OCT , 2006 COMSOL USERS CONF BOSTON, MA 1 Use of COMSOL Multiphysics for Optimization of an All Liquid PEM Fuel Cell MEA George H. Miley (Speaker), Nuclear, Plasma and Radiological Engineering E. D. Byrd Electrical & Computer Engineering University of Illinois at Urbana-Champaign Urbana, IL USA

OCT , 2006 COMSOL USERS CONF BOSTON, MA2 Outline  NaBH 4 /H 2 O 2 Fuel Cell  Description of Model Physical Layout Electrical Considerations Mass/Momentum Balance Considerations  COMSOL Application Mode coupling  Pressure Differential Simulations and Results  Land Area vs. Permeability and Conductivity Simulation and Results  Conclusions

OCT , 2006 COMSOL USERS CONF BOSTON, MA3 NaBH 4 /H 2 O 2 Fuel Cell  Use in fuel cells is a relatively new development  H 2 /H 2 O 2 and NaBH 4 /H 2 O 2 cells were investigated at NPL Associates, Inc., the University of Illinois (UIUC), and elsewhere  Have shown great results, demonstrating the general feasibility of a peroxide based cell  Excellent potential for space applications due to high power density and air (oxygen) independence.

OCT , 2006 COMSOL USERS CONF BOSTON, MA4 UIUC/NPL Direct Peroxide Fuel Cells  The sodium borohydride/hydrogen peroxide reactions. Anode: Cathode:

OCT , 2006 COMSOL USERS CONF BOSTON, MA5 15-W NaBH 4 /H 2 O 2 Test Fuel Cell as assembled. The 15-W cell shown here uses an integrated cooling channel to dissipate the waste heat generated in the relative small 25-cm 2 active area. An optimized version of this small cell generated 36-W at ~ 60ºC, representing the highest power density reported to date for a small fuel cell working at sub-100  C. Test Cells - Compact 1-30 W Power Units Flow rate of approximately 200 cm3/min Minimal pressure drop even with parallel flow due to low flow rate Temperature rise of approximately 15°C Heat flux is approximately equal to electrical power (500-W)

OCT , 2006 COMSOL USERS CONF BOSTON, MA6 The 500-W UIUC/NPL NaBH 4 /H 2 O 2 Fuel Cell Stack The active area per cell was 144 cm 2 and 15 cells were employed to provide a total stack active area of 2160 cm 2.

OCT , 2006 COMSOL USERS CONF BOSTON, MA7 UIUC/NPL Direct Peroxide Fuel Cells

OCT , 2006 COMSOL USERS CONF BOSTON, MA8 Objectives for COMSOL modeling  Gain insight into behaviors governing flow and current distributions  Determine space (diffusion layer parameters, conductivity effects, flow channel and land dimensions) for detailed optimization physics  Guide future design improvements

OCT , 2006 COMSOL USERS CONF BOSTON, MA9 Model Description- geometry Physical Layout  Based on repetitive cross section of MEA and flow channels.  Outlined area represents the physical model.  Portion of graphite plates included to see the current density in the plate and to be able to vary their conductivity.

OCT , 2006 COMSOL USERS CONF BOSTON, MA10 Model Description - electrical  Standard Electrical Model DC current conduction - applies to each section with different conductivity (graphite, diffusion layers, membrane) Butler-Volmer Equations Anode:Cathode:

OCT , 2006 COMSOL USERS CONF BOSTON, MA11 Modified Bulter-Volmer  The Butler-Volmer equation was modified to obtain an alternative version that is more robust when solving numerically in Comsol. In this version, the hyperbolic identity of Eq. 2-5 is used to form Eq  (2-5)  (2-6)

OCT , 2006 COMSOL USERS CONF BOSTON, MA12 Model Description – conservation equations  Mass Balance  Momentum Balance – Darcy’s Law

OCT , 2006 COMSOL USERS CONF BOSTON, MA13 COMSOL Application Mode Coupling

OCT , 2006 COMSOL USERS CONF BOSTON, MA14 Parameters used  Necessary parameters (other than exchange current and equilibrium potentials, discussed next) were acquired through experimental means or published values These include the conductivities, permeabilities, diffusion coefficients, and viscosities given in the following table.

OCT , 2006 COMSOL USERS CONF BOSTON, MA15 Parameter set 1

OCT , 2006 COMSOL USERS CONF BOSTON, MA16 Parameter set 2 - determination of the exchange current density and reversible potential  A Hydrogen half-cell was constructed and used to determine the exchange current density and additional parameters such as the Tafel slope in the Butler-Volmer eqns.. The reversible potential of each cell half was determined using the Gibb’s Free Energies applied to the reactants and products in each reaction.

OCT , 2006 COMSOL USERS CONF BOSTON, MA17 Model verification: I -V Curve calculated for the reference case agrees well with corresponding experiment – model next used to explore design changes

OCT , 2006 COMSOL USERS CONF BOSTON, MA18 Simulations – Pressure Differential  Vary the pressure differential between the two flow channels.  Reasons Different flow velocities create different pressure differences Different locations have different pressure drops

OCT , 2006 COMSOL USERS CONF BOSTON, MA19 Simulations – Pressure Differential- Higher values optimal  Results Low pressure drops cause less permeation in the diffusion layer, causing mass transport losses. High pressure drops allow reactants to easily reach under the land area.  Reactant permeation under flow channel depends on fluid velocity and location along channel.

OCT , 2006 COMSOL USERS CONF BOSTON, MA20 Simulations – Land Area selection  Current collector land area width to flow channel width ratio is varied (collector + channel widths = constant). Land Area Width varied while also varying diffusion layer permeability. Land Area Width varied while also varying diffusion layer conductivity.

OCT , 2006 COMSOL USERS CONF BOSTON, MA21 Simulations – Land Area – high permeability give flexibility in width  Max Power at different Permeability with varying land areas.  Low Permeability diffusion layers have optimum current collector land area to flow channel ratio.  High Permeability diffusion layers function well with wide current collector widths. Permeability Maximum Power vs. Land Area Width

OCT , 2006 COMSOL USERS CONF BOSTON, MA22 Simulations – Land Area – optimum with high conductivity and equal width design  Max Power at different Conductivity with varying land areas.  High conductivity diffusion layers have are optimum with equal width current collectors and flow channels.  Low conductivity diffusion layers function better with wider land areas and narrower flow channels. Conductivity Maximum Power vs. Land Area Width

OCT , 2006 COMSOL USERS CONF BOSTON, MA23 Conclusions  Simulations performed of all-liquid PEM fuel cell using COMSOL Multiphysics.  Normalization uses data from half cell for io and Vrev.  Pressure differentials, conductivities, permeabilities, and current collector widths varied in the simulations.  Cell performance varies with different flow velocities and along the flow channels.  Optimum current collector widths predicted for diffusion layers with known conductivities and permeabilities.  Model is very useful for optimization in region around normalization.  Simulations narrow region for experimental studies to zone near optimum performance. Greatly reduces time and expense of experimental studies.

OCT , 2006 COMSOL USERS CONF BOSTON, MA24 Acknowledgement We would like to thank: NPL Associates, Inc. for their support with starting the project. E. Byrd wishes to acknowledge fellow researchers N. Luo, J. Mather, G. Hawkins, and L. Guo for their help. This research was supported by DARPA SB Continuing studies are supported by DARPA/AFRL.

OCT , 2006 COMSOL USERS CONF BOSTON, MA25 Thank You Dr. George. H. Miley UIUC Phone: (217) Ethan D. Byrd UIUC For more information please contact: