FUEL CELL CATHODE AIR FILTERS: METHODOLOGIES 
FOR DESIGN AND OPTIMIZATION 
 
Except where reference is made to the work of others, the work described in this thesis is 
my own or was done in collaboration with my advisory committee. This thesis does not 
include proprietary or classified information. 
 
 
________________________________ 
Daniel Mabry Kennedy 
 
 
 
Certificate of Approval: 
 
 
_______________________                                                    _______________________ 
Gopal A. Krishnagopalan               Bruce J. Tatarchuk, Chair                             
Professor               Professor 
Chemical Engineering               Chemical Engineering                                   
  
                                                        
 
_______________________                                                    _______________________ 
R. Mark Nelms               Joe F. Pittman 
Professor                     Interim Dean  
Electical Engineering                                                              Graduate School 
 
 
 
 
FUEL CELL CATHODE AIR FILTERS: METHODOLOGIES 
FOR DESIGN AND OPTIMIZATION 
 
Daniel Mabry Kennedy 
 
 
A Thesis  
Submitted to 
the Graduate Faculty of 
Auburn University 
in Partial Fulfillment of the  
Requirements for the  
Degree of  
Master of Science 
 
 
 
 
 
Auburn, Alabama 
May 10, 2007
 iii
FUEL CELL CATHODE AIR FILTERS: METHODOLOGIES 
FOR DESIGN AND OPTIMIZATION 
 
Daniel Mabry Kennedy 
Permission is granted to Auburn University to make copies of this thesis at its discretion, 
upon the request of individuals or institutions and at their expense.  The author reserves 
all publication rights. 
 
 
 _____________________ 
 Signature of Author 
 
 _____________________ 
 Date of Graduation 
 iv
THESIS ABSTRACT 
FUEL CELL CATHODE AIR FILTERS: METHODOLOGIES 
FOR DESIGN AND OPTIMIZATION 
Daniel Mabry Kennedy 
Master of Science, May 10, 2007 
(B.S., Auburn University, 2004) 
 
98 Typed Pages 
Directed by Bruce Tatarchuk 
Platinum catalyst used in PEM fuel cells experience performance degradation 
such as reduction in efficiency and life as a result of airborne contaminants.  Research on 
these contaminant effects suggests that the best possible solution to allowing fuel cells to 
operate in contaminated environments is by filtration of the harmful contaminants from 
the cathode air.  A cathode air filter design methodology was created that considers the 
properties of the cathode air stream, fuel cell attributes, and filter options to optimize the 
filter design process.  Optimization of the filter requires an understanding of the balance 
that must be made between the loss in power due to poisoning of the platinum catalyst 
and a loss in fuel cell efficiency created by an increase in parasitic power required to 
operate the compressor.  The model was successfully applied to a 1.2kWe fuel cell.  
Results show that the optimal filter design is dependent on both the total logs of removal 
required and the total capacity required.  A novel filter media, microfibrous materials, 
 v
provides the thinnest possible bed depth and lowest parasitic power requirements for 
cases requiring high logs removal and low capacity.  Packed beds provide the best 
solution for high inlet concentrations and/or long breakthrough time applications.  If both 
a high contacting efficiency and capacity are required, an optimized composite bed 
provides the ideal solution by utilizing the advantages of both filter types. 
 
 
 
 vi
Style Manual Used:  Chemical Engineering Progress Style Guide, available at ___
www.cepmagazine.org/editorial/refstyle.htm____________________________________ 
Computer Software Used:  Microsoft Word, Microsoft Excel, Microsoft ________ 
PowerPoint, Pro-RAE Suite, Maple 7.0 ________________________________________ 
 
 vii
TABLE OF CONTENTS 
 
LIST OF TABLES...........................................................................................................ix 
 
LIST OF FIGURES .......................................................................................................... x 
 
I. INTRODUCTION ................................................................................................ 1 
 
II. BACKGROUND INFORMATION ..................................................................... 3 
 Fuel Cells .............................................................................................................. 3 
  Fuel Cell Applications .............................................................................. 4 
  Proton Exchange Membrane Fuel Cell Technology................................. 5 
  Contaminate Effects on PEM Fuel Cells .................................................. 6 
 Adsorption............................................................................................................. 9 
 Microfibrous Sorbent Supported Media (MSSM) ................................................ 9 
 
III. EXPERIMENTAL APPARATUS...................................................................... 14 
 Ballard NEXA Fuel Cell..................................................................................... 14 
 Compressor Testing Apparatus........................................................................... 16 
 Breakthrough Testing Apparatus ........................................................................ 19 
 
IV. DESIGN METHODOLOGY AND ALGORITHM ........................................... 22  
 Fuel cell system design discussion and design considerations ........................... 22 
  Fuel cell system discussion..................................................................... 22 
  Filter Design Considerations................................................................... 23 
   Inlet Air Properties...................................................................... 24 
   Filter Design Options.................................................................. 25 
   Filter Footprint............................................................................ 25 
   Adsorptive Filter Parameters ...................................................... 26 
   Compressor Attributes ................................................................ 26 
   Fuel Cell Stack Attributes........................................................... 27 
   Cathode Air Filter Optimization ................................................. 28 
 Design Equations for Describing Fuel Cell Attributes and Filter Parameters.... 29 
  Fuel Cell Air Flow Requirements........................................................... 30 
  Adsorptive Filter Breakthrough Equations ............................................. 30 
  Pressure Drop Relationships................................................................... 32 
  Compressor Power and Efficiency.......................................................... 34 
 viii
  Compressor Power...................................................................... 34 
  Compressor Efficiency................................................................ 35 
  Compressor Performance Charts ................................................ 35 
 Design Algorithm................................................................................................ 36 
 
V. DESIGN ALGORITHM APPLICATION TO BALLARD FUEL CELL ......... 38 
 Application Overview......................................................................................... 38 
 Filter Design Options and Footprint ................................................................... 39 
  Filter Design Options.............................................................................. 39 
  Filter Footprint........................................................................................ 39 
 Air Flow Rate and Stack Pressure Drop ............................................................. 40 
  Stack Air Flow Requirements................................................................. 40 
  Stack Inlet Pressure Requirements.......................................................... 41 
 Compressor Data Collection............................................................................... 41 
  A Compressor Study Using a Compressor Testing Apparatus............... 42 
  A Compressor Study within the Ballard NEXA System ........................ 46 
  Determination of Exact Breakthrough Equation Constants.................... 47 
  Determining Filter Pressure Drop Relationships .................................... 50 
 Filter Design Process Results.............................................................................. 52 
  Case I: Inlet concentration 100 ppm, outlet concentration 5 ppm .......... 54 
  Case II: Inlet concentration 10 ppm, outlet concentration 0.1 ppm........ 56 
  Case III: Inlet concentration 100 ppm, outlet concentration 0.1 ppm .... 57 
  Case Study for a Pleated Microfibrous Filter ......................................... 58 
  Detailed Composite Bed Configuration Analysis................................... 60 
 
VI. FILTER CONSTRUCTION AND TEST RESULTS ........................................ 62 
 
 Fuel Construction................................................................................................ 62 
  Filter Frames ........................................................................................... 63 
  Adsorbent Materials................................................................................ 64 
  Final Filter Design .................................................................................. 65 
 Filter Tests Results.............................................................................................. 68 
  Composite Bed Filter .............................................................................. 68 
  Pleated Microfibrous Filter..................................................................... 70 
 Recommendations for Improvement to Algorithm............................................. 71 
 
VII. CONCLUSIONS AND FINAL COMMENTS .................................................. 73 
 
REFERENCES ............................................................................................................... 75 
 
APPENDIX A.     AIR FLOW DERIVATION.............................................................. 78 
 
APPENDIX B.     SAMPLE MAPLE OUTPUT FOR A COMPOSITE BED............... 79 
 
 ix
LIST OF TABLES 
 
Table II.1: A list of contaminants and their steady-state effects on fuel cell 
power output during operation with contaminants air and operation 
after using clean air................................................................................... 8 
 
Table V.1: Values used in filter design algorithm .................................................... 50 
 
Table V.2: Quantitative fuel cell attributes used in algorithm.................................. 53 
 
Table V.3: Filter design input parameters collected from experimental data........... 53 
 
Table V.4: Predicted, optimized individual layer thicknesses of the packed bed 
(L
1
) and microfibrous layer (L
2
) used to create a composite bed ........... 61 
 
Table VI.1: Quantitative fuel cell attributes used in algorithm.................................. 62 
 
Table VI.2: Filter design input parameters collected from experimental data........... 63 
 
Table VI.3: Values of layer thicknesses for optimized composite bed...................... 65 
 
 
 x
LIST OF FIGURES 
 
Figure II.1: A picture of a proton exchange membrane............................................... 6 
 
Figure II.2: Three rolls of microfibrous materials using fibers made of ceramic, 
metal, and polymers................................................................................ 10 
 
Figure II.3: A micrograph showing the three types of microfibrous materials 
and supported sorbents............................................................................ 11 
 
Figure II.4: Diagram showing benefit of composite bed ........................................... 12 
 
Figure III.1: Picture of Ballard NEXA Fuel Cell with the cathode air filter 
circled...................................................................................................... 15 
 
Figure III.2: Compressor testing apparatus used for testing the compressor 
outside of the fuel cell............................................................................. 17 
 
Figure III.3: Compressor controller diagram............................................................... 18 
 
Figure III.4: Breakthrough testing apparatus............................................................... 19 
 
Figure III.5: Filter testing apparatus for testing filters that fit into a Ballard 
NEXA fuel cell ....................................................................................... 21 
 
Figure IV.1: Fuel cell mass and energy balance showing the flow of air and 
power though the system ........................................................................ 23 
 
Figure IV.2: Hierarchal design considerations for cathode air filters ......................... 24 
 
Figure IV.3: Power using a 12/64? Inlet Orifice and 11/64? Outlet Orifice ............... 34 
 
Figure IV.4: Final, programmable design algorithm................................................... 37 
 
Figure V.1: Predicted air flow model compared with actual data.............................. 40 
 
Figure V.2: Compressor efficiencies.......................................................................... 44 
 
Figure V.3: Parasitic power required by the compressor to operate at given 
pressure ratios ......................................................................................... 45 
 xi
Figure V.4: Breakthrough curves for a 0.95 cm packed bed and a 0.4 cm 
microfibrous layer used to determine values for tau and capacity.......... 49 
 
Figure V.5: A breakthrough equation that represents k? values as the slope ............. 50 
 
Figure V.6: Pressure drop as a function of face velocity for a 0.95 cm packed 
bed and a 0.4 cm microfibrous layer....................................................... 51 
 
Figure V.7: Predicted power required to meet breakthrough time requirements....... 54 
 
Figure V.8: Predicted thickness required to meet breakthrough time 
requirements............................................................................................ 55 
 
Figure V.9: Predicted power required to meet breakthrough time requirements....... 56 
 
Figure V.10: Predicted thickness required to meet breakthrough time 
requirements............................................................................................ 56 
 
Figure V.11: Predicted power required to meet breakthrough time requirements....... 57 
 
Figure V.12: Predicted thickness required to meet breakthrough time 
requirements............................................................................................ 58 
 
Figure V.13: Predicted sheet thickness of a microfibrous pleated filter at a 
challenge concentration of 10 ppm hexane and an outlet 
concentration of 0.1 ppm hexane required to meet breakthrough 
time requirements.................................................................................... 59 
 
Figure V.14: Predicted compressor power requirements to operate with a 
microfibrous pleated filter at a challenge concentration of 10 ppm 
hexane and an outlet concentration of 0.1 ppm hexane required to 
meet breakthrough time requirements .................................................... 60 
 
Figure VI.1: Cut metal frame dimensions used for constructing a cathode air 
filter......................................................................................................... 64 
 
Figure VI.2: Predicted composite bed thickness required for hexane removal 
used to design a composite bed filter...................................................... 65 
 
Figure VI.3: Materials of construction for composite bed cathode air filter............... 66 
 
Figure VI.4: Picture of pleated cathode air filter......................................................... 67 
 
Figure VI.5: Actual breakthrough curves for composite bed compared to a 
predicted breakthrough curve by the algorithm...................................... 68 
 xii
Figure VI.6: Pressure drop comparison between a composite bed filter and 
values predicted by the algorithm........................................................... 69 
 
Figure VI.7: Actual breakthrough curves for a pleated filter compared to a 
predicted breakthrough curve by the algorithm...................................... 70 
 
Figure VI.8: Pressure drop comparison between a microfibrous pleated filter 
and values predicted by the algorithm .................................................... 71 
 1
I. INTRODUCTION 
 
Fuel cells are being actively considered for a wide range of applications including 
automobiles, uninterruptible power supplies, and battery replacements in the military for 
dismounted soldiers [1].  The reason for the interest in fuel cells is owed to their high 
power densities, low operating temperatures, low emissions, quiet operation, and their 
potential for system robustness [2].  However, air in these operating environments may 
contain contaminants that are damaging to fuel cell performance such as carbon 
monoxide, sulfur compounds, and volatile organic compounds (VOCs).  It has been 
suggested that the best method for addressing fuel cell air contamination is by the 
inclusion of adsorptive filtration with a cathode air filter [3].  Filters used in fuel cells 
must be optimized to be durable and effective, as well as easily adaptable to each specific 
need.  Optimization of cathode filter design requires careful consideration of fuel cell 
operating variables and available filter technology.  These design considerations were 
studied in detail and organized to create a process that provides a blueprint for designing 
and optimizing cathode air filters capable of allowing fuel cells to be operated in 
contaminated environments. 
Research on the durability of proton exchange membrane fuel cells has typically 
focused on the effects and the remediation of the effects of chemical contaminants on the 
hydrogen electrode, or anode.  This focus has been driven primarily by the intent to use 
 2
reformed hydrogen from fossil fuels containing contaminants toxic to the sensitive 
platinum catalyst.  However, due to the higher complexity of the oxygen reduction 
reaction when compared to reduction of hydrogen at the anode, the oxygen electrode, or 
cathode, requires twice the platinum (0.1 mg cm
-2
 as compared to 0.05 mg cm
-2
) and is 
therefore potentially more sensitive [4].  Recent studies have confirmed that polluted or 
otherwise contaminated environments negatively affect the performance of proton 
exchange membrane fuel cell cathodes.  For example, 20 ppm of carbon monoxide causes 
a temporary 4% reduction in fuel cell output.  Other contaminants, for example sulfur 
compounds such as H
2
S and SO
2,
 cause a more permanent effect reducing performance to 
as low as 30% of original output [3, 5].   
 Two methods for dealing with contaminated air effects on fuel cell cathodes are to 
increase catalyst durability and/or to filter contaminants from the air.   Because of the 
wide variety of contaminants, increasing membrane resilience by modifying the catalyst 
or increasing the catalyst amount is a difficult and possibly expensive alternative.  
Therefore, the most effective and flexible method for operating fuel cells in contaminated 
environments is adsorptive filtration of contaminants from the ambient air stream.   
The wide variety of filtration technology options allows for tailoring of cathode 
air filters to different contaminant types and concentrations.  Optimization of filter design 
requires a methodology incorporating air properties, fuel cell attributes, and design 
options.  A better understanding of how a cathode air filter should perform is 
accomplished through a detailed study of the design considerations and how they affect 
each other.   
 
 3
 II. BACKGROUND INFORMATION 
 
1. Fuel Cells 
 Interest in fuel cell technology has increased dramatically in the last 20 years, 
driven primarily on concerns about the condition of the environment and sources of 
energy.  The increase in world population and per capita energy requirements has led 
society to search for sources of energy that are both efficient and environmentally 
friendly.   Goals of energy research focus on energy creation that produces fewer and less 
harmful gases than traditional combustion which currently provides most of the world?s 
power.  Fuel cells potentially meet both environmental and efficiency goals.  Fuel cells 
operate by continually converting chemical energy to electrical energy through a reverse-
electrolysis process as long as fuel and oxidant are supplied.  One of the most common 
types of fuel cells, proton exchange membrane fuel cells (PEMFCs), uses hydrogen as a 
fuel source and produces only water as a byproduct.  PEMFCs are therefore considered a 
zero emission engine.  Compared to combustion engines, fuel cells are far more efficient 
due to the fact that they are not limited by the Carnot Efficiency.  Due to their low 
emissions, quiet operation and high power densities that they provide, there is a large 
focus to apply fuel cells to a variety of applications ranging from laptops to unmanned 
aerial vehicles [6].   
 
 4
1.1. Fuel Cell Applications 
 Although the process by which fuel cells operate has been understood since 1839, 
the first practical application of fuel cells wasn?t until the space program during the 
1960?s.  NASA used fuel cells as power generators and as a water supply for the Gemini 
and Apollo space vehicles.  Over the last few decades development of fuel cells has 
reached a level where they are now being studied for numerous applications.  Phosphoric 
acid fuel cells (PAFCs), molten carbonate fuel cells (MCFCs) and solid oxide fuel cells 
(SOFCs) are all researched as a co-generator along side current electrical power plants 
[3].  Proton exchange membrane fuel cells, also known as polymer electrolyte membrane 
fuel cells, show promising potential for more varied applications such as transportation, 
uninterruptible power supplies (UPS), and small portable power devices (laptops, cell 
phones, PDAs,) as a battery replacement.  Because of their low environmental impact, 
high power density, high conversion efficiency (50-55%), and long runtime capabilities, 
PEMFCs also show promise as electrical generators for users requiring reliable and high 
quality energy like hospitals, computer networks, and remote applications such as 
satellite dishes and communication towers.  PEM fuel cells are also the most practical 
fuel cell type for application as domestic or residential generators.  Military interest in 
using fuel cells ranges from unmanned aerial and underwater vehicles, portable battery 
chargers, and a replacement for batteries as a power source for dismounted soldiers, 
soldiers who fight on foot.  Dismounted soldiers increasingly rely on electronic 
technologies such as computers, personal radios, global positioning systems, head up 
displays and thermal imaging.  As battery technology nears its limit of technological 
development, it can no longer provide soldiers with necessary power at an acceptable 
weight [7].   
 Several companies currently research and produce prototype PEM fuel cells for 
commercial and residential applications.  Ballard has recently introduced a hydrogen 
powered stationary fuel cell power generator for use as an uninterruptible power supply 
in telecommunication markets.  The NEXA RM series is a modular 1 kWe fuel cell 
designed to meet a range of power requirements depending on its specific application [1].  
The base model NEXA fuel cell is a 1200 watt DC power supply operating on 99.9% 
pure hydrogen and clean air, while producing only water as a by product [8].   
1.2. Proton Exchange Membrane Fuel Cell Technology 
 The chemical reaction in PEM fuel cells is the same as that in combustion of 
hydrogen.   
OHOH
222
22 ?+                        (II.1) 
The difference is instead of heat being produced causing thermal expansion, 
electrical current is generated.  By creating electricity directly instead of through 
utilization of a combustion/expansion process, limitations described by the Carnot 
efficiency are removed, creating an increase in overall efficiency. 
There are two steps for this reaction to complete and an electrical current to be 
generated.  First, at the anode, hydrogen is ionized forming H
+
 and electrons. 
?+
+? eHH 442
2
 (Eq. II.2) 
Second, at the fuel cell cathode, oxygen reacts with the hydrogen, using the electrons, to 
form H
2
O.  
OHHeO
22
244 ?++
+?
 (Eq. II.3) 
 5
The key to the production of electricity is the polymer membrane between the anode and 
cathode.  The membrane is specifically designed to allow only the proton and not the 
electrons to pass through [9].  The electrons are required to follow a circuit potentially 
through a load to reach the hydrogen and oxygen reaction. 
Anode Cathode
V
+-
2H
2
O
2
2H
2
O
4H
+
+ O
2 
+ 4e
-
4H
+
4e
-
Proton Exchange 
Membrane (PEM)
Catalyst 
layer
 
Figure II.1: A picture of a proton exchange membrane 
 
1.3. Contaminant Effects on Fuel Cell Performance 
Providing a clean fuel for fuel cells is important, especially for fuel cells using a 
sensitive polymer electrolyte membrane.  The amount of research and published papers 
focusing on clean fuels supports this idea.  In many cases the hydrogen for PEMFCs is 
provided by reforming heavy hydrocarbons.  This leaves a large amount of contaminants 
 6
 7
in the hydrogen stream that must be removed before supplied to the fuel cell.  However, 
in recent years, researcher attention has moved from an almost exclusive focus on 
purifying hydrogen for fuel cell anodes, to also attempting to understand and remediate 
effects of contaminants in the oxidant provided to fuel cell cathodes.  Contaminant effects 
on fuel cell cathodes becomes very important when considering the intended applications 
of fuel cells ranging from polluted city centers to dusty battlefields.  For this reason, 
research began studying the effects of impurities on fuel cell performance to access the 
durability and capability of fuel cells to operate in more harsh environments than those 
found in a laboratory.  The following table has been compiled from several sources on the 
effects of various contaminants on the current output of the fuel cell.  Values of the 
percentage of normal current output were recorded both during challenge of the 
contaminant gas and following a flow of clean air which was provided until steady state 
has been achieved.  
 
 
 
 
 
 
 
 
 
 
 8
Table II.1 
A list of contaminants and their steady-state effects on fuel cell power output during 
operation with contaminants air and operation after using clean air 
 
Contaminant Concentration Percentage 
Output During 
Challenge 
Percentage 
Output After 
Recovery 
Source
Carbon Monoxide 20 ppm 96% 100% [2] 
Sulfure Dioxide 0.5 ppm 90% 90% [2] 
 2.5 ppm 47%  [10] 
5.0 ppm 22% 35% [10] 
Benzene 50 ppm (50mA/cm
2
) 95% 95% [2] 
 50 ppm (100mA/cm
2
) 93%  [2] 
 50 ppm (200mA/cm
2
) 72%  [2] 
Sulfur Mustard 15 ppm 13% 13% [2] 
Sarin 170 ppm 30% 30% [2] 
H
2
S 200 ppm 10% 30% [10] 
 
 
The effect of sulfur and organic compounds on fuel cell output has been 
correlated not only to its concentration, but also total dosage.  This is primarily due to its 
more permanent, poisoning effect [11].  Recovery of the fuel cell from sulfur adsorbed on 
the platinum catalyst must be achieved by cyclic voltammetry which oxidizes the sulfur 
releasing it from the platinum catalyst.  Cyclic voltammetry is a possible method of 
dealing with the effect of sulfur compounds on fuel cells; however, the equipment is 
expensive making pre-emptive filtration a more suitable means of dealing with 
contaminants [11].  Research on the effects of heavier VOCs and nerve agents at levels 
that might be associated with the release of a chemical agent in warfare conditions has 
revealed a gradual yet non-recoverable effect of these contaminants on the catalyst.  The 
conclusion from research on the contaminant effects on fuel cells is that, if ambient air is 
used as an oxidant source, mitigation against exposure to contaminants will most likely 
 9
involve the use of some filtration device rather than an increase in tolerance of the fuel 
cell catalyst [2].   
2. Adsorption 
 Adsorption occurs when one or more components of a gas or liquid are attracted 
to a solid adsorbent become lightly and sometimes reversibly bonded.  This attraction is 
sometimes called van der Waals adsorption [12].  The fluid is continually passed through 
a fixed bed until the solid particles become saturated or are no longer able to remove 
desired components at an acceptable rate.  Applications of gas phase adsorption include 
removal of volatile organic compounds (VOC?s), sulfur compounds, and odors from air.   
 Adsorbents have a large surface area to volume ratio by containing pores that 
typically reach up to 50% of total volume.  This results in surface areas of 100 to 2000 
m
2
/g.  Activated carbon is a commonly used adsorbent for organic molecules.  It is a 
microcrystalline material made by thermal decomposition of wood, vegetable shells, and 
other organic solids.  The resulting material has surface areas of 300 to 1200 m
2
/g and 
average pore diameters of 20 to 60?.  Other common adsorbents include silica gel, 
activated alumina, molecular sieve zeolites, and synthetic polymers of resins [12].  
3. Microfibrous Materials 
 Microfibrous materials are a patented class of materials developed at Auburn 
University that possess several advantageous properties when used for catalyst or sorbent 
applications.  They are composed of a sinter-locked network of micron diameter fibers 
made out of either polymer, metal, or ceramic and used to entrap catalysts/sorbent 
particulates.  Particulates entrapped in the materials are far smaller (158 ?m) than those 
used in a packed bed (765 ?m) creating a more accessible surface area.  Static mixing 
adds to the contacting efficiency and is the result of flow around the fibers reducing 
intrapartical heat and mass transfer resistance.  Therefore, when microfibrous materials 
are applied as a reaction/sorption system, they exhibit superior contacting when 
compared to monoliths and packed beds [8].   
The process of preparing microfibrous materials involves mixing a slurry of 
micron diameter fibers with pulped cellulose and selected particulates (sorbents or 
catalyst supports).  This slurry is then cast into a sheet using a wet lay process and then 
sintered to remove cellulose binder and to entrap the particulates in a sinter-locked 
network of fibers.  This can be accomplished on well understood traditional high speed, 
low cost paper making equipment resulting in high uniformity of basis weight.  The end 
product exhibits unique properties in terms of void volume, conductivity, porosity, 
surface area, permeability, and particle size [8].  
 
 
Figure II.2:  Three rolls of microfibrous materials using fibers made of ceramic, metal, and polymers 
 
 10
The fibers can be made of metal, polymer or ceramic, each possessing its own 
advantages.  Metal fibers are conductive which allows the sheet to be applied as an 
electro catalyst or for electrical swing adsorption.  Metal fibers are also able to withstand 
temperatures high enough for temperature swing regeneration of sorbents.  Polymer 
fibers are low cost and an option for disposable filters such as home HVAC or cathode air 
filters; however, these fibers are for relatively low temperature applications only and 
unable to be temperature regenerated.  The flexibility of the polymer also makes them 
good candidates for pleating.  Ceramic fibers are resistant to caustic environments and 
heat.  Figure 2.4 shows a micrograph of the three different fiber types sintered to hold 
particulates. 
 
 
Figure II.3:  A micrograph showing the three types of microfibrous materials and supported sorbents 
 
 High voidage (70-95%) and extremely small particle size (as low as 50 micron in 
diameter) used in microfibrous materials creates static mixing thereby reducing 
intraparticle mass transport resistance [3].  With adjustments of fiber diameter and 
particle size, surface kinetics of the reaction can be matched with the heat and mass 
transport at the reaction site.  Optimization of the materials results in a higher contacting 
efficiency and lower pressure drops.   
 11
 The high contacting efficiency of microfibrous materials results in high sorbent 
utilization.  However, microfibrous materials have low sorbent loading when compared 
to a packed bed.  Packed beds have high volumetric loading and but have a comparably 
lower contacting efficiency [3].  By adding a layer of microfibrous materials downstream 
of a packed bed, a composite bed is created utilizing packed bed sorbent loading and 
microfibrous material contacting efficiency.  A composite bed synergistically combines 
the advantages of the a packed bed with those of the microfibrous ?polishing? layer to 
create a thinner more efficient filter bed capable of high logs of removal of contaminants.  
Further understanding and modeling of the composite bed can be accomplished with the 
equal areas design rule [13].   
 
Equal Areas 
 
5-Log Breakthrough  
of packed bed (PB) 
Time Elapsed (minutes) 
5-log breakthrough of Polishing Sorbent (PS) 
C
o Co 
 
Composite bed: 
5-Log Breakthrough of PB + PS 
 
Figure II.4:  Diagram showing benefit of composite bed 
 
 12
 13
 This plot shows the increase of breakthrough time by adding a microfibrous 
polishing layer to the exit of a packed bed.  The sigmoidal breakthrough curve of the 
packed bed shows its lack of sorbent utilization and inability to continue a high log 
removal once breakthrough occurs.  Conversely, the sharp breakthrough curve of 
microfibrous materials represents the high contacting efficiency and sorbent utilization 
before breakthrough.  By adding this layer to the end of a packed bed, the capacity of the 
microfibrous layer is used to remove contaminants breaking through the packed bed 
therefore causing a sharper breakthrough curve.  This causes a higher overall sorbent 
utilization and extends the high log breakthrough time two to three times.  This allows for 
an overall thinner filter design with lower catalyst/sorbent requirements [13]. 
 14
III. EXPERIMENTAL APPARATUS 
 
1. Ballard NEXA Fuel Cell 
A NEXA power module, produced by Ballard Power Systems, is a small 
automated fuel cell system designed for portable and back-up power applications.  The 
system uses a polymer membrane electrolyte, with a platinum catalyst layer.  It uses 
hydrogen as a fuel and oxygen from air as an oxidant.  This fuel cell design results in a 
system that is very quiet and produces zero harmful emissions (only water and heat), 
allowing for indoor operation.  Fuel cell operation is designed to be continuous as long as 
hydrogen is provided.  The system provides 1200 watts of DC power output at a nominal 
output voltage of 26 VDC.  Figure III.1 shows a picture of the system [8].   
 
 
 
Figure III.1: Picture of Ballard NEXA Fuel Cell with the cathode air filter circled 
 
The system is comprised of four parts: a Ballard fuel cell stack, hydrogen 
delivery, oxidant air supply, and cooling air supply.  Onboard sensors monitor system 
performance while a control board automates fuel cell operation. Battery power is 
required for start up and shutdown requirements [8].   
The NEXA fuel cell has been optimized for the portable power market.  It 
operates at low pressure, minimizing parasitic power loss, reducing noise, and enhancing 
system reliability.  Also, the stack does not require external fuel humidification.  A 
humidity exchanger controls humidity in the cathode air by exchanging water from the 
exhaust from the cathode [8].   
Oxygen is provided to the stack by a small compressor.  The compressor speed is 
variable, allowing for an optimized flow of oxygen to meet system power demand.  An 
 15
 16
intake filter is required to protect the compressor and fuel cell stack from particulates and 
chemical contaminates [8].   
Fuel cell stack temperature, hydrogen pressure, hydrogen leak concentrations, 
stack current, stack voltage, air mass flow, and purge cell voltage are monitored by input 
signals to the control board.  The controls then output a pass/fail check based on 
operations status.  Depending on the severity of the situation, the fuel cell will either 
attempt to remedy the situation, set off an alarm, or execute a shutdown sequence [8].   
 
2. Compressor Testing Apparatus 
 The compressor testing apparatus involves two replaceable orifice plates on each 
side of the compressor.  The upstream orifice represents a pressure drop across a filter 
and the downstream represents system pressure drop through the stack, from the 
compressor outlet to atmospheric pressure.  PVC pipes with a 1.5? ID were used to 
connect the compressor to hold the orifice plates.  At the exit, a 2? long PVC pipe with a 
0.5? hole in the 6 inches from the end of the pipe allows for insertion of an Extech hot 
wire anemometer.  The pressure was recorded at the inlet and outlet of the compressor by 
inserting a needle into the rubber hosing connecting the compressor to flow expanders.  
Flow expanders were then connected to the PVC pipes and sealed with ducting tape.  By 
recording values at the compressor inlet and outlet, rather than a traditional pressure tap 
at D and 0.5D before and after the orifice plates, pressure comparisons could be made 
with values recorded within the fuel cell system.  The following picture shows the testing 
apparatus.   
 
 
Orifice Plates
Hot Wire 
Anemometer
Controller (DC to 
BLDC Converter)
DC Power Supply
Air Inlet
Air Outlet
Volt MeterAmp Meter
Compressor
 
Figure III.2: Compressor testing apparatus used for testing the compressor outside of the fuel cell 
 
Pressure was recorded on a 36? Dwyer manometer.  Power to the motor was provided by 
a Lambda power supply model #LLS7060 capable of providing power at 0-60 volts at 4.8 
amps.  Power to the controller was provided by an Electro Industries Regulated power 
supply (model #3002A) capable of producing 0-30 volts at 2.5 amps.  Voltages and 
amperages were confirmed using two fluke 87 multimeters and one fluke 77 multimeter.  
Total power to the system was provided by adding the power to each component.   
 A bread board controller was built to control compressor speed while providing 
access to voltage and amperage values used to drive the compressor and power the 
controller.  A circuit diagram of the controller is shown in Figure III.3. 
 17
 
Figure III.3: Compressor controller diagram 
 
 
 18
3. Breakthrough Testing Apparatus 
Rotometer
Rotometer
House Air
Water Bath
 
Figure III.4: Breakthrough testing apparatus 
 Figure III.4 shows the testing apparatus used in testing 2? circular samples of 
packed bed and microfibrous sheets.  By modifying the exit portion of the apparatus, it 
was also possible to test the final filters used in the Ballard fuel cell.   
 The apparatus uses house air, which is split and controlled by two pairs of Omega 
rotometers.  The first part passes through one of either a 0-500 ml/min or 0-7 ml/min 
rotometer and then passes through a two stage bubbler containing hexane.  Only one 
rotometer was used at a time to maintain pressure and flow stability in the system.  The 
bubblers were placed in a constant temperature water bath at 20
o
C to ensure a constant 
hexane vapor pressure.   
 19
 20
 The other two rotometers control the bulk flow of air through the system using 
either a 0-150 SCFH or a 0-50 SCFH rotometer.  This air combines with the air flowing 
through the hexane before it is run through the filter apparatus.  The flow rate of the air 
through the hexane controls the final concentration of hexane in the system.  The addition 
of the flow rates from both rotometers is the total flow rate through the filter.   
 The filter apparatus is a 2? id pipe with a 5? lead section providing time for the air 
to be well mixed and achieve a laminar flow regime.  A 1? section after the filter provides 
a well mixed flow of air so that the concentration recorded at the exit represents and 
average concentration exiting the filter.  Hexane concentrations were measured using a 
MiniRae PID detector.   
For testing the full sized filters, a different filter holding device was constructed.  
Half of the device is shown is Figure III.5.  There is a one foot inlet section containing a 
flow straightening device before air is run through the filter.   
 
 
Figure III.5: Filter testing apparatus for testing filters that fit into a Ballard NEXA fuel cell 
 21
 22
IV.     DESIGN METHODOLOGY AND ALGORITHM 
 
1. Fuel cell system design discussion and design considerations 
1.1. Fuel cell system discussion 
Optimization of cathode air filtration requires addressing a unique set of 
challenges present in the fuel cell system, shown in Figure IV.1.  One such challenge is 
how to minimize the loss in efficiency caused by chemical contaminants poisoning the 
cathode section catalyst.  A cathode air filter can be used to remove these contaminants; 
however, it causes an increase in system pressure drop, presenting yet another challenge.  
This pressure drop causes an increase in the required amount of pressure-volume work 
that the compressor must perform.    An energy balance must be made between the loss in 
power due to poisoning of the platinum catalyst and a loss in fuel cell efficiency created 
by an increase in parasitic power required to pump air through additional filtration.  
 23
 
Power 
Net Output Power 
PV Work 
Contaminated Air 
Air Exhaust 
Compressor 
Cathode Air Filter
PEM Stack Fuel 
Other Parasitic Power 
e.g.:  Cooling Fan 
         Process Controls/Actuators 
         Power Conditioning/ Electronics 
         Battery Standby 
        Etc.
Figure IV.1: Fuel cell mass and energy balance showing the flow of air and power through the system 
 
1.2. Filter design considerations 
 Cathode air filter design considerations fall into six categories: inlet air properties, 
adsorptive filter parameters, compressor attributes, stack attributes, filter footprint, and 
filter design options.  Figure IV.2 shows how the design considerations are further broken 
down into individual variables within the filter and fuel cell system. 
From this methodology, there are three categories of variables to consider: inlet 
air properties, filter design options, and filter footprint.  Each of these categories can be 
further broken down into individual variables which can be studied for its effect on the 
compressor, stack, and ultimately the total fuel cell?s performance.  The effects of each of 
these categories are tied together with the adsorptive filter parameters: saturation 
capacity, removal efficiency, and pressure drop.  These parameters then influence the PV 
work required to operate the compressor at the desired air flow rate and the stack 
efficiency defined by the poison tolerance and lifetime. 
 24
 
Adsorptive Filter Parameters: 
     Saturation Capacity 
     Single Pass Removal Efficiency
     Pressure Drop 
Inlet Air Properties: 
     Contaminant Type 
     Contaminant Concentration 
     Flow Rate 
Fuel Cell Stack Attributes: 
     Poison Tolerance 
     Lifetime 
Particulate and Adsorptive 
Filter Design Options: 
     Packed Bed (PB + PF) 
     Microfibrous Materials (MFM + PF)
     Composite Bed (PB + MFM + PF) 
Compressor Attributes: 
     PV Work 
     Compressor Efficiency 
Filter Footprint: 
     Area 
     Thickness 
     Weight 
Figure IV.2: Hierarchal design considerations for cathode air filters 
 
1.2.1. Inlet air properties 
The inlet air properties, along with the stack attributes, dictate the efficacy and 
capability that the filters must possess.  Air properties affect filter parameters by defining 
the capacity and removal efficiency needed to remove contaminants to achieve a 
specified output level.  Single pass removal efficiency is increased by reducing adsorbent 
particle size thereby increasing surface area and eliminating intra-particle mass transport.  
High single pass removal efficiency results in a filter effective at removal of 
 25
contaminates at low concentrations.  An increase in either the capacity or the removal 
efficiency results in an increase in both filter pressure drop and the amount of power 
required by the compressor.   
 
1.2.2. Filter design options 
While many filter design options exist, for this research three options are 
considered: packed bed (PB), microfibrous materials (MFM), and composite bed (MFM 
+ PB).  A packed bed is the most widely used form of filtration.  Packed bed 
characteristics include high sorbent capacity, low single pass removal efficiency, and low 
pressure drop.  The second design option, microfibrous materials, has been proven 
effective with high log and high single pass removal of chemical contaminants, but has a 
lower capacity and higher pressure drop when compared to packed beds.  Microfibrous 
materials also add a new dimension to the problem, through application as a pleated filter 
media, or as a polishing sorbent in a composite bed formation [6].  Pleated microfibrous 
materials have an increased capacity and lower pressure drop than a flat sheet with a 
smaller area.  A composite bed effectively combines the capacity of a packed bed and the 
high contacting efficiency of microfibrous materials.  An optimized composite bed is 
capable of higher logs of removal, has a reduced total bed depth, and has a lower pressure 
drop than a packed bed. 
 
1.2.3. Filter footprint 
The last category that needs to be considered for its effect on filter parameters is 
the filter footprint.  An increase in area allows for a decrease in thickness and visa versa.  
 26
In many cases, the footprint may be set by the fuel cell system in order for the filter to be 
retrofitted into the existing particulate filter slot.  Filter weight affects the filter capacity 
and is required for analysis because of the differences in densities of the filter design 
options. 
 
1.2.4. Adsorptive filter parameters 
 Inlet air properties, filter design options, and filter footprint all relate to the 
adsorptive filter parameters.  Contaminant type, concentration, and flow rate affect the 
filter parameters by virtue of the differences in chemical properties that various 
contaminants have in relation to the filter adsorptive materials.  Filter design options 
control the mass and energy transport properties of the filter by affecting contacting 
regimes as well as pressure drop.  Lastly, the filter footprint affects all three filter 
parameters through bulk design. 
 
1.2.5. Compressor attributes 
 Once the adsorptive filter parameters have been established, their effect on the 
compressor can be established.  Pressure drop across the filter causes a decrease in inlet 
pressure (P
1
) to the compressor creating a vacuum.  The outlet pressure (P
2
) from the 
compressor is set by the required flow rate necessary to operate the fuel cell.  This 
increase in pressure ratio (P
2
/P
1
) across the compressor causes a decrease in compressor 
efficiency and an increase in compressor power.  Both of these effects cause an overall 
decrease in fuel cell operating efficiency and performance.   
 
 27
1.2.6. Fuel cell stack attributes 
 The fuel cell stack is also affected by variations in the filter parameters.  The stack 
has a certain tolerance and operating lifetime related to the concentration of contaminants 
passing through the cathode of the cells.  The filter controls this concentration by 
reducing the contaminant concentrations before they make contact with the cathode.    
The single pass removal efficiency required is dictated by the maximum concentration 
tolerable by the fuel cell.  The maximum concentration may be lower if the contaminant 
collects in the stack rather than passing through the exhaust.   
 Research on cathode durability has concluded that the effect that contaminants 
have on fuel cell performance varies depending on the contaminant type and 
concentration, leading to three different scenarios for contaminant remediation.  In the 
first scenario, contaminants have little to no permanent effect, and do not build up in the 
system, so can be ignored.   For example, it may be more acceptable to allow a small 
percentage of contaminant through, rather than to increase filtration, thereby resulting in 
a higher pressure drop.  An example of this would be CO, where the damaging effect is 
both small and temporary.   
In the second scenario, contaminants need to be completely removed, because any 
amount of exposure to the contaminant will cause a permanent degradation of fuel cell 
performance.  These contaminants build up in the system, causing more severe 
degradation with longer exposure times.  For example, a cathode air filter must be 
capable of removing sulfur compounds such as H
2
S and mustard gas(C
4
H
8
Cl
2
S).   
In the third scenario, contaminants must be removed to a point of maximum 
concentration to minimize a decrease in fuel cell performance.  A decision for further 
 28
removal beyond this defined maximum must take into account the effect the increase in 
filter pressure drop will have on the system.  For example, SO
2
 at concentrations of 2.5-5 
ppm has a significant effect on fuel cell performance, with as much as a 50 to 60% 
reduction in fuel cell power [4].  This situation would clearly require a cathode air filter 
for efficient fuel cell operation.  SO
2
 concentrations of lower than 0.5 ppm may be 
ignored completely as they have no noticeable effect [3], so a cathode air filter would not 
be necessary.  However, at moderate concentrations, between 1 and 2 ppm, filtration 
design can be optimized between fuel cell performance and the effects of the increased 
filter pressure drop from additional filtration. 
 
1.3. Cathode air filter optimization: a trade off between compressor power requirements 
and efficiency losses from catalysts poisoning 
 Research on cathode durability has concluded that the effect that contaminants 
have on fuel cell performance varies depending on the contaminant type and 
concentration, leading to three different scenarios for contaminant remediation.  In the 
first scenario, contaminants have little to no permanent effect, and do not build up in the 
system, so can be ignored.   For example, it may be more acceptable to allow a small 
percentage of contaminant through, rather than to increase filtration, thereby resulting in 
a higher pressure drop.  An example of this would be CO, where the damaging effect is 
both small and temporary.   
In the second scenario, contaminants need to be completely removed, because any 
amount of exposure to the contaminant will cause a permanent degradation of fuel cell 
performance.  These contaminants build up in the system, causing more severe 
 29
degradation with longer exposure times.  For example, a cathode air filter must be 
capable of removing sulfur compounds such as H
2
S and mustard gas(C
4
H
8
Cl
2
S).   
In the third scenario, contaminants must be removed to a point of maximum 
concentration to minimize a decrease in fuel cell performance.  A decision for further 
removal beyond this defined maximum must take into account the effect the increase in 
filter pressure drop will have on the system.  For example, SO
2
 at concentrations of 2.5-5 
ppm has a significant effect on fuel cell performance, with as much as a 50 to 60% 
reduction in fuel cell power [5].  This situation would clearly require a cathode air filter 
for efficient fuel cell operation.  SO
2
 concentrations of lower than 0.5 ppm may be 
ignored completely as they have no noticeable effect [3], so a cathode air filter would not 
be necessary.  However, at moderate concentrations, between 1 and 2 ppm, filtration 
design can be optimized between fuel cell performance and the effects of the increased 
filter pressure drop from additional filtration. 
 
2. Design equations for describing fuel cell attributes and filter parameters 
 The following design equations provide mathematical relationships that relate the 
design considerations to each other.  These sections include equations for determining the 
required air flow rate to meet fuel cell power requirements, breakthrough time based on 
capacity and adsorption efficiency, pressure drop based on thickness, and compressor 
power based on inlet and outlet pressures. 
 
 
 
2.1. Fuel cell air flow requirements 
The required oxygen flow rate to operate a fuel cell can be found by the number 
of faradays provided by a mole of oxygen.  Through substitution and simplification the 
following equation is derived [2].   
)(1082.1
2
SLPM
V
P
eAirFlowRat
c
e
???=
?
?         (IV.1) 
Where ? is the stoichiometric ratio defined as the total amount of oxygen flow 
divided by the oxygen used.  A good approximation for the minimum stoichiometric ratio 
to operate a fuel cell is 3 moles of oxygen for each mole of oxygen required.  P
e
 is the 
power output in watts, and V
c
 is the average voltage across a cell.  A value of 0.60 V can 
be used as an average voltage approximation if the value cannot be found through 
experimentation or is not given.  This equation can be used for all fuel cells regardless of 
size as an estimate for air flow rate requirements [2]. 
 
2.2. Adsorptive filter breakthrough equations describing capacity, adsorption rate, 
breakthrough time, and inlet and outlet concentrations 
 Filter breakthrough equations are required to relate filter attributes to inlet and 
outlet air properties.  The filter attributes include the chemical properties of sorbent 
capacity (N
0
) and rate of adsorption (k?), and the footprint which includes area (A), 
thickness (L) and weight (m).  Another attribute commonly used to describe filters is 
voidage (?).  The air properties are face velocity (v
0
), inlet concentration (C
0
), and outlet 
concentration (C
1
).  Several previously studied equations are effective at predicting 
breakthrough times for adsorbent beds. 
 30
 An exact breakthrough equation for non-reversible adsorption onto a surface can 
be applied to a single or dual layer bed.  Equation 2, derived by Neal Amundson [14], is 
based on mass balance across an adsorptive bed allowing for a time varying input 
concentration.   
()
() ()
0
-
0
0
0
z
0
0
0
-
0
0
0
0
0
0
,
1-exp)-(exp
exp-
??
?
0
0
v
k
t
dC
v
k
dnN
v
k
dC
v
k
v
z
tC
C
C
v
z
t
v
z
t
?
??????
????
?
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=     (IV.2) 
This equation is then integrated resulting in the following equations which can be 
used to calculate the outlet concentration of a gas after passing through a single or double 
layer filter.  A constant inlet concentration was used to integrate the outlet concentration 
of the first layer (C
1
).  C
1 
was then inserted back into Equation 2 and integrated to 
calculate C
2
. 
()
)/(
0
1
010111011
1-1
vLtCkCk
ee
C
C
???? ?
+
=
            (IV.3) 
()[]
() ( )11
22
0201110112022
11
22
020111011
1-11-
1-1
)/-(-
)/-(-
1
2
k
k
vztCkCkCk
k
k
vztCkCk
eee
eeC
C
?
?
??????
?
?
????
++
+
=
    (IV.4) 
 In Equation IV.4 z
2
=L
1
+L
2
.  Another equation used to determine outlet 
concentration from a single layer adsorptive bed, developed by Yoon and Nelson is based 
on probabilistic reasoning [6]. 
)('
0
1
1
tk
e
C
C
?
+
=
?            (IV.5) 
 31
Equation 5 allows for estimates to be made on the breakthrough constants used in 
the exact breakthrough equation.  By using experimental data, a regression analysis yields 
values for k?, (k*C
o
).  Also, from a breakthrough curve, saturation capacity (?) in units of 
time can be read off at the time corresponding to C = ? C
o
.  This allows for solving for 
capacity, N
0
, in grams contaminant being removed per cm
3
 sorbent, by using the 
following equation: 
00
0
Cv
LN
=?             (IV.6) 
Both the k? and N
0
 can be used for a prediction of breakthrough concentration at different 
flow rates, inlet concentrations, and filter thicknesses.  Face velocity is a function of flow 
rate and filter surface area.  Inlet concentration and time are user inputs related to the 
operating environment of the fuel cell.  Outlet concentration can be provided by the 
manufacturer or estimated based on experimental data on cathode durability.   
 
2.3. Pressure drop relationships with filter thickness and air face velocity  
 Pressure drop (?P) has been correlated with face velocity, bed depth, viscosity 
(?), void fraction, and gas density (?).  Pressure drop through both packed beds and 
microfibrous beds is also related to particle diameter.  However, in microfibrous beds 
there is an additional pressure drop related to fiber diameter.   
The Ergun equation is commonly used to describe flow through packed beds [15].   
()
3
2
3
2
2
1
4
71
150
?
??
?
?? ?
?
?
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
?
=
?
P
o
P
o
D
v
D
v
L
P
        (IV.7) 
 32
 In order to estimate the pressure drop across microfibrous materials (a high 
voidage mixture of fibers and particles), the Porous Media Permeability equation (PMP) 
is used to apply to beds that have porosity greater than 50%.  This equation considers 
form drag losses which are small in low porosity material (e.g. packed beds). 
PMP equation for high voidage microfibrous materials: 
()
()
()
?
?
?
?
?
?
+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
++
?
?
?
?
?
?
?
? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
?
?
?
??
?
?
?
??
4
1
2cos
61
1
cos
72
3
2
3
3
2
32
2
FD
f
o
FD
o
C
C
D
v
X
D
v
L
P
  (IV.8) 
 In addition to the variables used in the Ergun Equation, the PMP equation uses 
tortuosity (? = 1+0.5(1- ?)), shape factor (?), and the angle of flow paths through the bed 
(?).  The coefficient of drag is represented by C
D
 for a sphere (usually 0.6).  The 
coefficient of friction for turbulent flow is C
f
 and the coefficient of form drag of a sphere 
in turbulent flow is C
FD
 (C
FD
 = C
D
-C
f
).  Values corresponding to the Ergun equation can 
be found by equating the PMP equation to the Ergun and modified Ergun equations [15]. 
 Empirical pressure drop relationships can also be determined experimentally by 
recording pressure drop as a function of layer thickness and flow rate.  The advantage to 
this method is that it is more accurate; however, it prevents correlations to be made 
between properties of the filter such as the effects of particle size on pressure drop.  An 
example empirical equation is: 
2
01201112
vLCvLCPP +=?           (IV.9) 
where C
1
 and C
2
 are constants solved for from data, v
0
 is the face velocity and L
1
 is the 
thickness of the layer.  The second term in the equation accounts for inertial losses from a 
change in direction of air flow through the filter.  An example of these types of losses 
 33
would occur when studying a pleated filter.  In most cases, however, the losses are small 
and negligible. 
 
2.4. Compressor power and efficiency 
2.4.1. Compressor power 
A study of the compressor is necessary to correlate the pressure ratio and power.  
A simple experimental setup consists of a pressure tap on each side of the compressor, a 
means of adding pressure drop to each side of the compressor (e.g. orifice plates), and a 
controller to operate the compressor.  The compressor can be operated at different flow 
rates and pressure ratios while power provided to the compressor is recorded.  Figure 
IV.3 shows the resulting plot for a 12V compressor used in this study. 
Compressor Power as a Function of Outlet and Inlet 
Pressure Ratio
0
10
20
30
40
50
1 1.05 1.1 1.15
Pressure Ratio (P2/P1)
C
o
m
p
r
e
s
s
o
r Po
w
e
r U
s
e
d
 
(
w
a
tts
)
 
Figure IV.3: Power using a 12/64? Inlet Orifice and 11/64? Outlet Orifice. 
 
 
 
 34
 In this case the correlation is linear and will be in the form, 
b
P
P
mPowerCompressor +
?
?
?
?
?
?
?
?
=
1
2
       (IV.10) 
 
2.4.2. Compressor efficiency 
Several efficiency definitions exist that can be used to describe performance of a 
compressor.  The most common efficiency equation, due to its accuracy and ease of 
determining input values, is specific overall efficiency.   
eshaft
k
k
pe
W
P
P
Tcm
adoverall
,
1
1
2
1
1
,
&
&
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
=
?
?
     (IV.1) 
In this case, m
e
 is the experimental mass flow at the discharge, T
1
 is the ambient inlet 
temperature, P
2
/P
1
 is the pressure ratio, and k is a constant and is 1.4 for adiabatic [2].  A 
compressor study that yields efficiency measurements provides the ability to locate the 
surge line on a performance chart, identifiable as the points where efficiency approaches 
zero. 
 
2.4.3. Compressor performance charts 
Another method of obtaining data on compressors is with a performance curve.   
Centrifugal compressors have performance charts which give the efficiency and pressure 
ratios at which the compressor can be operated.  On this chart, efficiency is represented 
by constant efficiency contours.  Performance charts also show a surge line, which 
represents the point at which the compressor becomes unstable if operated at a higher 
 35
 36
pressure ratio, at the same flow rate, or at a lower flow rate at the same pressure ratio.  
The surge line represents the point air flow through the compressor is unable to keep the 
compressor from overheating.  This instability causes backflow and possible damage to 
the compressor.  The performance chart is important in understanding the operating 
capabilities of the compressor, as well as to help understand power requirements [2]. 
 
3. Design algorithm 
Using the previously discussed design equations and relationships, an algorithm 
was created to relate the design considerations to filter thickness and compressor parasitic 
power.  The design equations were organized such that each variable could be solved 
starting with inputs that are readily available from the manufacturer or able to be 
experimentally determined.  The linearity of the algorithm allows for a simple computer 
program to be built with few assumptions.  A completed algorithm is shown in Figure 
IV.4. 
 37
 
Equations IV.3  
and IV.4 
Equation IV.9 
Equation IV.10 
Equation IV.1 
Filter Thickness (T) 
Air Flow Rate  
Inputs: 
    Filter Capacity (N
o
) 
    Adsorption Rate (k?) 
    Voidage (?) 
    Inlet Concentration (Co) 
    Outlet Concentration (Cout) 
    Filter Area (A) 
Filter Pressure Drop (P
1
) 
Compressor Parasitic Power 
Required Inlet Stack Pressure (P
2
)
Inputs:  
    Desired Fuel Cell Power Output (P
e
)
    Average Cell Voltage (V
c
) 
    Minimum Stoichiometric Ratio (?) 
Figure IV.4: Final, programmable design algorithm 
 
 
 38
V. DESIGN ALGORITHM APPLICATION TO BALLARD FUEL CELL 
 
1. Application overview 
 A study was performed, applying the filter design algorithm to a Ballard NEXA 
fuel cell.  This study provides further detail on the algorithm through transgression of the 
individual steps of the algorithm and shows how the algorithm is applied to a real system.  
Application of the algorithm also allows for an analysis of the accuracy of assumptions 
and creation of recommendations for improvements.  The results of this study show how 
predictions may be made comparing filter thickness, pressure drop and parasitic power 
requirements of the compressor. 
The first step in applying the algorithm is choosing a filter type for study.  Four 
different options are considered for application to the NEXA fuel cell; packed bed, a 
microfibrous sheet, a pleated microfibrous sheet, and a composite bed.  
 Other design considerations require a study of the compressor, required air flow 
rates by the stack, and pressure drops in the fuel cell.  Inputs also include pressure drop 
relationships for filter materials and breakthrough constants for the filter options.  For 
application to the NEXA fuel cell, each step will be discussed and compared with actual 
data to provide an understanding of where future improvements can be made.  Once data 
has been collected, predictions are made about filter thickness, layer ratios in a composite 
bed, and power requirements to operate the compressor.   
 39
2. Filter design options and footprint 
2.1. Filter design options 
Filter design options include four different filter types; traditional packed bed, 
microfibrous materials, a composite bed, or a pleated microfibrous media filter.  Each 
design option possesses unique attributes allowing for a filter design that is capable of 
being tailored to specific applications.  Microfibrous materials have been proven effective 
at high log, low capacity removal.  Microfibrous materials can also be pleated which 
reduces pressure drop across the filter.  A packed bed has a low pressure drop and is 
effective at filtration requiring high capacity sorbents.  However, these materials have 
poor contacting and are therefore poor at high log removal applications.  By combining 
the two, a composite bed, provides another filter option capable of utilizing the strengths 
of microfibrous materials and packed beds to provide an overall thinner filter for high 
capacity high log removal applications. 
 
2.2. Filter footprint 
 Physical size constraints may exist in the fuel cell system that need to be 
considered both before and after the filter design analysis.  Filter area is required by the 
algorithm for calculation of face velocity and filter pressure drop.  Filter thickness is an 
output by the algorithm.  Thickness is studied at the end of the analysis to determine if 
the resulting filter thickness is feasible. 
 
 
 
3. Air flow rate and stack pressure 
3.1. Stack air flow requirements 
 Figure V.1 compares data taken on the NEXA FC with values predicted using the 
following equation. 
)(1082.1
2
SLPM
V
P
AirFlow
c
e
???=
?
?                     (V.1) 
The Ballard manual suggests using 0.6 volts for average voltage across a cell.  
Typical stoichiometric ratios based on values given by the NEXA software at higher 
power outputs were about 2.7.   
Air flow rates required to meet stack power requirements
0
10
20
30
40
50
60
70
80
90
100
0 200 400 600 800 1000 1200
Stack Power (Watts)
A
i
r F
l
ow (S
L
P
M
)
 
Figure V.1: Predicted air flow model compared with actual data 
 The results of the flow rate study using the fuel cell show a correlation between 
flow rate and power requirements that is nonlinear.  Accuracy of the model is highest 
between 200 and 800 slpm.  At low flow rates the flow rate is constant at around 20 
SLPM.  This is necessary in order to maintain membrane humidity and to prevent the 
 40
compressor from operating at low flow, inefficient levels.  It is evident that either other 
factors are involved in oxidant flow control than stoichiometric ratio and average cell 
voltage or that either the stoichiometric ratio or average cell voltage are not constant.  
Overall, the predicative model for air flow rate offers an effective means of predicting air 
flow when no other experimental data from the fuel cell is known. 
 
3.2. Stack inlet pressure requirements 
 In addition to flow rate, experimental data relating stack inlet pressure (same as 
the outlet pressure of the compressor) to flow rate must be obtained.  This data is similar 
to an orifice plot, and can be modeled or a trend line can be used to describe the system.  
This pressure should be plotted under the same experimental setup that the fuel cell will 
be operated in; using the same intake manifolds and exhausts systems.  A trend line of the 
data was taken, yielding the following flow-pressure relationship for the NEXA fuel cell: 
FlowFlowP ?+?= 2359.0005.0
2
2
         (V.2) 
Pressure is in ?H
2
O and flow rate is in SLPM.  P
2
 is necessary for calculation of a 
pressure ratio used in studying the compressor and predicting power loads required to 
operate the compressor. 
 
4. Compressor data collection 
 A study is required in order to understand the effects that a pressure drop resulting 
from additional filtration will have on compressor performance.  Two methods were used 
to study the compressor.  The first required a testing apparatus capable of operating the 
compressor outside of the fuel cell system.  The second method used to study the 
 41
 42
compressor was performed using the Ballard NEXA fuel cell by adding an inlet air flow 
pressure drop and determining its effect on fuel cell and compressor performance.  Both 
of these methods are outlined in the following sections. 
 
4.1. A compressor study using a compressor testing apparatus 
 If a cathode air filter is designed to fit into an existing fuel cell system, design 
must consider the operating capabilities of the compressor and the effects that a 
resistance to flow will have on system performance.  Compressor limitations may exist 
which limit the maximum flow restriction on the inlet or outlet air stream of the system.  
A study of compressor efficiency and parasitic power provides information for defining 
the compressor which allows for prediction of performance under different conditions.  
These predictions are inserted into the design algorithm and aid in optimization of 
cathode air filter design and comparison of different design options.  If compressor data 
is not available from the compressor or fuel cell manufacturer, a set of experiments may 
be performed to collect the data necessary for compressor analysis.  A data analysis of the 
compressor can then be used to create target filter thicknesses or maximum filter 
thickness and also to predict the required power output to provide required air flow rates 
at various filter thicknesses.   
Data collection on the compressor is accomplished with a detailed study on the 
type and performance of the compressor.  The most common type of compressor used in 
fuel cells is a centrifugal compressor.  One helpful tool for understanding centrifugal 
compressors and quantitatively representing them is a compressor performance curve.  
Compressor performance curves are helpful tools for understanding how the compressor 
 43
operates by showing efficiencies as well as the surge line and other operating limits.  In 
order to create a performance curve, an apparatus must be created which allows for the 
control of the power supplied to a compressor which will be operated at different flow 
rates and pressure ratios.  The power provided to the compressor and thermodynamic 
analysis of the PV work allows for calculation of efficiencies at different operating 
points.  This data analysis allows for the determination of the effect that pressure drop 
created by a cathode air filter will have on operation of and parasitic power consumed by 
the compressor.  
 A Ballard NEXA fuel cell uses a centrifugal compressor operated with a 12 volt 
3-phase brushless DC motor.  The advantages of brushless DC are its durability, 
operating efficiency, and few contacting parts.  Brushless DC motors require a controller 
that converts DC power to a 3-phase output.  Due to the complexity of the NEXA control 
board, it is difficult to accurately measure voltage and current values being supplied by 
the controller to the compressor. 
An attempt was made at studying the wave forms leading from the controller to 
the compressor by using an oscilloscope.  The monitor output from the oscilloscope was 
noisy and resulting in no reliable data able to be derived from this experiment.  It was 
also difficult to determine the power supplied to the controller due to the inaccessibility 
of the wires leading from the power supply to the controller.  For this reason, a controller 
was bread-boarded that was capable of operating the compressor separate from the 
NEXA fuel cell system.  This allowed for measuring voltage and current supplied to the 
controller rather than in the three phase wires leading to the compressor.   
 During the experiments, outlet face velocity, total current and voltage to the 
compressor and controller, and compressor inlet and outlet gauge pressure were recorded.  
Pressure at the apparatus inlet and outlet was assumed to be 1 atmosphere.  Compressor 
efficiency was calculated assuming adiabatic compressor efficiency.  The efficiencies 
were also calculated based on the total power supplied to the controller and the 
compressor.  This includes the efficiency of the controller, motor, and compressor. 
 Data points were collected starting at the lowest flow rate that the compressor 
could operate.  At this point the compressor efficiency is zero.  The power supplied to the 
compressor was then increased until face velocity increments of 10-15 fpm were 
observed.  Values were recorded at each increment until the upper flow rate was reached.  
The upper flow rates and pressure ratios were set by the fact that the controller 
overheated at higher power loads.  This limitation may exist in the fuel cell system and 
will be further investigated. 
 
Compressor Efficiencies found by increasing
compressor inlet orifice size
0
0.1
0.2
0.3
0.4
0.5
1 1.05 1.1 1.15 1.2
Pressure Ratio (P2/P1)
E
f
f
i
ci
e
n
cy
 (
%
)
11/64 in Orifice
12/64 in Orifice
13/64 in Orifice
15/64 in Orifice
96/64 in Orifice
 
Figure V.2: Compressor efficiencies 
 44
 The first observation from figure V.2 is that the efficiency of the compressor 
increases with the pressure ratio, at least up to a value of 1.15.  Higher pressure ratios 
could not be reached with the current apparatus.  This means that the upper operating 
regime cannot be studied and that the surge line cannot be found. The consequence of this 
is that the maximum filter pressure drop cannot be determined based on compressor data.  
However, it is possible to correlate the power requirements of the compressor to the inlet 
and outlet pressures.  This correlation is shown in figure V.3. 
 
Power requirements to meet pressure ratio demands
0
10
20
30
40
50
60
1 1.02 1.04 1.06 1.08 1.1 1.12 1.14 1.16
Pressure Ratio (P2/P1)
P
o
we
r Us
e
d
 (wa
t
t
s
)
11/64 in. Orifice
12/64 in. Orifice
13/64 in. Orifice
15/64 in. Orifice
96/64 in. Orifice
 
Figure V.3: Parasitic power required by the compressor to operate at given pressure ratios 
 Figure V.4 shows a linear correlation between total power and pressure ratio.  A 
trend line representing the data can be regressed allowing for prediction of parasitic 
power used to operate the compressor a different pressure ratios that result from various 
filter thicknesses.  This analysis results in the following equation for insertion in the 
algorithm program. 
74.31214.313 ??=
R
PPower          (V.3) 
 45
 46
4.2. A compressor study within the Ballard NEXA system 
 During experiments on the compressor using the testing apparatus, it was 
determined that higher pressure ratios could not be tested due to the possibility of over 
heating of the controller circuitry.  A comparison between the possible limitation 
discovered using the apparatus and a possible limitation in the fuel cell could not be made 
due to the differences in the setup between the bread boarded controller and the controller 
used in the Ballard fuel cell.  The possibility of the controller overheating brings to light 
the fact that limitations exist other than those measurable by the compressor testing 
apparatus.  These limitations may be in the circuitry or arise when the compressor is 
operated within the fuel cell from a combination of factors.  In order to further study 
possible system limitations, an in vitro study of the effect of restrictions placed on 
cathode air filter flow on system stability and performance was performed. 
Operating limitations of the compressor may arise when the compressor is 
challenged within the fuel cell system by addition of a resistance to flow.  A test for these 
limitations was designed to challenge the fuel cell allowing for the study of the effects of 
pressure drop on the compressor inlet on the fuel cell operating capability.  This was 
accomplished by building five orifice plates with orifice sizes comparable to the ones 
used in the compressor studies.  Orifice sizes of 10/64, 11/64, 12/64, 13/64, and 15/64? 
were drilled into plates the same size as the filter slot.  The fuel cell was then operated on 
a load starting at 0 watts of output and increasing to 1200.  A Dwyer manometer was 
used to measure pressure drop across the plate and the flow rate was measured by the 
flow detector built into the system and outputted onto the Ballard software.   
 The first test was done with the 15/64? orifice.  The pressure drop across the stack 
was initially observed to be low when compared to the same orifice size and flow rate 
tested in the compressor testing apparatus.  At high flow rates the pressure drop was still 
below 1? of water.  The smaller, 11/64?, orifice plate had the same results.  To test this 
inconsistency the orifice was completely covered, blocking all air flow through the 
system intake that the filter would be set in.  This brought about no noticeable effect on 
fuel cell performance as judged by changes in flow rate and stoichiometric ratio.  After 
completely closing off the inlet the pressure drop only rose 0.2? of water.   
 The results of the in vitro tests lead to two possible conclusions.  The Ballard 
manifold holding the compressor and intake filter contains leaks sufficient enough to 
completely bypass the intended inlet while causing little effect on the system or an intake 
bypass is built into the system.  It is not this projects intent to pursue defects in the 
Ballard NEXA system.  For this reason, the cathode air filter will be designed to optimize 
the breakthrough requirements and the filter effects on the power necessary to operate the 
compressor.  Predictions will be made assuming that the fuel cell intake manifold does 
not leak. 
 
5. Determination of Exact Breakthrough Equation Constants 
 For a breakthrough prediction values are needed for inputs into the following 
equations. 
()
)/(
0
1
010111011
1-1
vLtCkCk
ee
C
C
???? ?
+
=
        (V.4) 
 47
()[]
() ( )11
22
0201110112022
11
22
020111011
1-11-
1-1
)/-(-
)/-(-
1
2
k
k
vztCkCkCk
k
k
vztCkCk
eee
eeC
C
?
?
??????
?
?
????
++
+
=
      (V.5) 
 
In many cases, k and C
o
 are lumped to form k?.  This adsorption constant should 
be regressed from data at the same inlet concentration that other breakthrough estimates 
will be made.  For this experiment, the inlet concentration will be 100 ppm and the k 
values will be solved for at this same concentration.   
 A recipe for a polymeric microfibrous materials sheet supporting activated carbon 
will be used that has been optimized for effective application as a polishing sorbent or 
pleated media.  The sheet has a thickness of 0.4 centimeters with a basis weight of 870 
g/cm
2
.  The volume percents of carbon, polymer, and cellulose are, respectively, 18.30%, 
3.88%, and 0.62%.  This leaves a voidage of 77.2%.   
 The packed bed used activated carbon from PICA, sized 12-30 mesh.  The carbon 
has a voidage of 0.4.  The packed bed was 5 cm in diameter and 0.95 cm thick.  This 
resulted in a total carbon weight of 11.25 g.   
The experiment resulted in the following breakthrough curves for the 0.4cm 
microfibrous supported carbon sheet and 0.95cm of packed bed.  From Figure V.5, ? 
values can be read as the time corresponding to an outlet concentration equal to ? of the 
inlet concentration.  Solving for N
0
 in  
00
0
Cv
TN
=?             (V.6) 
 48
for a packed bed and microfibrous layer, values are 0.1 g/cm
3
 and 0.04 g/cm
3
 
respectively.   
Breakthrough Curces for Challenge at 130 ppm
0
20
40
60
80
100
120
140
0 20 40 60 80 100 120 140 160
Time (min)
O
u
t
l
et
 H
exan
e C
o
n
cen
t
r
at
io
n
 
(pp
m
)
Packed Bed
MFM
 
tau
Figure V.4: Breakthrough curves for a 0.95 cm packed bed and a 0.4 cm microfibrous layer used to 
determine values for tau and capacity 
 Adsorption rate (k) values can also be determined from breakthrough data by 
solving Yoon and Nelson?s breakthrough equation as a linear equation as follows. 
)('1ln
0
tk
C
C
?=?
?
?
?
?
?
? ?            (V.7) 
 By making the plot shown in figure V.6, and averaging the slope, an estimate of 
k? can be solved for.  For the most accurate values, k? should be solved for at the 
concentration at which the filter will be tested at.  However, a rough approximation of k? 
can be made by separating it into k*Co and solving for k.   
 49
Plot of rearranged Yoon and Nelson Eq.
-4
-2
0
2
4
6
8
-20 0 20 40 60 80 100
tau - t
l
n
(C
o
/
C
 - 
1
)
Packed bed
MFM
 
Figure V.5: A breakthrough equation that represents k? values as the slope  
 Adsorption rate constants for this case were determined to be 0.183 for a packed 
bed and 0.332 for microfibrous materials at 130 ppm hexane. 
 
Table V.1 
Values used in filter design algorithm 
 Microfibrous Layer Packed Bed 
                           ? 65 min 140 min 
Filter Capacity (N
o
) 0.04 mols Hexane/cm
3
 MFM 0.1 mols Hexane/cm
3
 PB 
Adsorption Rate (k?) 0.00332*C
0
0.00183*C
0
Voidage (?) 0.772 0.4 
 
 
6. Determining filter pressure drop relationships 
 Several methods exist for predicting pressure drop across packed beds and 
microfibrous materials, including Ergun?s equation and the porous media permeability 
equation.  However, the simplest and most accurate method involves a simple experiment 
 50
involving recording pressure drop across a known thickness of the material being tested.  
The correlation is nearly linear; however, more data points should be taken at multiple 
face velocities to increase accuracy.  The following plots show the results of a pressure 
drop experiment performed on 12-30 mesh PICA carbon and a polishing sorbent 
microfibrous media recipe. 
 The experiment was performed on the same apparatus used to test for 
breakthrough characteristics.  Pressure drop was recorded from a 0-3 inch Dwyer 
manometer.  A bed depth of 0.95 centimeters was used for the packed bed and 0.4 
centimeters was used for the microfibrous sheet (corresponding to a basis weight on 870 
g/m
2
).  The following plot shows the results of the experiment. 
0
0.5
1
1.5
2
2.5
3
102030405
Face Velocity (cm/s)
Pre
ssu
re
 D
ro
p
 
(
c
m
 H2
O
/
c
m
 
B
e
d De
pt
h)
 
0
12-30 Mesh PICA Carbon Pressure Drop
Microfibrous Pressure Drop (870 Basis Weight)
 
Figure V.6:  Pressure drop as a function of face velocity for a 0.95 cm packed bed 
 and a 0.4 cm microfibrous layer 
  
 
 
 51
This provides an estimate for pressure drop of each material: 
02
0664.0 vTdP
MFM
??=           (V.8) 
02
0514.0 vTdP
PB
??=           (V.9) 
 
7. Filter design process results 
 The cathode air filter design algorithm was applied to a 1.2 kWe fuel cell in order 
to compare filter design options for removal of hexane with activated carbon.  Hexane 
was used for this study because of its ease of use in the laboratory and well known 
adsorptive characteristics when used with activated carbon.   
Four different design options were considered for this fuel cell application.  The 
first three options were packed bed, microfibrous sheet, and composite bed.  For these 
options a filter area of 55 cm
2
 was studied for its ability to be retrofitted into the existing 
particulate filter location in the fuel cell when total thickness is less than 1 cm.  The 
fourth design option, a pleated microfibrous materials filter, will be studied separate from 
the other filters.  The pleated filter will be studied for the effects of breakthrough time 
and concentrations on the variable sheet thickness of the filter while the pleat depth is 
held constant at 1 cm, corresponding to the size required to retrofit the fuel cell.   
In order to understand the effects of different inlet and outlet concentrations on 
the optimum filter designs, three case scenarios were studied.  The three cases were: 
Case I: 100 ppm to 5 ppm 
Case II: 10 ppm to 0.1 ppm 
Case III: 100 ppm to 0.1 ppm 
 52
 53
Values shown in Table 2 were used to represent the fuel cell system and were 
obtained from the operating fuel cell manual. 
Table V.2 
Quantitative fuel cell attributes used in the algorithm 
 
Input Parameter Parameter Value 
Fuel Cell Power Output (P
e
) 1200 watts 
Average Cell Voltage (V
c
) 0.60 volts 
Minimum Stoichiometric Ratio (?) 2.7 
 
 The other inputs were experimentally determined.  Breakthrough tests were 
performed on both 12-30 mesh activated carbon and a microfibrous polishing layer with a 
basis weight of 870 g/m
2
 with 18.3% 60-140 mesh activated carbon.  Values for k? and 
N
0
 were regressed from experimental breakthrough data using the equation provided by 
Yoon and Nelson.  Pressure drop relationships were determined in the lab as a function of 
thickness and flow rate.  All resulting data used in describing the filter options are shown 
in Table 3.   
Table V.3 
Filter design input parameters collected from experimental data 
 
Design Parameter Microfibrous Layer Packed Bed 
Filter Capacity (N
o
) 0.04 mols Hexane/cm
3
0.1 mols Hexane/cm
3
Adsorption Rate (k?) 0.00332*C
0
0.00183*C
0
Voidage (?) 0.772 0.4 
Air Flow Rate  89.5 LPM 89.5 LPM 
Pressure Drop 0.0664*L
2
*v
0
0.0514*L
1
*v
0
Compressor Power 313.14 (P
2
/P
1
) ? 312.74 313.14 (P
2
/P
1
) ? 312.74 
 
 
The algorithm was used to predict an optimum filter thickness and the 
corresponding power requirements to operate the compressor.  By varying the inlet 
concentration, outlet concentration, and breakthrough times, the different design options 
were studied under different capacity requirements and removal efficiency requirements.  
 
7.1. Case I: Inlet Concentration of 100 ppm, Outlet Concentration of 5 ppm 
Compressor Power Required for
Hexane Removal from 100 to 5 ppm (1.3 logs)
38
39
40
41
42
43
44
45
46
0 100 200 300 400 500 600 700
Breakthrough Time (min)
Com
p
re
s
s
or P
a
ra
s
i
t
i
c
 
P
o
w
e
r (w
a
t
t
s
)
MFM (55.5 cm2)
PB (55.5 cm2)
Composite (55.5 cm2)
 
Figure V.7: Predicted power required to meet breakthrough time requirements 
 
 54
Filter Thickness Required for
Hexane Removal from 100 to 5 ppm (1.3 logs)
0
2
4
6
8
10
12
0 100 200 300 400 500 600 700
Breakthrough Time (min)
F
i
l
t
er
 T
hi
c
k
nes
s
 (
c
m
)
MFM (55.5 cm2)
PB (55.5 cm2)
Composite (55.5 cm2)
 
Figure V.8: Predicted thickness required to meet breakthrough time requirements 
 Figures V.7 and V.8 show the predicted thickness and compressor power 
requirements for different filter configurations optimized to meet breakthrough time 
requirements at an inlet concentration of 100 ppm and an outlet concentration of 5 ppm.  
This removal represents a 1.3 log removal and requires a high filter capacity.  Because of 
the high capacity, the packed bed represents the best solution based on total filter 
thickness and pressure drop/compressor power.  Because of the low logs of removal, 
microfibrous materials were not affective in reducing thickness or power requirements 
and therefore provided no benefit when used alone or in conjunction with a packed bed 
(composite bed).   
 
 
 
 
 
 55
7.2. Case II: Inlet Concentration of 10 ppm, Outlet Concentration of 0.1 ppm 
Compressor Power Required for
Hexane Removal from 10 to 0.1 ppm (2 logs)
39.35
39.45
39.55
39.65
39.75
39.85
0 100 200 300 400 500 600 700
Breakthrough Time (min)
C
o
m
p
r
e
ss
o
r
 P
a
r
a
s
i
ti
c 
P
o
w
e
r
 (
w
a
tts
)
MFM (55.5 cm2)
PB (55.5 cm2)
Composite (55.5 cm2)
 
Figure V.9: Predicted power required to meet breakthrough time requirements 
 
Filter Thickness Required for
Hexane Removal from 10 to 0.1 ppm (2 logs)
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 100 200 300 400 500
Breakthrough Time (min)
F
i
l
t
er
 T
hi
c
k
nes
s
 (
c
m
)
MFM (55.5 cm2)
PB (55.5 cm2)
Composite (55.5 cm2)
 
Figure V.10: Predicted thickness required to meet breakthrough time requirements 
 From Figure V.9 and Figure V.10 it is evident that microfibrous materials provide 
the thinnest filter option for a breakthrough time of less than 50 minutes; however, this 
solution also creates the most pressure drop, as evidenced by the increase in parasitic 
 56
power requirements.  If, in practice, power requirements are the priority, then a 
composite bed provides the best solution.  A composite bed also provides the best 
solution for higher breakthrough times.  Figure 8 shows that a composite bed at the same 
thickness as a packed bed provides about a 100 minute longer breakthrough time. 
 
7.3. Case III: Inlet Concentration of 100 ppm, Outlet Concentration of 0.1 ppm 
Compressor Power Required for
Hexane Removal from 100 to 0.1 ppm (3 logs)
39.5
40
40.5
41
41.5
42
0 100 200 300 400 500 600 700
Breakthrough Time (min)
C
o
mp
r
e
sso
r
 P
a
r
a
si
t
i
c 
P
o
w
e
r
 (
w
a
tts
)
MFM (55.5 cm2)
PB (55.5 cm2)
Composite (55.5 cm2)
 
Figure V.11: Predicted power required to meet breakthrough time requirements 
 
 57
Filter Thickness Required for
Hexane Removal from 100 to 0.1 ppm (3 logs)
0.5
1
1.5
2
2.5
3
3.5
4
0 100 200 300 400 500 600 700
Breakthrough Time (min)
Fi
lt
e
r
 Th
i
c
k
n
e
s
s
 (
c
m
)
MFM (55.5 cm2)
PB (55.5 cm2)
Composite (55.5 cm2)
 
Figure V.12: Predicted thickness required to meet breakthrough time requirements 
 The third scenario, shown in Figure V.11 and Figure V.12, tests filter options 
under a situation that requires both a high capacity filter that is also capable of high logs 
of removal.  In this situation, a composite bed outperforms the other design options by 
having a lower total thickness than the packed bed.  It is also evident from Figure 10 that 
the addition of a microfibrous layer in place of a small length of packed bed to make a 
composite bed causes little additional pressure drop. 
 
7.4. Case study for a pleated microfibrous filter 
For an analysis of a pleated filter as a design option, a pleat depth of 1 cm was 
assumed and the thickness of the sheet was variable like the previous design options.  
This resulted in a constant filter thickness of 1 cm even though the sheet thickness was 
variable.  Because of the pleats, a larger surface area was able to be fit into the same 
dimensions as the other options.  A pleated microfibrous sheet was represented in the 
algorithm as having an area of 105.5 cm
2
 (pleat factor of 1.9).  This pleat factor was only 
 58
attainable at layer thicknesses less than 0.5 cm if it is still to fit in the 1 cm pleat depth, 
and will only be studied up to that thickness. 
The Figure V.13 and Figure V.14 show the required sheet thickness and 
compressor power requirements if a pleated filter is used.  The filter was only studied at 
an inlet concentration of 10 ppm and an outlet concentration of 1 ppm.  
 
Sheet Thicknesses Required if Using a Pleated Microfibrous Filter at 
1 cm Pleat Depth (Filter Thickness)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 100 200 300 400 500
Breakthrough Time (min)
S
h
eet
 T
h
i
ckn
ess 
(
cm
)
 
Figure V.13: Predicted sheet thickness of a microfibrous pleated filter at a challenge concentration of 
10 ppm Hexane and an outlet concentration of 0.1 ppm 
 59
Compressor Power Requirements using a Pleated Microfibrous Filter
39.42
39.44
39.46
39.48
39.5
39.52
39.54
39.56
39.58
39.6
39.62
0 100 200 300 400 500
Breakthrough Time (min)
C
o
m
p
re
s
s
o
r
 P
a
ra
s
i
t
i
c
 P
o
w
e
r 
(w
atts
)
 
Figure V.14: Predicted compressor power requirements to operate a microfibrous pleated filter at a 
challenge concentration of 10 ppm Hexane and an outlet concentration of 0.1 ppm 
 An increase in sheet thickness results in only a small percentage increase in 
compressor power requirements from additional pressure drop.  Pleated filters provide an 
additional solution for low capacity, high log removal applications.  Another advantage to 
using a pleated filter rather than a packed or composite bed is the stability from the sinter 
locked network which adds resilience to the filter. 
 
7.5. Detailed composite bed configuration analysis 
For each case study and each breakthrough time tested, the composite bed was 
optimized to minimize total bed depth.  This was accomplished by varying the thickness 
of the packed bed, calculating the required thickness of the microfibrous layer, and 
calculating the total thickness.  Table 3 shows the individual layer thicknesses for each of 
the breakthrough times in the case study.  By studying these graphs, information can be 
learned about the capacity and removal efficiency requirements of the filter and how 
 60
 61
filter configurations are affected by inlet concentration, outlet concentration, and 
breakthrough time. 
Case I and III require the thickest packed bed portion due to the fact that they 
have the higher capacity requirements of 95 and 99.9 ppm removal respectfully, when 
compared to case II, at 9.9 ppm removal.  Secondly, the thickness of the microfibrous 
layer increases with log removal requirements.  This is a result of the higher single pass 
removal efficiency required for high log removal.   
Table V.4 
Predicted, optimized individual layer thicknesses of the packed bed (L
1
) and microfibrous 
layer (L
2
) used to create a composite bed 
 Case I Case II Case III 
 
100 ? 5 ppm  
1.3 logs removal 
10 ? 0.1 ppm  
2 logs removal 
100 ? 0.1 ppm  
3 logs removal 
Time (min) L
1
 (cm) L
2
 (cm) L
1
 (cm) L
2
 (cm) L
1
 (cm) L
2
 (cm) 
15 0.2 0.12 N/A N/A 0.2 0.37
60 0.52 0.09 0.09 0.27 0.51 0.34
180 1.26 0.09 0.23 0.22 1.25 0.34
300 2 0.09 0.32 0.21 2 0.34
420 2.74 0.09 0.4 0.2 2.74 0.34
600 3.86 0.09 0.5 0.2 3.85 0.34
6000 37.29 0.09 3.86 0.2 37.28 0.34
 
Further analysis is performed by transgressing down the columns and by 
increasing breakthrough time requirements.  By increasing breakthrough time, additional 
capacity is required to meet the breakthrough requirements.  This results in an increase in 
the packed bed layer of the filter.  However, at low breakthrough times, where capacity is 
less significant, a thicker microfibrous layer provides a shorter overall bed depth by 
reducing the critical bed depth (the minimum bed depth required to prevent immediate 
breakthrough). 
 
 62
VI. FILTER CONSTRUCTION AND TEST RESULTS 
 
1. Filter Construction 
 Two cathode air filters were constructed to fit into a Ballard NEXA fuel cell.  
Breakthrough and pressure drop tests on these filters allowed for comparing results from 
tests with values predicted using the design algorithm.  By making these comparisons, 
recommendations could be made on how the algorithm can be further modified to add 
detail and accuracy in its predictions.   
 For this test, filters will be constructed to fit into the Ballard NEXA fuel cell.  The 
filters were tested at an inlet concentration of 100 ppm so that data can be compared with 
algorithm predictions from cases I and III.  The filters were made of the same materials 
used to determine values for the algorithm.  Tables V.1 and V.2 show the values used for 
predicting the filter performance. 
Table VI.1 
Quantitative fuel attributes used in the algorithm 
 
Input Parameter Parameter Value 
Fuel Cell Power Output (P
e
) 1200 watts 
Average Cell Voltage (V
c
) 0.60 volts 
Minimum Stoichiometric Ratio (?) 2.7 
 
 
 
 63
Table VI.2 
Filter design input parameters collected from experimental data 
 
Design Parameter Microfibrous Layer Packed Bed 
Filter Capacity (N
o
) 0.04 mols Hexane/cm
3
0.1 mols Hexane/cm
3
Adsorption Rate (k?) 0.00183*C
0
0.00332*C
0
Voidage (?) 0.4 0.772 
Air Flow Rate  89.5 LPM 89.5 LPM 
Pressure Drop 0.0664*L
2
*v
0
0.0514*L
1
*v
0
Compressor Power 313.14 (P
2
/P
1
) ? 312.74 313.14 (P
2
/P
1
) ? 312.74 
 
The following sections describe the filter construction methods, testing results, and 
recommendations for improvements to the algorithm. 
 
1.1. Filter frames 
Two filters were constructed to retrofit into the existing slot for cathode air filters 
in the Ballard NEXA system.  The first one is a composite bed design, and the second, a 
pleated microfibrous materials sheet.  The filters are 5.87 cm high, 10.5 cm wide, and 1 
cm thick.  Aluminum, 1/32? thick, was used for the framing.  The dimensions for the four 
parts (2 sides, and a top and bottom) are shown in figure XXX.   
??
??
1 1/8?
1 1/8?4 3/4?
5/16?
5/16?
2 15/16?
4 1/8?
2 5/16?
 
Figure VI.1: Cut metal frame dimensions used for constructing a cathode air filter 
 A metal brake was used to accomplish all of the bends.  All four sides of the 
larger piece are bent towards the same direction at 90
o
 angles to make an open box.  The 
smaller piece has only the long sides bent.  The unbent tabs of the smaller pieces slide 
inside of the ends of the larger piece creating the filter frame.  
 
1.2. Adsorbent materials 
Microfibrous sheets were made on a one square foot paper making machine.  The 
basis weight used was 870 g/m
2
.  The final filter was 18.3% by weight 60-140 mesh 
PICA activated carbon.  The resulting filter had a thickness of 0.4 cm.  The packed was 
comprised of 12-30 mesh PICA activated carbon. 
 
 
 
 64
1.3. Final Filter Design 
Figure VI.2 shows the optimized filter thickness for a composite bed plotted at 
different breakthrough times.   This plot predicts a 100 minute breakthrough for an 
optimized 1.2 cm composite bed.  The thicknesses of the individual layers to make the 
composite bed are shown in Table VI.3. 
Composite Bed Filter Thickness Required for
Hexane Removal from 100 to 0.1 ppm (3 logs)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 50 100 150 200
Breakthrough Time (min)
F
i
l
t
er
 T
hi
c
k
nes
s
 (
c
m
)
 
Figure VI.2: Predicted composite bed thickness required for hexane removal used to design a 
composite bed filter 
 
Table VI.3 
Values of layer thicknesses for optimized composite bed 
 100-.1 (3 log) 
minutes L
1
L
2
15 0.2 0.37
60 0.51 0.34
180 1.25 0.34
300 2 0.34
420 2.74 0.34
600 3.85 0.34
6000 37.28 0.34
 
 65
 The optimized composite bed configuration in this case has a packed bed of about 
.9 cm and a microfibrous bed of about 0.34 cm.  The filter that will be constructed will 
utilize a 0.8 cm packed bed and a 0.4 cm microfibrous polishing layer. 
 Through a hands on trial and error process it was determined that a maximum 
basis weight of about 870 g/m
2
, at a thickness of about 0.4 cm, can be pleated to fit into a 
filter 1.2 cm thick.  For this reason, the same sheet used in the composite bed was used in 
a microfibrous pleated filter. 
 The following pictures show the construction materials for the composite bed as 
well as a final filter. 
 
 
Figure VI.3:  Materials of construction for composite bed cathode air filter 
 66
 
Figure VI.4: Picture of pleated cathode air filter 
 
 
 
 
 
 
 
 
 
 
 
 
 
 67
2. Filter Tests Results 
2.1. Composite Bed Filter 
Composite Filter Breakthrough Curve
0
20
40
60
80
100
120
0 20 40 60 80 100 120 140 160 180
Time (min)
C
o
n
cen
t
r
at
io
n
 (
p
p
m
)
Actual
Predicted (No = 0.1)
 
Figure VI.5:  Actual breakthrough curves for composite bed compared to a predicted breakthrough 
curve by the algorithm 
Figure VI.5 shows the results of the 100 ppm challenge breakthrough test on the 
composite bed filter.  Breakthrough at 0.1 ppm was 27 minutes.  Predicted value was 100 
minutes.  The predicted curve shows a sharper breakthrough.  A shallower breakthrough 
curve results from a lower contacting efficiency than predicted.  Since contacting 
efficiency is primarily a factor of adsorption rates the inaccuracy in the breakthrough 
curve is caused by inaccuracy in the determination of the adsorption rate values.  The 
filter capacity is only slightly overestimated by the algorithm and as a result the 
breakthrough is sooner than predicted.  
 68
Composite Filter Pressure Drop
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
20406080
Flow Rate (SLPM)
P
r
essu
r
e D
r
o
p
 
(
"
H
2O
)
10
Actual
Predicted
 
Figure VI.6: Pressure drop comparison between a composite bed filter and values predicted by the 
algorithm 
 Pressure drop predictions, shown in Figure VI.6, were accurate at flow rates lower 
than 70 slpm when compared to experimentally determined data.  At higher flow rates 
actual pressure drop was higher than predicted.  This may be due to inertial losses which 
cause a non-linear pressure drop relationship when operating at higher flow rates and was 
assumed negligible when the pressure drop formula was determined for a packed bed.   
 
 
 
 
 
 
 
 
 69
 
2.2. Pleated Microfibrous Filter 
Pleated Filter Breakthrough Curve
0
20
40
60
80
100
120
1020304050607
Time (min)
C
o
n
cen
t
r
at
io
n
 (
p
p
m
)
0
Actual
Predicted (No = 0.04)
 
Figure VI.7: Actual breakthrough curves for a pleated filter compared to a predicted breakthrough 
curve by the algorithm 
 
 Figure VI.7 shows breakthrough curves predicted by the algorithm compared with 
those found experimentally.  In this case, the breakthrough was 15 minutes sooner than 
predicted, indicating that the capacity was overestimated.  Adsorption rate values were 
more accurate, evident by the similar shape of the sigmoidal breakthrough curve for both 
the actual and predicted curves. 
 Pressure drop predictions for the pleated filter, shown in Figure VI.8, are higher 
than actual values.  A small percentage of this drop in pressure drop is a result of leaks in 
the filter and can be resolved with better construction techniques.  A small curvature in 
the actual pressure drop is the result of inertial losses due to the air having to pass 
through the pleats.  Inertial losses were assumed negligible in original pressure drop 
 70
calculations used to predict the pressure drop; therefore, the predictions were less 
accurate.   
 
Pleated Media Pressure Drop
0
0.05
0.1
0.15
0.2
0.25
0 20 40 60 80 100
Flow Rate (SLPM)
P
r
essu
r
e D
r
o
p
 
(
"
H
2O
)
Actual
Predicted
 
Figure VI.8: Pressure drop comparison between a microfibrous pleated filter and values predicted 
by the algorithm 
 Pressure drop predictions for the pleated filter are higher than actual values.  This 
results from a higher surface area of the constructed filter that estimated when making the 
pressure drop predictions.  The area above the lips of the filter was not accounted for 
when calculating the filter pressure drop.  This effect was not significant in the composite 
bed because the lip seals off air flow from reaching this space.  However, in a pleated 
filter air may move under the lip and into the area covered by the frame, pass through the 
media, and move back under the lip on the other side.   
 
3. Recommendations for Improvement to Algorithm 
 71
 72
 The current algorithm assumes that the adsorption rate (k) is constant and obtains 
a value based on an average from data collected at one inlet concentration.  This value is 
then used to calculate k values for other concentrations.  A more accurate model of k 
could provide more accurate estimates on filter performance; however, this complicates 
the algorithm adding a new dimension to the program.  Currently, the algorithm is kept 
one dimensional by solving for the concentration only at the exit of the filter leaving time 
as the only variable.  A more accurate estimate of k would be calculated at an infinite 
number of points moving through the filter (perpendicular to the face) and at all points in 
time.  In other words, k is a function of distance into the filter and time.  A more accurate 
k estimate is increasingly important as filter thickness is increased due to the fact that 
there is a wider range of adsorption rate values through the filter.  A well modeled 
adsorption rate also allows for a larger range of estimating filter performance while using 
less experimental data.  This decreases the time necessary to design multiple filters for 
different applications using the same algorithm inputs. 
 Improvements to the understanding and modeling of pleated media would make 
the algorithm for accurate for pleated design.  These changes could be made in both the 
pressure drop estimates and the breakthrough predictions.  Pressure drop across pleated 
media contains inertial losses caused by a change in direction of flow of the bulk fluid.  
This causes an uneven distribution of pressure drop across the filter.  An uneven pressure 
drop distribution may result in an uneven flow regime causing some areas to be 
challenged at a higher flow rate; hence, experiencing breakthrough at an earlier time. 
 
 
 
 73
VII. CONCLUSIONS AND FINAL COMMENTS 
 
Optimization of a cathode air filter design to a variety of different fuel cell 
applications involves inputting design considerations into an algorithm predicting 
compressor power and filter breakthrough characteristics, both of which reflect fuel cell 
performance and net power output.  Results from application of a programmable design 
algorithm to a 1.2kWe fuel cell show that the algorithm and design methodology are 
effective means of predicting filter effects on compressor power requirements.  The 
predictions also show trends useful in tailoring a filter to specific requirements of an 
application. 
Microfibrous materials provide the lowest pressure drop for cases requiring high 
log removal due to their high contacting efficiency.  Packed beds have the lowest 
pressure drop for applications requiring a high capacity.  However, the bed depth of a 
packed bed becomes large when meeting high log removal requirements.  An optimized 
composite bed provides the most favorable solution for cases requiring both high 
contacting efficiency and high capacity.  A filter must be optimized for each contaminant 
type and concentration to achieve the required operating life of the fuel cell and to 
minimize pressure drop which can consume as much as 1% of total output power.   Use 
of a design algorithm provides a means of optimization by comparing the effectiveness of 
 74
different filter configurations, ultimately yielding higher fuel cell efficiency and 
operating life.  
 75
REFERENCES 
 
1. Moore, Jon M., Lakeman J. Barry, Mepsted, Gary O.  ?Development of a PEM Fuel  
     Cell Powered Portable Field Generator for the Dismounted Soldier? Journal of Power  
     Sources 106 (2002) 16-20 
2. Larminie, James and Dicks, Andrew. ?Fuel Cell Systems Explained?, SAE  
     International; 2nd edition, May 2003, Chapter 1 and Appendix A 2.2. 
3. Moore, Jon M., Adcock, Paul L., Lakeman, J. Barry, and Mepsted, Gary O.  ?The  
     effects of Battlefield Contaminants on PEMFC Performance?  Journal of Power  
     Sources 85 (2000) 254-260 
4. Costamagna, Paola and Srinivasan, Supramaniam.  ?Quantum Jumps in the PEMFC  
     Science and Technology from 1960s to the Year 2000: Part I. Fundamental Scientific  
     Aspects?. Journal of Power Sources 102 (2001) 242-252 
5. Mohtadi, R., Lee W. and Van Zee, J.W.  ?Assessing Durability of Cathodes Exposed  
     to Common Air Impurities?  Journal of Power Sources 138 (2004) 216-225 
6. Cahela, Donald R., and Bruce J. Tatarchuk. ?Design of Polishing Filter Adsorbents  
     Using Sintered Microfibrous Metallic Networks as Carriers for High Effectiveness       
     Sorbent Particulates.? Proceedings of the AIChE 2003 Annual Meeting, Nov. 16 ? 21,    
     2003: Fundamentals of Adsorption and Ion Exchange II. San Francisco, CA. 
 
 76
7. Cacciola, G., Antonucci, V., Freni, S., ?Technology up date and new strategies on fuel  
     cells?  Journal of Power Sources 100 (2001) 67-79 
8. Ballard NEXA Fuel Cell Manual, Ballard Power Systems, Inc (2003) 
9. Hoogers, Gregory.  ?Fuel Cell Technology Handbook? CRC Press LLC 2003 1-5 
10. Jakobert, B.J., Veldhuis, Frank A de Bruijn, and Ronald K.A.M. Mallant.  ?Air  
Pollution: A Problem for the SPFC Cathode?? 1998 Fuel cell seminar, Palm Springs, 
California, USA, 16-19 November 1998 
11. Ruthven, Douglas M. Principles of Adsorption and Adsorption Processes. New York: 
Wiley-Interscience, 1984. 
12. Harris, Daniel K., Cahela, Donald R., Tatarchuk, Bruce J. ?Wet Layup and Sintering 
of Metal-Containing Microfibrous Composites for Chemical Processing 
Opportunities.?  Composites: Part A 32 (2001) 117-1126 
13. Queen, Andy ?High Efficiency Adsorbent Filters via Packed Bed + Polishing Sorbent 
Architectures for Regenerable Collective Protection Equipment? 
14. Amundson, Neal R. ?A Note on the Mathematics of Adsorption in Beds? J.Phys 
Colloid Chem. 52 (1948) 1153-1157 
15. Tatarchuk, Bruce J. and Cahela, Donald R. ?Permeadbility of Sintered Microfibrous 
Composites for Heterogeneous Catalysis and Other Chemical Processing 
Opportunities.?   Catalysis Today 69 (2001) 33-39 
16. Ueno, K., Bye, R.E., Hunter, K.S., VAIREX Corporation, ?Compressor Efficiency 
Definitions? May 2003. http://www.vairex.com/pdf/Efficiency-0523.pdf 
 77
17. Jakobert, B.J., Veldhuis, Frank A de Bruijn, and Ronald K.A.M. Mallant.  ?Air 
Pollution: A Problem for the SPFC Cathode?? 1998 Fuel cell seminar, Palm Springs, 
California, USA, 16-19 November 1998 
18. Mohtadi, R., Lee W. and Van Zee, J.W.  ?Assessing Durability of Cathodes Exposed 
to Common Air Impurities?  Journal of Power Sources 138 (2004) 216-225 
19. Kosanovic, Lisa.  ?Running on Air? Mechanical Engineering January (2004) 40- 
42 
20. Costamagna, Paola and Srinivasan, Supramaniam.  ?Quantum Jumps in the PEMFC 
Science and Technology from 1960s to the Year 2000: Part I. Fundamental Scientific 
Aspects?. Journal of Power Sources 102 (2001) 242-252 
21. Lister S., McLean, G. ?PEM fuel cell electrodes? Journal of Power Sources130 
(2004) 61-76 
22. Ashley, Steven. ?On the road to fuel cell cars?  Scientific American March 2005 62-
69 
23. Betournay, Marc C., Bonnell, Gary, Edwardson, Eric, Paktunc, Dogan, Kaufman, 
Arthur, Lomma, A. Timothy.  ?The effects of mine conditions on the performance of 
a PEM fuel cell.?  Journal of Power Sources 134 (2004) 80-87 
24. Qi, Zhigang, He, Chunzhi, Kaufman, Arthur. ?Effect of CO in the anode fuel on the 
performance of PEM fuel cell cathode? Journal of Power Sources  111 (2002) 239-
247 
 
APPENDIX A. AIR FLOW DERIVATION 
 
 
Referenced from Fuel Cell Systems Explained 
 
Starting with: 
 
2
4 OFcharge ?=  
 
?
?
?
?
?
?
=
s
moles
F
I
O
4
2
 
 
For multiple stacks: 
 
 
?
?
?
?
?
?
?
?
=
s
moles
F
nI
O
4
2
 
 
To substitute in for current: 
 
  nIVP
ce
??=
 
 
nV
P
I
c
e
?
=  
  
Substituting for current and converting from moles/s to kg/s 
 
 
?
?
?
?
?
?
??
??
=
?
s
kg
FV
P
O
c
e
4
1032
3
2
 
 
Dividing by 0.21 to find total air flow, converting to the more commonly used units of 
SLPM and multiplying by the stoichiometric ratio (? = (total oxygen supplied)/(total 
oxygen used by FC)): 
 
 )(1082.1
2
SLPM
V
P
AirFlow
c
e
???=
?
?  
 
 
 78
 79
APPENDIX B. SAMPLE MAPLE OUTPUT FOR A COMPOSITE BED 
 
 
 
> restart; 
> Coppm := 100; #ppm 
 
                             Coppm := 100 
 
> Coutppm := 5; #ppm 
 
                             Coutppm := 5 
 
> Co := Coppm/(1000000*22.4*1000); #mol/cm3 
 
                                            -8 
                        Co := .4464285714 10 
 
> Cout := Coutppm/(1000000*22.4*1000); #mol/cm3 
 
                                             -9 
                       Cout := .2232142857 10 
 
> BT := 2*60; #in minutes 
 
                              BT := 120 
 
> Area := 55.5; #sq cm 
 
                             Area := 55.5 
 
> stoic := 2.7; #Minimum Stoichiometric Ratio Required 
 
                             stoic := 2.7 
 
> Vcell := 0.60; #Average Voltage Across a Cell  FROM BALLARD MANUAL 
 
                             Vcell := .60 
 
> FCPower := 1200; #Operating Power Output 
 80
 
                           FCPower := 1200 
 
> O2used := 3.48E-3*FCPower/Vcell; #in SLPM 
 
                        O2used := 6.960000000 
 
> AirFlow := O2used*stoic/0.21; #AirFlow in SLPM 
>  
 
                        AirFlow := 89.48571429 
 
> AirFlowKg := AirFlow*2.38*10^(-5); #AirFlow in kg/s 
 
                      AirFlowKg := .002129760000 
 
> P2 := 0.6453*AirFlow - 7.1332; #AirFlow to Stack Inlet Pressure #in "H2O 
 
                          P2 := 50.61193143 
 
> P1 := (406.782-((406.782+P2)/1.135))*2.54; #cm H2O 
 
                           P1 := 9.6310501 
 
> vo := AirFlow*1000/(Area*60); #cm/sec 
 
                          vo := 26.87258687 
 
> dP2 := 0.0664*T2*vo; #cm of H2O 
 
                        dP2 := 1.784339768 T2 
 
> dP1 := 0.0514*T1*vo; #cm of H2O 
 
                        dP1 := 1.381250965 T1 
 
> solve(dP1+dP2 = P1, T2); 
 
                    -.7740963855 T1 + 5.397542706 
 
> T2 := %; 
 
                 T2 := -.7740963855 T1 + 5.397542706 
 
 81
> plot(T2, T1=0..5, 0..5, labels = ["Thickness of Packed Bed (cm)", "Total Thickness of 
MFM (cm)"], font=[COURIER, 12], labeldirections=[HORIZONTAL,VERTICAL], 
numpoints=1000); 
 
> vo :=AirFlow*1000/Area; #cm/min 
 
                          vo := 1612.355212 
 
> tau1 := 0.1*T11/(vo*Co*86); 
 
                       tau1 := 161.5432595 T11 
 
> tau2 := 0.04*T22/(vo*Co*86); 
 
                       tau2 := 64.61730377 T22 
 
> k1 := 0.183; 
 
                              k1 := .183 
 
> k2 := 0.332; 
 
                              k2 := .332 
 
> epsilon1 := 0.4; 
 
                            epsilon1 := .4 
 
> epsilon2 := 0.772; 
 
                           epsilon2 := .772 
 
> Conc1 := Co/(1+(exp(epsilon1*k1*tau1)-1)*exp(-epsilon1*k1*BT)); 
 
                         -8 
  Conc1 := .4464285714 10 
 
                                1 
        ------------------------------------------------- 
        .9998468358 + .0001531642014 exp(11.82496660 T11) 
 
> Conc2 := (Conc1*(1+exp(-epsilon1*k1*tau1)*(exp(epsilon1*k1*BT)-
1))^((epsilon2*k2)/(epsilon1*k1)))/((exp(epsilon2*k2*tau2)-1)+(1+exp(-
epsilon1*k1*tau1)*(exp(epsilon1*k1*BT)-1))^((epsilon2*k2)/(epsilon1*k1))); 
>  
 82
 
                         -8 
  Conc2 := .4464285714 10 
 
                                               3.501420765   / 
        (1 + 6527.940774 exp(-11.82496660 T11))             /  ( 
                                                           / 
 
        (.9998468358 + .0001531642014 exp(11.82496660 T11)) ( 
 
        exp(16.56167343 T22) - 1 
 
                                                  3.501420765 
         + (1 + 6527.940774 exp(-11.82496660 T11))           )) 
 
> solve(Cout = Conc2, T22); 
 
                               / 
                               | 
                               |               13                 14 
  .06038037184 ln(-.3333333333 \-.4999234179 10   - .9500076582 10 
 
                                                /700284153\ 
                                                |---------| 
                                                \200000000/ 
        (1. + 6527.940774 exp(-11.82496660 T11)) 
 
                        9                                     9 
         - .765821007 10  exp(11.82496660 T11) + .765821007 10 
 
        exp(11.82496660 T11) 
 
                                                /700284153\\ 
                                                |---------|| 
                                                \200000000/|   / 
        (1. + 6527.940774 exp(-11.82496660 T11))           /  /  ( 
                                                             / 
 
                      13                9 
        .1666411393 10   + .255273669 10  exp(11.82496660 T11))) 
 
> T22 := %; 
 
                                      / 
                                      | 
 83
                                      |               13 
  T22 := .06038037184 ln(-.3333333333 \-.4999234179 10   - 
 
                      14 
        .9500076582 10 
 
                                                /700284153\ 
                                                |---------| 
                                                \200000000/ 
        (1. + 6527.940774 exp(-11.82496660 T11)) 
 
                        9                                     9 
         - .765821007 10  exp(11.82496660 T11) + .765821007 10 
 
        exp(11.82496660 T11) 
 
                                                /700284153\\ 
                                                |---------|| 
                                                \200000000/|   / 
        (1. + 6527.940774 exp(-11.82496660 T11))           /  /  ( 
                                                             / 
 
                      13                9 
        .1666411393 10   + .255273669 10  exp(11.82496660 T11))) 
 
> plot(T22, T11=0..2, numpoints=1000, labels = ["Thickness of Packed Bed (cm)", 
"Thickness of MFM (cm)"], font=[COURIER, 12], 
labeldirections=[HORIZONTAL,VERTICAL]); 
 
> T11 := T1; 
 
                              T11 := T1 
 
> plot([T22 , T2], T1=0..5, 0..8, labels = ["Thickness of Packed Bed (cm)", "Total 
Thickness of MFM (cm)"], font=[COURIER, 12], 
labeldirections=[HORIZONTAL,VERTICAL], numpoints=1000); 
 
> Ttotal2 := T22+T1; 
 
                                          / 
                                          | 
                                          |               13 
  Ttotal2 := .06038037184 ln(-.3333333333 \-.4999234179 10   - 
 
                      14 
 84
        .9500076582 10 
 
                                               /700284153\ 
                                               |---------| 
                                               \200000000/ 
        (1. + 6527.940774 exp(-11.82496660 T1)) 
 
                        9                                    9 
         - .765821007 10  exp(11.82496660 T1) + .765821007 10 
 
        exp(11.82496660 T1) 
 
                                               /700284153\\ 
                                               |---------|| 
                                               \200000000/|   / 
        (1. + 6527.940774 exp(-11.82496660 T1))           /  /  ( 
                                                            / 
 
                      13                9 
        .1666411393 10   + .255273669 10  exp(11.82496660 T1))) + T1 
 
> Ttotal1:= T2+T1; 
 
               Ttotal1 := .2259036145 T1 + 5.397542706 
 
> plot([Ttotal1, Ttotal2], T1=0..6, 0..6, labels = ["Thickness of Packed Bed (cm)", "Total 
Thickness of Composite (cm)"], font=[COURIER, 12], 
labeldirections=[HORIZONTAL,VERTICAL], numpoints=1000); 
 
> Ttotaldiff := diff(Ttotal2, T1); 
 
                             /             / 
                             |             | 
                             |             | 
  Ttotaldiff := -.1811411155 |-.3333333333 \ 
                             \ 
 
                                              /500284153\ 
                                              |---------| 
                      20                      \200000000/ 
        .2567719253 10   (1. + 6527.940774 %2)            %2 
 
                         10                    10 
         - .9055807829 10   %1 + .9055807829 10   %1 %3 
 
 85
                                                    /500284153\   \ 
                                                    |---------|   | 
                         15                         \200000000/   | 
         - .2069892097 10   %1 (1. + 6527.940774 %2)            %2/ 
 
           /                13                9 
          /  (.1666411393 10   + .255273669 10  %1) + 
         / 
 
                      10                 13                 14 
        .1006200870 10   (-.4999234179 10   - .9500076582 10   %3 
 
                        9                   9             / 
         - .765821007 10  %1 + .765821007 10  %1 %3) %1  / 
                                                        / 
 
                                               \ 
                                               | 
                       13                9    2| 
        (.1666411393 10   + .255273669 10  %1) | 
                                               / 
 
                       13                9       /                 13 
        (.1666411393 10   + .255273669 10  %1)  /  (-.4999234179 10 
                                               / 
 
                         14                   9 
         - .9500076582 10   %3 - .765821007 10  %1 
 
                        9 
         + .765821007 10  %1 %3) + 1 
 
  %1 := exp(11.82496660 T1) 
 
  %2 := exp(-11.82496660 T1) 
 
                             /700284153\ 
                             |---------| 
                             \200000000/ 
  %3 := (1. + 6527.940774 %2) 
 
> fsolve(0=Ttotaldiff, T1); 
 
                             .8901518117 
 
 86
> T1:=%; 
 
                          T1 := .8901518117 
 
> T2 := T22; 
 
                          T2 := .09063040950 
 
> Ptotal := dP1+dP2; 
 
                        Ptotal := 1.391238493 
 
> Pratio := (406.782+P2)/(406.782-Ptotal); 
 
                        Pratio := 1.128279119 
 
> Tpower := 303.56*Pratio - 302.13; #Watts 
 
                         Tpower := 40.3704094