve is known as the kinetic controlled regime. The linear drop in the second region of the curve is attributed to the ohmic losses incurred through the cell as an increasing number of hydrogen ions travel from anode to cathode through the resistance of the electrolyte. This region is known as the ohmic controlled regime. The final drop in the polarization curve is attributed to the lack of sufficient reactant supply to the cell reaction sites. This final region is known as the mass-transport controlled regime. Comprehension of this polarization curve is critical in understanding the performance of any fuel cell. 1.3 Hydrocarbon Reforming The hydrogen generation step of any fuel processing system is referred to as reforming. Currently, three main methods exist for on-board reforming: steam reforming, partial oxidation, and a combination of the two, commonly known as autothermal reforming. By far, the largest portion of all hydrogen generated industrially in the U.S. is 7

26 via steam reforming (SR) of natural gas (Kirk-Othmer, 1996). The endothermic steam reforming reaction, given by Trimm et al. (2001), is shown in Equation 1.7. CH o kj + H 2O CO + 3H 2, H 298K [1.7] mol 4 = Simultaneously in the same reactor, the water gas shift (WGS) mechanism is used: o kj CO + H 2O CO2 + H 2, H 298K = [1.8] mol This reaction, which not only generates more hydrogen, also helps to rid the feed gas of CO. To fuel the endothermic reaction, energy is typically supplied via external combustion of fuel gas or oil. Industrially, the reforming process has a thermal efficiency of approximately 78.5% (Kirk-Othmer, 1996). While similar to the industrial process, on-board SR has many differences. Equation 1.9 gives the steam reforming decomposition reaction for methanol (Kordesch et al., 1996). o kj CH 3 OH CO + 2H 2, H 298K = [1.9] mol As with methane reforming, the WGS reaction is simultaneously occurring, to give an overall reaction of: CH OH + H H 3 2 o 298K = O CO kj mol 2 + 3H 2, [1.10] The endothermic nature of SR requires burning 15-25% of the feed fuel. Despite this drawback, the ease of methanol SR at low temperatures makes it an excellent candidate to provide hydrogen either on-board or off-board a vehicle. In fact, for methanol, SR is the leading contender (Berlowitz, 2000). This method has been demonstrated numerous 8

27 times, both in the lab and in functional vehicles. Prototype vehicles using on-board methanol SR include the Daimler-Chrysler NECAR 5 and Jeep Commander 2 and the Honda FCX-V2. Reforming efficiencies as high as 70% have been reported for these implemented, in-vehicle reformers (Tamura, 2000). The other basic method for reforming a hydrocarbon into hydrogen is known as partial oxidation reforming (POX). In this method, the fuel is introduced into the reactor with a controlled amount of oxygen. POX reactors generally have faster start-up times than SR reactors (Berlowitz, 2000). The high temperatures inside the reactor generated by the POX reaction result in thermal decomposition of the hydrocarbon. Equation 1.11 gives the expression for the POX reaction of a general hydrocarbon C n H m. C 1 m + H 2 [1.11] 2 2 n H m no2 nco + If water is injected into the reactor vessel, then, like steam reforming, a large portion of the carbon monoxide is converted to hydrogen via the WGS mechanism. POX has been demonstrated on the lab to the industrial scale, with hydrocarbon fuels ranging from methane to fuel oil. The overall POX/WGS reaction for methanol is given by Equation o kj CH 3OH + O2 2H 2 + CO2, H 298K = 155 [1.12] 2 mol Huang (1986) found that although SR yields higher concentrations of H 2 in the feed gas, POX is superior to SR in terms of moles of hydrogen that can be produced per hour per kilogram of catalyst. Reforming at temperatures as low as 200 C, with 80% hydrogen conversion, has been achieved over electrically activated indium/tin oxide catalysts (Miremadi, 2000). 9

28 Autothermal reforming (ATR) is a combination of SR and POX. In SR, the endothermic reaction must be fueled by combustion of part of the fuel external to the reactor. In POX reforming, the exothermic oxidation of a portion of the reactants inside the reactor is used to fuel the thermal decomposition of the rest. In ATR, a portion of the fuel is partially oxidized inside the reactor, but the heat generated is used to fuel the endothermic steam reforming reaction. Inside the reactor, a balance is struck so that just enough fuel is oxidized to satisfy the heat requirements of the SR reaction. When both POX and SR are combined, but an exact heat balance is not struck, the process is often referred to as oxidative steam reforming (OSR). In practice, ATR is very similar to POX, except that extra steam is injected into the reactor to initiate the SR process. ATR produces effluent more concentrated in hydrogen than POX alone, but not as concentrated as straight SR reactors. Thermodynamic calculations have shown that ATR theoretically yields higher reforming efficiencies than POX (Docter, 1999). Because of its complexity, ATR is usually reserved for hydrocarbons that are difficult to reform, like gasoline. However, ATR has been applied to methanol over copper catalysts. The ATR reaction (Equation 1.13) is utilized to balance overall heat at 300 C (Geissler, 2001): 1 4CH 3OH + O2 + 3H 2O 4CO2 + 11H 2 [1.13] 2 Successful application of ATR technology to methanol has been demonstrated by the Johnson Matthey HotSpot reformer (Carpenter, 1999). 1.4 CO Clean-up 10

29 Despite the presence of the WGS reaction inside any methanol or gasoline reformer, carbon monoxide levels in the effluent are usually far too high to be fed directly into a PEM fuel cell. The PEM fuel cell s catalysts have been found to be extremely sensitive to deactivation by carbon monoxide adsorption at low temperatures (Trimm, 2001). Thus, to avoid poisoning the fuel cell s catalyst, separate unit operations to remove CO must be incorporated into a reforming system. These methods include the use of an additional WGS reactor, preferential oxidation of the CO (PROX), and membrane separation. Other methods of CO cleanup (i.e. pressure swing adsorption) are used industrially, but are extremely expensive and require far too much space to be implemented at the in-vehicle level. Due to equilibrium constraints, additional WGS reactors can only lower the CO concentration to levels near 1% (10,000 ppm). However, PROX can lower the CO concentration to much lower levels (Hagan, 2000 and Qi, 2000). PROX, as the name suggests, involves the preferential oxidation of CO to CO 2. The term preferential is used because the goal is to oxidize CO only. The desired reaction is given by Equation CO + O2 CO2 [1.14] 2 The undesired reactions include the oxidation of hydrogen, the methanization reactions of CO and CO 2, and the reverse WGS reaction. Methanol SR, if controlled appropriately, can reduce the CO levels to 1% in the effluent from the reformer. Thus, a PROX reactor can be used directly after a methanol SR reactor to reduce the concentration to below 10 ppm (Dudfield, 2001). Presently, PROX is considered the leading technology for the final feed gas cleanup of CO to acceptable levels. 11

30 The method of membrane separation essentially involves the use of hydrogenselective membranes to clean the effluent gas. These membranes are typically manufactured from palladium-based alloys, which have sufficiently low permeability to only allow hydrogen through (Han, 2000). The flux of hydrogen across the membrane is proportional to the pressure differential across it. This method has the advantage of producing exceptionally pure hydrogen (>99.99%), with undetectable amounts of CO. Thus, using membrane separation, the fuel cell is fed pure H 2, which means an increase in cell performance. This is as opposed to other methods which do remove CO, but still feed the fuel cell hydrogen gas diluted with CO 2, N 2 from the air, etc. However, hydrogen membrane separation is a relatively new technology, requires high efficiency compressors to create the needed pressure differential, and costs a great deal due to the palladium which must be used to achieve the desired permeability. Lastly, as with all membrane separation processes, a certain amount of the hydrogen is lost during the purge of the positive pressure side of the membrane. This typically results in a loss of 25% of all the hydrogen produced (Han, 2000). However, this bleed gas can be combusted to provide the heat for a reforming reaction. With this background understanding of fuel cells, reforming, and CO clean-up, the technical challenges of operating a fuel cell on reformate gas can be better appreciated. Any on-board reforming system or hydrogen storage system will not only occupy space, but will add to the vehicle s weight and complexity. Thus, optimization of the fuel cell engine itself is critical to system wide harmony. In an automobile, the compressors and humidifiers associated with maintaining cell pressure and humidity can get quite large and complex. To this end, Chapter 2 examines how low cell pressure and 12

31 humidity operation affect performance. Optimization of these operations is critical in meeting the space and weight budgets that an on-board hydrocarbon reforming FCV or hydrogen storing FCV would require. Chapter 3 then goes on to examine the exact effects than hydrocarbon reformate gas have on cell performance. Specifically, the effects of hydrogen dilution combined with trace quantities of CO will be studied. Chapter 4 presents a method for remediation of a CO poisoned catalyst layer, known as air bleed. The usefulness of this method, specifically for dealing with reformate gas, will be analyzed. Chapter 5 looks at the performance in an integrated methanol reforming PEM fuel cell system. Using the performance data from Chapter 6, Chapter 7 analyzes the system wide efficiencies associated with the reforming of methanol, and compares this with a DMFC system. 13

32 Figure 1.1. Fuel cell flow plate. 14

33 Flow Plates Figure 1.2. Disassembled fuel cell. MEA 15

34 Figure 1.3. Fully assembled fuel cell. 16

35 Figure 1.4. Typical fuel cell polarization curve. 17

36 CHAPTER 2 PRESSURE AND HUMIDITY EFFECTS ON CELL PERFORMANCE 2.1 Advantages of Lower Cell Pressure and Humidity As noted in Chapter 1, an on-board hydrocarbon reforming FCV or hydrogen storing FCV would certainly entail a complex system, and thus pose packaging challenges, as well as contributing to vehicle weight. One possible area for improving the overall vehicle packaging and weight is to focus on the fuel cell engine itself, and not just the energy storage system. One area of improvement in the fuel cell engine, with the potential for large gains, is through cell pressurization and humidification optimization. On-board compressors and humidifiers not only take up space and weight, but they also contribute to a fuel cell engine s parasitic loses, and thus lower overall system efficiency. The year 2010 DOE technical targets for a fuel cell engine s specific volume and weight are 550 W/liter and 550 W/kg respectively. However, current fuel cell technology can only achieve a specific volume and weight of 200 W/liter and 200 W/kg (U. S. DOE, 2002). Higher pressure is advantageous to a cell as it increases the relative humidity within the cell and increases reactant concentrations in the anode and cathode flow channels. Increased reactant concentration translates to higher open circuit potentials and reduced kinetic losses. High cell relative humidity is critical in maintaining the membrane ionic conductivity (Zawodzinski, 1993 and Anantaraman, 1996). This in turn greatly reduces ohmic losses within the cell and helps to maintain membrane durability. 18

37 Thus, both high pressure and cell humidification are important to increased cell performance. However, much work has been done towards the possibility of running a fuel cell at low pressures, and with little or zero external humidification. This possibility translates to a fuel cell engine needing much smaller compressors and humidifiers. For example, by using proprietary technology that enabled them to eliminate the need for external humidification in their HydroGen 1, General Motors was able to remove 10 different water management devices from the vehicle s fuel cell engine system (Carney, 2001). From a system wide fuel cell engine perspective, a trade off exists between higher cell pressures and relative humidifies and vehicle space and weight constraints. Typical vehicle fuel cell engines operate between 1 and 3 bars above atmospheric pressure (Adams, 2000, Ashley, 2001, and Fronk, 2000). Anantaraman et al. showed the conductivity of Nafion TM 117 polymer electrolyte membranes to be a strong function of relative humidity. Ge et al. showed that although cell output is reduced, internally humidified fuel cell operation is possible with the anode and cathode feed stream running in a counter flow pattern (Ge, 2003). Similar findings have also been published suggesting that although performance is reduced, complete elimination of cathode humidification is possible (Natarajan, 2003). Staschewski et al. studied cell performance under atmospheric pressure and complete internal cell humidification (Staschewski, 1999). However, a large performance drop was noted for this type of operation. The effect of relative humidity in lower temperature (i.e. < 40 C) PEM fuel cells operating at atmospheric pressure has also been studied (Chu, 1999 and Jiang, 2001). This past work suggests that combined low pressure and humidity operation is possible for PEM fuel cells. However, several previous studies fix either pressure or 19

38 relative humidity, and vary the other. Consequently, the optimal pressure and humidity combination is difficult to ascertain. In addition, the majority of previous investigations studied the cell performance under constant flow rate conditions, and did not vary the flow rate with the instantaneous current draw. Constant flow rate operation results in low overall efficiency because fuel utilization is low when the cell is operated at any point away from the cell s peak power. Thus, although constant flow rate is easier to control, variable flow rate operation, known as constant stoichiometry, is the industrial standard for how fuel cells are operated. For these reasons, a parametric investigation of fuel cell performance at various pressure and humidity levels, with the cell operating in a constant stoichiometry fashion, is presented here. After the parametric study, a theoretical analysis of bulk cell exit relative humidity, and how it relates to cell performance, is presented. 2.2 Experimental Study of Low Cell Pressure and Humidity The experimental tests were performed in a 50 cm 2 titanium fuel cell fixture supplied by Lynntech Industries, Ltd. (College Station, TX). Both anode and cathode flow fields consisted of 6 parallel channels following a serpentine path to cover the 50 cm 2 of active area. The cell flow plates used are shown in Figure 1.1. The plates were manufactured from titanium, and coated with a platinum layer to promote durability and surface electrical conductivity. The cell was arranged to have the anode and cathode run in a counter flow manner. The membrane electrode assembles (MEAs), also supplied by Lynntech Industries, were composed of 40 wt.% platinum on carbon, Nafion 112, with a 20

39 platinum catalyst loading of 0.5 mg/cm 2 on each side. The MEA also had an ELAT carbon cloth GDL pressed on the catalyst surface. The specifications for the MEA and fuel cell are given in Tables 2.1 and 2.2 respectively. The polarization measurements were taken using an Arbin Instruments (College Station, TX) fuel cell test station and electronic load bank. A diagram of the experimental set-up is shown in Figure 2.1. As mentioned, the flow rates of both streams were continually modified with respect to the instantaneous current draw to maintain a molar flow level corresponding to a constant stoichiometry condition of 1.5 and 2.5 on the anode and cathode respectively. Thus, for any current draw level (i.e. electron generation rate), 150% of the hydrogen required to complete the reaction given by Equation 1.1 is supplied to the cell. Similarly, 250% of the air required to complete Equation 1.2, for any given reaction rate, is provided to the cell. The cell itself was kept at 80 C. According to Equation 1.2, water is generated on the cell s cathode side. Because of this internal water generation, the cathode side has potential for reduction or elimination of external humidification. So, for this study, the anode was kept fully saturated and only the cathode side was investigated for humidification optimization. The cathode relative humidity was varied from 0 to 100% at 5 distinct levels (0, 25, 50, 75, and 100%). The cell pressure, which was kept equal on both sides, was also varied at 5 distinct levels above atmospheric pressure (0, 10, 20, 30, and 45 psig). Thus, with 5 humidity levels and 5 pressure levels, a total of 25 experiments were required to complete the parametric study. During each experiment, the cell was allowed to reach a steady state condition, and the cell polarization was measured. 21

40 Comparing 25 polarization curves taken under different conditions, and trying to ascertain overall trends in performance, can be a quite complex task. In order to simplify the data and analysis, it was decided to plot the cell current density at specific voltages instead of showing the entire polarization curve. Thus, the effects of pressure and cathode relative humidity can be understood more easily. Figure 2.2 displays the current available, at a cell voltage of 0.6V, versus cell pressure and cathode relative humidity. The surface was obtained using a cubic interpolation function between the 25 experimental data points (one data point from each polarization curve). Figure 2.2 displays the surface in an isometric view, while Figure 2.3 shows an overhead view of the surface. Figures 2.4 and 2.5 are similar to the first two, but they show the current obtainable at a cell potential of 0.4V. The isometric views of the current obtainable at 0.6V and 0.4V (Figures 2.2 and 2.4) clearly show the performance trend as a function of cell pressure and cathode relative humidity. As shown, the current is strong function of cell pressure, while displaying only a mild dependence on cathode humidity. This mild dependence on cathode relative humidity indicates that anode gas humidification and water generation on the cathode side are sufficient to maintain the membrane at an adequate ionic conductivity level. As a result of sufficient internal humidification of the cathode feed stream, at 45 psig the cell performance is reduced by only 10% with completely dry cathode feed gas as compared to the maximum performance obtainable with a wet cathode fed gas. At 45 psig, the maximum performance is obtainable with a cathode inlet relative humidity in the 50-75% range, indicating that full saturation of the cathode feed gas results in an excess of water within the cell, and leads to some degree cell flooding. 22

41 Also, as shown, only at atmospheric pressure operation does a dry cathode feed result in a significant drop in cell current. The trends observable in Figures 2.2 through 2.5 also show that while there exists a strong dependence of current on cell pressure, the dependence is not linear. A sharp rise in performance is seen as pressure is increased from atmospheric to roughly 20 psig. After that, increasing cell pressure results in only marginal gains in available current. For example, with a dry cathode feed, the current at 0.4V and atmospheric pressure is only 53% of the current at 45 psig. However, at 20 psig, 85% of the current at 45 psig is available. These findings demonstrate that the potential exists for low cell pressure operation (i.e. 20 psig) and complete elimination of the cathode humidifier, while only suffering a marginal loss in cell performance. 2.3 Analysis of Bulk Cell Exit Relative Humidity As found experimentally in Section 2.2, cathode feed gas relative humidity has little impact on overall cell performance, except at very low cell pressures. This finding suggests that water generation on the cathode side and electro-osmotic drag of water from the anode to cathode is sufficient in keeping the membrane at a fully saturated state, and thus maintaining its ionic conductivity. This finding can be backed up with a simple bulk model of water balance in a fuel cell based on the inlet gas conditions and water generation rate. The result yields an approximation for the average relative humidity within the cell. 23

42 Figure 2.6 shows a simple model of water balance in a fuel cell. The molar flow rate of the inlet feed gases on the anode and cathode side, n a and n c, along with the inlet relative humidities, is specified. The molar flow rates are calculated knowing the cell current draw and respective stoichiometries on both sides. The production of water within the cell is also known from the current draw. Thus, knowing the amount of water coming into the cell and the water generated within the cell, an overall water balance can be done to determine the exit amount of water. The exit molar flow rate of gaseous hydrogen and air can also be determined knowing the inlet flow rates and the consumption of the gases within the cell due to the electrochemical reaction. Knowing both the exit molar flow rate of water and gases, the exit relative humidity can be determined. The details of this calculation are given in Appendix C. This analysis assumes the fuel cell to behave as a continuously stirred reactor, and the membrane to be sufficiently thin that water transport across it ensures the relative humidities at the exit on both the anode and cathode sides to be equal. The simulation was conducted under the same operating conditions as the experimental tests (i.e. temperature maintained at 80 C, fully humidified anode, anode stoichiometry of 1.5, and cathode stoichiometry of 2.5). Figure 2.7 shows the bulk cell relative humidity as a function of cathode inlet relative humidity. To maintain ionic conductivity of the MEA, the cell must maintain a bulk relative humidity of at least unity. A relative humidity greater than unity means that liquid water is condensing within the cell, and must be entrained with the gas flow to exit the cell. As shown in Figure 2.7, at 3 and 4 atmospheres absolute, an absolutely dry cathode feed still ensures full bulk humidification of the cell. In fact, it was calculated that a cell pressure of

43 atmospheres absolute (23.25 psig) and a completely dry cathode inlet will result in a bulk cell humidity of exactly unity. This supports the experimental results (Figure 2.5), which show very little dependence of performance on cathode inlet relative humidity above cell pressures of 30 psig (3 atm absolute). The water balance model was also used to calculate the minimum pressure required to have complete internal humidification of the cell (i.e. both anode and cathode feed streams are dry, and the cell relies on water generation for humidification). At 80 C, with an anode and cathode stoichiometry of 1.5 and 2.5 respectively, complete internal humidification is possible with a cell pressure of 3.25 atmospheres absolute (33 psig). This of course assumes that the fuel cell behaves as a well-stirred reactor. Although the well-stirred reactor assumption is not true in practice, novel cell flow field designs and ultra-thin MEAs can make the possibility of completely eliminating external humidification, on both anode and cathode sides, a reality. 2.4 Conclusion A parametric study of fuel cell performance with low pressure and low cathode humidity has been conducted. It was found that while performance is affected by both parameters, the current available is a steeper function of cell pressure, and only a slight function of cathode inlet relative humidity. This finding was confirmed by performing a simple bulk water balance analysis on the cell. The analysis showed that full saturation of the cell is possible with moderate pressures, a fully humidified anode feed, and a dry cathode feed. Saturation of the cell, and therefore the MEA, is critical in maintaining its ionic conductivity, and thus minimizing ohmic losses. The aforementioned data and simple humidity analysis is an excellent method for comparing cell performance under 25

44 various pressure and inlet humidity operating conditions. Also, it allows the designer of FCV architecture to make trade-off decisions between maximum fuel cell engine power and the size/weight savings associated with compressor and humidifier elimination and/or downsizing. This type of fuel cell engine component optimization or elimination is critical in meeting the space and weight constraints of an FCV. Those constraints on the fuel cell engine become even more demanding and stringent if an on-board hydrocarbon reforming system is used. The effects of reformate gas on the fuel cell engine are analyzed in Chapter 3. 26

45 Table 2.1. Specifications for the Lynntech Membrane Electrode Assembly Active Area 50 cm 2 Membrane Type Dupont Nafion 112 Membrane Thickness Equivalent Weight 2 mil (50.8 µm) 1100 g/equivalent Anode Catalyst Loading 0.5 mg Pt per cm 2 Cathode Catalyst Loading 0.5 mg Pt per cm 2 Catalysts Support Gas Diffusion Layer 40% wt. Pt on Carbon ELAT hydrophobic carbon cloth 27

46 Table 2.2. Specifications for the Lynntech Fuel Cell. Endplate Material Endplate Coating Titanium Platinum Active Area 50 cm 2 Channel Width Channel Depth Rib Width 1.5 mm 1.5 mm 0.9 mm Number of Parallel Channels 6 Number of Passes 5 Heating Inlet/Outlet Connections Electric via 0.2 Cartridge Heaters 1/8 Female NPT Number of Bolts 8 Bolt Size 1/4 20 Bolt Torque 35 in-lbs 28

47 Figure 2.1. Experimental Set-up. 29

48 Figure 2.2. Cell performance map at 0.6 V versus cell pressure and cathode relative humidity, isometric view. 30

49 Figure 2.3. Cell performance map at 0.6 V versus cell pressure and cathode relative humidity. 31

50 Figure 2.4. Cell performance map at 0.4 V versus cell pressure and cathode relative humidity, isometric view. 32

51 Figure 2.5. Cell performance map at 0.4 V versus cell pressure and cathode relative humidity. 33

52 Figure 2.6. Diagram for analysis of cell exit relative humidity. 34

53 P=4 atm Bulk Cell Exit Relative Humidity P=3 atm P=2 atm P=1 atm Cathode Inlet Relative Humidity Figure 2.7. Bulk cell exit relative humidity versus cathode inlet relative humidity. 35

54 CHAPTER 3 TRANSIENT CARBON MONOXIDE POISONING 3.1 Theoretical Background on CO Poisoning As noted earlier, all hydrocarbon reforming methods, whether it be steam reforming, partial oxidation, or autothermal reforming, generate an effluent which is dilute in hydrogen and contains varying amounts of trace carbon monoxide (CO). Depending on the hydrocarbon feedstock, reformer effluent can contain hydrogen as dilute as 32% (Brown, 2001). The effect of dilution with a CO-free feed gas has been studied, and is widely understood at a fundamental level (Um, 2000). The decrease in cell polarization at increased current draw is attributed to the reduction in average hydrogen concentration in the anode flow field, thus requiring a larger anode overpotential to maintain the hydrogen oxidation reaction at a specified current density. Fortunately, due to the anode normally exhibiting very fast kinetics, the inlet hydrogen concentration has to be very dilute to have an appreciable effect on cell polarization. Carbon monoxide (CO), however, is a known fuel cell catalyst poison even in trace amounts, and is preferentially adsorbed on the catalyst surface. CO concentrations as low as 10 ppm are extremely detrimental to fuel cell performance. Gottesfeld et al. (1988) conducted some of the earliest work on CO poisoning of fuel cells. In that work, the poisoning phenomenon was documented with CO levels varying from 10 to 100 ppm. A catalyst layer remediation technique using oxygen injection into the anode feed stream, called air bleeding, was also proposed (Gottesfeld, 36

55 1988). Springer et al. (2001) have developed a kinetic model for hydrogen and CO adsorption and subsequent electro-oxidation. This model was then solved under steady state conditions for the fractional surface coverage of hydrogen and CO, as well as the cell current. It was calculated that under conditions of CO-free feed gas, the performance loss should not exceed 10% of full stack power with hydrogen concentrations as low as 40% (Springer, 2001). However, in the presence of CO levels as low as 10 ppm, the losses start to become significant, and were calculated to be exaggerated even further under the combined conditions of anode feed gas having a low hydrogen concentration and trace amounts of CO. This is in agreement with findings by Divisek et al., who showed that with a 75% hydrogen, 25% CO 2, and 100 ppm CO fuel feed, the steady state cell performance was lower than that with a 100% hydrogen, 100 ppm CO feed (Divisek, 1998). The model developed by Springer et al. has also been extended and modified several times by other researchers (Chan, 2003 and Zhang, 2002). The transient process of poisoning in a hydrogen/oxygen fuel cell has also been experimentally studied. Oetjen et al. (1996) found that for Pt catalysts and feed gas containing 100 ppm CO, performance degradation was observable even after 5 minutes of exposure, with the cell reaching the fully steady state poisoned condition after roughly 2 hours (Oetjen, 1996). However, the feed consisted of 100% hydrogen with trace CO. Moreover, oxygen was used on the anode instead of air. Murthy et al. (2001) found that a small amount of air injection in the anode feed stream can significantly reduce the transient decay rate of fuel cell performance during the poisoning process. Aside from cell polarization measurements, other methods, such as electrochemical impedance spectroscopy, have been used to study the CO poisoning process (Wagner, 2003). 37

56 This past work shows that while much is known about steady state poisoning with a 100% hydrogen feed containing trace amounts of CO, there is a lack of experimental data to quantify the transient CO poisoning process with diluted hydrogen, which is the actual case for fuel cells being fed a hydrocarbon reformate gas. In addition, even less is known fundamentally about the transient process of fuel cell poisoning with reformate gas. Understanding this process is critical in determining the minimum purity requirements for anode feed gas as well as developing any sort of poisoning remediation method. The transient CO poisoning work presented below has also been outlined in the professional literature (Bhatia et al., 2004). 3.2 Theoretical Analysis Springer et al. (2001) present a set of reactions (Equations ) to describe the adsorption, desorption, and electro-oxidation of hydrogen and CO on the catalyst surface, where M represents a free catalyst site. These assume that any inert species, such as nitrogen or carbon dioxide, which may be diluting the anode feed stream, do not participate in the surface adsorption chemistry. H 2 H 2 k fh 2M 2 + b k fh fh + 2M 2 ( M H ) ( M H ) [3.1] [3.2] CO k fc M M CO [3.3] + 38

57 b k fc fc CO + M M CO [3.4] k [3.5] ( ) eh + M H H + e + M k H O ( M CO) ec M + CO + 2H + e [3.6] From this, a set of kinetic equations describing the rate of change of hydrogen and CO coverage on the catalyst surface in terms of the rates of adsorption, desorption, and electro-oxidation can be written. Springer et al. (2001) drop the rate of change of surface coverage with time in order to find the steady state cell polarization. However, here we are interested in the transient cell behavior, and thus include these terms in our calculations. These kinetic equations developed by Springer et al. (2001) have been modified to assume that the adsorption and desorption of species are first order in nature. Also, the charge transfer coefficient, α, is assumed to be equal to 0.5, which reduces the general Butler-Volmer equation for electro-oxidation into a hyperbolic sine relationship. dθ ρ h dt = k fh x h P 1 [3.7] ( ) η θ h θc b fh k fh θ h 2k eh θ sinh a h RT αf dθ ρ c dt = k fc x c P 1 [3.8] ( ) η θ h θc b fc k fc θc 2k a ec θ sinh c RT αf Equations 3.7 and 3.8 balance the rate of change of hydrogen and CO fractional surface coverage, θ h and θ c, with respect to time against the respective rates of adsorption, desorption, and electro-oxidation from the catalyst surface. The terms ρ, x h, x c, and η a 39

58 represent the molar area density of catalyst sites times Faraday s Constant, hydrogen mole fraction, CO mole fraction, and anode overpotential respectively. The hydrogen electro-oxidation term was then re-written in terms of hydrogen current, i. The CO electro-oxidation term was dropped due to its relatively small magnitude at the cell voltages considered here as compared to the CO adsorption and desorption terms. An additional equation for cell voltage in terms of current was also needed to close the system. This zero-dimensional, lumped model for cell voltage assumes Tafel kinetics on the cathode and linear ohmic losses through the membrane electrode. This results in a simplified set of equations (Equations ): ( θ θ ) b k i dθ ρ h = k fh x h P 1 h c dt fh fh θ h [3.9] dθ ρ c dt V cell ( θ θc ) b k θc = k fc x c P 1 h fc fc [3.10] = V η η η [3.11] o a c ohmic Where the terms in Equations 3.11 are given by Equations η a RT 1 i = sinh αf 2k θ [3.12] eh h RT i η = ln c αf i [3.13] oc η ohmic = ir [3.14] ohmic 40

59 Equations can be solved numerically for the time variation of fractional surface coverage of hydrogen, CO, and the cell current at a constant cell voltage. This theoretical model will be used later to compare and explain the experimental results. Values for the constants in Equations are given in Table 3.1. Values for pressure, temperature, and cell voltage represent the actual operating conditions that the experiments were conducted under. Values for the ohmic resistance, R ohmic, and cathode exchange current density, i oc, were found from curve fits to baseline cell performance data. Values for all kinetic parameters were borrowed from those used previously in the literature (Springer et al., 2001), except for b fc and k fh. Springer et al. (2001) note that these two kinetic parameters are functions of the fractional CO coverage; however, for the purposes of this study, they are assumed to be constant. The values for these two parameters were chosen to best match the transient cell performance data. These numeric values for these two parameters fall within the range of variability that has been previously reported by Springer et al. 3.3 Experimental Analysis Like the experiments described in Chapter 2, the experiments here were performed in a 50 cm 2 titanium fuel cell fixture supplied by Lynntech Industries, Ltd. (College Station, TX). Both anode and cathode flow fields consisted of 6 parallel channels following a serpentine path to cover the 50 cm 2 of active area. The membrane electrode assembles (MEAs), also supplied by Lynntech Industries, were composed of 40 wt.% platinum on carbon, Nafion 112, with a platinum catalyst loading of 0.5 mg/cm 2 41

60 on each side. The MEA also had an ELAT GDL pressed on top of the catalyst surface. Specifications for this MEA are given in Table 2.1. The polarization measurements were taken using an Arbin Instruments (College Station, TX) fuel cell test station and electronic load bank. A diagram of the experimental set up is shown in Figure 2.1. Both anode and cathode feed streams were fully saturated with water at 80 C, and maintained at that temperature while being fed to the cell. All values for the hydrogen dilution in the anode feed streams are based on the dry gas condition before saturation with water at 80 C. The flow rates of both streams were continually modified with respect to the instantaneous current draw to maintain a molar flow level corresponding to a constant stoichiometry condition of 1.5 and 2.5 on the anode and cathode respectively. The cell itself was kept at 80 C and pressurized to 30 psig on both anode and cathode sides. Experiments were performed under a variety of hydrogen dilution levels and CO concentrations. During these tests, the cell polarization was measured at specific times throughout the CO poisoning process. Between polarization scans, the cell was maintained at a current draw corresponding to a constant cell potential of 600 mv. 3.4 Results and Discussion The 50 cm 2 fuel cell was tested under a variety of conditions to simulate a wide range of actual reformate gases. First, the cell was tested with CO-free feed gas consisting of 100% hydrogen and then 40% hydrogen, balanced with nitrogen (Figure 3.1). Both COfree cases do not exhibit any transient performance loss, and thus only the steady state 42

61 cell polarization is shown. As shown, almost no performance variation is observable even with the low hydrogen concentration feed gas. Then, the cases of anode feed gas having varying levels of CO and hydrogen dilution were considered. Figures 3.2 and 3.3 show the entire cell polarization, at various times throughout the poisoning process, for an anode feed of 100% and 40% hydrogen respectively, with 10ppm CO. The curves at 120 minutes represent the fully poisoned, steady state cell polarization. Similar curves for 100% and 40% hydrogen with 100ppm CO are shown in Figures 3.4 and 3.5 respectively. Figure 3.6 shows the current obtainable at a cell voltage of 0.6V vs. time for all four feed gas cases considered here, as well as the calculated current from the model described by Equations The details of this calculation are given in Appendix D. In all four cases, after the poisoning process was complete, the catalyst was recovered by feeding the cell with a CO-free, 100% hydrogen anode feed for a period of 2 hours. The current obtainable at 0.6V during the recovery process for all four cases is shown in Figure 3.7. For all cases, the cell performance was revived after the poisoning process with the use of a pure hydrogen anode feed stream. As shown in Figures 3.2 through 3.5, the detrimental effects of CO can be measured even after just 10 minutes of exposure to the cell, with the fully poisoned steady state condition reached in roughly 120 minutes. It is clear from these polarization curves that while the presence of 10ppm CO in 100% hydrogen has a noticeable effect (Figure 3.2), 10ppm CO in 40% hydrogen has an extremely detrimental effect (Figure 3.3). The same is true at the 100ppm level (Figures 3.4 and 3.5). Figure 3.6, which plots the current at a cell voltage of 0.6V vs. time, shows the transient performance drop for all four cases very clearly. At 10ppm of CO in a 100% hydrogen fed stream, the steady state 43

62 current obtainable is 26% less than that before the poisoning process began. With 40% hydrogen and 10ppm CO, the drop is 60%. At the 100ppm level, the respective drops in current for 100% vs. 40% hydrogen are 67% and 86% respectively. The experimental poisoning data was collapsed even further in Figure 3.8. In this graph, the ratio of CO to hydrogen in the anode feed stream was multiplied by a dimensionless time factor. This factor was calculated by dividing the poisoning time by a characteristic time (in this case, total poisoning time for the test). This combined dimensionless number of CO to hydrogen ratio multiplied by dimensionless time is plotted verses normalized current density in Figure 3.8. The normalized current density is simply the measured current divided by the current achievable with neat hydrogen feed stream. Although there appears to be some scatter at high dimensionless ratios, the data collapses quite well at lower values. Utilizing this graph, it is possible to determine the current density available during the transient poisoning process by simply knowing the anode feed composition and the total time elapsed. It is clear from both experimental and computational results that while hydrogen dilution alone has almost no effect on cell performance, and CO alone has a detrimental effect on cell performance, the combined effects of trace quantities of CO and hydrogen dilution have an extremely detrimental effect. This can be explained using the zerodimensional model developed earlier. Under normal cell operating conditions with pure hydrogen, the anode kinetic losses, η a, are very low as compared to the kinetic losses at the cathode and the ohmic losses through the membrane. These normally small anode losses are given by Equation 3.12, and are a function of fractional surface coverage of hydrogen, θ h. Because θ h appears in the denominator of an inverse hyperbolic sine 44

63 relationship, the hydrogen fractional coverage must be very low for the anode overpotential losses to have a significant detrimental effect on cell performance. Thus, dilution alone, which reduces the hydrogen mole fraction, x h in Equation 3.9, and subsequently θ h, does not have an appreciable affect on cell performance (Figure 3.1). The existence of CO in the anode feed stream, and its preferential adsorption on the catalyst surface, does have a substantial effect on cell performance. The presence of CO slows the hydrogen adsorption to such a degree that the fractional coverage θ h falls by an order of magnitude, and thus the anode kinetic losses become significant in effecting overall cell voltage. Since the presence of CO reduces θ h to a degree that the anode losses are now significant, hydrogen dilution, which further reduces θ h, now causes an additional decrease in cell performance. This finding has enormous implications in terms of minimum purity requirements for anode feed gas. As Figure 3.6 clearly illustrates, the commonly quoted number of 10ppm CO as the limit for platinum catalyst based PEM fuel cells is highly dependent on the associated hydrogen dilution level of the feed gas. The performance loss from 40% hydrogen with 10ppm CO is almost equal to the loss associated with 100% hydrogen and 100ppm CO. Although the simulation does not match the experimental data exactly for all cases, it does predict the general trend of transient poisoning, and demonstrates the combined effects that hydrogen dilution and CO can have on the cell performance. The differences between the data and the simulation can be attributed to various causes. First, the CO poisoning model here is coupled with a zero-dimensional fuel cell performance model. Thus, no consideration is given for the spatial variations in fuel cell performance 45

64 resulting from gradients in hydrogen, CO, water, and oxygen concentration that may exist within the cell flow field and the GDL. Secondly, as mentioned earlier, the CO poisoning model here assumes the CO electro-oxidation term to be negligible as compared to CO surface adsorption and desorption terms. While true for low anode overpotential values, this assumption becomes decreasingly valid as the anode overpotential rises. This may explain to some degree the steady state error between the data and model. Lastly, the model is highly dependent on the hydrogen and CO adsorption, desorption, and electrooxidation parameters chosen. Exact values for these parameters are unknown, and vary from membrane to membrane. Also, as noted earlier, several of these parameters are believed to vary with the fractional CO coverage rather than being constant, as assumed here. Despite these drawbacks, the model still predicts the profile of the resulting current vs. time curve of the transient poisoning process, as well as simulates the mutually detrimental effects that CO and dilute hydrogen can have. Using the model, simulation of cell behavior under a host of different hydrogen dilution levels and CO concentrations is possible. Two of these such parametric simulations are shown in Figures 3.9 and The first shows the steady state current obtainable at a cell voltage of 0.6V versus CO concentration. The current has been normalized with respect to the maximum current obtainable with CO-free feed gas. This is plotted for various hydrogen dilution levels. Figure 3.10 compares the time constant for the poisoning process, for various hydrogen dilution levels, versus CO concentration. This time constant was found by calculating the period required for the cell current to suffer 90% of its steady state losses. The inverse of this time constant is plotted versus CO concentration in Figure Thus, a condition that poisons the cell faster has a 46

65 higher inverse of time constant. The CO-free cases never gets to a poisoned state, and thus take infinite time. Consequently, the CO-free condition has an inverse time constant equal to zero. Figures 3.9 and 3.10 reinforce the finding that hydrogen concentration has an effect on both the extent and the rate of CO poisoning. 3.5 Conclusion The transient polarization of a PEM fuel cell undergoing the CO poisoning process has been experimentally measured. This process was observed under variable CO and hydrogen dilution levels. The transient poisoning model developed by Springer et al. (2001), which was modified and solved here, agrees well with the experimentally observed results of transient CO poisoning for both pure and dilute hydrogen feed streams. It was found that while hydrogen dilution alone lowers the fractional coverage on the catalyst surface, it is only when CO is present that the coverage is lowered to a degree that affects cell voltage. Under this condition, the addition of hydrogen dilution will compound the low surface coverage problem even further, and thus cause very poor cell performance. Even with low CO levels normally considered safe for cell operation (i.e. 10ppm), hydrogen dilution can cause an extremely severe loss of cell polarization. These results are easily explainable by the hydrogen and CO adsorption, desorption, and electro-oxidation model presented. With this knowledge, the development of a CO poisoning prevention method can be tackled in Chapter 4. 47

66 Table 3.1. Constants used for solving transient CO poisoning kinetics model. b fc 2.75 x 10-7 atm R J mol -1 K -1 b fh 0.5 atm R ohmic 0.3 ohm cm 2 F C mol -1 T 353K i oc 7.0 x 10-4 A cm -2 V cell 0.6V k eh 4 A cm -2 V o 1.2 V k fc 10 A cm -2 atm -1 α ½ k fh 100 A cm -2 atm -1 ρ 0.1 mol cm -2 P 3 atm absolute 48

67 Figure 3.1. Steady state cell polarization for 100% H 2 and 40% H 2 anode fed gas. 49

68 Figure 3.2. Cell polarization at various time steps (in minutes) throughout the poisoning process, with 100% H 2, 10ppm CO anode feed. 50

69 Figure 3.3. Cell polarization at various time steps (in minutes) throughout the poisoning process, with 40% H 2, 10ppm CO anode feed. 51

70 Figure 3.4. Cell polarization at various time steps (in minutes) throughout the poisoning process, with 100% H 2, 100ppm CO anode feed. 52

71 Figure 3.5. Cell polarization at various time steps (in minutes) throughout the poisoning process with, 40% H 2, 100ppm CO anode feed. 53

72 Figure 3.6. Current at 0.6V during the poisoning process vs. time for all four different anode feed gas compositions. The points represent actual experimental results, and the curves represent simulations based on the model developed. 54

73 Figure 3.7. Current at 0.6V during the recovery process vs. time. 55

74 Figure 3.8. Normalized current density at 0.6V verses CO to hydrogen ratio multiplied by dimensionless time. 56

75 Figure 3.9. Computed normalized steady state current at 0.6V vs. CO concentration for various hydrogen dilution levels. 57

76 Figure Computed poisoning time constant vs. CO concentration for various hydrogen dilution levels. 58

77 CHAPTER 4 AIR BLEED REMEDIATION OF CARBON MONOXIDE POISONING 4.1 Advantages of Anode Air Injection As found in Chapter 3, the combined effects of hydrogen dilution and trace quantities of CO have an extremely detrimental effect on cell performance. Although the oxidative methanol steam reformer to be integrated (Chapter 5) was designed to have very low CO concentrations (on the order of ppm), the dilution of hydrogen will exaggerate the poisoning process. Thus, to successfully generate a desired amount of power without having to increase the cell active area, a technique must be developed for dealing with CO in the anode feed stream. Under steady state conditions, the cleanup methods outlined in Section 1.4 can adequately rectify the feed gas of CO, but many methods fail during sharp transients and start-up periods. Depending on the dynamics of the fuel processor and the control scheme, CO spikes as high as 3,000 ppm may occur during these transients (Murthy, 2001). Without a suitable method for treating CO during these transient intervals, permanent CO poisoning would seem inevitable. In the Nissan methanol reforming FCV, transients from the reformer were dealt with by using a membrane separation process for hydrogen purification (Tamura, 2000). However, as outlined in Section 1.4, there are many efficiency disadvantages to relying on membrane separation as a CO treatment method. In Chapter 3, it was found that by switching to a pure hydrogen feed stream, cell performance can be recovered. Unfortunately, while the CO poisoning process can occur very quickly, the recovery is a much slower process. 59

78 Thus, with a combined reformer and fuel cell system, the cell would be quickly poisoned during a reformer transient. Once the feed gas was relatively free of CO after the transient, the recovery process would take quite some time. This method of MEA remediation is unacceptable, and would be extremely difficult to manage for a vehicle power plant that is constantly undergoing transients. Also, this method deals with the problem after the catalyst layer has been poisoned, and does nothing to prevent CO poisoning in the first place. However, additional options exist for dealing with CO and for remediation of a poisoned MEA. One technique employs the injection of small quantities of air into the CO containing feed stream just after the cleanup steps, but before the gas enters the fuel cell. The oxygen in the air chemically oxidizes the CO on the catalyst layer surface. This catalyst layer remediation technique using heterogeneous catalysis of CO to CO 2 via oxygen injection, commonly called air bleed, was first proposed by Gottesfeld et al. (Gottesfeld, 1988). Murthy et al. conducted air bleed experiments by first feeding a fuel cell pure hydrogen, then switching to a feed containing 500 ppm of CO until the catalyst layer had become fully poisoned. Then, a 5% air bleed was introduced into the feed stream along with the 500 ppm of CO. The air bleed brought the cell s performance back to the levels that a cell operating on pure hydrogen would have. In spite of the continued presence of 500 ppm CO in the feed gas, which would normally devastate performance, the 5% air bleed was able to keep the cell s output at levels comparable to a cell being fed pure hydrogen (Murthy, 2001). When repeated with 3,000 ppm CO, a 15% air bleed was found to slightly remediate the catalyst. However, complete recovery in performance was unachievable with these high levels of CO. Various different catalysts, along with 60

79 anode air bleed, have also been shown to increase the CO oxidation rate within the anode electrode of a PEM fuel cell (Rohland, 1999). A drawback to introducing air directly into the feed steam is the uncontrollable oxidation of other chemical species, namely hydrogen. Because the CO comprises such a small portion of the feed gas, the majority of the air injected into the stream oxidizes hydrogen to water instead of carbon monoxide to carbon dioxide. Thus, although air injection has been found to remediate a catalyst layer of CO, it lowers the overall efficiency by consuming an appreciable portion of the hydrogen fuel. Also, the localized heating caused by air injection into the anode can result in decreased membrane life (Knights, 2004). However, despite any temporary efficiency losses that may result, the controlled use of air bleed during fuel processor transients has great potential for prevention of CO poisoning. Several of the studies outlined above used oxygen injection instead of air. Also, these studies used 100% hydrogen gas as opposed to diluted hydrogen, which would be the case with an actual reformer. Therefore, despite their ability to study the fundamental process of air bleed, these studies do little to identify the practical feasibility of an air injection system with actual hydrocarbon reformate gas. If integrated and optimized correctly, this air injection system has the potential to both increase cell performance and reduce the stringent CO requirements currently placed on any reforming system. Based on this previous work, it was decided to integrate an air bleed system with the fuel cell system to treat the excessively high CO concentration in an anode feed gas resulting from the methanol reforming process. It was chosen to study the performance of a fuel cell 61

80 operating on reformate gas and using an air bleed system for CO rectification. This was done using simulated reformate before using an actual methanol reformer. 4.2 Theoretical Background The process of CO oxidation over platinum is one of the oldest and most studied systems in heterogeneous catalysis (Chorkendorff, 2003). This fundamental catalytic process, as applied to a fuel cell for CO oxidation within the anode electrode, has been studied computationally using oxygen (Baschuk, 2003). Baschuk et al. present a model for oxygen adsorption, desorption, and reaction with both hydrogen and CO on the anode catalyst layer. Equations are the set of reactions modeled by Baschuk et al. H 2 H 2 k fh 2M 2 + b k fh fh + 2M 2 ( M H ) ( M H ) [4.1] [4.2] CO k fc M M CO [4.3] + b k fc fc CO + M M CO [4.4] k [4.5] ( ) eh + M H H + e + M k H O ( M CO) ec M + CO + 2H + e [4.6] O 2 k fo 2M 2( M O) [4.7] + 62

81 O 2 b k fo fo + 2M 2( M O) [4.8] k ( M CO) + ( M O) oc CO + 2M [4.9] 2 k ( M H ) + 2( M O) oh H O + 3M [4.10] 2 Nitrogen, argon, and other species present in air are assumed to be inert, and do not interfere with the surface kinetics other than in their effect in diluting the effective hydrogen and oxygen injection concentration. Note that Equations are the same as those presented in Chapter 3 for hydrogen and CO adsorption, desorption, and electrooxidation. Equations represent oxygen adsorption, oxygen desorption, CO oxidation, and hydrogen oxidation respectively. These additional reactions are used to account for the anode air bleed process. These last 4 reactions were integrated with the model described in Chapter 3 to come up with a set of kinetic equations describing the fractional coverage, on the catalyst layer, for hydrogen, CO, and oxygen (Equations ). dθ ρ h dt dθ ρ c dt = k fh x h ( ) b k ηa θ h θc θ fh k o fh θ h eh θ h k θ θ 2 sinh oh h o P 1 [4.11] RT αf ( ) b ηa θ h θc θ fc k fc θc k o ec θ c k θ θ 2 sinh oc c o = k fc x c P 1 [4.12] RT αf dθ o ρ dt ( ) b 1 θ h θ c θ fo k fo θ o k θ θ oh h o k θ θ oc c o = k fo x o P 1 [4.13] o 2 63

82 Baschuk et al. assume the hydrogen adsorption, desorption, and oxidation reactions to be second order in nature. However, to be consistent with the model presented in Chapter 3, and for simplicity, these reactions are assumed to be first order with respect to surface coverage. Rewriting Equations in terms of hydrogen current yields Equations dθ ρ h dt ( θ h θc θ ) b fh k fh θ h i o k θ θ oh h o = k fh x h P 1 [4.14] dθ ρ c dt dθ o ρ dt ( ) k b 1 θ θ ec c h θc θ fc k fc θc i k θ θ oc c o = k fc x c P 1 o [4.15] 2 k θ eh h ( ) b 1 θ h θ c θ fo k fo θ o k θ θ oh h o k θ θ oc c o = k fo x o P 1 [4.16] o 2 These equations, along with the model for cell current as a function of hydrogen coverage used in Chapter 3, were applied in determining the effects of variable amounts of CO, hydrogen dilution, and air injection on cell performance. The parameters used for this model are given in Table 4.1. The basic membrane performance parameters were estimated using pure hydrogen data for a Gore MEA. The specifications for this MEA are given in Table 4.2. The CO poisoning parameters are the same as values given in the literature (Springer, 2001) and in Table 3.1. The value for CO desorption was modified for this specific MEA, but is still within the range given by Springer et al. The kinetic parameters for the anode air injection (i.e. oxygen adsorption, desorption, and chemical oxidation with hydrogen and CO) were derived from the literature (Baschuk, 2003). The kinetic parameters presented by Baschuk et al. were modified to have dimensional consistency with the model presented here. The only exception was for the chemical rate 64

83 of CO oxidation, which was varied from the value given by Baschuk et al. to fit the observed experimental data. 4.3 Experimental Study of Air Injection To substantiate the ability of anode air injection to recover the performance of a CO poisoned anode catalyst layer, an air bleed system was integrated into the fuel cell test station described in Chapter 2.2. Lynntech Industries, Ltd. (College Station, TX) supplied the cell for these experiments. The specifications for this cell are given in Table 2.2. W.L. Gore and Associates (Newark, DE) supplied the MEA for these experiments. The specifications for this MEA are given in Table 4.2. Please refer to Appendix B for scanning electron microscope images of this MEA. Gas flow, pressure, humidity, and temperature regulation, along with voltage and current regulation, were done using a 2- channel fuel cell test stand supplied by Arbin Instruments (College Station, TX). When mixing hydrogen and air directly, the possibility exists for the creation of a flammable mixture. This prospect raises safety concerns with any anode air injection system. The lean and rich flammability limits of H 2 in air are 4% and 75% respectively (Glassman, 1996). Thus, a mixture of less than 4% H 2 in air has too little hydrogen to be dangerous whereas a mixture of greater than 75% hydrogen in air has too little oxidizer to be hazardous. For the purposes of these experiments, percent air bleed was defined as the moles of air divided by the moles of hydrogen in the stream. Thus, using the rich flammability limit of 75% hydrogen as an extreme, the maximum safe percent air bleed is 33%. For the safety of these experiments, and to prevent excessive efficiency loss due to 65

84 hydrogen oxidation, the percent air bleed was restricted to an upper limit of 10%. Careful regulation of anode bleed air was done using a rotameter. The necessary check valves on both the hydrogen and air supplies were used to ensure system safety. The entire air bleed remediation system was developed, integrated, and optimized using simulated reformate gas. To closely simulate methanol reformate, many of these experiments were conducted with diluted hydrogen at the 40% level, and CO at the 100ppm level. For many of the graphs presented here, the performance under various conditions is compared to that of a CO free, pure hydrogen feed stream. 4.4 Results and Discussion Figure 4.1 shows the cell polarization curves, at atmospheric pressure, with and without the air injection system for various anode feed gases. As shown, without air injection, the poisoning effect with 100ppm of CO is prominent with both a 100% and 40% hydrogen feed stream, with the 40% case showing increased CO poisoning. This finding was also established in Chapter 3. However, with a 10% anode air injection (mole basis relative to hydrogen), cell performance was partially recoverable. Similar trends were observed at higher cell pressures. Figures 4.2 and 4.3 show the results at 2 atmospheres (15 psig) and 3 atmospheres (30 psig) respectively. In both high-pressure cases, the 10% anode air bleed was sufficient to return to the cell back to, or near, the performance with a pure hydrogen feed stream. However, excessive anode air injection has the side effect of reducing overall efficiency in two ways. First and foremost, in addition to oxidizing carbon monoxide from the catalyst layer, oxygen injection also 66

85 oxidizes hydrogen fuel to water. This consumption of hydrogen on the anode reduces overall fuel utilization, and thus efficiency. Secondly, in an integrated system, excessive air injection can lead to increased parasitic pumping losses due to additional air feed required. Thus, an optimized methanol reformer/hydrogen fuel cell system would try to not only minimize CO generation from the reformer, but also minimize the amount of air injection required to deal with the CO. Therefore, the effect of low anode air injection rates was studied. With a simulated reformate feed of 40% hydrogen and 100ppm CO, the effects of varying air injection levels from 0 to 10% were investigated. The cell polarization curves are shown in Figure 4.4, while the power density curves are shown in Figure 4.5. As observed, even a 2% air injection can help to recover performance, but full recovery is not possible without larger amounts of air bleed. Similar trends were observed at the 600ppm level (Figures 4.6 and 4.7), with a 2% injection contributing to a large recovery, but with 5 and 10% injection rates providing only marginal additional gains in performance. It is important to note that at the 600ppm level, even a 10% air injection level is insufficient to fully rectify the catalyst layer of CO. This performance data at the 100 and 600ppm CO levels verses amount of air injection is shown in Figure 4.8. As illustrated, there are diminishing performance gains in increasing air bleed beyond 2%. There is also an observable limit to the extent of performance recovery. With a 40% hydrogen, 100ppm CO feed stream, this limit appears to be roughly 75%. However, at the high CO level of 600ppm, this maximum performance limit is roughly 60% of that from a pure hydrogen feed. In order to sort through this multitude of fuel cell data, it was decide to collapse the polarization curves by looking at the cell performance at a fixed voltage (0.6V). 67

86 Figure 4.9 shows the normalized current available, relative to pure hydrogen, verses the ratio of CO to hydrogen in the anode feed stream. Once again, it is shown that without air injection, even low CO to hydrogen ratios have an extremely detrimental effect. However, a small amount of air bleed contributed to a great deal of cell recovery, with increasing bleed rates contributing marginal gains. Like Figure 4.8, Figure 4.9 illustrates the inability of large quantities of air injection to rectify cell performance at high CO to hydrogen ratios. This graph in particular can be very useful for determining the amount of air bleed necessary to maintain a certain cell performance level despite varying reformer CO to hydrogen ratio output conditions. This figure also illustrates clearly the dependency of cell performance on the ratio of CO to hydrogen in the anode feed stream, and not just simply the ppm level of CO. These experimental findings were confirmed using the joint CO poisoning and air bleed model outline in section 4.2. As shown in Figure 4.10, the model predicts similar trends to that of the experimental data for the case without air injection. However, for the case with air injection, the model overestimates the current available. As shown, there is quite a bit of disagreement between the current density values predicted by the model and those measured experimentally. This divergence between data and simulation can be attributed to many of the same causes that differentiate the experiments from the model outlined in section 3.4. Like the CO poisoning model presented earlier, the joint air bleed and CO poisoning model here is coupled with a zero-dimensional fuel cell performance model. Thus, no consideration is given for the spatial variations in fuel cell performance resulting from gradients in hydrogen, CO, water, and oxygen concentration that may exist within the cell flow field and the GDL. More importantly however, the model presented 68

87 here is highly dependent on the hydrogen and CO adsorption, desorption, electrooxidation, and chemical oxidation parameters chosen. It is also highly dependent on the rates of oxygen adsorption and desorption on the anode catalyst layer due to the air injection. Exact values for these parameters are unknown, and vary from membrane to membrane. Also, several of these parameters are believed to vary with the fractional CO and oxygen coverage rather than being constant, as assumed here. Thus, although Baschuck et al. were able to obtain good agreement between their experimental data and a similar version of this model, the kinetic parameters used then are not necessarily valid for all conditions. Despite these drawbacks, the experimental data and attempt at validation presented here provide a good foundation for future model refinement. 4.5 Conclusion The use of anode air injection as a means of controlling the extent of CO poisoning has been studied both experimentally and theoretically. The usefulness of air bleed with simulated reformate gas, which contains diluted hydrogen and trace quantities of CO, has been shown. It was found that even low levels of air injection (i.e. 2%) have a significant effect on improving the performance with fuel dilution levels as low as 40% and CO levels as high as 100ppm. With simulated reformate gas and a 2% air bleed, roughly 70% of the current available from a pure hydrogen feed is obtainable. Higher air bleed percentages were found to increase the obtainable current up to only 75%. At very high CO levels in the 600ppm range, which are typical of a reformer undergoing a transient load change, only 60% of the current available from pure hydrogen is 69

88 obtainable. This was found with air bleed percentages as high as 10%. While the theoretical joint model for CO poisoning and anode air injection confirms the general trends observed in the experimental data, this simple model deviates enough to make it unreliable for accurate current density predictions. Further model refinement, based on the observed data, is necessary for exact forecasts of current density. However, the experimentally observed performance using air bleed to prevent CO poisoning with simulated reformate gas is extremely valuable in the transition to using an actual methanol reformer for fueling the cell. 70

89 Table 4.1. Constants used for solving air injection kinetics model. b fc 1.74 x 10-6 atm k oc 1.8 x 10-2 A cm -2 b fh 0.5 atm k oh 2.8 x 10-5 A cm -2 b fo 2.0 x 10-3 atm P 3 atm absolute F C mol -1 R J mol -1 K -1 i oc 7.0 x 10-4 A cm -2 R ohmic 0.3 ohm cm 2 k ec 1.0 x 10-8 A cm -2 T 353K k eh 4 A cm -2 V cell 0.6V k fc 10 A cm -2 atm -1 V o 1.2 V k fh 100 A cm -2 atm -1 α ½ k fo 7.9 x A cm -2 atm -1 ρ 0.1 mol cm -2 71

90 Table 4.2. Specifications for the W.L. Gore Membrane Electrode Assembly. Active Area 50 cm 2 Membrane Thickness Equivalent Weight 0.71 mil (18 µm) <1000 g/mol Anode Catalyst Loading 0.15 mg Pt per cm 2 Cathode Catalyst Loading 0.35 mg Pt per cm 2 Gas Diffusion Layer Hydrophobic carbon paper with hydrophilic micro porous layer 72

91 Figure 4.1. Cell polarization curves at atmospheric pressure with different anode feed gases and 10% anode air injection. 73

92 Figure 4.2. Cell polarization curves at 15 psig with different anode feed gases and 10% anode air injection. 74

93 Figure 4.3. Cell polarization curves at 30 psig with different anode feed gases and 10% anode air injection. 75

94 Figure 4.4. Cell polarization curves at 30 psig with 100ppm CO simulated reformate gas feed and varying anode air injection levels. 76

95 Figure 4.5. Cell power curves at 30 psig with 100ppm CO simulated reformate gas feed and varying anode air injection levels. 77

96 Figure 4.6. Cell polarization curves at 30 psig with 600ppm CO simulated reformate gas feed and varying anode air injection levels. 78

97 Figure 4.7. Cell power curves at 30 psig with 600ppm CO simulated reformate gas feed and varying anode air injection levels. 79

98 Figure 4.8. Percent available current (as compared to a pure hydrogen feed) verses air injection levels for two different simulated reformate gas feeds. 80

99 Figure 4.9. Normalized current density at 0.6 V verses CO to H 2 ratio for different air injection levels. 81

100 Figure Theoretically determined normalized current density at 0.6 V verses CO to H 2 ratio for different air injection levels. 82

101 CHAPTER 5 INTEGRATED METHANOL REFORMING FUEL CELL SYSTEM 5.1 Integrated System Development In Chapter 2, the effects of low cell pressure and humidity were studied. In Chapter 3, the effects of simulated reformate gas and the resulting transient CO poisoning process was investigated. In Chapter 4, an anode air injection system was developed to treat a poisoned catalyst layer. This system was found to be effective in oxidizing CO from the catalyst layer, and therefore prevent extensive poisoning. Once cell performance under various operating conditions with simulated reformate gas was known, a transition to studying cell behavior with actual reformate gas became possible. With this integrated system, the efficiency, from liquid methanol fuel to useful energy, can be studied. A methanol reformer, constructed by the research groups of Dr. Chunshan Song and Dr. André Boehman of The Pennsylvania State University, was integrated with the fuel cell test system shown in Figure 2.1. The final integrated system schematic is shown in Figure 5.1. An image of the entire system is shown in Figure 5.2. The reformer utilizes the OSR reaction discussed in Section 1.3. The reformer also utilizes a secondary Pt/Al 2 O 3 catalyst bed to reduce CO content of the effluent gas. This reformer generates a hydrogen anode feed gas which is diluted by nitrogen and carbon dioxide, and contains trace quantities of CO. The reformer effluent contains hydrogen and CO concentrations 83

102 similar to the simulated reformate gas used in the transient CO poisoning and air bleed studies of Chapter 3 and 4. As noted, the anode gas in Chapter 3 used nitrogen as the diluent, whereas actual reformer output has both nitrogen and carbon dioxide as diluents. However, carbon dioxide does not participate in the surface adsorption chemistry in any appreciable manner. Thus, like nitrogen, the presence of carbon dioxide in the anode fed gas is not expected to have any effect on cell performance other than lowering the hydrogen concentration. All of the critical reformer specifications are given in Table 5.1. Several intermediate reactions occur within this oxidative steam reformer. However, a general overall reaction is given by Equation 5.1 (Gu, 2003). 1 CH 3OH + O2 4 H o 298K = H 2 kj mol 2 O CO H 2 [5.1] The reformer is fed a methanol/water solution. The methanol concentration of this solution, in terms of molarity, is approximately The flow rate of methanol solution to the reformer was controlled using a Waters Reagent Manager pump. The reformer also required an air supply to not only complete the OSR reaction, but to aid the CO clean-up process within the reactor as well. Airflow rates for both the main reforming reaction and CO clean-up reaction were controlled using dual rotameters. The reforming reaction was conducted at 230 C, while the CO clean-up reaction was carried out at 150 C. Due to the high temperatures of reformer effluent, the gas was cooled down to 80 C before entering the fuel cell. In addition, a knock-out system was employed to catch any water that condensed during the cooling process. This was done to avoid feeding liquid water to the fuel cell along with the anode gas. The knock-out 84

103 system was positioned to not only thwart water from entering the cell, but to also prevent water from backing up into the reformer and flooding the CO clean-up catalyst bed. 5.2 Experimental Investigation Baseline tests were conducted shortly after integrating the reformer with the hydrogen fuel cell test stand. The group at Pennsylvania State University who built the reformer claimed its effluent composition would be at least 50% hydrogen and less than 30ppm of CO (Gu, 2003). Nevertheless, a gas chromatograph (GC) was employed to corroborate these figures. Utilizing an Agilent 3000A MicroGC, the composition of the reformer effluent was confirmed to contain 60-63% hydrogen. CO concentrations were confirmed to reside below 100ppm. However, accurate measurements below the 100ppm level proved difficult due to the high level of GC resolution required at these low CO concentrations. With the output composition confirmed, testing of the reformer in combination with a fuel cell began. At various pressure levels, cell polarization was measured. Experiments were conducted with a W.L. Gore MEA and carbon paper GDLs. The specifications for this MEA are given in Table 4.2. For these experiments, cell temperature was maintained at 80 C. For each case shown, steady state polarization curves were taken using three different anode feed gases. First, as a reference, the performance was measured using a pure hydrogen anode feed. Next, the reformer feed alone was used to fuel the cell. Lastly, the reformer feed in combination with anode air injection was utilized. 85

104 5.3 Results and Discussion Figure 5.3 shows the performance of the fuel cell, at atmospheric pressure, with a pure hydrogen feed, the reformer feed alone, and the reformer feed plus 10% anode air injection. As shown, the use of air bleed was able to recover a great deal of performance, which would otherwise be lost due to the CO poisoning effect. The corresponding power curves at atmospheric pressure are given in Figure 5.4. Similar trends in the data were observed at 15 and 30 psig (2 and 3 atmospheres) of cell pressure, and are illustrated in Figures While the reformer feed alone yields extremely inferior performance, approximately 70 76% of the peak power obtainable with pure hydrogen is available utilizing the reformer feed in combination with anode air injection. This capability was observed at all three pressure levels. In terms of total peak power density, as expected, the 30 psig case yielded the highest performance. A comparison of the peak power available, normalized to that of pure hydrogen at 30 psig of pressure, is shown in Figure 5.9. As illustrated, air injection improved the peak performance greatly at all pressures. However, as noted earlier, there exists a noticeable different between the pure hydrogen case and the reformer feed plus 10% air injection case. Like Figures 8 and 9 in Chapter 4, this figure clearly illustrates the limit of performance recovery possible with air bleed. As noted in Chapter 4, this limit becomes increasing unfavorable as the CO to hydrogen ratio grows. For the reformer fed cases, a significant drop in cell performance is clearly visible at current densities between 1200 and 1300 ma/cm 2. This drop is most noticeable at 3 86

105 atm (30 psig) of pressure (Figures 5.7 and 5.8). This is due to a mass transport limitation on the fuel cell s anode side. Essentially, hydrogen requirements at these high current draw levels exceed the maximum hydrogen production rate from the reformer, which is roughly 555 ml/min. For a 50-cm 2 cell operating at stoichiometric conditions, 555 ml/min corresponds to a current density of roughly 1480 ma/cm 2. Therefore, a drop in cell performance as current densities approach 1480 ma/cm 2 was expected. A higher capacity reformer would have enabled production of power at higher current densities. This could have pushed the obtainable peak power closer to the pure hydrogen case. However, without a high capacity reformer, it is not possible to determine the exact potential for increases in peak power density. It is important to note that while the reformer effluent was fed directly into the fuel cell without humidification, the cathode air feed was humidified in certain cases. At 15 and 30 psig, the cathode gas was fed dry. However, at atmospheric pressure, the air was fully humidified. As discussed in Chapter 2, humidified gas is critical to maintaining membrane conductivity at low pressures. As demonstrated in Chapter 2, water generation on the cathode at higher pressures is sufficient to maintain internal cell humidification. However, this is not the case at atmospheric pressure. Thus, for the atmospheric pressure test, the cathode feed gas was fully humidified. This need to fully humidify the cathode gases introduces further parasitic losses to the system. This led to an investigation concerning reduced cathode humidification as a method of improving overall system efficiency. Figure 5.10 shows the cell polarization, at atmospheric pressure, utilizing a reformer feed plus anode air injection. The different curves represent varying cathode 87

106 relative humidity levels. This data was then analyzed to determine the peak cell power available in each case. Figure 5.11 shows the peak power available, relative to a pure hydrogen feed, verses cathode relative humidity. As shown, with a dry cathode feed gas, only 44% of the power is obtainable. This is as opposed to a fully humidified cathode stream, where 76% of the pure hydrogen anode feed power is obtainable. Thus, combining low cathode relative humidity levels with a reformer fed anode greatly reduces overall cell power density. However, at certain mid-range humidity levels, this drop in cell power density may be acceptable due to the reduced humidification requirement for the cathode air stream. 5.4 Conclusion An oxidative steam reforming methanol unit was incorporated with a PEM fuel cell. The effluent of this reformer was confirmed to have hydrogen concentrations of at least 60%. CO concentrations were confirmed to be below 100ppm, although the actual concentration is believed to be closer to 30ppm. Anode air injection was used to decrease the CO poisoning effect of reformate gas. The reformate gas fed fuel cell s performance then was then evaluated. It was found that without the use of air injection, cell performance was greatly deteriorated at all pressure levels. However, with air injection, approximately 70-76% of the peak power density available from a pure hydrogen feed was obtainable with the reformate feed. This was confirmed at pressures from 0 to 30 psig. At higher pressures and current draws, power densities may increase with the use of a scaled-up reformer. At atmospheric pressure, it was found that low cathode humidity 88

107 had an extremely detrimental effect on peak power output. Using reformate gas, 76% of the peak power obtainable with pure hydrogen was available with a fully humidified cathode. However, with a dry cathode, only 44% of the peak power was available. Knowing the cell performance with methanol reformate, it is now possible to determine overall efficiencies for producing electricity from methanol via a reformer and hydrogen fuel cell. It is also possible to compare this efficiency to that of a fuel cell utilizing methanol directly as a fuel. This IDMFC and DMFC system comparison is carried out in Chapter 6. 89

108 Table 5.1. Reformer Specifications and Operating Conditions. Catalyst Composition Catalyst Properties Solution Composition Reformer Feed Rates Reformer Output (approximate) CuO ZnO Al 2 O 3 C (graphite) Total Loading Surface Area Bulk Density Catalyst Bed Temperature Molarity MeOH to Water Molar Ratio Methanol Concentration (vol. %) Solution Flow Rate Air Feed Flow Rate CO Clean-up Air Flow Rate Oxygen to Methanol Molar Ratio Clean-up Air to Air Feed Molar Ratio Hydrogen Production Hydrogen Nitrogen Carbon Dioxide Carbon Monoxide 58 +/- 2.0 % 31 +/- 3.0 % 11 +/- 1.5 % % 3 grams 60 +/- 20 m 2 /gram 67 +/- 5 lbs/ft C 15.2 M 1: % 0.6 ml/min 319 ml/min 159 ml/min 0.3:1 0.5:1 ~555 ml/min >60 % 20 % 30 % <100 ppm 90

109 Figure 5.1. Integrated methanol reformer and fuel cell system. 91

110 Fuel Cell Feed Line from Reformer Reformer Figure 5.2. Integrated methanol reformer, fuel cell, and CO remediation system. 92

111 Figure 5.3. Cell polarization curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at atmospheric pressure. 93

112 Figure 5.4. Cell power curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at atmospheric pressure. 94

113 Figure 5.5. Cell polarization curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 15 psig. 95

114 Figure 5.6. Cell power curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 15 psig. 96

115 Figure 5.7. Cell polarization curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 30 psig. 97

116 Figure 5.8. Cell power curves for the reformer feed and reformer feed plus air injection as compared to a pure hydrogen feed at 30 psig. 98

117 Figure 5.9. Normalized peak power density verses pressure for pure hydrogen, the reformer feed alone, and reformer feed plus air injection. 99

118 Figure Cell polarization curves, for the reformer feed plus air injection as compared to a pure hydrogen feed, at atmospheric pressure with varying levels of cathode relative humidity. 100

119 Figure Peak cell power, as compared to a pure hydrogen feed, verses cathode relative humidity. This data was taken utilizing a reformer feed plus air injection. 101

120 CHAPTER 6 COMPARISON OF INDIRECT AND DIRECT METHANOL FUEL CELL SYSTEMS 6.1 Introduction The development of an integrated methanol reformer and fuel cell system, as outlined in Chapter 5, opens up many additional possibilities. Most notably, the ability to conduct system-wide comparisons between the methanol reforming system (IDMFC) and the direct methanol fuel cell (DMFC) exists. Unlike the IDMFC, which requires a methanol reformer to generate hydrogen, the DMFC utilizes dilute liquid methanol directly in a PEM fuel cell. As noted in Chapter 1, the voltaic efficiency of current DMFC technology is very low. However, the DMFC may be preferred due to variations between IDMFC and DMFC system size, weight, complexity, and cost. In addition to the development of a methanol reforming fuel cell system, a DMFC was also integrated into the same fuel cell test station (Gonder, 2003). A study comparing the efficiency of the two methods, direct and indirect use of methanol, was then undertaken using methanol as the base feedstock for both systems. Because the requirements for efficiency, size, and complexity for various applications differ greatly, the two methanol fuel cell systems were compared at three different power levels: 20W for portable power, 5 kw for auxiliary power, and 80 kw for automotive applications. Considering the complexities and efficiencies of both electrochemical methods for 102

121 generating power from methanol, this comparative study will determine the usefulness of both systems for all three power applications. It is important to note that, depending on the application, the total system efficiency may include additional losses not examined here. For example, in portable applications, direct current (DC) power is desirable. However, for stationary power applications, this DC power would need to be inverted to alternating current (AC), thus introducing another source of efficiency reductions. In automotive applications, the DC power would need to be translated into rotary motion (usually through the use of an inverter and electric motor), which would also introduce sources for power losses. In addition, the portable and auxiliary power calculations do not include the production efficiency of methanol, which may vary depending on the base fuel (i.e. natural gas) and the process for converting this base fuel into methanol. For modern combined reforming systems, the efficiency in converting the higher heating value of natural gas into methanol is only 72% (Allard, 2000). However, the comparison presented here is valid for determining efficient paths for conversion of methanol into DC power. 6.2 Experimental Conditions and Methods of Comparison For this comparison, the DMFC was run at 75 C and atmospheric pressure. It utilized a 1 molar methanol solution with an anode stoichiometry of 27 and cathode stoichiometry of 20. These conditions were found give peak performance in previous work (Gonder, 2003). The IDMFC was run at 80 C and 3 atm (30 psig) of pressure. It utilized a methanol reformer and 10% anode air injection to compensate for the CO 103

122 poisoning effect. The cell polarization curves for both systems are shown in Figure 6.1, whereas the power density curves are shown in Figure 6.2. As shown in Figure 6.2, the DMFC system has far less peak power density per unit cell area than the IDMFC system. Thus, to generate equal quantities of power, the DMFC would need to be sized much larger. Also, the DMFC reaches its peak power a much lower cell voltage than the IDMFC. In determining the feasibility of the two systems for different applications, cost is one factor. Although a complete cost comparison between the two systems would be extremely difficult, a platinum catalyst cost analysis is relatively easy. For this analysis, a platinum cost of $30 per gram is utilized. The historic cost of platinum is shown in Figure 6.3 (Johnson Matthey Inc, 2004). Platinum catalyst costs for the three different applications are presented in the sections below. Additionally, for certain applications, size is an additional consideration. The IDMFC system requires the use of a reformer, which does not occupy a trivial amount of space. Finally, for automotive applications, the overall efficiency under dynamic conditions (i.e., with frequent load transients) becomes important. Thus, in section 6.5, in addition to cost calculations, simulation results for the efficiency of the two different systems under automotive conditions are presented. 6.3 Portable Power (~20W) Portable electronics, such as cell phones and notebook computers, require a longlasting electrical supply. These devices power supplies put a premium on energy density and low weight. For these applications, an output of 20W was chosen to cover the entire 104

123 range of power supplies desired for portable electronics. Based on the DMFC and IDMFC power curves in Figure 6.2, it is possible to calculate the minimum active cell area needed to generate 20W. For the DMFC, with its low peak power density of 66 mw/cm 2, approximately 300 cm 2 of active cell area would be required to generate 20W of DC power. For the IDMFC, the required cell area would be roughly 35 cm 2 based on a peak power density of 571 mw/cm 2. Despite the DMFC having a much larger cell area, the overall system is far simpler than the IDMFC. For portable power applications, the integration of a compressor, high temperature oxidative steam reformer, and air injection system in such a small space would be prohibitively difficult. Table 6.2 shows the cost of platinum for both the DMFC and IDMFC at the 20W power level. This is based on a projected platinum cost of $30 per gram (Figure 6.3). The costs in Table 6.2 were established using the MEA platinum loadings and peak power densities achieved in this study. As shown, the DMFC s platinum cost DMFC would be roughly $55, as compared to a cost of $0.53 for the IDMFC. Despite the DMFC s much higher catalyst costs, it is still desirable over the IDMFC system for portable applications. As mentioned earlier, integration of an IDMFC system at such a small scale would be extremely challenging. Although a platinum cost of $55 for the DMFC system may appear high, it is not prohibitively high when compared to cost of current high energy density lithium ion notebook computer and portable electronics batteries. 105

124 6.4 Auxiliary Power (~5kW) Auxiliary power applications are those requiring a small, fuel efficient, and cost effective method for generating small quantities of electricity. A small internal combustion engine mated to a generator often takes up this role. By contrast with portable power applications, where the foremost concern is size and simplicity, in auxiliary power systems, energy efficiency and cost are of increasing importance. Therefore, the IDMFC and DMFC systems were compared for both cost and efficiency in converting methanol into power. As mentioned earlier, a fuel cell stack generates DC power. Unlike portable power applications, auxiliary power applications often require AC power. Thus, when comparing the efficiency numbers presented here to those of an internal combustion engine/ AC generator system, the efficiency in inverting the fuel cell s DC power to AC must be included. However, in this study, we are concerned with comparing one fuel cell system to another. Therefore, the efficiency in converting methanol to DC power will suffice. Figure 6.4 shows the IDMFC system s power density and raw conversion efficiency. This raw conversion efficiency includes the reformer and cell efficiencies (parasitic losses are ignored). As shown, the efficiency varies depending on the cell current draw. A similar chart for the DMFC system is shown in Figure 6.5. For both the DMFC and IDMFC systems, the cell efficiencies are compared at the peak power density location along the polarization curves shown in Figures 6.4 and 6.5. These DMFC and IDMFC systems efficiencies are 1.95% and 25.90% respectively. 106

125 Table 6.3 shows the total efficiency for both systems. The Conversion Efficiency represents the actual measured cell power output divided by the power available in the methanol feed stream to each system. This is efficiency obtained from Figures 6.4 and 6.5. The Current System Efficiency listed in Table 6.3 represents the inclusion of projected parasitic losses for both systems in the efficiency calculation. As shown, the DMFC system was only 1.75% efficient in converting methanol to electricity. This is in sharp contrast to the IDMFC system, which was 21.23% efficient. However, looking at the breakdown of losses given in Table 6.3, it is clear that low efficiency is the result of low DMFC fuel utilization. This low fuel utilization is due the DMFC running at a high anode stoichiometry of 27. In practical applications, the remaining methanol, not reacted in the fuel cell, would be simply re-circulated back into the system. During re-circulation, a small amount of pure methanol would be added to maintain a constant concentration level. Therefore, a much higher practical DMFC efficiency than 1.75% is possible due to increased fuel utilization. For the IDMFC system, an anode stoichiometry of 1.5 was used. This led to an IDMFC fuel utilization efficiency of only 66.67%. Like the DMFC, in practical applications, this number would be nearly 90%. Thus, for both systems, a projected efficiency employing a realistic fuel utilization figure is also given at the bottom of Table 6.3. For both systems, projected efficiencies in the mid to high 20% range are easily possible. The IDMFC resulted in a slightly higher efficiency than the DMFC, with an overall effectiveness of 28.67% in converting methanol to electricity. Mizsey and co-workers calculated a possible efficiency of 38.9% for a methanol reforming fuel cell system (Mizsey, 2001); however, these researchers assumed a high 107

126 cell voltaic efficiency of 50%, as opposed to 40.54% measured in this study. Mizsey et al. also estimate parasitic loses as only 13%, whereas the current investigation used 18%. Advancements in MEA technology and reduced parasitic loses can make these higher efficiency numbers reported by Mizsey et al. achievable. On the DMFC side, improvements in membrane technology can push these numbers higher as well. For the current study, it is interesting to note that the two very different systems are surprisingly close in terms of overall possible efficiency in converting methanol to electricity. Unlike portable power applications, auxiliary power applications can afford the space necessary for IDMFC subsystems. Considering this, and the roughly equal efficiencies in generating electricity from methanol for the two systems, it appears that either would be suited for auxiliary power applications. However, the two systems vary greatly in terms of platinum costs. According to Table 6.2, $132 of platinum is needed for the IDMFC system. However, to generate 5kW of power, over $13,000 of platinum is needed for the DMFC system. For portable power applications, high cost per unit power can be tolerated. However, as power requirements increase, the cost per unit power becomes increasing stringent. Over $13,000 in catalysts costs alone would make a 5kW DMFC auxiliary power system far too expensive. Considering this, an IDMFC is seen as favorable for auxiliary power applications despite the comparable system efficiencies. 108

127 6.5 Automotive Power (~80kW) Automotive power applications have many analogous constraints as portable and auxiliary power applications. Size, weight, efficiency, and cost are all of major concern. However, unlike portable power applications, size and weight constraints are far less stringent. As mentioned previously, as power levels increase, the acceptable cost per unit power generally becomes lower. Thus, at the 80kW level, cost becomes an extremely important factor. In order to be competitive with the internal combustion engine, current DOE research goals set a target mass production cost of $35/kW for fuel cell in 2010 (U. S. DOE, 2002). Of this $35/kW, the DOE sets a platinum loading target for hydrogen fuel cells of 0.2 mg/cm 2, or $6/kW based on current precious metal prices. According to Table 6.2, the platinum cost per kilowatt is $2,727 and $26 for the DMFC and IDMFC systems respectively. Thus, while the current IDMFC system requires a little over a 4 fold decrease in platinum loading to be competitive, the DMFC system requires over a 450 fold decrease in catalyst loading. In terms of overall platinum cost for an 80 kw system, the current DMFC and IDMFC systems are estimated at $218,182 and $2,102 respectively, whereas DOE goals, based on $6/kW, yield a $480 target cost. Despite efficiencies of 23.71% measured here, and over 30% reported in the literature (Moore, 2000 and Moore et al., 1999), the outrageously high cost of platinum for a DMFC system makes it unattractive for automotive applications. Based on this platinum cost analysis, the IDMFC has an extremely clear advantage over the DMFC system for automotive applications. However, knowing this fact does not lead to any information concerning the efficiency of an IDMFC system in automotive applications. 109

128 The auxiliary power analysis results, shown in Table 6.3, illustrate the efficiency of an IDMFC system in generating steady state DC power from a methanol feedstock. However, this does not indicate the IDMFC s efficiency in automotive applications, where frequent load transients exist. Although a hybrid IDMFC/battery system could be envisioned to somewhat reduce the drive cycle load transients, and load following IDMFC system is considered here. To that end, a simulation of IDMFC performance in automotive applications at the 80 kw level was developed in a Simulink programming environment. Simulink was chosen because of its ability to easily and accurately model dynamic systems. The complete Simulink model is presented in Appendix D. Using the experimental data taken, fuel cell and reformer performance maps were integrated into Simulink. A simulated vehicle with a fuel cell engine power plant was run through a Federal Test Procedure (FTP) cycle. The specific cycle chosen was the FTP-75, which is used to measure the tailpipe emissions of light duty vehicles in the United States. This cycle, which emulates mild city driving, is approximately 11 miles long, takes just over 31 minutes complete, and has an average vehicle speed of 21.2 mph. The top speed in the cycle is 56.7 mph. The cycle s speed trace is shown in Figure 6.6. Vehicle parameters were chosen to emulate a 4 passenger sedan of approximately 4.5 meters (15 ft) in length. The vehicle was given a drag coefficient of 0.35, a frontal area of 2 m 2, a rolling resistance coefficient of 0.015, and a weight of 2500 lbs. This weight is considered the rolling chassis weight, and includes the chassis, electric motors, and drivetrain, but does not include the fuel cell power plant. The 2010 DOE target for a combined fuel cell stack and fuel processing system is 325 W/kg (U. S. DOE, 2002). Utilizing this for an 80kW (107 hp) system, the total vehicle weight arrives at 3041 lbs. 110

129 At 28.4 lbs/hp, this vehicle is by no means a sports car. However, it has roughly the same weight-to-power ratio as a fully charged Honda Insight Hybrid, which can complete a zero to 60 mph run in just under 11 seconds. All critical vehicle and fuel cell specifications are given in Table 6.4. Inverter/motor and drivetrain efficiencies were the same as those used in previous studies (Mizsey, 2001). The methanol reformer efficiency was obtained from the study outlined in Chapter 5. Based on the literature, a fuel utilization efficiency of 90% and parasitic loss of 18% were assumed (Larminie et al., 2000). An upstream efficiency of 72%, for generating methanol fuel from natural gas, was used for the well-to-tank efficiency analysis (Allard, 2000). The vehicle s fuel cell stack is based on the reformer fed cell performance, at 30psig, from Chapter 5. The power required to drive the FTP-75 cycle, for the simulated vehicle, is shown in Figure 6.7. As shown, just over 30kW is required to maintain the city cycle s speedtime profile. Therefore, an 80kW methanol reforming fuel cell power plant is of sufficient power. Figure 6.8 shows the voltage of an individual cell in the stack. As shown by the relatively shallow drops, the fuel cell power plant is sufficient in size to prevent the cell from reaching extremes of low cell voltage during this city cycle. These shallow voltage drops help in maintaining a high cell voltaic efficiency. Figure 6.9 illustrates the cell current density during the drive cycle. Knowing the cell efficiency during the drive cycle, and the additional efficiency losses outlined in Table 6.4, the instantaneous well-to-wheel efficiency can be calculated. This efficiency includes all losses incurred from initial generation of methanol from natural gas, to actually moving the vehicle through a drive cycle. This efficiency is illustrated in Figure As shown, 111

130 during the entire cycle, the efficiency varies between 20% and 24%. The average cumulative efficiency, as measured from the cycle s commencement, is shown in Figure The simulation revealed a total well-to-wheel efficiency, at the FTP-75 cycle s conclusion, of 22.3%. Obviously, this is the result of the many steps involved in energy transfer: converting natural gas to methanol, reforming methanol to hydrogen, generating DC power from this hydrogen using a fuel cell, inverting DC power to AC, converting AC power to mechanical motion, and transmitting this mechanical motion down the driveline of a vehicle. To put this well-to-wheel efficiency in perspective, Figure 6.12 is presented. The well-to-wheel efficiency for several different vehicle configurations, from four different studies, is presented. As illustrated, both the current study and the study of Mizsey et al. yield well-to-wheel efficiencies for the methanol reforming FCV in the low 20% range. While this is a marked improvement over the conventional gasoline internal combustion engine (ICE), it is only slightly competitive over current diesel ICE technology. Furthermore, the methanol reforming FCV is inferior to hybrid gasoline, hybrid diesel, and direct hydrogen FCV technology. In particular, the direct hydrogen hybrid FCV shows very high well-to-wheel efficiencies. It is clear from this analysis that while the IDMFC system shows great promise in certain applications, for automotive power, considerable improvements are essential to be competitive and cost effective against other vehicle technologies. 112

131 Table 6.1. Operating Conditions for the DMFC and IDMFC. DMFC IDMFC Pressure 0 psig 30 psig Temperature 75 C Cell: 80 C Reformer: 230 C CO Clean-up: 150 C Anode Humidification N/A Direct from Reformer Cathode Humidification 0 % 0% Anode Stoichiometry Cathode Stoichiometry Membrane Electrode Assembly Lynntech 115 with Carbon Cloth GDL Gore MEA with Carbon Paper GDL Total Platinum Loading 6 mg/cm mg/cm 2 Methanol Solution Strength Notes: 1 M 15.2 M Requires an air supply to reformer and 10% anode air injection 113

132 Table 6.2. Cost comparison for both systems various power levels. Portable Electronics DMFC IDMFC Power, W Power Density, W/cm Cell Area, cm Platinum Loading, mg/cm Total Pt Loading, g Platinum Cost, $ $55 $0.53 Auxiliary Power DMFC IDMFC 5,000 5, ,758 8, $13,636 $131 Automotive DMFC IDMFC 80,000 80, ,212, , , $218,182 $2,102 Per kw DMFC IDMFC 1,000 1, ,152 1, $2,727 $26 *Based on a platinum cost of $30 per gram. 114

133 Table 6.3. System efficiency comparisons: Methanol to DC Electric Power. DMFC IDMFC Methanol Feed ml/min ml/min Methanol Feed W 110 W Total Peak Cell Power 3.3 W 28.6 W Conversion Efficiency 1.95 % % Current System Efficiency (including projected parasitic losses) Efficiency Breakdown: Cell Voltaic Efficiency Fuel Utilization Reformer Efficiency Parasitic Efficiency 1.75 % % % N/A 90.0 % % % % 82.0 % Possible System Efficiency (assuming MeOH fuel recirculation for the DMFC, and 90% utilization for the IDMFC) % % 115

134 Table 6.4. Vehicle specifications for IDMFC automobile. Vehicle mass 1382 kg Frontal Area 1.5 m 2 Coefficient of Drag 0.35 Rolling Resistance Fuel Cell Stack Number of Cells Active Area per Cell Pressure Anode Air Injection Peak Power Density Gross Peak Power cm 2 30 psig 10% W/cm 2 80 kw Reformer Efficiency 95.8 % Drivetrain Efficiency 83.6 % Inverter/Motor Efficiency 92 % Fuel Utilization 90 % Parasitic Losses 18 % Well-to-Tank Efficiency 72 % 116

135 Figure 6.1. Comparison of the cell polarization curves for the direct (DMFC) and indirect (IDMFC) fuel cell systems. 117

136 Figure 6.2. Comparison of the cell power density curves for the direct (DMFC) and indirect (IDMFC) fuel cell systems. 118

137 Figure 6.3. Historic cost of platinum. 119

138 Figure 6.4. Power density and efficiency (not including parasitic losses) for the indirect methanol fuel cell system (IDMFC) across the entire polarization curve. 120

139 Figure 6.5. Power density and efficiency (not including parasitic losses) for the direct methanol fuel cell system (DMFC) across the entire polarization curve. 121

140 Figure 6.6. FTP-75 Driving Cycle. 122

141 Figure 6.7. Vehicle power requirements during the FTP-75 driving cycle. 123

142 Figure 6.8. Cell voltage during drive cycle. 124

143 Figure 6.9. Cell current density during drive cycle. 125

144 Figure Instantaneous well-to-wheel efficiency during drive cycle. 126

145 Figure Cumulative well-to-wheel efficiency during drive cycle. 127

146 Figure Comparison of well-to-wheel efficiencies from various studies. 128

147 CHAPTER 7 CONCLUSION AND FUTURE WORK 7.1 Conclusion Hydrocarbon reforming, whether on-board a vehicle, at a fueling station for hydrogen generation, or as a sub-system with a stationary power fuel cell, has great potential over direct hydrogen distribution and storage. To that end, a comprehensive study of a methanol reforming polymer electrolyte fuel cell system has been conducted. As summary of the work, and the major findings, are as follows: 1. To facilitate the vehicle packaging constraints caused by an on-board reformer, the performance of a fuel cell operating at low humidity and pressure conditions was investigated. It was found that while cell current density output is a strong function of pressure, the dependence on cathode humidification is much less. In fact, cathode humidification can be eliminated without any appreciable performance loss. This low pressure and humidity operation has the potential for eliminating compressors and humidifiers from a fuel cell system, and thus liberating room for reforming systems. 2. The generation of an effluent dilute in hydrogen and containing trace quantities of carbon monoxide (CO) is a resultant of any hydrocarbon reforming process. The transient process of CO poisoning was investigated, and found to be not only dependent 129

148 on the level of CO within the anode feed stream, but also on the level of hydrogen dilution. The combined effect of hydrogen dilution and CO was found to be much more detrimental on cell performance than either of the two individually. Hydrogen dilution was found to increase both the rate and extent of CO poisoning. This finding was confirmed using a transient model of hydrogen and CO adsorption, desorption, and electro-oxidation on the anode catalyst surface. 3. To deal with the negative effects of CO, an air bleed remediation system was integrated with the fuel cell. The usefulness of air bleed with simulated reformate gas was studied. In this investigation, the poisoning effect was greatly reduced through the use of a 2% air bleed. This was despite hydrogen concentrations as low as 40% and CO quantities as high as 100ppm. At higher CO levels, air bleed was found effective, but not as successful as with the 100ppm case. A dependency on both the amount of air bleed and the CO to hydrogen ratio was found. This dependency was confirmed utilizing a CO poisoning model combined with an air bleed kinetics model. This model described oxygen adsorption and desorption, as well as oxidation of CO and hydrogen on the anode catalyst layer. 4. With a comprehensive knowledge of the CO poisoning process, and how to prevent it, a single stage oxidative methanol steam reformer was integrated with the PEM fuel cell and air bleed system. Utilizing the reformer feed alone, peak power densities were very low. However, when combined with an air bleed system, 70-76% of the peak power obtainable from pure hydrogen was achievable with the reformer feed. With the 130

149 methanol reformer fed fuel cell, the efficiency in converting methanol to electricity was evaluated. Neglecting parasitic losses, approximately 25% of the HHV available in methanol was extracted as DC power. 5. The efficiency of this indirect methanol fuel cell system (IDMFC) was compared to that of a direct methanol fuel cell (DMFC). It was found that for portable applications, where size is of primary concern, the DMFC proved far superior. However, for auxiliary power applications, where efficiency is of prime concern, both systems have roughly equal potential for high efficiency operation. However, due to the high platinum costs associated with the DMFC, the IDMFC system is desirable. For automotive applications, where efficiency, size, and cost are all of concern, the IDMFC is extremely favorable over the DMFC. The IDMFC system was analyzed for efficiency in completing an FTP- 75 driving cycle. Over the entire drive cycle, a well-to-wheel efficiency of 22.3% was achieved. 7.2 Recommendations for Future Work There exist several promising opportunities for future inquiry based on the current investigation of a methanol reforming fuel cell system. The present study investigated the effects of simulated reformate gas across a wide range of hydrogen and CO concentrations. However, methanol was the only hydrocarbon actually reformed and fed to the fuel cell. As opposed to methanol, gasoline and diesel reformate typically contain lower hydrogen concentrations and higher CO ppm levels. A study investigating the 131

150 transient effects of reforming these higher hydrocarbons, as well as the resulting fuel cell performance, would be of great interest. Also, these higher hydrocarbons typically contain appreciable concentrations of sulfur compounds. A fundamental study of anode catalyst layer poisoning due to sulfur compounds would be of great importance. The air bleed system developed here, and investigated for effectiveness with methanol reformate gas, also has potential for further examination. For example, platinum-ruthenium anode catalysts show great promise for improved CO tolerance over existing Pt catalysts. In addition, work is being conducted to resolve the more complex CO poisoning mechanism on such a catalyst surface. As a method for CO remediation, a fundamental study could be undertaken to investigate the effects of anode air injection with Pt-Ru catalysts. An air bleed model could be built upon current CO adsorption, desorption, and electro-oxidation models on Pt-Ru catalysts. The addition of an oxygen adsorption, desorption, and chemical oxidation model to the existing theory would greatly increase the fundamental understanding of both poisoning and recovery processes. An experimental study could be undertaken to support or disprove this model. On the more practical side, a closed loop system can be envisioned which would actively monitor CO concentrations and cell performance. This system could be optimized to inject specific quantities of air based on the measured conditions. Such a system would be of great practical importance to the implementation of fuel cells. 132

151 REFERENCES Adams, J., Yang, W., Oglesby, K., Osborne, K., 2000, The Development of Ford s P2000 Fuel Cell Vehicle, Society of Automotive Engineers, Allard, M., Issues Associated with Widespread Utilization of Methanol, Society of Automotive Engineers, An, F., Santini, D., Assessing Tank-to-Wheel Efficiencies of Advanced Technology Vehicles, Society of Automotive Engineers, Anantaraman, A., Gardner, C., 1996, Studies on ion-exchange membranes, Journal of Electroanalytical Chemistry, 414, 115. Ashley, S., Fuel Cells Start to Look Real, Automotive Engineering International, March Baschuk, J., Xianguo, L., 2003, Modeling CO poisoning and O 2 bleeding in a PEM fuel cell anode, International Journal of Energy Research, 27, Berlowitz, P., Darnell, C., 2000, Fuel Choices for Fuel Cell Powered Vehicles, Society of Automotive Engineers,

152 Bhatia, K.K., Wang, C.Y., 2004, Transient carbon monoxide poisoning of a polymer electrolyte fuel cell operating on diluted hydrogen feed, Electrochimica Acta, 49, Brown, L., 2001, A comparative study of fuels for on-board hydrogen production for fuel-cell-powered automobiles, International Journal of Hydrogen Energy, 26, 381. Carney, D., Jost, K., GM fuel-cell-vehicle advances, Automotive Engineering International, September Carpenter, I, Edwards, N., Ellis, S., Frost, J., Golunski, S., van Keulen, N., Petch, M., Pignon, J., Reinkingh, J., 1999, On-board Hydrogen Generation for PEM Fuel Cells in Automotive Applications, Society of Automotive Engineers, Chan, S., Goh, S., Jiang, S., 2003, A mathematical model of polymer electrolyte fuel cell with anode CO kinetics, Electrochimica Acta, 48, Chorkendorff, I., Niemantsverdriet, J., Concepts of Modern Catalysis and Kinetics, Wiley-VCH, Weinheim, Chu, D., Jiang, R., 1999, Performance of polymer electrolyte membrane fuel cell (PEMFC) stacks, Journal of Power Sources, 83,

153 Divisek, J., Oetjen, H., Peinecke, V., Schmidt, V., Stimming, U., 1998, Components for PEM fuel cell systems using hydrogen and CO containing fuels, Electrochimica Acta, 43, Docter, A., Lamm, A., 1999, Gasoline fuel cell systems, Journal of Power Sources, 84, 194. Dudfield, C., Chen, R., Adcock, P., 2001, A carbon monoxide PROX reactor for PEM fuel cell automotive application, International Journal of Hydrogen Energy, 26, 763. Fronk, M., Wetter, D., Masten, D., Bosco, A., 2000, PEM Fuel Cell System Solutions for Transportation, Society of Automotive Engineers, Ge, S., Yi, B., 2003, A mathematical model for PEMFC in different flow modes, Journal of Power Sources, 124, 1. Geissler, K., Newson, E., Vogel, F., Truong, T., Hottinger, P., Wokaum, A., 2001, Autothermal methanol reforming for hydrogen production in fuel cell applications, Physical Chemistry Chemical Physics, 3, 289. Glassman, I., Combustion, 3 rd Edition, Academic Press,

154 Gonder, J., Direct Methanol Fuel Cells for Automotive Applications: Experimental Analysis of Transient Characteristics, Periodic Fuel Injection, and Fuel Quality Requirements, Pennsylvania State University, Gottesfeld, S., Pafford, J., 1988, A new approach to the problem of carbon monoxide poisoning in fuel-cells operating at low temperatures, Journal of the Electrochemical Society, 135, Gu, W., Shen, J., Song, C., 2003, Hydrogen production from integrated methanol reforming over Cu-ZnO/Al 2 O 3 and Pt/Al 2 O 3 catalysts for PEM fuel cells, American Chemical Society, Division of Fuel Chemistry, 48, 804. Hagan, M., Northrop, W., Bowers, B., Rumsey, J., Prabhu, S., 2000, Automotive Fuel Processing Systems for PEM Fuel Cells, Society of Automotive Engineers, Han, J., Kim, I., Choi, K., 2000, Purifier-integrated methanol reformer for fuel cell vehicles, Journal of Power Sources, 86, 223. Huang, T., Wang, S., 1986, Hydrogen Production via Partial Oxidation of Methanol Over Copper-Zinc Catalysts, Applied Catalysis, 24, 287. Jiang, R., Chu, D., 2001, Stack design and performance of polymer electrolyte membrane fuel cells, Journal of Power Sources, 93,

155 Johnson Matthey Inc., Kirk-Othmer Encyclopedia of Chemical Technology, 4 th Edition, New York, Knights, S., Colbow, K., St-Pierre, J., Wilkinson, D., 2004, Aging mechanisms and lifetime of PEFC and DMFC, Journal of Power Sources, 127, 127. Kordesch, K., Simader, G., Fuel Cells and their Applications, VCH, Weinheim, Larminie, J., Dicks, A., Fuel Cell Systems Explained, Wiley, Chichester, McNicol, B., Rand, D., Williams, K., 2001, Fuel cells for road transportation purposes yes or no?, Journal of Power Sources, 100, 47. Miremadi, B., Yadav, T., Zhang, J., Falconer, J., 2000, Low-temperature hydrogen production using electrically activated catalysts, Chemical Communications, Mizsey, P., Newson, E., 2001, Comparison of different vehicle power trains, Journal of Power Sources, 102, 205. Moffat, R., 1988, Describing the Uncertainties in Experimental Results, Experimental Thermal and Fluid Science, 2,

156 Moore, R., Direct Methanol Fuel Cells for Automotive Power Systems, Society of Automotive Engineers, Moore, R., Gottesfeld, S., Zelenay, P., A Comparison Between Direct-Methanol and Direct-Hydrogen Fuel Cell Vehicles, Society of Automotive Engineers, Murthy, M., Esayian, M., Hobson, A., MacKenzie, S., Lee, W., Van Zee, J., 2001, Performance of a Polymer Electrolyte Membrane Fuel Cell Exposed to Transient CO Concentrations, Journal of the Electrochemical Society, 148, Natarajan, D., Nguyen, T., 2003, Three-dimensional effects of liquid water flooding in the cathode of a PEM fuel cell, Journal of Power Sources, 115, 66. Oetjen, H., Schmidt, V., Stimming, U., Trila, F., 1996, Performance Data of a Proton Exchange Membrane Fuel Cell Using H 2 /CO as Fuel Gas, Journal of the Electrochemical Society, 143, Qi, C., Pollica, D., Hagan, M., Cole, T., Rizzo, V., 2000, Transient CO Preferential Oxidation (PrOx) for PEM Fuel Cell, Society of Automotive Engineers, Rohland, B., Plzak, V., 1999, The PEMFC-integrated CO oxidation a novel method of simplifying the fuel cell plant, Journal of Power Sources, 84,

157 Rousseau, A., Ahluwahlia, R., Deville, B., Zhang, Q., Well-to-Wheels Analysis of Advanced SUV Fuel Cell Vehicles, Society of Automotive Engineers, Springer, T. E., Rockward, T., Zawodzinski, T. A., Gottesfeld, S., 2001, Model for Polymer Electrolyte Fuel Cell Operation on Reformate Feed, Journal of the Electrochemical Society, 148, A11. Staschewski, D., Mao, Z., 1999, Hydrogen-air PEMFC operation with extraordinarily low gas pressures and internal humidification conception and experimental prototype stack, International Journal of Hydrogen Energy, 24, 543. Stodolsky, F., Gains, L., Marshall, C., An, F., Eberhardt, J., 1999, Total Fuel Cycle Impacts of Advanced Vehicles, Society of Automotive Engineers, Tamura, H., Ura, M., Sasaki, M., Imaseki, T., Oshiage, K., Hirota, T., 2000, Development of the Nissan Fuel Cell Vehicle, Society of Automotive Engineers, Trimm, D., Onsan, Z., 2001, Onboard Fuel Cell Conversion for Hydrogen-Fuel-Cell- Driven Vehicles, Catalyst Reviews, 43,

158 Um, S., Wang, C. Y., Chen, K. S., 2000, Computational Fluid Dynamics Modeling of Proton Exchange Membrane Fuel Cells, Journal of the Electrochemical Society, 147, U.S. Department of Energy, 2002 Annual Progress Report, Hydrogen, Fuel Cells, and Infrastructure Technologies Program, Office of Transportation Technology, Wagner, N., Schulze, M., 2003, Change of electrochemical impedance spectra during CO poisoning of the Pt and Pt-Ru anodes in a membrane fuel cell (PEFC), Electrochimica Acta, 48, Yamaguchi, J., Concepts from Tokyo, Automotive Engineering International, December Zawodzinski, T., Derouin, C., Radzinski, S., Sherman, R., Smith, V., Springer, T., Gottesfeld, S., 1993, Water Uptake and Transport through Nafion 117 Membranes, Journal of the Electrochemical Society, 140, Zhang, J., Thampan, T., Datta, R., 2002, Influence of Anode Flow Rate and Cathode Oxygen Pressure on CO Poisoning of Proton Exchange Membrane Fuel Cells, Journal of the Electrochemical Society, 149, A

159 APPENDIX A: Experimental Uncertainty The issue of experimental uncertainty in fuel cell performance measurement is rarely discussed in the literature. Establishing the range of possible uncertainty is critical in determining what variations in measured cell performance can be attributed to different operating conditions, and what variation can be attributed to measurement and experimental error. To address this issue, it was decided to conduct a series of experiments to establish a baseline estimation of the total possible uncertainty in fuel cell performance measurements. For simplicity s sake, rather than analyzing the entire system step-by-step and conducting a single point uncertainty analysis (Moffat, 1988), it was decided to conduct a multiple sample statistical analysis of uncertainty. This method works well in determining the total so-called random and variable but deterministic error, but the relative amounts of each are still unknown. Also, this method has the drawback of not being able to quantify the so-called fixed error (i.e. consistent and repeatable voltage measurement could be observed, but an incorrect calibration of the voltage measuring devices will result in an incorrect reading). Despite these drawbacks, the technique is still valuable in getting a handle on the total random and variable but deterministic error. In terms of the fixed error, the Arbin system itself is assumed to be in good calibration. Furthermore, the system has a voltage and current measurement resolution of much less than 1 mv and 1 ma respectively. Therefore, for this study, the fixed error in the Arbin current and voltage measurement system is assumed to be extremely small. 141

160 A 50 cm 2 cell, shown in Figure 1.2, was run under the operating conditions given in Table A.1. Once the cell reached a quasi-steady state condition under constant cell operating conditions, roughly 1000 quick scans of the entire cell polarization curve were performed over a 48 hour period. This was done to ensure a large sample size on which to base the statistical analysis. All of these polarization curves are shown in Figure A.1. The data was analyzed assuming a normal distribution for the variation in measured voltage at any specific current. The average polarization curve is shown in Figure A.2, along with error bars indicating the total 4 sigma range (+/- 2 standard deviations) for specific current values. According to a normal distribution, this range should encompass 95% of all the observed values. The size of half this range (2 sigma), measured as a percent of the average value, is plotted verses cell current density in Figure A.3. Due to the very large sample size, the 99% confidence interval for the mean polarization curve varied between 1.7 and 2.4 mv, depending on the location within the curve. As shown in Figure A.3, at low current densities, the variation is roughly +/- 2 % of the average value (and assuming no fixed uncertainties, the true value). However, at higher current draws, the uncertainty approaches nearly +/- 13 %. From this basic statistical uncertainty analysis, it is clear that as a cell approaches high current draw, variations in performance of less than 13% can not be blindly attributed to, say, an experimentally varied operating condition or gas composition. For example, the data shown in Figure 3.1 compares the performance of pure hydrogen to that of hydrogen diluted down to 40% by nitrogen. Based on the experiment alone, it is not possible to attribute the slight difference between the two curves to the single experimentally varied parameter, namely hydrogen concentration. All that is known is the two performance 142

161 curves fall within the known experimental uncertainty range, and thus no conclusions about the positive or negative effect of hydrogen dilution should be drawn. Likewise, all performance data presented here should be analyzed with this level of scrutiny in mind. 143

162 Table A.1. Experimental conditions used for uncertainty analysis. Membrane Electrode Assembly Lynntech 112 Cell Temperature 80 C Cell Pressure 30 psig Anode Relative Humidity 100% Cathode Relative Humidity 100% Anode Stoichiometry 1.5 Cathode Stoichiometry

163 Current Density, ma/cm 2 Figure A.1. Variation in cell polarization curve for 1000 runs. 145

164 Figure A.2. Typical cell polarization curve showing the +/- 2σ uncertainty range. 146

165 Figure A.3. Variation of 2σ range, as a percentage of average value, with cell current density. 147

166 APPENDIX B: Scanning Electron Photographs of Membrane Electrode Assembly The images displayed here are of various regions of a hydrogen fuel cell membrane electrode assembly (MEA). The specifications for the MEA, supplied by W.L. Gore and Associates (Newark, DE), are given in Table 4.2. The images were captured using a Philips XL20 Scanning Electron Microscope. 148

167 Figure B.1. Fuel Cell Membrane Catalyst Layer. 149

168 Figure B.2. Fuel Cell Membrane Catalyst Layer, High Magnification. 150

169 Figure B.3. Gas Diffusion Layer. 151

170 Figure B.4. Gas Diffusion Layer, High Magnification. 152

171 Figure B.5. Micro-porous Structure. 153

172 Figure B.6. Micro-porous Structure, High Magnification. 154

173 Micro-porous Structure, Catalyst Later, and Membrane GDL GDL Figure B.7. Cross Section of Entire Membrane Electrode Assembly (Gas Diffusion Layer, Micro-porous Structure, Catalyst Layer, and Membrane). 155