Professional Documents
Culture Documents
L.O. VASQUEZ
EDITOR
All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.
For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com
Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.
This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.
Preface vii
Expert Commentary 1
Qualification of Fuel Cell Membrane Electrode Assemblies 3
Zhigang Qi
Research and Review Studies 7
Chapter 1 Power and Energy Efficiency Analysis of Direct Methanol 9
Fuel Cell (DMFC) from Single Cell, Fuel Cell Stack
to DMFC System
Rongzhong Jiang
Chapter 2 Experimental Activity on a Large SOFC Generator 71
M. Santarelli, P. Leone, M. Cal and G. Orsello
Chapter 3 Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 135
Angelo Basile and Adolfo Iulianelli
Chapter 4 A Potential Alternative in the Electric Utility 161
Francisco Jurado
Chapter 5 Microfabrication Techniques: Useful Tools for Miniaturizing 211
Fuel Cells
Tristan Pichonat
Chapter 6 Technical H2 Electrodes for Low Temperature Fuel Cells 247
Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
Chapter 7 PEM Fuel Cell Modeling 273
Maher A.R. Sadiq Al-Baghdadi
Chapter 8 Surface Functionalization of Carbon Catalyst-Support for PEM 381
Fuel Cells: A Review
Zhigang Qi
vi L.O. Vasquez
Chapter 9 A Study on the Role of Carbon Support Materials for Fuel-Cell 411
Catalysts
Seok Kim and Soo-Jin Park
Chapter 10 Developments of Alkaline Solid Polymer Electrolyte Membranes 445
Based on Polyvinyl Alcohol and Their Applications
in Electrochemical Cells
G.M. Wu, S.J. Lin and C.C. Yang
Index 473
PREFACE
will significantly increase the power density of the DMFC system, but may not increase the
energy efficiency. Increasing oxygen partial pressure will increase both the power density and
the energy efficiency significantly. It is apparent that the energy efficiency of a DMFC
system is sensitive to the power or current output. With increasing power or current output,
the energy efficiency increases significantly until reaching a peak value, at which a voltage or
current limit is met.
As presented in Chapter 2, the Multiscale Analysis Group of the Politecnico di Torino
(Italy) is involved in the experimental analysis and modeling of the CHP-100 SOFC Field
Unit built by Siemens Power Generation Stationary Fuel Cells (SPG-SFC). The experimental
analysis of a large SOFC generator in operation is a complex task, due to the large number of
variables which affect its operation, the limited number of measurements points in the
generator volume, the necessity to avoid malfunctions in the real operation. As a
consequence, the experimental analysis of the CHP-100 SOFC Field Unit has been developed
with methods of Design of Experiments, and with a statistical analysis of the collected data.
The experimental sessions have been designed in order to investigate the effect of two
important operation factors (the overall fuel consumption FC and the air stoichs ox), in order
to characterize the operation of the single sectors of the SOFC generator, and to obtain the
sensitivity maps of the main investigated dependent variables. Furthermore, the main result is
the estimation of the local values of fuel utilization of the various sectors of the generator,
through the combination of the experimental voltage sensitivity analysis to overall FC and an
analytical model of polarization, to outline the distribution of fuel inside the generator.
Finally, the sectors of the generator, of different pedigree and position, are compared in terms
of the polarization effects, showing how the local fuel utilization and temperature affect the
estimated local anode exchange current density values.
Commonly, polymer electrolyte membrane fuel cells operate at temperature <100C
because above this temperature the electrochemical performances of the Nafion drop down.
Therefore, many scientists have studied different types of no-fluorinated polymers as
alternative to the Nafion. According to the literature, comparable performances to the Nafion
in terms of proton conductivity and thermo-chemical properties, lower crossover and cost are
the characteristics that can be obtained by using treated polyetheretherketone (PEEK).
Different methods are used for producing electrolyte membranes from PEEK: a) PEEK
electrophilic sulfonation (S-PEEK); b) S-PEEK and no-functional polymers blending; c) S-
PEEK, heteropolycompounds and polyetherimide doping with inorganic acids; and so on. In
particular, the sulfonation of polyetheretherketone/cardo-group (PEEK-WC) by using
sulphuric acid is also presented in Chapter 3.
A fully mature fuel cell industry constitutes a potential opportunity to electric utilities. It
could meliorate the technical and financial performance of existing distribution lines by
improving service quality and reliability.
When fuel cells connect to the power system, both the owner of the energy resource and
the central power system benefit. Reliability increases for both because they can support each
other.
However, the interconnection of fuel cell plants to the grid is still hindered by restrictive
conditions and procedures for grid connection. Problems arise with regard to determination of
the point of connection, safety and stability issues. Most important it is needed to establish a
standardized technical interface for allocation of connection that take into account possible
positive effects of distributed generation on transmission and distribution losses.
Preface ix
At last, the aim would be the substitution of conventional coal-fired generation with
Integrated Gasification Combined Cycle plants that integrate fuel cells and gas turbines.
Chapter 4 studies the use of distributed resources for ancillary services and simulating the
impact of distributed resources on utility distribution networks.
Chapter 5 will introduce the reader to the reasons for miniaturizing fuel cells and to the
specifications required by this miniaturization. It will then show what kinds of fuel cells can
fit to these specifications and which fuels can be employed to supply them. The techniques
presently used for the realization of miniature fuel cells will be described, underlining
particularly the growing part of the microfabrication techniques inherited from
microelectronics. It will present an overview on the applications of these latter techniques on
miniature fuel cells by presenting several solutions developed throughout the world. It will
finally detail, as an example, the complete fabrication process of a particular microfabricated
fuel cell based on a silane-grafted porous silicon membrane as the proton-exchange
membrane instead of a common ionomer such as Nafion.
Hydrogen electrode reaction has been widely studied regarding to its applications in fuel
cell technology. In the last years, a lot of studies concerning kinetic and mechanistic aspects
about smooth and well defined surfaces have been recently published, but less is known about
the electrochemical behavior of technical electrodes, in half-cells and complete cells. In
Chapter 6, the main manufacture methods of technical hydrogen electrodes reported in the
literature for low temperature fuel cells such as alkaline fuel cells, phosphoric acid fuel cells
(PAFCs), and polymer electrolyte fuel cells (PEFCs) is examined. The kinetics of hydrogen
electrode oxidation reaction both, in liquid and solid electrolytes is also reviewed. Previous
work constitutes a significant background that can help to develop technical hydrogen
diffusion anodes for application in practical fuel cells. In particular, the electrochemical
behavior of such anodes is correlated with that observed on well characterized surfaces.
Fuel Cells are growing in importance as sources of sustainable energy and will doubtless
form part of the changing program of energy resources in the future. Two key issues limiting
the widespread commercialization of fuel cell technology are better performance and lower
cost. The development of physically representative models that allow reliable simulation of
the processes under realistic conditions is essential to the development and optimization of
fuel cells, the introduction of cheaper materials and fabrication techniques, and the design and
development of novel architectures. The difficult experimental environment of fuel cell
systems has stimulated efforts to develop models that could simulate and predict multi-
dimensional coupled transport of reactants, heat and charged species using computational
fluid dynamic (CFD) methods. The strength of the CFD numerical approach is in providing
detailed insight into the various transport mechanisms and their interaction, and in the
possibility of performing parameters sensitivity analyses.
Among all kinds of fuel cells, proton exchange membrane (PEM) fuel cells are compact
and lightweight, work at low temperatures with a high output power density and low
environmental impact, and offer superior system startup and shutdown performance. These
advantages have sparked development efforts in various quarters of industry to open up new
field of applications for PEM fuel cells, including transportation power supplies, compact
cogeneration stationary power supplies, portable power supplies, and emergency and disaster
backup power supplies.
This chapter of "PEM Fuel Cell Modeling" looks at how engineers can model PEM fuel
cells to get optimal results for any application. A review of recent literature on PEM fuel cell
x L.O. Vasquez
modeling was presented. A full three-dimensional, non-isothermal CFD model of a PEM fuel
cell with straight flow field channels has been developed in this chapter. The model was
developed to improve fundamental understanding of transport phenomena in PEM fuel cells
and to investigate the impact of various operation parameters on performance. This
comprehensive model accounts for the major transport phenomena in a PEM fuel cell:
convective and diffusive heat and mass transfer, electrode kinetics, transport and phase
change mechanism of water, and potential fields. In addition, the hygro and thermal stresses
in membrane, which developed during the cell operation, were modeled and investigated. The
new feature of the algorithm developed in this model is its capability for accurate calculation
of the local activation overpotentials, which in turn results in improved prediction of the local
current density distribution. Fully three-dimensional results of the velocity flow field, species
profiles, liquid water saturation, temperature distribution, potential distribution, water content
in the membrane, stresses distribution in the membrane, and local current density distribution
are presented and analyzed with a focus on the physical insight and fundamental
understanding. The model is shown to be able to understand the many interacting, complex
electrochemical, and transport phenomena that cannot be studied experimentally. Chapter 7 is
a practical summary of how to create CFD models, and how to interpret results.
In order to achieve high performance and low catalyst loading, a proton-exchange
membrane (PEM) fuel cell typically employs noble metal catalysts that are dispersed on a
support such as carbon. The chemical and physical properties of carbon largely affect the
dispersion of the catalyst, the strength of the interaction between the carbon and the catalyst
particles, the making of catalyst ink formulations, and the utilization of the catalyst. An
interesting and useful aspect of carbon is that its surface can be chemically modified to render
it with certain desired properties. For example, proton conducting groups can be covalently
bonded onto the surface of carbon black such as Vulcan XC-72 to make it possess some ionic
conductance, which in turn significantly increases the catalyst utilization and the fuel cell
performance. Accompanying all the benefits, a carbon-type support also raises some potential
problems. One serious concern is the corrosion of an amorphous carbon support during the
operation of a fuel cell, which subsequently results in the loss of the catalyst-electrolyte-
reactant three-phase sites. This factor alone may prevent a PEM fuel cell from achieving a
target of 40,000 hours of operation for stationary applications. Chapter 8 reviews various
aspects of surface functionalization of carbon supports for PEM fuel cells.
The ideal support material for fuel-cells catalysts should have the following
characteristics: high electrical conductivity, adequate water-handling capability at the
electrode, and also good corrosion resistance under oxidizing conditions. Whereas carbon
blacks are the common support materials for electrocatalysts, new forms of carbon materials
such as graphite nanofibers (GNFs), carbon nanotubes (CNTs), ordered mesoporous carbons
(OMC) had been investigated as catalyst supports.
In Chapter 9, the size and the loading efficiency of metal particles were investigated by
changing the preparation method of carbon-supported platinum catalysts. First, the effect of
acid/base treatment on carbon blacks supports on the preparation and electroactivity of
platinum catalysts. Secondly, binary carbon-supported platinum (Pt) nanoparticles were
prepared using two types of carbon materials such as carbon blacks (CBs) and graphite
nanofibers (GNFs) to check the influence of carbon supports on the electroactivity of catalyst
electrodes. Lastly, plasma treatment or oxyfluorination treatment effects of carbon supports
Preface xi
Zhigang Qi
Plug Power Inc.
A membrane electrode assembly (MEA), where the fuel cell anode and cathode half-
reactions occur, is the heart and the most delicate part of a fuel cell system. The performance,
stability, and durability of a fuel cell largely depend on the quality of the MEAs. Therefore,
MEA qualification should be critical in developing durable, high performance fuel cell
systems.
Ideally, each MEA should be qualified according to the following categories before it is
assembled into a fuel cell stack:
5. Pinholes
Pinholes refer to small holes through the thickness of the membrane. These pinholes
allow more reactants to cross over, resulting in lower open circuit voltage (OCV), more waste
of fuels, and faster decrease in MEA life. Combustion may occur between hydrogen and air
around the pinholes and the heat generated could enlarge the pinholes quickly, which in turn
leads to higher reactant crossover rates and more severe combustion. Pinholes may arise from
the original defects in the membrane, or be created during the MEA fabrication process, or
gradually form due to localized decay of the membrane materials. It is critical to have
pinhole-free virgin MEAs for longer fuel cell lifetime.
6. Electrode shorting
If electrons can move from the anode to the cathode directly through the membrane, the
electrodes are then shorted. If the electronic resistance between the anode and the cathode of
an MEA is not high enough, it is an indication that the anode and the cathode may be in
contact with each other at some points. The fuel cell OCV will be lower due to shorting, a
symptom similar to that caused by pinholes. One effective way to distinguish shorting from
pinholes is to monitor the OCV change with the pressure difference between the fuel and the
air. If the OCV drops significantly with an increase in the pressure difference, it is more likely
to be due to pinholes. If the OCV decreases little with the increase in the pressure difference,
it is more likely to be shorting. Shorting is most likely caused by carbon fibers or catalyst
clumps being pushed through the membrane during the MEA fabrication process. So, it could
be very helpful to smooth the surfaces of electrodes before they are bonded onto a membrane.
11. Performance
One of the goals of these above characterizations is the achievement of high and uniform
fuel cell performance. These characterizations provide the necessary knowledge about what
parameters should and can be tailored in order to achieve higher performance. Once a good
and uniform performance is consistently achieved, some or all of these characterizations can
be eliminated. The best way to gauge the performance of an MEA is to collect voltage
current polarization curves.
12. Durability
Another goal of these characterizations is the achievement of durable fuel cell
performance. Durability can be gauged by the fuel cell performance decay rate and the
likelihood of premature failures. If a fuel cell suffers frequent premature failures, which are
mostly due to the breach of the membrane, it becomes impossible to even discuss durability.
Once premature failures are eliminated, the performance decay rate can be used to predict
how many hours the fuel cell can run before it hits the low performance or efficiency limit.
Performance decay is mostly related to increases in catalyst particle size and increases in the
mass transport resistance of the catalyst layer. Testing an MEA under a properly chosen
accelerated condition can offer some guidance on its decay rate under regular fuel cell
conditions.
MEA manufacturing process and assume that future MEAs from that process are good unless
testing in a fuel cell indicates that they are not. At that time, the entire fuel cell stack has to be
opened to replace the bad MEAs. This is a time-consuming process and frequently damages
some of the previously good MEAs.
Before a simple universal method that can quickly and non-destructively evaluate the
quality of an MEA is established, the best practical approach in the moment is to optimize the
MEA manufacturing process so that MEAs with consistent and controlled properties are made
every time.
RESEARCH AND REVIEW STUDIES
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 9-70 2007 Nova Science Publishers, Inc.
Chapter 1
Rongzhong Jiang
Sensors and Electron Devices Directorate, U.S. Army Research Laboratory
2800 Powder Mill Road, Adelphi, MD 20783-1197
Abstract
Direct methanol fuel cell (DMFC) is a device to directly convert the chemical energy of
methanol into electricity through the electrochemical reaction between methanol and oxygen.
For practical application, a number of single fuel cells connected in series form a fuel cell
stack to gain higher voltage and power. A practical fuel cell device is called fuel cell system,
which is built by integration of multiple complex parts, including fuel cell stacks, pumps,
batteries, sensors, fuel cartridge and electronic controller. The theoretical energy density of
methanol in DMFC is 6081 Wh/Kg. So far only one sixth of the theoretical value can be
obtained for an optimized DMFC system. This article gives a detailed analysis of power and
energy efficiency in DMFC single cell, stack and system by experimental research and by
simulation with semi-empirical equations. In a DMFC single cell, the energy efficiency is
dependent on operating conditions, such as methanol concentration, cell temperature and
partial pressure of oxygen. About 1700 Wh/Kg was obtained for DMFC single cell at 60 oC
under the operating conditions of 0.5M methanol and 1 atmosphere of air pressure. From a
single cell to a DMFC system the energy density is decreased by about 20 percent due to the
energy consumption by the auxiliary components, such as air and fuel pumps, heat dissipation
and internal system impedance. About 1400 Wh/Kg of energy density can be obtained for a
DMFC system by optimization of the operating conditions. Increasing operating temperature
will significantly increase the power density of the DMFC system, but may not increase the
energy efficiency. Increasing oxygen partial pressure will increase both the power density and
the energy efficiency significantly. It is apparent that the energy efficiency of a DMFC system
is sensitive to the power or current output. With increasing power or current output, the
energy efficiency increases significantly until reaching a peak value, at which a voltage or
current limit is met.
10 Rongzhong Jiang
Introduction
Energy consumption plays an important role in our modern civilization and daily life, which
is heavily dependent on burning fossil fuels. The increasing threat by the fast depletion of the
resources of petroleum, coal and natural gas, and, in turn, the green house effect by burning
fossil fuels, forces people to seek regenerative energy sources, such as solar, wind,
geothermal and hydroelectric energies. An alternative way to save valuable natural resources
and solve the environmental problem is to develop cleaner and more efficient energy
conversion devices. In recent years, fuel cell research and development have received much
attention [1-8] for its higher energy conversion efficiency and lower or non greenhouse-gas
emissions than thermal engines in the processes of converting fuel into usable energies. The
power and energy efficiency of a fuel cell is highly dependent on the thermodynamics,
electrode kinetics, reactant mass transfer, as well as materials and components for assembling
the fuel cell. These factors have been addressed throughout the fuel cell history, and are now
still the major challenges for fuel cell research and development.
The concept of a fuel cell was proposed about 170 years ago when William Robert Grove
conceived the first fuel cell in 1839, which produced water and electricity by supplying
hydrogen and oxygen into a sulfuric acid bath in the presence of porous platinum electrodes.
Unfortunately, there were no practical fuel cells developed for the following 120 years until
Dr. Francis Bacon demonstrated a 5kW fuel cell for powering a welding machine in 1959,
where an inexpensive nickel electrode and less corrosive alkaline electrolyte were used. One
of the most important milestones in fuel cell history is an invention of polymer electrolyte
membrane (PEM) in 1955 when Willard Thomas Grubb in General Electric (GE) modified
the original fuel cell design with a sulfonated polystyrene ion-exchange membrane as the
electrolyte. Better polymer electrolyte material, sulfonated tetrafluorethylene copolymer
(Nafion) [9-10], was discovered in late 1960s by Walther Grot at DuPont. With its excellent
thermal and mechanic stability, Nafion became the most widely used electrolyte material for
PEM fuel cells. The use of solid polymer electrolyte membrane has established the base for
the modern fuel cell technology because the fuel cell with polymer electrolyte membrane is
much simpler and more reliable than that of using liquid electrolyte. The second important
factor in fuel cell history is the development of electrode catalysts for oxygen reduction and
fuel oxidation. Much effort has been made to seek non-platinum catalysts, such as metallo-
porphyrins and metallo-phthalocyanines [11-14] for catalytic oxygen reduction. Anson and
Collman et al [15-16] have synthesized and studied dimeric cofacial cobalt porphyrins, which
demonstrated the capability of catalyzing oxygen 4-electron reduction to water. However, the
catalytic activity of these transition metal macrocycles is not stable in acidic electrolytes.
Although an approach of heat-treatment [17-19] was proposed to enhance the stability, these
transition metal macrocycles have not actually been used as catalysts in a fuel cell because of
uncompetitive catalytic activity and stability as compared to platinum based catalysts. Up to
the present, electrode catalysts and electrolyte membranes are still the major challenges in
fuel cell research and development.
Since the beginning of the invention of fuel cells, hydrogen has primarily been
considered as the fuel. The difficulty in safe storage and transportation of hydrogen has
limited its wide applications in automobiles and portable electronics. Under the motivation of
US department of Defense in 1960s, the earliest work to develop fuel cells that could operate
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 11
on various types of liquid hydrocarbons started, such as direct methanol fuel cell (DMFC)
[20-21]. However, the research on DMFC was relatively slow during the first 30 years [22-
26] due to two main technical barriers, i.e., slow catalytic kinetic rate of methanol oxidation
at the anode electrode, and methanol crossover from the anode to the cathode through the
electrolyte membrane, which causes cathode electrode depolarization. The application of
Nafion perfluorosulfonic acid as solid polymeric electrolyte membrane [10, 27-30] has
relatively blocked the methanol crossover in comparison to using liquid electrolytes. The
discovery of PtRu binary metal alloy [26] has significantly overcome the anode poisoning of
DMFC caused by the formation of an intermediate CO in the process of methanol oxidation
on the Pt catalyst surface.
With rapidly increasing demands for portable power sources by telecommunications,
computers, and portable electronics in the last decade, research on DMFC has received much
attention since 1990 [3, 31-44]. Most of the research articles focus on better electrode
catalysts, electrolyte membranes, and understanding the catalytic mechanisms of methanol
oxidation. Many other publications are dealing with the transport phenomena and methanol
crossover within a DMFC [10, 45-49]. There is much less attention for study of an actual fuel
cell power source. For an actual application, a number of single fuel cells have to be
connected in series to form a fuel cell stack to obtain higher voltage and power. In order to
make the fuel cell stack be operable, we have to supply fuel, air and water to the stack; and
remove the products (CO2 and water) from the stack; as well as optimize the temperature and
humidity. Therefore, many complex components have to be integrated together in a fuel cell
box including air blower, fuel pump, sensors, heat exchanger, fuel reservoir, fuel mixer,
startup batteries and electronic controller. A practical fuel cell power source is a hybrid
system [50-53], which is called a fuel cell system. A fuel cell systems performance is not
only dependent on these materials and components; but also dependent on the art of the
integration. With the progress in fuel cell design and manufacture, several DMFC system
prototypes have emerged in the domestic and global markets [54-60] in recent years.
The energy efficiency of a fuel being converted into electrical energy is one of the major
concerns for a fuel cell system. However, most efforts have been spent in separately pursuing
higher power or higher energy density in fuel cell research and development. The relationship
between power and energy efficiency has not received much attention [7, 60-62] in the
research and development of fuel cells. Especially, a quantitative analysis of power and
energy efficiencies among single cell, fuel cell stack and fuel cell system, has not been
reported in the literature so far. The present article explores the relevance between power and
energy efficiency, and studies how the power and energy efficiency are varying from single
cell, fuel cell stack, to DMFC system based on experimental data and simulation with semi-
empirical equations.
Experimental
Nafion 117 (purchased from DuPont) was used as electrolyte membrane for the DMFC single
cell, which was pretreated in mildly boiling water with 3% H2O2 for 2 hours, then boiled in 2
M H2SO4 for 2 hours. For each treatment the membrane was washed in de-ionized water
several times. After these treatments it was stored in water for preparation of membrane
electrode assembly (MEA). Johnson Matthey's unsupported Pt black (2mg/cm2) and Pt-Ru
12 Rongzhong Jiang
black (2 mg/cm2) were used as catalysts for preparing electrodes. The BET surface areas of Pt
and PtRu were 27 and 60 m2/g, respectively. The MEA was made by a hot-press at 125 oC to
combine the cathode, anode, and the Nafion 117 membrane together. E-Tek carbon cloths
were then attached to both the cathode and the anode by a secondary hot press at 125 oC. The
gas diffusion layer for the cathode had higher Teflon content than that for the anode. A single
cell was fabricated by simply placing a freshly made MEA into a commercially available
single cell test device (Fuel Cell Technology) that consisted of two graphite plates. On the
inner sides of the graphite plates there were two sets of micro channels for gas and fuel flows,
respectively. On the outer sides of the graphite plates, there were two copper plates for
inserting two heaters and a thermocouple, collecting current and voltage from the fuel cell,
and sending current and voltage to a fuel cell test instrument through electric leads. The fuel
was automatically preheated by passing through a copper plate before entering the fuel flow
channel. The active area of the DMFC single cell was 5 cm2. The methanol flow rate to the
anode was controlled with a HPLC pump; and was kept at 2.5 ml/min. The cathode flow rate
was controlled by flow meters at 600 ml/min air or 200 ml/min O2, respectively. Methanol
crossover was determined using a gravimetric method [63-64] with Ba(OH)2 to precipitate
CO2 that was generated in the process of operating the DMFC single cell. The CO2 from the
anode and the cathode outlets was separately trapped with CO2 getters that contained clear but
saturated Ba(OH)2 solution to insure precipitating the CO2 completely. The BaCO3 precipitate
was separated from the liquid by a centrifuge, washed with de-ionized water, and then dried
at 70 oC for 20 hours. After cooling on dry silicon gel in a desiccator, it was weighed. The
reliability of the method for the quantitative analysis of CO2 was examined by using a CO2
standard. The getters for CO2 collection were calibrated by a method of a standard addition of
CO2 into the outlets of the single cell. The methanol crossover rate can be calculated with the
previous method [63-64]. Energy efficiencies of the DMFC systems and stacks at various
operating conditions were obtained by simulation, in order to compare with the experimental
results of the DMFC single cells by a method of data normalization. The simulated DMFC
system, of using air as oxidant and 1.0M methanol as fuel operating at 60 oC, has a power
output of 20W for long-term operation, with a peak power of 35W by a few minutes of power
pulse. The DMFC system is supposed to mix pure methanol from a fuel reservoir with the
water produced in the cathode of the fuel cell stack. The methanol concentration in the
internal DMFC stack is automatically controlled to about 1.0M by a methanol sensor in the
DMFC system. Even if the DMFC system is operated at different environmental
temperatures, the internal temperature of the DMFC stack must be maintained at a constant
status. A rechargeable battery pack must be installed in the system for starting the fuel cell,
and maintaining operation during the idle time.
method. The objective of this article is to understand the variations of fuel energy efficiencies
with operating conditions, and with the external power and current loading conditions among
DMFC single cell, stack and system, in order to provide a useful reference for DMFC
designers and manufacturers. The quality of MEAs used to build a DMFC system plays a
primary role in the determination of the energy efficiency of the system. The proposed
method is based on experimental single cell data in order to ensure the accuracy and
reliability of the calculated performance data of the DMFC stacks and systems, which is very
different from those of common modeling and simulation, with which some significant
deviations may result due to some inaccurate parameters being chosen or there are lack of
reliable parameters. The following describes some basic concepts of DMFC single cell, stack
and system in the present study.
DMFC is a type of PEM fuel cell that directly uses liquid methanol as fuel and air or oxygen
as oxidant. Figure 1 shows the structure of a common PEM fuel cell, which consists of
multiple layers, including a PEM, a cathode, an anode, a fuel diffusion layer, a gas diffusion
layer, and two end-plates containing fuel and air flow channels. In the present study, the fuel
is supplied from the anode flow channel through the fuel diffusion layer into the anode
catalyst layer, where it is electrochemically oxidized at the electrode/electrolyte interface. The
air or oxygen is supplied from the cathode flow channel through the gas diffusion electrode to
the cathode catalyst layer, where it is reduced at the interface of the electrode/electrolyte. The
protons generated at the anode migrate toward the cathode through the PEM layer. Water is
produced at the cathode by combining the protons and the oxygen ions. Figure 2 shows the
electrode reactions and reactant transports at the electrode/electrolyte interfaces and within
the PEM layer in a DMFC. The electrode reactions are given by
3
At the cathode O 2 + 6 H + + 6e 3 H 2 O Eo = 1.23V (2)
2
The theoretical cell potential of a DMFC single cell is 1.21V. However, the actual cell
voltage is only about 0.7V at open circuit, and 0.4V or less at operating conditions due to a
large over potential loss by slow kinetic processes at both the anode and the cathode. These
overpotentials cause about two thirds of the energy efficiency to be lost by operating at 0.4V
for a DMFC single cell. Furthermore, there are additional losses of energy efficiency caused
by mass transfer of fuel and air, as well as fuel crossover from the anode through the PEM
membrane to the cathode, where mixed reactions occur for oxygen reduction and methanol
oxidation. The problem of fuel crossover further lowers the potential at the cathode. As
shown in Figure 2, the electrochemical reactions and reactant transports are complex. Here,
not only protons, but also the reactants and products are transported through the PEM layer in
the DMFC. For example, at the anode, the un-reacted methanol crosses over; the produced
CO2 permeates; the produced proton migrates; and water transports by electro-osmotic drag
14 Rongzhong Jiang
[47] of protons through the electrolyte membrane toward the cathode. At the cathode, the
methanol, which comes from the anode by crossover, is oxidized catalytically by directly
contacting with the oxygen; and some of the produced CO2 at the cathode permeates in the
reverse direction back to the anode if the concentration of CO2 at the cathode is higher than
that at the anode. The produced water at the cathode is transported in two ways, the majority
flows into the cathode exhaust and the remainder permeates in reverse direction through the
PEM layer toward the anode.
- V+
Fuel In Air In
4
3 2 1 5 6 7
Figure 1. Schematic view of the structure of a polymer electrolyte membrane fuel cell. (1) Polymer
electrolyte membrane; (2) anode catalyst; (3) fuel diffusion layer; (4) anode end-plate with flow
channel; (5) cathode catalyst; (6) air diffusion layer; (7) cathode end-plate with flow channel.
E 0 = E r + b 0 log i 0 (4)
In equation (3), Ei (V) is the experimentally measured potential at current i; Eact (V) is
the activation overpotential; ER (V) is Ohmic potential loss at current i; Emass (V) is
overpotential caused by mass transfer, Ecross is an additional potential loss at the cathode
because of methanol crossover. In equation (4), Er (V) is the reversible potential for the
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 15
cell; and i0 (A) and b0 (mV/dec) are the exchange current and the Tafel slope at i0,
respectively.
e-
H+
Methanol Air
MeOH
H2O CO2
CH3OH + H2O + -
H2O 3/2 O2 + 6H + 6e
O2
Figure 2. Schematic view of electrode reactions at the electrode/electrolyte interfaces and reactants
transport through a polymer electrolyte membrane in a direct methanol fuel cell.
Because the overpotential caused by mass transfer occurs mainly at very high
currents, it can be avoided in an actual DMFC system by setting up an appropriately
high-current limit and adjusting the flow rates of fuel and air. The equation (3) can be
rewritten by neglecting mass transfer over potential and combining E0 and Ecross to a new
constant of Eopen,
Rcell icell
Ecell = Eopen bLog (icell ) (5)
1000
Here, Ecell (V) and icell (mA/cm2) are the experimentally measured cell voltage and
current; Eopen (V) is the open circuit voltage of the fuel cell; b (V/dec) is Tafel slope at
icell; and Rcell () is Ohmic resistance of the fuel cell.
16 Rongzhong Jiang
Where, ieo (mA/cm2) is the equivalent current of methanol crossover rate at open circuit
voltage; and B is a constant.
With equations (5) and (6), we can calculate the Faradic efficiency (cell ), energy
efficiency (cell), and energy density (cell, Wh/Kg) of a DMFC single cell from the
experimental voltage-current curves.
icell
cell =
icell + i e
(7)
icell E cell
cell =
1.21 (icell + i e ) (8)
Here, the number 1.21 is a constant of the theoretical voltage of a DMFC single cell;
MMeOH is the molecular weight of methanol; and F is the Faradic constant. The energy
efficiency, cell, in eq. (8) is the product of voltage efficiency, Ecell/1.21, and Faradic
efficiency, cell.
In the present study, the energy density of a DMFC single cell is defined as the
experimental energy density of fuel neglecting the energies provided by the external
equipment to pump air and fuel, and to maintain the cells temperature and humidity.
2. DMFC Stack
A DMFC stack consists of a number of DMFC single cells connected in series. Bipolar plates
are widely used to connect the single fuel cells and to provide channels for gas and fuel flows.
A bipolar plate must be electrically conductive, chemically and electrochemically inert, and
no micro mechanic holes to avoid liquid and gas leaking from one cell to the other through
the bipolar plate. A bipolar plate separates two adjacent single cells, which is the anode
current collector for one single fuel cell and the cathode current collector for the other. Figure
3 shows how a DMFC stack is assembled from single cells with bipolar plates. In a DMFC
stack, there is an additional Ohmic resistance due to the material of the bipolar plates and the
imperfect connections between the interfaces. An additional mass transfer is also possible in a
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 17
fuel cell stack. The mass transfer must be avoided by using a short stack, good air and fuel
flow channels, and setting up an appropriate current limit to avoid running the DMFC at high
current. A PEM fuel cell stack operated under poor mass transfer conditions may damage the
electrodes due to fuel or air starvation that could cause the catalyst to be oxidized [68] at the
electrodes. The voltage of a DMFC stack (Estack) can be expressed as
Fuel In Air In
Single
Cell
Air Out Fuel Out
2-Cell
Stack
28-Cell
Stack
Figure 3. Schematic view of how a fuel cell stack is built from single cells.
Here, n is the number of cells in the DMFC stack; Rstack () is the resistance caused by
bipolar plates and the connections between the cells; and istack are the stack current. Because
these single cells are connected in series in a fuel cell stack, the istack must be equal to the icell.
The value of Estack decreases with increasing Rstack. The maximum stack current, or the
maximum cell current in the stack, is lower than that of the maximum cell current of an
independent single cell (i.e, not in a stack), due to the additional voltage loss by the term of
Rstackistack in eq. (10).
The equations of Faradic efficiency, energy efficiency and energy density for a DMFC
stack are similar to that of the DMFC single cell.
18 Rongzhong Jiang
istack
stack = (11)
istack + ie stack
istack E stack
stack =
1.21n (istack + i e stack )
(12)
Where, ie-stack is the equivalent current of methanol crossover of the DMFC stack. The
energy density of a DMFC stack is defined as the experimental energy density of the fuel that
has included the energy loss caused by the Rstack, but has neglected the energies provided by
the external equipment to pump air and fuel, and to maintain the stack temperature and
humidity.
3. DMFC System
A DMFC system contains one or more DMFC stacks, and the other auxiliary components of
air compressors (or blowers), fuel pumps, sensors, fuel and water containers, heat exchanger,
sensors, electronic controllers and a rechargeable battery pack. Figure 4 shows a schematic
view of possible components and their relationships in a DMFC system. The thick arrow lines
indicate how the fuel and air flow from one component to the others in the fuel cell system.
The air is obtained from an air compressor, and goes through an air-flow-controller, and an
air humidifier to the fuel cell stack. The fuel is pumped from a fuel reservoir into a fuel mixer,
where it is mixed with water to obtain 0.5 to 1.0M methanol, and then it is pumped again
through a fuel-flow-controller into the fuel cell stack. In the fuel cell stack, the fuel and air are
consumed to produce electricity and heat. The exhaust of the fuel and water at the anode
outlet are sent back into the fuel mixer, where the CO2 is removed to ambient. The exhaust of
air and water at the cathode outlet are separated; and the water is recycled to the fuel mixer.
Overall, a DMFC system can use pure methanol as fuel by recycling the cathode water. A
methanol sensor is used to assist the central controller to maintain the methanol concentration
between 0.5 and 1.0M. Several thermometers are used to monitor and control the operating
temperatures. Fans and heat exchangers are used to dissipate the excess heat. A rechargeable
battery pack is used to start the DMFC system and to store the electric energy obtained from
the fuel cell stack. The central electronic controller controls the temperatures, humidity, stack
current, stack voltage, battery current, battery voltage, and digital data acquisition. The thin
solid lines in Figure 4 shows the relationships between the electronic controller and the other
components in the DMFC system.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 19
Fuel Flow
Thermometers
Controller
Fuel Pump 2
Methanol Heat Electric
Sensor Exchanger Fans
Fuel Mixer
Electronic Air
Controller Flow Controller
Fuel
Reservoir
Rechargeable Battery Air
Compressor
Figure 4. Schematic view of the components and their relationships in a direct methanol fuel cell
system. The thick arrow lines indicate how the air and fuel flow from one part to others; and the thin
solid lines express the relationships between two components connected by a line.
+ i i1
Fuel Cell Stack i2 Load
- DCDC converter
Figure 5. Power generation and power distribution in a DMFC system. The current i from the DMFC
stack is divided into two: i1 to the external load, and i2 to the auxiliary components.
20 Rongzhong Jiang
The auxiliary components in the DMFC system will consume a part of the electric energy
generated by the fuel cell stack. The percentage of power usage varies from system to system
with different DMFC system design. Generally, the internal auxiliary components consume
about 15 25% power generated by the DMFC stack. Figure 5 shows a schematic view of
power generation and distribution in a general DMFC system, where all the auxiliary
components except the rechargeable battery are expressed by using only one text-box. A
DMFC system must use a battery pack to provide the initial power to pump fuel and air to the
fuel cell stack for its startup. After startup there are three possible operating conditions: (1)
the load is off and fuel cell stack charges the battery; (2) the load is on and the fuel cell stack
charges the battery; (3) the load is on and the fuel cell stack does not charge the battery. In the
condition (2), the rechargeable battery becomes a secondary internal load, and the actual
power output becomes smaller than that of common operating conditions. Therefore, we are
only interested in condition (3) (the load is on and the fuel cell stack does not charge the
battery) in the present study. In some actual DMFC systems, a DCDC converter is connected
between the fuel cell stack and the battery, and also between the fuel cell stack and the
external load. The input voltage of the DCDC converter varies in a certain range, but the
output voltage of a DCDC converter is constant. A DCDC converter may consume about 5%
of the input energy from the fuel cell stack. For convenience of calculation, the 5% energy
consumption by the DCDC converter is added into the internal power users in the fuel cell
system for a total power consumption of about 20%. With the addition of a DCDC converter
between the fuel cell stack and the load, the voltage-current curves will be very different from
that without a DCDC converter, however, the power-current curve will not be affected.
In the present study, the voltage-current curves of a DMFC system are calculated by
assuming the absence of a DCDC converter between the fuel cell stack and the load. We use
the experimental data of DMFC single cells to simulate the power-current curves of a 20W
DMFC system based on a long-term operation. Here, we assume that the DMFC system
contains a DMFC stack with 28 single cells, and each cell in the stack has 22 cm2 electrode
area. In order to avoid poor mass transfer, a low-voltage limit is set in the DMFC system to
cut off the stack voltage at 8.0V, which corresponds to about 0.3V for each of the single cells.
As shown in Figure 5, if there were zero energy consumption by the auxiliary components,
the system voltage output would be equal to the voltage of the fuel cell stack due to the
parallel connections. However, the current output of the DMFC system must be smaller than
that of the fuel cell stack due to parasitic losses to auxiliary components. Consequently, the
actual system voltage output is smaller than that of the DMFC stack by standing alone.
Assuming a consumption of 20% of the specific power, there is 4W power consumed by the
auxiliary components for a 20W DMFC system operated at normal conditions (i.e., 60 oC,
1.0M methanol, and 1 atmosphere pressure with constant power output), and 0.05cm-2cell-
1
resistance is generated by the graphite based bipolar plates in the fuel cell stack. The
resistance caused by the bipolar plates is given by
n
Rstack = (14)
A
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 21
Here, (cm-2cell-1) is the resistivity of bipolar plates; and A (cm2) is the electrode area
of a single cell in the stack. We assume that all the single cells have the same electrode area
and linked in series in the fuel cell stack.
The output current of DMFC system is given by
Pint ernal
isystem = istack = istack iint ernal (15)
Estack
Here, isystem and iinteral are the output current to the external load by the DMFC system and
the total current consumed by the auxiliary components, respectively. Pinternal (W) is the power
consumed by the auxiliary components.
The equations of energy efficiency and energy density of a DMFC system are similar to
that of a DMFC stack
isystem E system
system =
1.21n (isystem + iint ernal + ie stack ) (16)
Here, the energy density of a DMFC system is defined as the experimental energy density
of fuel that has included the energy loss caused by Rstack and the energy loss by the auxiliary
components, which is the net energy density generated by the DMFC system, and consumed
by the external load.
Figure 6 shows experimental and simulated voltage-current curves of a DMFC single cell at
60 oC at various methanol concentrations. The simulated results (solid lines) fit the
experimental points very well within the experimental current ranges. The voltage-current
curves for higher current ranges are not calculated in order to avoid the deviations caused by
poor mass transfer. Actually such high current range or low cell voltage below 0.3V is cut off
in an actual DMFC system in order to avoid the catalyst layer oxidation on the electrodes
caused by fuel or air starvation [68]. The highest cell performance is obtained for methanol
concentration as 1.0M. At low current range the voltage-current curve for 0.5M methanol is
similar to that for 1.0M methanol. With further increasing of the cell current the voltage-
22 Rongzhong Jiang
current curve for 0.5M methanol is lowered because the discharge current is limited by the
methanol mass transfer. Increasing methanol concentration to 2.0 and 3.0M, causes the cell
performance to decrease significantly due to a higher rate of methanol crossover. Figure 7
shows power-current curves under the same operating conditions as shown in Figure 6. The
peak power for 1.0M methanol is 73 mW/cm2 at 60 oC. Table 1 lists the electrochemical
parameters used for calculation of the data in Figure 6, and the peak powers of the DMFC
single cell under different methanol concentrations. The open circuit voltage decreases with
increasing methanol concentration. The Tafel slope, b value, also decreases with increasing
methanol concentration except for the condition of 3.0M methanol. The cell resistance, Rcell,
has no apparent change due to the operating temperature is constant at 60 oC. The peak power
decreases with increasing methanol concentration if it is higher than 1.0M.
0.7
0.5M MeOH
0.6
1.0M MeOH
0.5 2.0M MeOH
Voltage (V)
3.0M MeOH
0.4
0.3
0.2
0.1
0
0 50 100 150 200 250 300 350 400
2
Current (mA/cm )
Figure 6. Voltage-current curve of a DMFC single cell at 60 oC using air as oxidant and methanol as
fuel with various concentrations. The points and lines are experimental and calculated results,
respectively.
80
70
60
Power (mW)
50
40
30 0.5M MeOH
1.0M MeOH
20
2.0M MeOH
10 3.0M MeOH
0
0 50 100 150 200 250 300 350 400
2
Current (mA/cm )
Figure 7. Power-current curve of a DMFC single cell at 60 oC using air as oxidant and methanol as fuel
with various concentrations. The points and lines are experimental and calculated results, respectively.
500
3.0M MeOH
Equivalent Current of Methanol
2.0M MeOH
400
Crossover (mA/cm 2 )
1.0M MeOH
0.5M MeOH
300
200
100
0
0 50 100 150 200 250 300 350 400
2
Cell Current (mA/cm )
Figure 8. Plots of equivalent current of methanol crossover versus cell discharge current in a DMFC at
60 oC using different methanol concentrations. The points and the lines are experimental and calculated
results, respectively.
Figure 8 shows experimental and simulated results of the equivalent current of methanol
crossover versus discharge cell current. The simulated results (solid lines) fit the experimental
points well. At open circuit voltages there are the highest rates of methanol crossover. With
increasing discharge current, the rate of methanol crossover becomes smaller. The methanol
concentration plays a significant role in methanol crossover. With increasing methanol
concentration the rate of methanol crossover becomes remarkably higher. The methanol
crossover rate at 3.0M methanol is almost 7 times higher than that at 0.5M methanol. The
electrochemical parameters used for calculation of the data in Figure 8 are listed in Table 2.
24 Rongzhong Jiang
0.7 1.2
Cell Voltage
Fa ra dic Eff
A
0.6 1.0
Egergy Eff
0.4
0.6
0.3
0.4
0.2
0.1 0.2
0.0 0.0
0 50 100 150 200
2
Cell Current (mA/cm )
0.7 1.2
Cell Voltage
0.6 Faradic Eff
B
1.0
Faradic Eff and Energy Eff
Egergy Eff
0.5
0.8
Cell Voltage (V)
0.4
0.6
0.3
0.4
0.2
0.1 0.2
0.0 0.0
0 50 100 150 200 250 300
2
Cell Current (mA/cm )
0.7 1.0
Cell Voltage
0.6 Faradic Eff
C
Egergy Eff
0.8
0.3
0.4
0.2
0.2
0.1
0.0 0.0
0 50 100 150 200 250 300 350
Cell Current (mA/cm2)
0.6 1.0
Cell Voltage
0.4
0.6
0.3
0.4
0.2
0.2
0.1
0.0 0.0
0 100 200 300 400
Cell current (mA/cm2)
Figure 9. Plots of cell voltage, Faradic efficiency and energy efficiency versus discharge current for a
DMFC single cell at 60 oC using air as oxidant. The methanol concentrations are: A, 0.5M; B, 1.0M; C,
2.0M; and D, 3.0M, respectively.
The Faradic efficiency, energy efficiency and energy density can be calculated using
equations (7), (8) and (9). Figures 9A through 9D show the plots of cell voltage, Faradic
efficiency and energy efficiency versus cell current, respectively. With increasing discharge
current, the cell voltage decreases but both the Faradic and energy efficiencies increase. A
peak value of energy efficiency is obtained at a relatively low cell voltage. The peak energy
efficiency and energy density are also listed in Table 2. Both of them decrease substantially
with increasing methanol concentration from 0.5 to 3.0M. At 0.4V the DMFC obtains 96% of
its peak energy efficiency (0.272, the 100% of its peak energy efficiency is 0.283) for 0.5M
methanol at current 104 mA/cm2; 88% of its peak energy efficiency (0.214, the 100% of its
peak energy efficiency is 0.243) for 1.0M methanol at current 108 mA/cm2; 56% of its peak
energy efficiency (0.079, the 100% of its peak energy efficiency is 0.142) for 2.0M methanol
26 Rongzhong Jiang
at current 68 mA/cm2; and 25% of its peak energy efficiency (0.016, the 100% of its peak
energy efficiency is 0.063) for 3.0M methanol at current 20 mA/cm2, respectively. In order to
obtain 90% peak energy efficiency, the DMFC must be discharged at 0.42V for 0.5M
methanol, at 0.39V for 1.0M methanol; at 0.32V for 2.0M methanol; and at 0.25V for 3.0M
methanol, respectively. Figure 10 shows plots of energy density versus cell current for a
DMFC single cell at 60 oC using air as oxidant with different methanol concentrations. The
energy density increases with increasing discharge current, and gradually reaches a peak
value at a relatively high current. For 0.5M methanol the DMFC has the highest peak energy
density of 1726 Wh/Kg. For 3.0M methanol the peak energy density is decreased to only 374
Wh/Kg.
1800
1600
Energy Density (Wh/Kg)
1400
Energy Density
(0.5M MeOH))
1200
Energy Density
1000 (1.0M MeOH)
200
0
0 50 100 150 200 250 300
2
Cell Current (mA/cm )
Figure 10. Effect of methanol concentration on fuel energy density for a DMFC single cell at 60 oC
using air as oxidant and methanol as fuel with various concentrations.
There is an additional energy loss related to the Rstack in a DMFC stack, which is caused by
the materials of the bipolar plates and the connections between two bipolar plates. According
to equations (5), (10), (14) and the electrochemical parameters in Table 1, the voltage-current
and power current curves are calculated for a DMFC stack at 60 oC using air as the oxidant
with various methanol concentrations. Figure 11 shows voltage-current curves of a 28-cell
DMFC stack with 22 cm2 electrode area using different methanol concentrations. These
voltage-current curves are limited at 8.0V except for a methanol concentration of 0.5M,
where the voltage-current curve decreases rapidly at high current density. Figure 12 shows the
corresponding power-current curves for the data shown in Figure 11. The peak powers are
obtained at the points of the voltage limit (8.0V, corresponding to about 0.3V for a single
DMFC cell). Continuing to decrease the voltages to lower than 8.0V may increase the peak
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 27
powers; but it is unsafe for the fuel cell stack because of some side effects, such as anode
catalyst oxidation and ruthenium crossover [67]. With increasing methanol concentration
from 0.5M to 3.0M, the open circuit voltage of the DMFC stack decreases from 18.4 to
15.7V. The DMFC stack has the highest discharge peak power (40W) for methanol
concentration at 1.0M. Increasing or decreasing methanol concentration from 1.0M will
decrease the discharge peak power. The discharge peak powers of the DMFC stack and open
circuit voltages under different methanol concentrations are listed in Table 3. The resistances
of the PEM, electrode materials and bipolar plates, as well as the oxidation of the crossed over
methanol at the cathode will contribute to the production of heat in the DMFC stack. At low
discharge current density the energy loss is mainly caused by methanol crossover. At high
discharge current density and low methanol concentration, the energy loss is mainly caused
by the resistances. However, at high discharge current and high methanol concentration the
energy loss is caused by both the methanol crossover and the resistances. Figure 13 shows
plots of energy efficiency and energy density versus stack current. Here, each of the curves
has two Y-coordinates to express the energy density and the energy efficiency
simultaneously. With increasing discharge current, both the energy efficiency and energy
density increase until reaching the peak values. The peak energy efficiency and peak energy
density are also listed in Table 3. The DMFC stack has the highest energy efficiency and
energy density for methanol concentration at 0.5M. With increasing methanol concentration
both the energy efficiency and energy density decrease significantly.
20
16
Volatge (V)
12
4
Stack Vo ltag e (0.5M M eOH) Stack Voltage (1.0M M eOH)
Stack Vo ltag e (2.0MM eOH) Stack Voltage (3.0M M eOH)
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Stack Current (A)
Figure 11. Voltage-current curves of a 28-cell DMFC stack with 22 cm2 electrode area at 60 oC using
air as oxidant and methanol as fuel with various concentrations. The results are obtained by simulation
from DMFC single cell data.
28 Rongzhong Jiang
50
40
Stack Power (W)
30
20
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Stack Current (A)
Figure 12. Power-current curves of a 28-cell DMFC stack with 22 cm2 electrode area at 60 oC using air
as oxidant and methanol as fuel with various concentrations. The results are obtained by simulation
from DMFC single cell data.
0.28
Energy Density (Wh/Kg)
1600
0.24
Energy Efficiency
1200 0.20
0.16
800
0.12
0.08
400
0.04
0 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Stack Current (A)
Figure 13. Plots of energy efficiency and energy density versus stack discharge current for a 28-cell
DMFC stack with 22 cm2 electrode area at 60 oC using air as oxidant and methanol as fuel with various
concentrations. The results are obtained by simulation from DMFC single cell data.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 29
Table 3. Peak stack power, energy efficiency, energy density and open circuit voltages of
a 28-cell DMFC stack with 22 cm2 electrode area operated at 60 oC using air as oxidant
and methanol as fuel with different concentrations.
50
Power, DMFC Stack
40
A
30
Power (W)
10
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Output Current (A)
0.6
0.4
0.3
0.2
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Figure 14. (A) Comparison of power outputs between DMFC stack and system. (B) Plot of the
internally consumed current versus output current. The simulated DMFC system contains a 28-cell
stack with 22 cm2 electrode area at 60 oC using air as oxidant and 1.0M methanol as fuel.
methanol. There is a 4W power loss from the DMFC stack to DMFC system because part of
the current is bypassed into the system auxiliary components. The peak powers for the DMFC
stack and system are 40W and 36W; and the peak currents from the stack to the system are
5.1A and 4.6A, respectively. Figure 14B shows plot of the internally consumed current versus
output current of the DMFC system. There is a loss of 0.5A current output from the DMFC
stack to the system. With increasing external output current, the internal current increases
noticeably, which is attributed to the decrease of stack voltage in order to maintain about a
constant value of the internal power consumption. The simulated power-current curve of the
DMFC system shown in 14A is compared with that of an actual 20W DMFC system [69],
which had 20W for continuous operation, and peak power 36W for a few minutes of power
pulse. Satisfactory results are obtained either for long-term operation (20W) or for peak
power (36W), which let us believe that the simulated results of the DMFC system are reliable
and accurate.
Table 4. Peak system power, energy efficiency, energy density and open circuit voltages
of a DMFC system that contains a 28-cell DMFC stack with 22 cm2 electrode area
operated at 60 oC using air as oxidant and methanol as fuel with different
concentrations.
18
16
14
12
Voltage (V)
10
0
0.0 1.0 2.0 3.0 4.0 5.0
Output Current (A)
Figure 15. Voltage-current curves of a DMFC system that contains a 28-cell stack with 22 cm2
electrode area at 60 oC using air as oxidant and methanol as fuel with various concentrations. The
results are obtained by simulation from DMFC single cell data.
Since we are interested in not only the normal operating conditions of the common
DMFC systems that use 0.5 to 1.0M methanol as fuel and air as oxidant operated at 60 oC, but
also other operating conditions such as varied methanol concentration, cell temperature and
oxygen partial pressure, the discharge performances under other operating conditions for the
DMFC system are also calculated. Figure 15 shows voltage-current curves of a simulated
DMFC system operated at 60 oC with different methanol concentrations. The open circuit
voltage decreases with increasing methanol concentration. For methanol concentrations of
0.5M, 1.0, 2.0 and 3.0M the open circuit voltage are 15.3, 14.8, 13.5 and 11.5V, respectively.
The voltage-current curve for 0.5M methanol terminates at 1.3A, which is much lower than
that of the others (3.6 to 4.6A) due to a mass transfer limitation under low methanol
concentration. On the other hand, the peak discharge current (limited by the cell voltage or
poor mass transfer of fuel or air) of the DMFC system for all operating conditions is
apparently lower than that of the DMFC stack under the same operating conditions, which is
attributed to the parasitic current losses of the auxiliary components. For 0.5, 1.0, 2.0 and
3.0M methanol the parasitic currents are 0.45, 0.50, 0.53 and 0.51A, respectively, which are
obtained by subtracting the peak system output currents from the peak stack currents. Because
the auxiliary components consume a constant power and the stack voltage decreases with
increasing methanol concentration, the parasitic current will increase with increasing
methanol concentration. Figure 16 shows plots of output power versus output current for the
DMFC system at 60 oC with various methanol concentrations. The system has the highest
peak output power (36W) for 1.0M methanol. According to equations (16) and (17), the
system energy efficiency and energy density can be calculated. Figure 17 shows plots of
system energy efficiency and energy density versus output current. With increasing output
current the system energy efficiency increases until reaching a peak value. With increasing
32 Rongzhong Jiang
methanol concentration the system energy efficiency decreases significantly. The peak power,
peak energy efficiency and peak energy density of the DMFC system are listed in Table 4.
40
35
30
Power (W)
25
20
0
0.0 1.0 2.0 3.0 4.0 5.0
Figure 16. Power-current curves of a DMFC system that contains a 28-cell stack with 22 cm2 electrode
area at 60 oC using air as oxidant and methanol as fuel with various concentrations. The results are
obtained by simulation from DMFC single cell data.
1200 0.2 0
Energy Efficiency
0.1 6
800
0.1 2
0.0 8
400
0.0 4
0 0.0 0
0.0 1.0 2.0 3.0 4.0 5.0
Output Current (A)
Figure 17. Fuel energy efficiency and energy density of a DMFC system that contains a 28-cell stack
with 22 cm2 electrode area at 60 oC using air as oxidant and methanol as fuel with various
concentrations. The results are obtained by simulation from DMFC single cell data.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 33
One of the most important objectives of this study is to understand how one can achieve the
highest power density and energy efficiency by a DMFC system design and choosing
appropriate operating conditions. The simulated results under constant current, voltage and
power are used to determine how the energy efficiency changes with operating conditions.
Figure 18A through 18C show plots of energy density versus output current, power and
voltage of the simulated DMFC system that contains a 28-cell stack with 22 cm2 electrode
area operated at 60 oC with 1.0M methanol, respectively. The system energy density increases
with increasing output current (Figure 18A) or output power (Figure 18 B) until reaching a
peak value; but decreases with increasing output voltage (Figure 18C). The decrease in
energy density with decreasing power or current output for the DMFC system is attributed to
the increased methanol crossover that can be observed in an individual cell (see Figures 9 and
10). A DMFC system can achieve the highest energy density and efficiency as long as
operated at an appropriately high level of power or current conditions [60]. About 1280
Wh/Kg energy density is obtained at the plateau of the energy-power or energy-current
curves.
1600
1400
A
Energy Density (Wh/Kg)
1200
1000
800
600
400
200
0
0.0 1.0 2.0 3.0 4.0 5.0
Output Current (A)
1400
1200
B
Energy Density (Wh/Kg)
1000
800
600
400
200
0
0 10 20 30 40
Output Power (W)
1400
1000
800
600
400
9.0 9.5 10.0 10.5 11.0 11.5 12.0 12.5 13.0
Figure 18. Plots of fuel energy density versus output current (A), power (B) and voltage (C) for a
simulated DMFC system, respectively, which contains a 28-cell stack with 22 cm2 electrode area at 60
o
C using air as oxidant and 1.0M methanol as fuel.
1.5. Comparison of Power and Energy Densities among DMFC Single Cell, Stack
and System
It is difficult to compare the discharge performances among single cell, fuel cell stack and
fuel cell system, unless the electrochemical parameters are normalized, such as a unit power
of mWcm-2cell-1. Although the same single fuel cell standing alone, assembled in a stack and
operated in a fuel cell system should have the same performance under the same operating
conditions, the actually obtained apparent current, voltage and power outputs may be different
because an additional stack resistance is produced and additional power is consumed by the
auxiliary equipment in the fuel cell system. Figure 19 shows normalized voltage-current
curves of a DMFC single cell, stack and system. Here, the DMFC system contains a 28-cell
stack with 22 cm2 electrode area operating at 60 oC with 1.0M methanol. Comparing the
DMFC stack with the single cell at the same current, the stack voltage becomes lower than
that of the single cell with increasing discharge current due to the presence of an additional
stack resistance. Comparing the DMFC system with the stack at the same voltage, the output
current of the DMFC system becomes smaller because a part of current in the DMFC system
is lost to the auxiliary components. Comparing the DMFC system with the single cell, both
the voltage and current outputs for the DMFC system are decreased. Figure 20 shows plots of
normalized power output versus normalized current for a DMFC single cell, stack and
system. In the top three curves, i.e., cell power versus cell current, stack power versus stack
current, and system power versus system current, the power output becomes higher with
increasing discharge current; but the deviations among them become more apparent.
Comparing the single cell with the system (the top points with the bottom curve), the power
loss of the DMFC system varies from 6 to 13 mWcm-2cell-1. If the current is the higher,
more power will be lost. Table 5 lists the normalized peak power outputs among DMFC
single cell, stack and system operated at 60 oC using air as the oxidant and methanol as fuel at
different concentrations. The peak power output decreases with increasing methanol
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 35
concentration if it is higher than 1.0M for single cell, stack or DMFC system, respectively.
Table 6 summarizes the peak energy efficiency for single cell, stack and DMFC system. The
operating condition at 0.5M methanol exhibits the highest energy efficiency 0.284 for the
single cell. Increasing the methanol concentration decreases the energy efficiency. At 3.0M
methanol, the peak energy efficiency is only 0.058. From single cell to DMFC system the
peak energy efficiency is decreased by 14% to 40% depending on the methanol concentration.
For example, at 0.5M methanol, if the single cells peak energy efficiency (0.284) is
considered as 100%, the DMFC systems peak energy efficiency (0.243) is decreased by
about 14%. At 3.0M methanol, the energy efficiency from single cell to system is decreased
by 40% (from 0.058 to 0.035). Table 7 lists the peak energy density for single cell, stack and
DMFC system. At 0.5M methanol, the peak energy density is 1726 Wh/Kg for single cell, but
only 1438 Wh/Kg for DMFC system. At 3.0M methanol, the DMFC system has an energy
density of only 204 Wh/Kg.
0.60
Normalized Voltage (V/Cell)
0.55
0.50
0.45
DMFC Stack
0.40
DMFC Single Cell
0.35
DMFC
0.30 System
0.25
0 50 100 150 200 250 300
2
Normalized Output Current (mA/cm )
Figure 19. Comparison of voltage-current curves among DMFC single cell, stack and system. The
simulated DMFC system contains a 28-cell stack with 22 cm2 electrode area using air as oxidant and
1.0M methanol as fuel at 60 oC. The points are obtained from a DMFC single cell result under the same
operating conditions. The cell voltage is limited at 0.28V.
Table 5. Comparison of peak power outputs among DMFC single cell, stack and system
operated at 60 oC using air as oxidant and methanol as fuel with different
concentrations.
80
2
Sys Output Power vs.
60
Sys Output Current
50
Cell Power vs.
Cell Current
40
30
20
Sys Output Power vs.
Cell Current
10
0
0 50 100 150 200 250 300
2
Normalized Output Current (mA/cm )
Figure 20. Comparison of power-current curves among DMFC single cell, stack and system. The
simulated DMFC system contains a 28-cell stack with 22 cm2 electrode area using 1.0M methanol at 60
o
C. The points are obtained from a DMFC single cell result. The cell voltage was limited at 0.28V.
Table 6. Comparison of peak energy efficiencies among DMFC single cell, stack and
system operated at 60 oC using air as oxidant and methanol as fuel with different
concentrations.
Table 7. Comparison of peak energy densities among DMFC single cell, stack and
system operated at 60 oC using air as oxidant and methanol as fuel with different
concentrations.
Generally, there is a serious energy efficiency loss for DMFC single cell, stack and
system by increasing methanol concentration from 0.5 to 3.0M due to methanol crossover.
The optimal methanol concentration is between 0.5 and 1.0M. From single cell to DMFC
system there is an additional energy efficiency loss (e.g., a loss of 288 Wh/Kg for 0.5M
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 37
methanol; and a loss of 196 Wh/Kg for 1.0M methanol) due to an additional stack resistance
and the system parasitic loss for operating the fuel cell stack. There are good suggestions for a
DMFC design. For a DMFC with high stack resistance, such as using mono-polar or strip cell
stack design [70], we need to choose a short stack (with less number of cells) and large
electrode area to achieve less power and energy efficiency loss. For bipolar stack design with
low stack resistance, we need to lower the internal power consumptions in order to achieve
the lowest power and energy efficiency loss. In order to avoid an energy efficiency loss
caused by methanol crossover, we need to monitor and control the methanol concentration
with a methanol sensor and to use the recycled water from the cathode to dilute the methanol
to a value between 0.5 and 1.0M.
Figure 21 shows the effect of operating temperature on the behaviors of voltage-current curve
for a DMFC single cell using 1.0M methanol as fuel and air as oxidant. The points and lines
are experimental and simulated results, respectively. The simulated curves fit the
experimental points very well. With increasing temperature the electrode kinetic rates for
both methanol oxidation and oxygen reduction are enhanced significantly. Consequently, the
discharge performance of the DMFC single cell is much improved. From 40 to 80 oC the open
circuit voltages are only a little different (Eopen from 0.63V to 0.66V). However, with
increasing discharge current the difference in the cell voltages becomes more significant. For
0.7
0.6 V oltage at 80 oC
V oltage at 60 oC
0.5
Voltage (V)
V oltage at 40 oC
0.4
0.3
0.2
0.1
0
0 100 200 300 400 500 600
2
Current (mA/cm )
Figure 21. Voltage-current curve of a DMFC single cell using air as oxidant and 1.0M methanol as fuel
at various operating temperatures. The points and the lines are experimental and calculated results,
respectively.
38 Rongzhong Jiang
example, at 200 mA/cm2 cell current, there is 168 mV difference in cell voltage for
temperatures between 40 (at 0.213V) and 80 oC (at 0.381V). In addition to electrode kinetics,
increasing cell temperature will enhance the ionic conductance of the PEM in the DMFC.
Consequently, the discharge voltage increases, which can be seen on the voltage-current
curve in the high current range. Figure 22 shows plots of discharge power versus cell current
for the same data in Figure 21. At 40, 60 and 80 oC the peak powers are 43, 73 and 101
mWcm-2cell-1, respectively. Table 8 summarizes the electrochemical parameters for
simulation of the voltage-current curves and the peak powers obtained. With increasing
temperature, the value of Pmax-cell becomes larger, but the b value and Rcell become smaller.
120
100
Power (mW)
80
60
40
Power at 80 oC
20
Power at 60 oC
Power at 40 oC
0
0 100 200 300 400 500 600
2
Current (mA/cm )
Figure 22. Power-current curve of a DMFC single using air as oxidant and 1.0M methanol as fuel at
various operating temperatures. The points and the lines are experimental and calculated results,
respectively.
Temperature (oC)
Parameters
40 60 80
Eopen (V) 0.63 0.63 0.66
b (V/Dec) 0.095 0.080 0.075
Rcell () 0.85 0.63 0.53
Pmax-cell (mW/cm2) * 43 73 101
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 39
200
60 oC MeOH Crossover
Crossover (mA/cm )
150
2
40 oC MeOH Crossover
100
50
0
0 50 100 150 200 250 300 350 400
2
Cell Current (mA/cm )
Figure 23. Equivalent current of methanol crossover in a DMFC using 1.0M methanol at various
operating temperatures. The points and the lines are experimental and calculated results, respectively.
In order to carry out energy efficiency analysis, we need to accurately measure the
methanol crossover rate through the PEM. Figure 23 shows plots of equivalent current of
methanol crossover rate versus cell current. The points and lines are experimental and
simulated results, respectively. The simulated curves fit the experimental points very well.
The highest rate of methanol crossover is obtained at the open circuit voltage. With increasing
discharge current the methanol crossover rate becomes smaller. With increasing temperature
the methanol crossover rate increases significantly. The methanol crossover rate from 60 to
80 oC seems apparently higher than that from 40 to 60 oC. The origin of the higher methanol
crossover rate from 60 to 80 oC is unknown, which is probably relevant to the boiling point of
methanol (65 oC). The advantage of using the simulated curves, instead of the experimental
points, is because of convenience for calculation of energy efficiency with equations (5), (6)
and (8). Table 9 lists the electrochemical parameters for simulation of the equivalent current
of methanol crossover rate for a single DMFC cell using 1.0M methanol as fuel and air as
oxidant at various cell temperatures.
Temperature (oC)
Parameters
40 60 80
ieo (mA/cm2) 74 105 181
B 0.47 0.44 0.40
cell 0.218 0.243 0.217
cell (Wh/Kg) 1327 1481 1322
40 Rongzhong Jiang
0.7 1.2
Voltage at 40 oC
0.6
A
0.1 0.2
0.0 0.0
0 20 40 60 80 100 120 140 160
Cell current (mA/cm 2 )
0.7 1.2
Voltage at 80 o C
B
0.8
0.4
0.6
0.3
0.4
0.2
0.1 0.2
0.0 0.0
0 100 200 300 400
Figure 24. Plots of cell voltage, Faradic efficiency and energy efficiency versus discharge current for a
DMFC single cell using air as oxidant and 1.0M methanol as fuel operated at 40 (A) and 80 oC (B),
respectively.
Figures 24A and 24B show plots of energy efficiency, Faradic efficiency and cell voltage
versus cell current for a DMFC single cell using 1.0M methanol as fuel and air as oxidant at
40 and 80 oC cell temperatures, respectively. The relationship of energy efficiency, Faradic
efficiency and cell discharge performance can be more readily understood by putting these
data into the same figure. The Faradic efficiency is directly related to the methanol crossover.
The energy efficiency is not only dependent on the Faradic efficiency but also on the cell
discharge voltage. The optimum cell discharge voltages can be determined from these figures
in order to obtain the highest energy efficiency. A too low cell operating voltage should be
avoided because it may cause problems of poor mass transfer and anode catalyst oxidation. It
was reported [67] that some forms of ruthenium, dissociated from the PtRu catalyst on the
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 41
anode electrode at a too low cell voltage may migrate through the electrolyte membrane to the
cathode, inhibiting the catalytic reduction of oxygen by the presence of ruthenium on the
platinum catalyst. If the DMFC is operated at 0.40V, the energy efficiency is 0.160 (73% of
its peak energy efficiency) for cell temperature at 40 oC; 0.214 (88% of its peak energy
efficiency) for cell temperature at 60 oC; and 0.201 (93% of its peak energy efficiency) for
cell temperature at 80 oC, respectively. The peak cell energy efficiency and energy density are
also listed in Table 9. At 60 oC the highest peak energy efficiency is obtained (0.243) with
corresponding energy density of 1481 Wh/Kg. Increasing or decreasing cell temperature will
decrease the peak energy efficiency. However, considering power output, we may choose to
operate at 80 oC, instead of at 60 oC. From 60 to 80 oC, the peak energy efficiency is
decreased by only 10.7%, but the peak discharge power is increased by 38.4% (from 73 to
101 mW/cm2). Figure 25 shows plots of energy density versus cell current for a DMFC using
1.0M methanol as fuel and air as oxidant at different cell temperatures. With increasing cell
current density, the cell energy density increases until reaching a peak value. It is observed
that the DMFC reaches its peak energy density at a lower current density if the cell
temperature decreases. The required cell currents to obtain the peak energy densities at 40, 60
and 80 oC are 126, 188 and 256 mA/cm2, respectively. In order to obtain the same energy
density, the DMFC at 80 oC has to operate at a current density as high as twice that at 40 oC.
1600
1400
Energy Density (Wh/Kg)
1200
1000
800
600
En ergy Den sity (40 o C)
0
0 50 100 150 200 250 300
2
C ell C urren t (m A /cm )
Figure 25. Plots of energy density versus current density of a DMFC single cell using air as oxidant and
1.0M methanol as fuel at various operating temperatures.
The operating temperature may directly affect the single cell performance, stack heat
dissipation, and bipolar plate resistance. The difference in resistance of the bipolar plates
from 40 to 80 oC is not noticeable. Good heat dissipation can be achieved by appropriate
stack and system designs. Actually, the single cell performance plays a main role in the
DMFC stack performance. According to equations (5) and (10), the voltage-current and
42 Rongzhong Jiang
power-current curves can be calculated. Figures 26 and 27 show voltage-current and power-
current curves of a 28-cell DMFC stack with 22 cm2 electrode area using 1.0M methanol as
fuel and air as oxidant at various temperatures, respectively. With increasing temperature, the
open circuit voltage has no significant change. For 40, 60 and 80 oC, the value of Eopen is
almost constant at about 18V. With increasing discharge current the stack voltages at 40, 60
and 80 oC deviate. If the temperature is higher, correspondingly, the discharge power will be
higher due to a significant improvement of electrode kinetic rates for oxygen reduction and
for methanol oxidation. The stack power increases almost proportionally with increasing
operating temperature from 40 to 80 oC. The values of peak stack power (Pmax-stack) and Eopen
are summarized in Table 10. The peak stack powers at 40, 60 and 80 oC are 23, 40 and 57W,
respectively.
20
16
Volatge (V)
12
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Figure 26. Voltage-current curves of a 28-cell DMFC stack with 22 cm2 electrode area using air as
oxidant and 1.0M methanol as fuel at various operating temperatures. The results are obtained by
simulation from DMFC single cell data.
Table 10. Peak stack power, energy efficiency, energy density and open circuit voltage of
a 28-cell DMFC stack with 22 cm2 electrode area using 1.0M methanol as fuel and air as
oxidant operated at different temperatures.
Temperature (oC)
Parameters
40 60 80
Pmax-stack (W) 23 40 57
stack 0.214 0.237 0.209
stack (Wh/Kg) 1299 1438 1272
Eopen (V) 17.6 17.6 18.5
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 43
60
50
Stack Power (W)
40
30
20
Stack Power (80 oC)
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Stack Current (A)
Figure 27. Power-current curves of a 28-cell DMFC stack with 22 cm2 electrode area using air as
oxidant and 1.0M methanol as fuel at various operating temperatures. The results are obtained by
simulation from DMFC single cell data.
0.28
1600
0.24
Energy Density (Wh/Kg)
0.16
800
0.12
Figure 28. Plots of Fuel energy efficiency and energy density versus stack current of a 28-cell DMFC
stack with 22 cm2 electrode area using air as oxidant and 1.0M methanol as fuel at various operating
temperatures. The results are obtained by simulation from DMFC single cell data.
44 Rongzhong Jiang
The stack energy efficiency and energy density can be calculated according to equations
(6), (10), (12) and (13). Figure 28 shows the stack energy efficiency and energy density of a
28-cell DMFC stack with 22 cm2 electrode area using 1.0M methanol as fuel and air as
oxidant, respectively. With increasing stack current, the energy efficiency increases until
reaching a peak value. At the higher temperature, higher current is needed to reach the peak
energy efficiency. The operating temperature does not appear to significantly affect the peak
energy efficiency and energy density. At 60 oC the energy efficiency is 0.237, which is only a
slightly higher than that at 40 oC (0.214) or that at 80 oC (0.209). The peak energy efficiency
and energy density are also listed in Table 10. Interestingly, the benefit of increasing stack
temperature is only for achieving higher power but not for obtaining higher energy efficiency
and energy density.
Table 11. Peak system power, energy efficiency, energy density and open circuit voltage
of a DMFC system that contains a 28-cell DMFC stack with 22 cm2 electrode area using
1.0M methanol as fuel and air as oxidant operated at different temperatures.
Temperature (oC)
Parameters
40 60 80
Pmax-system * (W) 19 36 53
system 0.176 0.211 0.193
system (Wh/Kg) 1070 1285 1174
Eopen (V) 13.4 14.8 15.8
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 45
20
16
Voltage (V)
12
Figure 29. Voltage-current curves of a DMFC system that contains a 28-cell stack with 22 cm2
electrode area using air as oxidant and 1.0M MeOH as fuel at various operating temperatures. The
results are obtained by simulation from DMFC single cell data.
60
50
40
Power (W)
30
20
Sys Power (80 oC)
0
0.0 2.0 4.0 6.0 8.0
Output Current (A)
Figure 30. Power-current curves of a DMFC system that contains a 28-cell stack with 22 cm2 electrode
area using air as oxidant and 1.0M MeOH as fuel at various operating temperatures. The results are
obtained by simulation from DMFC single cell data.
46 Rongzhong Jiang
1400
1200 0.20
Energy Efficiency
800
0.12
600
0.08
400 Sys Energy Density (40 oC)
Figure 31. Plots of energy efficiency and energy density versus output current of a DMFC system that
contains a 28-cell stack with 22 cm2 electrode area using air as oxidant and 1.0M MeOH as fuel at
various operating temperatures. The results are obtained by simulation from DMFC single cell data.
The energy efficiency and energy density can be calculated for a DMFC system
according to the equations (16) and (17). Figure 31 shows plots of energy efficiency and
energy density versus output current for a DMFC system that contains a 28-cell stack with 22
cm2 electrode area using 1.0M methanol as fuel and air as oxidant operated at different
temperatures. With increasing output current, the energy efficiency increases quickly until
reaching a peak value. With increasing temperature the peak energy efficiency changes only
slightly. At 60 oC the highest peak energy efficiency is obtained (0.211). A DMFC can
achieve 1285 Wh/Kg energy density for operating at 60 oC using 1.0M methanol as fuel and
air as oxidant. The peak energy efficiency and energy density of the DMFC system are also
listed in Table 11. Although increasing operating temperature will result in increasing output
power significantly, the energy efficiency and energy density are changed only slightly. At
higher temperature, there is a higher rate of methanol crossover, which limits the increase of
energy efficiency and energy density for a DMFC system. As compared with the DMFC
stack, the DMFC system has 7.7% (at 80 oC) to 17.8% (at 40 oC) energy efficiency loss
depending on operating temperatures.
2.4. Comparison of Power and Energy Densities among DMFC Single Cell, Stack
and System
The performances of DMFC single cell, stack and system can be compared as long as the
electrochemical parameters are normalized. Table 12 lists normalized peak powers of DMFC
single cell, stack, and system using 1.0M methanol as fuel and air as oxidant operated at
different temperatures. There are appreciable power losses from single cell to fuel cell stack,
and from fuel cell stack to fuel cell system. The percentage of power loss varies with
operating temperature. From single cell to fuel cell stack, lower percentages of peak power
are lost with increasing operating temperature. For example, the losses of peak power from
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 47
single cell to stack at temperatures of 40, 60 and 80 oC are 14%, 11% and 8%, respectively.
The lower percentages of power loss at higher temperatures are attributed to higher total
power output. From fuel cell stack to fuel cell system, the same phenomenon is observed as
that from single cell to fuel cell stack. For example, the decreased peak power from stack to
system at temperatures 40, 60 and 80 oC are 16%, 11% and 8%, respectively. Table 13
summarizes peak energy efficiencies of DMFC single cell, stack and system using 1.0M
methanol as fuel and air as oxidant at different operating temperatures. From single cell to
fuel cell stack, increasing the operating temperature changes the energy efficiency only
slightly. For example, at 40, 60 and 80 oC, the decreases of energy efficiency from single cell
to stack are 2.3%, 2.9% and 3.7%, respectively, which is attributed to an iR caused energy
efficiency loss. However, with decreasing temperature, the loss of energy efficiency from fuel
cell stack to fuel cell system is more significant. The energy efficiency losses from fuel cell
stack to fuel cell system at temperatures of 40, 60 and 80 oC are 17.4% , 10.6% , and 7.7%,
respectively. A lower percentage of energy efficiency is lost at higher temperature because
significantly higher power is generated with increasing temperature. Table 14 lists peak
energy density of DMFC single cell, stack and system using 1.0M methanol as fuel and air as
oxidant at different operating temperatures. The losses of energy density from the single cell
to the fuel cell stack at operating temperatures of 40, 60 and 80 oC are 29, 42 and 50 Wh/Kg,
respectively. However, the losses of energy density at 40, 60 and 80 oC from the fuel cell
stack to the fuel cell system are remarkably higher, which are 229, 154 and 98 Wh/g,
respectively. The effect of operating temperature on the discharge power and energy
efficiency of the DMFC single cell, stack and system gives us some good suggestions for
DMFC system designs. If we want to achieve higher power output, we need to set the system
at a higher temperature limit, at which the energy efficiency decreases only slightly.
However, if we want to obtain only higher energy efficiency, we need to set the system at a
lower temperature limit to extend the operational life for the DMFC system.
Table 12. Comparison of peak power outputs among DMFC single cell, stack and
system using air as oxidant and 1.0M methanol as fuel at different operating
temperatures.
Table 13. Comparison of peak energy efficiencies among DMFC single cell, stack and
system using air as oxidant and 1.0M methanol as fuel at different operating
temperatures.
Table 14. Comparison of peak energy densities among DMFC single cell, stack and
system using air as oxidant and 1.0M methanol as fuel at different operating
temperatures.
Generally, for an actual fuel cell system, air is used as an oxidant simply because of its
abundance. The oxygen content in air is only about 21%, which implies that the O2 partial
pressure of air is 0.21 at 1 atmosphere pressure air. The other 79% of air is mainly nitrogen.
In order to better understand the effect of oxygen partial pressure on DMFCs performance
we used pure oxygen (O2 partial pressure is 1.0 at 1 atmosphere oxygen) to replace air as
oxidant to continue the present study. The main role of increasing oxygen partial pressure is
to improve the mass transfer at the cathode. The second role is to quickly remove the
methanol at the cathode, which comes from the anode through methanol crossover, in order to
recover the activity of the cathode catalyst. Figure 32 shows voltage-current curves of a
DMFC single cell operated at 60 oC using O2 as oxidant and methanol as fuel with different
concentrations. Here, the points and lines are experimental and simulated results,
respectively. As compared with the results of using air as oxidant in Figure 6, the voltage-
current curves with O2 as oxidant are bunched even if the methanol concentrations are
different; and the discharge current ranges for all these curves are significantly extended to
higher current, indicating that the performance of DMFCs are cathode limited in air. Figure
33 shows power-current curves for the same conditions as those in Figure 32. At 2.0M
methanol the highest power is obtained, which is different from that of using air as the
oxidant, with which the highest power is obtained at 1.0M methanol. Table 15 lists the
electrochemical parameters for simulation of the voltage-current curves of the DMFC single
cell at 60 oC using O2 as oxidant and methanol as fuel with different concentrations. As
compared with the results of using air as oxidant in Table 1, the b values for O2 as oxidant are
apparently smaller; and the peak powers become much higher. It is interesting that the
DMFCs performance at high methanol concentration for O2 as oxidant is significantly
improved. For example, the peak power is increased by 41 mW/cm2 (61%) for the condition
of 2.0M methanol, which implies that the DMFC is more tolerant to methanol crossover when
using O2 to replace air as oxidant. The enhanced tolerance of the DMFC to high methanol
concentration is attributed to an improvement of cathode performance in the condition of high
oxygen partial pressure.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 49
0.7
O2 and 0.5M MeOH
0.6 O2 and 1.0M MeOH
0.4
0.3
0.2
0.1
0
0 100 200 300 400 500
2
Current (mA/cm )
Figure 32. Voltage-current curve of a DMFC single cell at 60 oC using O2 as oxidant and methanol as
fuel with various concentrations. The points and lines are experimental and calculated results,
respectively.
120
100
Power (mW)
80
60
40
O2 and 0.5M Me OH
O2 and 1.0M Me OH
20 O2 and 2.0M Me OH
O2 and 3.0M Me OH
0
0 100 200 300 400 500
2
Current (mA/cm )
Figure 33. Power-current curve of a DMFC single cell at 60 oC using O2 as oxidant and methanol as
fuel with various concentrations. The points and lines are experimental and calculated results,
respectively.
50 Rongzhong Jiang
0.35
2000
0.30
Energy Density (Wh/Kg)
1600
Energy Efficiency
0.25
1200 0.20
0.15
800
0.10
O2 and 0.5M MeOH
400 O2 and 1.0M MeOH
O2 and 2.0M MeOH
0.05
O2 and 3.0M MeOH
0 0.00
0 50 100 150 200 250 300 350 400
2
Cell Current (mA/cm )
Figure 34. Plots of energy efficiency and energy density versus current density of a DMFC single cell at
60 oC using O2 as oxidant and methanol as fuel with various concentrations.
Figure 34 shows energy efficiency and energy density of a DMFC single cell operated at
60 oC using O2 as oxidant and methanol as fuel with different concentrations. As compared
with the results of using air as oxidant, the energy efficiency and energy density of the DMFC
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 51
are increased significantly, especially under high methanol concentration. However, the
variation order of energy efficiency with methanol concentration for O2 as oxidant is still the
same as that for using air as oxidant, i.e., with increasing methanol concentration, the energy
efficiency decreases significantly. Table 16 summarizes the peak energy efficiency and
energy density of the DMFC single cell using O2 as oxidant. For methanol concentrations as
0.5, 1.0, 2.0 and 3.0M there are 16%, 21%, 47% and 101% increase in energy efficiency as
compared with those of using air as oxidant, respectively. The highest energy density of 1996
Wh/Kg (about one third of the theoretical energy density) is obtained for methanol
concentration at 0.5M. As a consequence of using O2 to replace air, the DMFC has
significantly improved its overall performance, such as the increased power output, energy
efficiency and energy density, as well as the apparently enhanced tolerance to methanol
crossover.
According to equations (5) and (10), the voltage-current and power-current curves of DMFC
stack can be calculated. Figure 35 shows voltage-current curves of a 28-cell DMFC stack
with 22 cm2 electrode area operated at 60 oC using O2 as oxidant and methanol as fuel with
different concentrations. Compared with those of using air as oxidant shown in Figure 11, the
voltage-current curves for using O2 as oxidant in Figure 35 are gathered so close that the three
curves for methanol concentrations as 0.5, 1.0 and 2.0M are almost overlapped. However, the
upper ranges of discharge current of these voltage-current curves vary with methanol
concentration. If the methanol concentration is higher, the range of discharge current is more
extended to a higher current value, which is attributed to better mass transfer and higher
kinetic rate of pure oxygen than that of air. Figure 36 shows power-current curves of the
DMFC stack under the same conditions as those in Figure 35. Correspondingly, three of the
power-current curves overlap for methanol concentration from 0.5 to 2.0M. With increasing
discharge current, the stack power increases. The peak powers are obtained at 8.0V stack
voltage, where the current is turned off to protect the DMFC system from running at too low a
cell voltage. Table 17 lists peak powers for a 28-cell DMFC stack with 22 cm2 electrode area
operated at 60 oC using O2 as oxidant and methanol as fuel with different concentrations. The
highest peak power, 57W, is obtained for methanol concentration at 2.0M, which is 73%
higher than that of 33W for using air as oxidant at the same methanol concentration.
Table 17. Peak stack power, energy efficiency, energy density and open circuit voltages
of a 28-cell DMFC stack with 22 cm2 electrode area operated at 60 oC using O2 as
oxidant and methanol as fuel with different concentrations.
20
16
Volatge (V)
12
8
O2 and 0.5M MeOH
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Stack Current (A)
Figure 35. Voltage-current curves of a 28-cell DMFC stack with 22 cm2 electrode area at 60 oC using
O2 as oxidant and methanol as fuel with various concentrations. The results are obtained by simulation
from DMFC single cell data.
60
50
Stack Power (W)
40
30
20
O2 and 0.5M MeOH
O2 and 1.0M MeOH
10
O2 and 2.0M MeOH
O2 and 3.0M MeOH
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Figure 36. Power-current curves of a 28-cell DMFC stack with 22 cm2 electrode area at 60 oC using O2
as oxidant and methanol as fuel with various concentrations. The results are obtained by simulation
from DMFC single cell data.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 53
Stack Energy (O2 and 0.5M MeOH) Stack Energy (O2 and 1.0M MeOH)
Stack Energy (O2 and 2.0M MeOH) Stack Energy (O2 and 3.0M MeOH)
2000 0.32
0.28
Energy Density (Wh/Kg)
1600
Energy Efficiency
0.24
1200 0.20
0.16
800
0.12
0.08
400
0.04
0 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Stack Curre nt (A)
Figure 37. Plots of energy efficiency and energy density versus stack current of a 28-cell DMFC stack
with 22 cm2 electrode area at 60 oC using O2 as oxidant and methanol as fuel with various
concentrations. The results are obtained by simulation from DMFC single cell data.
Compared with the DMFC single cell, the energy efficiency and energy density of the
DMFC stack are decreased only slightly due to the resistance of bipolar plates and the
connections between the electrodes. With increasing discharge current, the loss of energy
efficiency caused by stack resistance is more apparent. Figure 37 shows energy efficiency and
energy density of a 28-cell DMFC stack operated at 60 oC using O2 as oxidant and methanol
as fuel with different concentrations. Compared with the results of using air as oxidant the
energy efficiency and energy density of the DMFC stack are increased significantly. Table 17
summarizes the peak energy efficiency and energy density of the DMFC stack using O2 as
oxidant. For methanol concentrations as 0.5, 1.0, 2.0 and 3.0M there are 16%, 20%, 49% and
154% increase in energy efficiency as compared with those of using air as oxidant,
respectively. The DMFC stack using O2 as oxidant has significantly higher tolerance against
methanol crossover than that of using air. The highest energy efficiency of 0.322 and energy
density of 1959 Wh/Kg are obtained for methanol concentration as 0.5M with O2 as oxidant.
It is reasonable to consider that there is the same amount of internal energy consumption in a
DMFC system using O2 as oxidant as that of using air as oxidant, since it does not need to put
additional auxiliary components into the DMFC system. According to equations (10), (14)
and (15) the voltage-current and power-current curves of a DMFC system are calculated, and
shown in Figures 38 and 39, respectively. The simulated DMFC system contains a 28-cell
stack with 22 cm2 electrode area operated at 60 oC using O2 as oxidant and methanol as fuel
54 Rongzhong Jiang
with various concentrations. Compared with the DMFC stack, there are apparent voltage
decreases on the voltage-current curves for the DMFC system, especially at the beginning of
the operation, which is attributed to the parasitic losses by the auxiliary components. Such
losses decrease the total output power significantly. Apparently, the discharge current of a
DMFC system can not go as high as those of a DMFC stack under the same operating
conditions, which results in lowering the upper discharge current ranges. For example, the
upper discharge current limit is 5.1A for a DMFC stack, but only 4.6A for the DMFC system
for methanol concentration at 3.0M. With increasing discharge current, the output power of
the DMFC system increases until reaching a peak value, at which the current is terminated.
Table 18 summarizes the electrochemical parameters of the DMFC system that contains a 28-
cell stack with 22 cm2 electrode area operated at 60 oC using O2 as oxidant and methanol as
fuel with various concentrations. With increasing methanol concentration, the open circuit
voltage of the DMFC system decreases slightly, but the output power increases. When the
DMFC system reaches the highest power 53W at 2.0M methanol, the output power decreases
again by further increasing methanol concentration to 3.0M due to an increased methanol
crossover. Compared with that of using air as oxidant, the discharge performance of the
DMFC system using O2 as oxidant is improved significantly. For methanol concentrations at
0.5, 1.0, 2.0 and 3.0M, there are 0.7 (4.6%), 1.2 (8.1%), 1.8 (13.3%) and 2.6V (22.6%)
increase in open circuit voltage; and 12 (37.5%), 15 (41.7%), 24 (82.8%) and 26W (236.4%)
increase in output power, respectively. The improvement in system discharge performance for
high methanol concentration is attributed to the improved cathode performance by using O2
as oxidant.
20
16
Voltage (V)
12
Figure 38. Voltage-current curves of a DMFC system that contains a 28-cell stack with 22 cm2
electrode area at 60 oC using O2 as oxidant and methanol as fuel with various concentrations. The
results are obtained by simulation from DMFC single cell data.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 55
60
50
40
Power (W)
30
20
Sys Power (O2 and 3.0M MEOH)
Sys Power (O2 and 0.5M MEOH)
10 Sys Power (O2 and 1.0M MEOH)
Sys Power (O2 and 2.0M MEOH)
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
Output Current (A)
Figure 39. Power-current curves of a DMFC system that contains a 28-cell stack with 22 cm2 electrode
area at 60 oC using O2 as oxidant and methanol as fuel with various concentrations. The results are
obtained by simulation from DMFC single cell data.
Sys Energy (O2 and 0.5M MeOH) Sys Energy (O2 and 1.0M MeOH)
Sys Energy (O2 and 2.0M MeOH) Sys Energy (O2 and 3.0M MeOH)
2000 0.32
1800
0.28
Energy Density (Wh/Kg)
1600
0.24
Energy Efficiency
1400
1200 0.20
1000 0.16
800
0.12
600
0.08
400
200 0.04
0 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
Output Current (A)
Figure 40. Plots of energy efficiency versus output current of a DMFC system that contains a 28-cell
stack with 22 cm2 electrode area at 60 oC using O2 as oxidant and methanol as fuel with various
concentrations. The results are obtained by simulation from DMFC single cell data.
56 Rongzhong Jiang
Table 18. Peak system power, energy efficiency, energy density and open circuit voltages
of a DMFC system that contains a 28-cell DMFC stack with 22 cm2 electrode area
operated at 60 oC using O2 as oxidant and methanol as fuel with different
concentrations.
Figure 40 shows plots of energy efficiency and energy density for a DMFC system that
contains a 28-cell stack with 22 cm2 electrode area operated at 60 oC using O2 as oxidant and
methanol as fuel with various concentrations. With increasing output current, the energy
efficiency increases until reaching a peak value. The peak energy efficiency decreases
significantly with increasing methanol concentration. As compared with that of using air as
oxidant, the energy efficiency is increased by 18.9%, 24.2%, 58.6% and 211.4% for methanol
concentrations as 0.5, 1.0, 2.0 and 3.0M, respectively. It is noticeable that for higher methanol
concentration there is a higher rate of increase in energy efficiency for using O2 as oxidant in
replace of air. The peak energy efficiency and energy density are also summarized in
Table 18.
3.4. Comparison of Power and Energy Densities among DMFC Single Cell, Stack
and System
methanol the peak energy density is 1996 Wh/Kg for a DMFC single cell, but only 1756
Wh/Kg for the DMFC system. At 3.0M methanol the energy density of the DMFC system is
decreased to only 642 Wh/Kg. Apparently, low methanol concentration is preferred for
achieving the highest energy efficiency and energy density. However, increasing methanol
concentration to 2.0M is good for achieving the highest power output if O2 is used as the
oxidant.
Table 19. Comparison of peak power outputs among DMFC single cell, stack and
system operated at 60 oC using O2 as oxidant and methanol as fuel with different
concentrations.
Table 20. Comparison of peak energy efficiencies among DMFC single cell, stack and
system operated at 60 oC using O2 as oxidant and methanol as fuel with different
concentrations.
Table 21. Comparison of peak energy densities among DMFC single cell, stack and
system operated at 60 oC using O2 as oxidant and methanol as fuel with different
concentrations.
Figure 41 shows voltage-current curves of a DMFC single cell using O2 as oxidant and 1.0M
methanol as fuel at various operating temperatures. The points and lines are experimental and
calculated results, respectively. Compared with those of using air as oxidant in Figure 21, the
open circuit voltages have no apparent differences. However, with increasing discharge
current, the voltage differences become more pronounced. As a result of using O2 as oxidant
58 Rongzhong Jiang
in the DMFC, higher discharge voltages are obtained at each of the corresponding discharge
currents; and the voltage-current curves are extended to higher current ranges. Apparently, the
enhancement of the DMFCs performance results from the improved cathode performance by
using O2 as oxidant. Table 22 lists the electrochemical parameters for simulation of the
voltage-current curves of the DMFC single cell. Compared with the data in Table 8 for using
air as oxidant, the values of Rcell have no apparent differences since the conductivity of
polymer electrolyte membrane has not changed by using O2 to replace air. A significant
difference can be seen of the Tafel slopes (b values), which become much smaller due to O2
replacing air as oxidant. With increasing operating temperature, the b values decrease
appreciably using either O2 or air as oxidant. Figure 42 shows the power-current curves of the
DMFC under the same conditions as that shown in Figure 41. As O2 is used, the peak powers
at 40, 60 and 80 oC are increased by 16, 22 and 46 mW/cm2, respectively, which implies that
the effect of oxygen partial pressure on the discharge performance is more pronounced at
higher operating temperature.
0.7
O2 and 80 oC
0.6 O2 and 60 oC
O2 and 40 oC
0.5
Voltage (V)
0.4
0.3
0.2
0.1
0
0 100 200 300 400 500 600 700 800
2
Current (mA/cm )
Figure 41. Voltage-current curves of a DMFC single cell using O2 as oxidant and 1.0M methanol as
fuel at various operating temperatures. The points and the lines are experimental and calculated results,
respectively.
Temperature (oC)
Parameters
40 60 80
Eopen (V) 0.61 0.65 0.68
b (V/Dec) 0.068 0.058 0.058
Rcell () 0.85 0.66 0.46
Pmax-cell (mW/cm2) 59 95 147
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 59
160
140
120
Power (W)
100
80
60
40 O2 and 80 oC
O2 and 60 oC
20
O2 and 40 oC
0
0 100 200 300 400 500 600 700 800
2
Current (mA/cm )
Figure 42. Power-current curve of a DMFC single cell using O2 as oxidant and 1.0M methanol as fuel
at various operating temperatures. The points and the lines are experimental and calculated results,
respectively.
2000
0.30
1600
Energy Density (Wh/Kg)
0.25
Energy Efficiency
1200 0.20
0.15
800
0.10
Energy Density (O2 and 40 oC)
400
Energy Density (O2 and 60 oC) 0.05
Energy Density (O2 and 80 oC)
0 0.00
0 50 100 150 200 250 300
2
Cell Current (mA/cm )
Figure 43. Plots of energy efficiency and energy density versus current density of a DMFC single cell
using O2 as oxidant and 1.0M methanol as fuel at various operating temperatures.
60 Rongzhong Jiang
Temperature (oC)
Parameters
40 60 80
2
ieo (mA/cm ) 74 105 181
B 0.47 0.44 0.40
cell 0.271 0.294 0.275
cell (Wh/Kg) 1651 1786 1629
Figure 43 shows plots of energy efficiency and energy density versus cell current for a
DMFC single cell using O2 as oxidant and 1.0M methanol as fuel at different operating
temperatures. Compared with the data of using air as oxidant shown in Figure 25, the energy
efficiency and energy density increase significantly. With increasing discharge current, the
energy efficiency and energy density both increase until reaching peak values. Table 23
summarizes the peak energy efficiency and energy density of the DMFC using O2 as oxidant,
and 1.0M methanol as fuel at various operating temperatures. Using O2 to replace air, the
enhancements of energy efficiency at 40, 60 and 80 oC are 24%, 21% and 27%, respectively.
At 60 oC the highest energy density 1786 Wh/Kg is obtained, which is 305 Wh/Kg more than
that of using air as oxidant.
Figures 44 shows voltage-current curves of a 28-cell DMFC stack with 22 cm2 electrode area
using O2 as oxidant and 1.0M methanol as fuel at different operating temperatures. Compared
with those of using air as oxidant, the discharge voltages increase, and the upper ranges of
discharge current (i. e., the current at 8.0V stack voltage) are extended to significantly higher
current. For example, at 80 oC the maximum discharge current can go to 10.3A using O2 as
oxidant; but can only go to 7.1A for air as oxidant. It is interesting to compare the variation of
the upper discharge current range at different operating temperatures. The upper discharge
current increases 2.6 to 3.5A for every 20 degree increase of the operating temperature by
using O2 as oxidant; but increases only 2.0 to 2.2A by using air as oxidant. Figure 45 shows
the corresponding power-current curves of the same data shown in Figure 44. With increasing
stack current, the discharge power increases until reaching a peak value. The peak powers are
Table 24. Peak stack power, energy efficiency, energy density and open circuit voltages
of a 28-cell DMFC stack with 22 cm2 electrode area using 1.0M methanol as fuel and O2
as oxidant operated at different temperatures.
Temperature (oC)
Parameters
40 60 80
*
Pmax-stack (W) 34 55 83
stack 0.266 0.285 0.259
stack (Wh/Kg) 1614 1736 1574
Eopen (V) 17.1 18.2 18.5
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 61
limited by a voltage limit at 8.0V. Table 24 lists the peak powers of a 28-cell DMFC stack
with 22 cm2 electrode area using O2 as oxidant and 1.0M methanol as fuel at different
operating temperatures. The peak power increases with operating temperature. At 80 oC the
peak power is 83W, which is 26W (31%) more than that of using air as oxidant.
20
16
Volatge (V)
12
0
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Figure 44. Voltage-current curves of a 28-cell DMFC stack with 22 cm2 electrode area using O2 as
oxidant and 1.0M methanol as fuel at various operating temperatures. The results are obtained by
simulation from DMFC single cell data.
100
80
Stack Power (W)
60
40
Stack Power (O2 and 80 oC)
0
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Stack Current (A)
Figure 45. Power-current curves of a 28-cell DMFC stack with 22 cm2 electrode area using O2 as
oxidant and 1.0M methanol as fuel at various operating temperatures. The results are obtained by
simulation from DMFC single cell data.
62 Rongzhong Jiang
1800
0.28
1600
0.24
Energy Efficiency
1200 0.20
1000 0.16
800
0.12
600
0.08
400 Stack Energy (O2 and 40 oC)
Stack Energy (O2 and 60 oC) 0.04
200
Stack Energy (O2 and 80 oC)
0 0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0
Stack Current (A)
Figure 46. Plots of energy efficiency and energy density versus stack current of a 28-cell DMFC stack
with 22 cm2 electrode area using O2 as oxidant and 1.0M methanol as fuel at various operating
temperatures. The results are obtained by simulation from DMFC single cell data.
Figure 46 shows energy efficiency and energy density of a 28-cell DMFC stack with 22
cm2 electrode area using O2 as oxidant and 1.0M methanol as fuel at different operating
temperatures. With increasing discharge current the energy efficiency and energy density
increase until reaching peak values. From 40 to 80 oC, there are only small differences in the
peak energy efficiency and energy density. However, compared with those of using air as
oxidant, a large increase in energy efficiency and energy density are observed for using O2 as
oxidant. For example, at 60 oC, the peak energy efficiency for using O2 as oxidant is 0.285,
and peak energy density is 1736 Wh/Kg; but using air as the oxidant, the peak energy
efficiency is 0.237, and peak energy density is 1438 Wh/Kg, respectively. Table 24
summarizes the peak energy efficiency and energy density of the DMFC stack, which has 28-
cells and each cell has a 22 cm2 electrode area using O2 as oxidant and 1.0M methanol as fuel
at different operating temperatures. The variations of energy efficiency and energy density
with operating temperature for using O2 and air as oxidants approximately follow the same
order. For both conditions, from 40 to 60 oC the peak energy efficiency and energy density
increase by a small value, and from 60 to 80 oC the peak energy efficiency and energy density
decrease slightly. At 60 oC the highest energy efficiency and energy density are obtained.
Figure 47 shows voltage-current curves of a DMFC system that contains a 28-cell stack with
22 cm2 electrode area using O2 as oxidant and 1.0M methanol as fuel. Compared with the
results of DMFC stack, there is a remarkable decrease in open circuit voltage and discharge
current. For 40, 60 and 80 oC from the DMFC stack to system, the open circuit voltages are
decreased by 13.4%, 12.1% and 8.1%; and the upper discharge current ranges are lowered by
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 63
11.9%, 7.4% and 4.9%, respectively. Compared with the results of using air as oxidant shown
in Figure 29, there is a 1.2 to 1.4V (or 8% to 10%) increase in open circuit voltage using O2
as oxidant; and 1.3 to 3.2A (or 37% to 54%) increase in upper discharge current range
depending on operating temperature. With increasing operating temperature, the enhancement
of the upper discharge current range becomes more significant. Figure 48 shows power-
current curves of the same data shown in Figure 47. Using O2 to replace air as oxidant, the
system powers are significantly increased; and the power enhancements at 40, 60 and 80 oC
are 58%, 42% and 59%, respectively. The peak powers are listed in Table 25. At 80 oC the
highest peak power 79W is obtained, which is 26W (33%) more than that of using air as
oxidant.
20
16
Voltage (V)
12
Figure 47. Voltage-current curves of a DMFC system that contains a 28-cell stack with 22 cm2
electrode area using O2 as oxidant and 1.0M methanol as fuel. The results are obtained by simulation
from DMFC single cell data.
Table 25. Peak system power, energy efficiency, energy density and open circuit voltages
of a DMFC system that contains a 28-cell DMFC stack with 22 cm2 electrode area using
1.0M methanol as fuel and O2 as oxidant operated at different temperatures.
Temperature (oC)
Parameters
40 60 80
*
Pmax-system (W) 30 51 79
system 0.231 0.262 0.242
system (Wh/Kg) 1403 1591 1476
Eopen (V) 14.8 16.0 17.0
64 Rongzhong Jiang
100
80
Power (W)
60
40
0
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Figure 48. Power-current curves of a DMFC system that contains a 28-cell stack with 22 cm2 electrode
area using O2 as oxidant and 1.0M methanol as fuel. The results are obtained by simulation from
DMFC single cell data.
0.28
1600
0.24
Energy Density (Wh/Kg)
Energy Efficiency
1200 0.20
0.16
800
0.12
0.08
400 Sys Energy Density (O2 and 40 oC)
Figure 49. Plots of efficiency and energy density versus output current of a DMFC system that contains
a 28-cell stack with 22 cm2 electrode area using O2 as oxidant and 1.0M methanol as fuel at various
operating temperatures. The results are obtained by simulation from DMFC single cell data.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 65
Figure 49 shows energy efficiency and energy density for a DMFC system that contains a
28-cell stack with 22 cm2 electrode area using O2 as oxidant and 1.0M methanol as fuel. At
low current range the energy efficiency increases significantly with increasing discharge
current; and then it does gradually until reaching a maximum value. With increasing
operating temperature, the discharge current increases noticeably, and the peak energy
efficiency of the DMFC system is obtained at the higher current range. Table 25 summarizes
energy efficiency and energy density of the DMFC system using O2 as oxidant and 1.0M
methanol as fuel. Because part of the current in the DMFC system is lost to the auxiliary
components, there is a large decrease in energy efficiency and energy density as compared
with the DMFC stack. At 40, 60 and 80 oC, the energy efficiency from DMFC stack to system
decreases by 13.2%, 8.1% and 6.6% respectively. However, compared with those of using air
as oxidant, the energy efficiency is increased remarkably. The energy efficiency for using O2
to replace air at 40, 60 and 80 oC, increase by 31.3%, 24.2% and 25.4%, respectively. The
highest energy efficiency, 0.262, and energy density, 1591 Wh/Kg, are obtained at 60 oC with
1.0M methanol, respectively. A DMFC system at 60 oC using air as oxidant with 1.0M
methanol can obtain a maximum energy density up to 1285 Wh/Kg. However, the same
DMFC system at 60 oC using O2 as oxidant with 1.0M methanol can achieve energy density
as high as 1591 Wh/Kg.
4.4. Comparison of Power and Energy Densities among DMFC Single Cell, Stack
and System
The peak powers of the DMFC stack and system are normalized in order to compare with
those of single cell results. Table 26 summarizes the peak powers of DMFC single cell, stack,
and system using 1.0M methanol as fuel, and O2 as oxidant operated at different
temperatures. There are appreciable power losses from single cell to fuel cell stack, and from
fuel cell stack to fuel cell system. The peak powers from single cell to stack at temperatures
of 40, 60 and 80 oC are decreased by 6.8%, 6.3% and 8.2%, respectively, which is attributed
to the power loss caused by the resistance of the fuel cell stack. From DMFC stack to system
the peak power losses at 40, 60 and 80 oC are 10.9%, 6.7% and 5.2%, respectively, which is
attributed to parasitic current losses due to the auxiliary components. Table 27 summarizes
peak energy efficiencies of DMFC single cell, stack and system using 1.0M methanol as fuel
and air as oxidant at different operating temperatures. From single cell to fuel cell stack, with
increasing operating temperature the energy efficiency is only slightly changed. For example,
at 40, 60 and 80 oC, the losses of energy efficiency from single cell to stack are 2.6%, 2.4%
and 6.2%, respectively, which is attributed to an iR caused energy efficiency loss. However,
from fuel cell stack to fuel cell system, the loss of energy efficiency is more significant with
decreasing operating temperature. The energy efficiency losses from DMFC stack to system
at temperatures of 40, 60 and 80 oC are 13.2%, 8.4% , and 6.2%, respectively, which is
attributed to the power loss caused by parasitic current losses due to the auxiliary
components. Table 28 lists peak energy density of DMFC single cell, stack and system using
1.0M methanol as fuel and O2 as oxidant at different operating temperatures. The losses of
energy density from single cell to fuel cell stack at 40, 60 and 80 oC are 37 (2.2%), 50 (2.8%)
and 54 Wh/Kg (2.3%), respectively. However, the losses of energy density from DMFC stack
to system at 40, 60 and 80 oC are 211 (13.1%), 146 (8.4%) and 98 Wh/g (6.2%), respectively.
66 Rongzhong Jiang
Table 26. Comparison of peak power outputs among DMFC single cell, stack and
system using O2 as oxidant and 1.0M methanol as fuel at different operating
temperatures.
Table 27. Comparison of peak energy efficiencies among DMFC single cell, stack and
system using O2 as oxidant and 1.0M methanol as fuel at different operating
temperatures.
Table 28. Comparison of peak energy densities among DMFC single cell, stack and
system using O2 as oxidant and 1.0M methanol as fuel at different operating
temperatures.
Conclusion
Based on the experimental DMFC single cell results of discharge performance and methanol
crossover, and the proposed models for assembling the DMFC stacks and systems, we have
calculated the power densities, energy efficiencies and energy densities for a series of DMFC
single cells, stacks, and systems at various methanol concentrations, operating temperatures
and oxygen partial pressures (0.21 atmosphere for air, and 1.0 atmosphere for O2). The
advantages of this method are its simplicity and accuracy due to all of the calculations being
based on the actual experimental data, where it is not necessary to set various uncertain and
unknown parameters. The disadvantages are its limitations to the availability of the
experimental data and the ranges and scopes of these experimental data. The optimal
performances of a DMFC system can be predicted as long as the minimum single cell
experimental data are known. In other words, the DMFC single cells performances play an
important role in the performance of a DMFC system. This study has originally analyzed the
interrelationship of power and energy efficiencies of DMFC single cell, stack and system.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 67
Different DMFC system may have different parasitic losses and configurations. The present
results are obtained by simulations in comparison to an actual DMFC system [69] that is
recognized as an optimized DMFC system.
Both the output power and energy efficiency of a DMFC decrease in the order of single
cell, stack and system due to an additional stack resistance for connecting a number of single
cells together to form a DMFC stack, and the internal power consumption in a DMFC system.
When air is used as oxidant, the optimal methanol concentration for obtaining the highest
power is 1.0M. The peak powers for DMFC single cell, stack and system at 60 oC using 1.0M
methanol are 73, 65 and 58 mWcm-2cell-1, respectively. With increasing or decreasing
methanol concentration from 1.0M methanol, the power density decreases. With increasing
temperature, the power density increases. At 80 oC the peak powers for the DMFC single cell,
stack and system using 1.0M methanol are 101, 93 and 86 mWcm-2cell-1, respectively. The
energy efficiency decreases with increasing methanol concentration. The highest energy
efficiencies are obtained by using 0.5M methanol as fuel. The peak energy densities for
DMFC single cell, stack and system at 60 oC using 0.5M methanol are 1726, 1690 and 1438
WhKg-1, respectively. The optimal temperature for obtaining the highest energy efficiency is
60 oC. Increasing or decreasing the operating temperature from 60 oC will result in slightly
decreasing energy efficiency. It is noticed that the energy efficiency of a DMFC system is
sensitive to power or current output. With increasing power or current output, the energy
efficiency increases significantly until reaching a peak value, at which a voltage or current
limit is met.
Although pure oxygen is not commonly used in DMFC systems, we have calculated the
possible results of power density, energy efficiency and energy density of the DMFC single
cell, stack and system, in order to understand how much benefit can be obtained by increasing
oxygen partial pressure, such as applying an air pressure on the cathode in a DMFC system.
With O2 as oxidant, the discharge performance of the DMFC increases significantly as
compared with that of using air as oxidant. The optimal methanol concentration for obtaining
the highest power using O2 as oxidant is 2.0M, which is significantly different from that of
using air as oxidant. The peak powers for DMFC single cell, stack and system using 2.0M
methanol at 60 oC are 102, 93 and 86 mWcm-2cell-1, respectively. Increasing or decreasing
methanol concentration from 2.0M, decreases the power density. With increasing
temperature, the power density increases remarkably. The peak powers for DMFC single cell,
stack and system at 80 oC using 1.0M methanol as fuel and O2 as oxidant are 147, 135 and
128 mWcm-2cell-1, respectively. The highest energy efficiencies are obtained by using 0.5M
methanol as fuel. The peak energy densities for DMFC single cell, stack and system at 60 oC
using 0.5M methanol as fuel and O2 as oxidant are 1996, 1959 and 1756 WhKg-1,
respectively. The energy efficiency decreases with increasing methanol concentration.
Increasing or decreasing operating temperature from 60 oC will result in slightly decreased
energy efficiency. It is demonstrated that increasing oxygen partial pressure will increase not
only the power density but also the fuel energy efficiency significantly.
68 Rongzhong Jiang
Acknowledgement
The author wishes to thank the U. S. Department of Army and the Army Materiel Command
for their support to this work, and Dr. Cynthia Lundgren for reviewing the manuscript and
helpful suggestions.
References
[1] Steele, B. C. H., Heinzei, A. Nature 2001, 441 (6861), 345-352.
[2] Appleby, A. J. Energy 1996, 21 (7-8), 521-653.
[3] Ren, X. M., Zelenay, P., Thomas, S., Davey, J., Gottesfeld, S. J. Power Sources 2000,
86 (1-2), 111-116.
[4] Conte, M., Iacobazzi, A., Ronchetti, M., Vellone, R. J. Power Sources 2001, 100 (1-2),
171-187.
[5] Dyer, C. K. J. Power Sources 2002, 106 (1-2), 31-34.
[6] Rikukawa, M., Sanui, K. Progress in Polymer Science 2000, 25 (10), 1463-1502.
[7] Jiang, R. Z., Rong, C., Chu, D. R. J. Power Sources 2004, 126 (1-2), 119-124.
[8] Chalk, S. G., Miller, J. E. J. Power Sources 2006, 159 (1), 73-80.
[9] Yeo, S. C., Eisenberg, A. J. Applied Polymer Science 1997, 21 (4), 875-898.
[10] Springer, T. E., Zawodzinski, T. A., Gottesfeld, S. J. Electrochem. Soc. 1991, 138 (8),
2334-2342.
[11] Scherson, D. A., Gupta, S. L., Fierro, C., Yeager, E. B., Kordesch, M. E., Eldridge, J.
Hoffman, R. W., Blue, J. Electrochimica Acta 1983, 28 (9), 1205-1209.
[12] Chang, C. K., Liu, H. Y., Abdalmuhdi, I. J. Am Chem. Soc. 1984, 106 (9), 2725-2726.
[13] Jiang, R. Z., Dong, S. J. Phys. Chem. 1990, 94 (9), 7471-7476
[14] Jiang, R. Z., Dong, S. J. Electroanal. Chem. 1988, 246 (1), 101-117.
[15] Collman, J. P., Marrocco, M., Denisevich, P., Koval, C., Anson, F. C. J. Electroanal.
Chem. 1979, 101 (1), 117-112.
[16] Collman, J. P., Denisevich, P., Konai, Y., Marrocco, M., Koval, C., Anson, F. C. J. Am.
Chem. Soc. 1980, 102 (19), 6027-6036.
[17] Vanveen, J. A. R., Colijn, H. A., Vanbaar, J. F. Electrochimca Acta 1988, 33 (6),
801-804.
[18] Dong, S, Jiang, R. Z. Ber. Bunsenges. Phys. Chem. 1987, 91 (4), 479484.
[19] Gojkovic, S. L., Gupta, S, Savinell, R. F. J. Electroanal. Chem. 1999, 462 (1), 63-72.
[20] Boies, D. B., Dravnieks, A. J. Electrochem. Soc. 1962, 109 (8), C198-C198.
[21] Cairns, E. J., Bartosik, D. C. J. Electrochem. Soc. 1964, 111 (11), 1205-1210.
[22] Hampson, N. A., Willars, M. J., Mcnicol, B. D., J. Power Sources 1979, 4 (3), 191-201.
[23] Abens, S., Jacobs, R. J. Electrochem. Soc. 1983, 130 (8), C331-C331.
[24] Goodenough, J. B., Hamnett, A., Kennedy, B. J., Weeks, S. A. Electrochimica Acta
1987, 32 (8), 1233-1238.
[25] Kanda, T., Kawashima, A., Asami, K., Hashimoto, K. J. Electrochem. Soc. 1987, 134
(8B), C418-C418.
[26] Hamnett, A., Kennedy, B. J. Electrochimica Acta 1988, 33 (11), 1613-1618.
[27] Grot, W. G. F., Mehra, V., Munn, G. E., Solenberger, J. C. J. Electrochem. Soc. 1975,
122 (3), C104-C104.
Power and Energy Efficiency Analysis of Direct Methanol Fuel Cell (DMFC) 69
[28] Hora, C. J., Maloney, D. E. J. Electrochem. Soc. 1977, 124 (8), C319-C319.
[29] Huba, F., Yeager, E. B, Olah, G. A. Electrochimica Acta 1979, 24 (5), 489-494.
[30] Nakajima, H. J. Chemical Technology And Biotechnology 1991, 50 (4), 555-561.
[31] Bronoel, G., Besse, S., Tassin, N. Electrochimica Acta 1992, 37 (8), 1351-1353.
[32] Burke, L. D., Casey, J.K. Electrochimica Acta 1992, 37 (10), 1817-1829.
[33] White, J. H., Sammells, A. F. J. Electrochem. Soc. 1993, 140 (8), 2167-2177.
[34] Meli, G., Leger, J. M., Lamy, C., Durand, R. J. Appl. Electrochem. 1993, 23 (3),
197-202.
[35] Surampudi, S., Narayanan, S. R., Vamos, E., Frank, H., Halpert, G., Laconti,A., Kosek,
J., Prakash, G. K. S., Olah, G. A. J. Power Sources 1994, 47 (3), 377-385.
[36] Ren, X. M., Wilson, M. S., Gottesfeld, S., J. Electrochem. Soc. 1996, 143 (1), L12-L15.
[37] Fan, Q. B., Pu, C., Smotkin, E. S., J. Electrochem. Soc. 1996, 143 (10), 3053-3057.
[38] Chu. D., Gilman, S. J. Electrochem. Soc. 1996 143 (5), 1685-1690.
[39] Wang, J. T., Wainright, J. S., Savinell, R. F., Litt, M., J. Appl. 1996, Electrochem., 26
(7), 751-756.
[40] Witham, C. K., Chun, W., Valdez, T. I., Narayanan, S. R. Electrochemical and Solid
State Letters 2000, 3 (11), 497-500.
[41] Mehta, V., Cooper, J. S. J. Power Sources 2003, 114 (1), 32-53.
[42] Deluca, N. W., Elabd, Y. A. J. Polymer Science Part B-Polymer Physics 2006, 44 (16),
2201-2225.
[43] Chu, D., Gilman, S., J. Electrochem. Soc. 1994, 141 (7), 1770-1773.
[44] Jiang, R. Z., Chu, D. J. Electrochem. Soc.2000, 147 (12), 4605-4609.
[45] Kauranen, P. S., Skou, E. J. Appl. Electrochem. 1996, 26 (9): 909-917.
[46] Wang, J. T., Wasmus, S., Savinell, R. F, J. Electrochem. Soc.1996, 143 (4), 1233-1239.
[47] Ren, X. M., Henderson, W., Gottesfeld, S. J. Electrochem. Soc. 1997, 144 (9),
L267-L270.
[48] Ravikumar, M. K., Shukla, A. K. J. Electrochem. Soc.1996, 143 (8), 2601-2606.
[49] Cruickshank, J., Scott, K. J. Power, Sources 1998, 70 (1), 40-47.
[50] Han, J., Park, E. S. J. Power Sources 2002, 112, 477-483.
[51] Lee, B. D., Jung, D. H., Ko, Y. H. J. Power Sources 2004, 131, 207-212.
[52] Gao, L. J., Jiang, Z. H., Dougal, R. A., IEEE Transaction on Aerospace and Electronic
System 2005, 41, 346-355.
[53] Gao, W. Z. IEEE Transaction on Vehicular Technology 2005, 54, 846-855. (2005).
[54] Bostic, E., Sifer, N., Bolton, C., Ritter, U., Dubois, T., J. Power Sources 2004, 137 (1),
76-79.
[55] Dohle, H., Schmitz, H., Bewer, T., Mergel, J., Stolten, D. J. Power Sources 2002, 106,
313-322.
[56] Peled, E., Blum, A., Aharon, A., Philosoph, M., Lavi, Y., Electrochemical and Solid
State Letters 2003, 6, A268-A271.
[57] Xie, C. G., Bostaph, J., Pavio, J. J. Power Sources 2004, 136, 55-65.
[58] Dillon, R., Srinivasan, S., Arico, A. S., V. Antonucci, V. J. Power Sources 2004, 127,
112-126 (2004).
[59] Qian, W. M., Wilkinson, D. P., Shen, J., Wang, H. J. Zhang, J. J. J. Power Sources
2006, 154, 202-213.
[60] Jiang, R. Z., Chu, D. J. Power Sources 2006, 161, 1192-1197.
70 Rongzhong Jiang
[61] Srinivasan S., Velev, O. A., Parthasarathy, A., Manko, D. J., Appleby, A. J. J. Power
Sources 1991, 36 (3), 299-320.
[62] Gurau, B., Smotkin, E. S. J. Power Sources 2002, 112 (2), 339-352.
[63] Jiang, R. Z., Chu, D. J. Electrochem. Soc. 2004, 151 (1), A69-A76.
[64] Jiang, R. Z., Chu, D., Electrochem. Solid-State Lett. 2002, 5 (7), A156-A159.
[65] Kim, J., Lee, SM., Srinivasan, S., Chamberlin, C. E. J. Electrochem. Soc. 1995, 142
(8), 2670-2674.
[66] Rho, Y. W., Velev, O. A., Srinivasan, S., Kho, Y. T. J. Electrochem. Soc. 1994, 141
(8), 2084-2088.
[67] Piela, P., Eickes, C., Brosa, E., Garzon, F., Zelanay, P. J. Electrochem. Soc. 2004, 151
(12), A2053-A2059.
[68] Patterson, T. W., Darling R. M. Electrochemical and Solid State Letters 2006, 9 (4),
A183-A185.
[69] Jiang, R. Z., Chu, D. Proceedings of the 42nd Power Sources Conference 2006, Page
411-414, Philadelphia, U.S.A.
[70] Jiang, R. Z., Chu, D. J. Power Sources 2001, 93 (1-2), 25-31.
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 71-134 2007 Nova Science Publishers, Inc.
Chapter 2
Abstract
The Multiscale Analysis Group of the Politecnico di Torino (Italy) is involved in the
experimental analysis and modeling of the CHP-100 SOFC Field Unit built by Siemens Power
Generation Stationary Fuel Cells (SPG-SFC). The experimental analysis of a large SOFC
generator in operation is a complex task, due to the large number of variables which affect its
operation, the limited number of measurements points in the generator volume, the necessity
to avoid malfunctions in the real operation. As a consequence, the experimental analysis of the
CHP-100 SOFC Field Unit has been developed with methods of Design of Experiments, and
with a statistical analysis of the collected data. The experimental sessions have been designed
in order to investigate the effect of two important operation factors (the overall fuel
consumption FC and the air stoichs ox), in order to characterize the operation of the single
sectors of the SOFC generator, and to obtain the sensitivity maps of the main investigated
dependent variables. Furthermore, the main result is the estimation of the local values of fuel
utilization of the various sectors of the generator, through the combination of the experimental
voltage sensitivity analysis to overall FC and an analytical model of polarization, to outline
the distribution of fuel inside the generator. Finally, the sectors of the generator, of different
pedigree and position, are compared in terms of the polarization effects, showing how the
local fuel utilization and temperature affect the estimated local anode exchange current
density values.
*
E-mail address: massimo.santarelli@polito.it
72 M. Santarelli, P. Leone, M. Cal et al.
1. Introduction
The Multiscale Analysis Group of the Politecnico di Torino (Italy) is involved in the
experimental analysis and modeling of the CHP-100 SOFC Field Unit built by Siemens
Power Generation Stationary Fuel Cells (SPG-SFC). The generator has been installed in a
SOFC laboratory at TurboCare S.p.A. (a subsidiary of SPG-SFC) since June 19, 2005, and to
date has shown the record availability of 99.5%. The SOFC CHP-100 is the first to utilize the
commercial prototype air electrode supported cells and in-stack reformers. The generator is
fed with natural gas from the grid. The stack is composed of 1152 single tubes, arranged in 48
cell bundles (4 cell bundles are connected in series to form a bundle row, and 12 bundle rows
are aligned side by side, interconnected with an in-stack reformer between each bundle row).
The experimental analysis of a real large SOFC generator in operation is a complex task.
The generator is a big plant, with many variables which affect its operation: therefore, the
experimental environment is completely different compared to a laboratory, because it is
difficult to control the many variables involved during the development of the experiment;
moreover, the data acquired in the experimental session have to be carefully analysed, in
order to isolate the main effects which have to be detected, to take care of the interactions,
and to quantify the significance of every effect. This causes the necessity of a careful design
of the experiment coupled with a consequent statistical analysis of the collected data, in order
to be able to outline the significance of every observed effect.
Moreover, the large plant is characterized by a not uniform distribution of its physical
and chemical variables (e.g. temperatures, chemical composition of mass flows, etc.), and at
the same time the measurement points are discrete: therefore, in many parts of the plant the
data are not available. As a consequence, a combination of some experimental techniques and
of analytical models could be used to deduce the distribution, inside the generator volume, of
some important variables.
Finally, the generator is in real operating conditions, and therefore its experimental
analysis has to avoid malfunctions and dangerous operations; usually, one possibility is the
perturbation of some independent variables in a safety experimental range, and the analysis of
the sensitivity of the other variables (the dependent variables) to the imposed perturbation.
As a conclusion, the experimental analysis of the CHP-100 SOFC Field Unit has been
developed with methods of Design of Experiments, and with a statistical analysis of the
collected data.
A design of experiments procedure has been used to study in a rigorous manner the
collected experimental data and also to relate, through analytical expressions, the investigated
dependent variables with the control factors (independent variables). The design of
experiments approach allows one to study the main effect of the factors and also their
interaction effects. The aim is the description of the analytical relations which express the
dependent variables as a function of the examined control factors through first (simple
factorial) or second-order (spherical CCD) regression models.
The experimental sessions have been designed in order to investigate the effect of two
important operation factors: the overall fuel consumption (FC) and the air stoichs (ox). The
main expectation has been the characterization of the operation of the single sectors of the
SOFC generator, pointing out the analysis on the distribution of the local voltages (and
Experimental Activity on a Large SOFC Generator 73
temperatures). Particular attention has been addressed in the description of the operation in
terms of sensitivity maps of the main investigated dependent variables (voltages).
The sensitivity analysis of the local voltage to the overall fuel consumption modifications
can be used as a powerful procedure to deduce the local distribution of fuel utilization (FU)
along the single sectors of the generator: in fact, through an analytical model obtained by
deriving the polarization curve respect to FU, it is possible to link the distribution of voltage
sensitivities to the overall FC to the distribution of the local FU. Therefore, the main purpose
of the previous sensitivity analysis is to estimate the local values of fuel utilization of the
stack sectors, through the combination of the experimental voltage sensitivity analysis to FC
and the analytical model. This is useful to outline its drawbacks on the distribution of fuel
inside the generator: the results are used to debate on the issue of fuel distribution systems in
the generator.
Finally, the sectors of the generator, of different pedigree and position, are compared in
terms of the main polarization effects. The polarization behavior has been investigated at
fixed setpoint temperature and overall fuel consumption. On this basis, a parameter
estimation, using a polarization model of a single cell, has been performed and used for the
analysis of the local activation effects. It has been shown how the local fuel utilization and
temperature affect the estimated local anode exchange current density values.
Figure 1. Picture of the SOFC CHP-100 test site in TurboCare (Torino, Italy).
74 M. Santarelli, P. Leone, M. Cal et al.
cm active length, 834 cm2 active area) and in-stack reformers. The generator is fed with
natural gas from the grid distribution. In Figure 1 the picture representing the CHP 100 test
site in TurboCare (Torino, Italy) is shown.
The upper level of the system hierarchy after the single cell is the cell bundle, which
consists of a 24-cell array arranged as 8 cells in electrical series by 3 cells in electrical
parallel. Two bundles in series form a sector; two sectors are connected in series to form a
row, and 12 rows are aligned side by side, interconnected in serpentine fashion with an in-
stack reformer between each row (for a total of 1152 single tubes). The schematic of the cell
stack arrangement is shown in Figure 2 [1], outlining the position of the fuel ejectors and the
power leads. In this Chapter, the results will be discussed dividing ideally the primary
generator in four main zones (North; South; Power Leads side and Ejectors side), and also in
the central zone (in the middle between the Power Leads and the Ejectors sides).
NORTH SOUTH
The system is completed by the balance of plant (BoP), with five major skids: Generator
Module, Electrical Control System, Fuel Supply System (FSS), Thermal Management System
(TMS), and Heat Export System (HES) [2,3].
From the electrical point of view, the module is made up of four parts: (a) the
electrochemical generator; (b) the Power Conditioning System (PCS): an inverter, which
operates the DC/AC conversion; (c) the SOFC auxiliaries (blowers); (d) the Board Fuel Cell
(QFC), planned from Politecnico of Torino, which allows the electrical connection of the
SOFC CHP-100 to the net. From the thermal point of view, the exhausts from the stack,
passing in a cross-finned tubes gas-water exchanger, provide, in nominal conditions,
approximately 60 kWt of thermal energy, used for the winter and summer conditioning
(through a absorption refrigerator cycle water-lithium bromide fed with warm water) of some
offices of the TurboCare factory.
The simplified flow schematic of the SOFC CHP-100 BoP is shown in Figure 3, while in
Figure 4 the simplified schematic of the primary generator structure is reported.
The commissioning date in TurboCare was June 19, 2005. The operational data @
December 2006 report a number of run hours in TurboCare of around 12,000, with a total run
hours of around 32,000 (including run hours in the Netherlands and Germany); a average
stack temperature of 954C; a DC generated power of 123.6 kWe, at 246.1 VDC and 502.3 A;
a AC generated power of 113 kWe; a power to TurboCare workshop grid of 103 kWe (20% of
the workshop requirement); a heat generation of 60 kWt (hot water @85C). Other
characteristics are: a reliable operation also with significant LHV changes of the natural gas;
no measurable voltage or power degradation; a very high availability of 99.5% (actual, on
annual basis); 4 stops (one operation error, three inverter failures); 24 successful operation
transition to power dissipator due to utility (AEMD, Azienda Energetica Metropolitana
76 M. Santarelli, P. Leone, M. Cal et al.
Torino Distribuzione) or TurboCare grid failures; one complete thermal cycle (August 2006);
a very dependable automatic operation (no operator in control room); a remote control
capability via modem; a easy maintenance (replacement of air filters and of reactant
desulphurization).
In one year of operation the SOFC CHP-100 fuelled by natural gas avoids the production
of 272 tons of CO2, the environmental pollution of 1035 kg of NOx, the import of 121 tons of
oil equivalent, if compared to a gas turbine plant of higher size (in the order of 100 MWe;
therefore, if compared to a similar size gas turbine, the advantages would be even higher).
Green Certificates are provided by the Italian Authority (GRTN) for electrical energy
produced by a fuel cell generator (even if using a fossil fuel): the energy production estimate
in 2006 is 900 MWh (18 certificates).
Politecnico of Torino has developed several activities, like the design of the SOFC
thermal plant [4], the modeling of the generator and of the BoP [5-8], the safety design and
analysis [9]. In [10] a model of the operation of the SOFC CHP-100 has been developed
using a 0D approach and validated through a first session of experimental tests. In [11,12] the
effect of the setpoint generator temperature and fuel consumption factors on several
dependent variables (i.e. DC and AC electric power, recovered heat, electric and global
efficiency, efficiency of the pre-reforming process) is analyzed in form of screening tests and
the process responses are treated in form of response surface plots. In [13] the regression
models have been used in constrained optimization procedures to maximize different
objective functions (AC electric power, recovered heat, etc.).
To develop the study, the factorial analysis and the response surface method (RSM) have
been applied, and first and second-order regression models linking the dependent variables to
the control factors have been found and analyzed with an ANOVA [14,15].
In many physical or engineering problems, two or more variables are related and it is of
interest to model and explore this relationship. Suppose that there is a single dependent
variable or response Y that depends on k independent or regressor variables x1 , x2 ,..., xk . The
relationship between these variables is characterized by a mathematical model, called a
regression model, that fits a set of sample data. In some instances, the experimenter knows the
exact form of the true functional relationship between Y and x1 , x2 ,..., xk . However, in most
cases, the true functional relationship is unknown and the experimenter can choose an
appropriate function in order to approximate it. Low-order polynomial models are widely
used as approximating functions. Thus, a regression model can be written in the following
form:
Experimental Activity on a Large SOFC Generator 77
Y = 0 + 1 x1 + 2 x 2 + ... + k x k + (1)
It is called a multiple linear regression model with k regressor variables. The parameters
j (j=0,1,k) are called the regression coefficients. The parameter j represents the expected
change in the response Y per unit change in xj when all the remaining independent variables xi
(ij) are held constant. Thus, the regression coefficient of a regressor variable represents the
sensitivity parameter of the dependent variable to changes in the regressor variable
(independent variable). The method of least squares is typically used to estimate the
regression coefficients in a multiple linear regression model.
If an experimental session is designed by applying the methodology of the Design of
Experiments, then the investigated dependent variables can be described by regression
models. Different forms of regression models can be obtained from an experimental session
according to the chosen design of experiments, for example first (equation (1)) or second
order regression models (equation (2)):
k k
Y = 0 + j x j + ij xi x j + jj x 2j + (2)
j =1 i< j j =1
Because of the reduced experimental domain (due to safe operation of the real plant), the
analysis described in this Chapter will deal with results obtained by a first-order design.
The collected data are first analyzed through the Yates method (analysis of the factor
significance on the dependent variable); after, first-order regression models are obtained.
Particular attention has been addressed in the description of the operation in terms of
sensitivity maps.
To better understand the methods of Design of Experiment used in the experimental
sessions, in the Appendix the statistical concepts and the methods used in the analysis are
briefly resumed. The analysis is described in the following Paragraphs.
The experimental session has been designed in order to evaluate the main and interaction
effects of two independent variables (factors), the overall fuel consumption (FC) and the air
stoichs (ox), on the distribution of voltages in different sectors of the generator.
G f ,ech
FU = (3)
G f ,tot
78 M. Santarelli, P. Leone, M. Cal et al.
The control variable overall fuel consumption (FC) factor is defined in the same way, but
at the numerator it takes into account also the fuel mass flow which is chemically oxidized on
the anode side by the air leakage from cathode to anode, plus the fuel leakage through the
electrolyte layer (in the order of 2030 cc/min according to Gopalan and DiGiuseppe [16]),
plus the fuel by-pass through the insulation package.
The overall fuel consumption value is imposed by the operator, while the overall fuel
utilization value is evaluated by the control system of the generator. An indicative difference
between overall fuel utilization and overall fuel consumption is as follows: imposing an
overall fuel consumption value of 83.75%, the control overview shows an overall fuel
utilization value of 79%.
Air stoichs
The air stoichs factor is defined as the ratio of the total air mass flow which enters in the
generator and the air mass flow actually used in the electrochemical reaction on the cathode
surface.
Gair ,tot
ox = (4)
Gair ,ech
In other words, 4.5 stoichs means an introduction inside the generator of the 450% of air
mass flow compared to the actually used value; an increase of 0.1 stoichs means an increase
of 10% of the air mass flow compared to the used value of air mass flow.
In the case of the generator, the variable controlled to modify the air stoichs is the speed
of the air blowers, named AIRO (speed of air primary blower, expressed as percentage of its
maximum value).
Procedure
In the experimental session, the average current density has been kept constant at the nominal
value of 0.2 A/cm2 (500 A generator current).
The generator setpoint temperature (TGEN) is the highest temperature measured by five
thermocouples placed in the central zone of the generator. This variable can be controlled by
the operator, but in the experimental sessions it has not been considered as a independent
variable (factor). Its value depends on the value imposed to the air stoichs factor.
The investigated dependent variables are evaluated as a function of the two factors:
overall fuel consumption (FC) and air stoichs (ox).
A design of experiments procedure has been used in the planning of the experimental
session; this has been done with the aims to study in a rigorous manner the collected
experimental data and also to relate, through analytical expressions, the investigated
dependent variables with the control factors (independent variables). The design of
experiments approach has been the factorial design: since two factors are involved, a 22
factorial design has been applied. The experimental session has been designed by applying
simple 22 factorial analysis. The aim is the description of the analytical relations which
express the dependent variables as function of the two control factors through first-order
(simple 22) regression models.
Experimental Activity on a Large SOFC Generator 79
Each single combination of the factors has been replicated twice to evaluate the
significance of the factor modification on every dependent variable. Furthermore, the factorial
analysis of variance (ANOVA) and the response surface method (RSM) have been applied,
and the first-order regression models linking the dependent variable to the two control factors
have been found and analyzed with an ANOVA.
The ANOVA both of the experimental data and of the first-order regression models have
been developed using the factorial of Table 1, where the treatment combinations and the
range of variation of each factor are reported:
ox = 4.6 4.8
FC = 81.75 84.25%
The range of variation of each factor has been imposed in agreement with TurboCare and
SPG-SFC, in order to avoid malfunctions of the generator.
Due to the slow thermal modification of the generator, the air stoichs were imposed every
day at 6 p.m., to reach a stabilized temperature of the generator during the night. During the
day, the tests with constant ox and variable FC were performed, thanks to the fast
adaptation of the generator to the overall fuel consumption factor modifications.
Imposed the FC value, the data collection started after 60 min (to stabilize the signal
data) for a length of 30 min, with a rate of 1 min. The single value of a variable is an average
of the 30 values collected in every test length.
In the experimental session, all the data monitored by the SOFC Field Unit are collected:
voltage of every single sector, and distribution of temperatures inside the generator.
First, the data have been analyzed through the Yates method (analysis of the factor
significance on the dependent variable).
Second, first-order regression models have been obtained, with the consequent response
surfaces and contour plots.
80 M. Santarelli, P. Leone, M. Cal et al.
Third, particular attention has been addressed in the description of the operation in terms
of sensitivity maps of the investigated dependent variables (voltages) to the air stoichs (ox),
and the overall fuel consumption (FC).
3.2. Analysis of the Polarization Effects of the Various Sectors at Variable Local
Fuel Utilization and Temperature
In the experimental session we have obtained also the polarization curves that describe the
electrochemical behaviour of the various sectors of the SOFC generator. The curves have
been obtained varying the values of the overall fuel consumption in two different sequential
ranges of current:
The tests have been performed maintaining constant the generator setpoint temperature at
the value of 967C: this value of temperature is automatically maintained from the system
acting on the air flow. A current variation causes a variation of the inlet air temperature: this
is controlled by a by-pass valve of the low temperature heat recuperator. Besides, a current
variation causes a variation in the thermodynamics condition of the stack: to maintain
constant TGEN, the control system operates on the mass flow of the inlet air flow.
The control system changes the input value of current with step of 5 A at regular time
steps of 30 minutes (to obtain nearly constant data).
As already noticed, at the current value of 475 A a reduction of 0.50% of the overall fuel
consumption was imposed, and after the load was increased with the same regular ramp (5 A
every 30 min), up to the value of 495 A.
The process data are averaged out on the whole time interval where the current is
constant. The control system acquires the process data: temperatures, pressures, voltages,
current output, oxidant and fuel flow. The acquisition time is 1 minute, but it is possible to
acquire the data every second. The length of acquisition for the test has been 30 minutes: on
this interval of 30 data the mean value and the variance are evaluated.
In the case of this test session, the fixed variables have been the generator setpoint
temperature, and the overall fuel consumption in two different values. All the other variables
were free to vary, in particular the air flow and the inlet air temperature (Tair,in).
The polarization behavior has been related to the voltage sensitivity of each sector to the
overall fuel consumption (Paragraph 3.1), and this has outlined how the fuel distribution
affects the polarization behavior.
Then, a deeper analysis has been made by performing a parameter estimation, allowing
an analysis of the local anode activation effects with variable local fuel utilization and
temperature: the anode exchange current density of every sector has been the estimated
parameter, and its strict relation to the local fuel utilization of every sector (Paragraph 3.1)
has been outlined.
Experimental Activity on a Large SOFC Generator 81
Finally, the multi-pedigree of the stack arrangement has been characterized by comparing
the different estimated anode exchange current densities of the various sectors of different
pedigree and position inside the generator.
0.67
0.66
C ellvoltage/V
0.65
0.64
0.63
0.62
0.61
2 4 6 8 10 12 14 16 18 20 22 24
Sectors
Figure 5. Distribution in the stack sectors of the voltage of the single cell (I=435 A, setpoint
temperature TGEN=967C, FC=84.25%, ox=4.6)
In order to better understand the discussion about voltage sensitivity tests, some results
are presented about the voltage distribution in the stack [17]. In Figure 5 the distribution in
the stack sectors of the voltage of the single cell is shown for the following operating
condition (I=435 A, setpoint temperature TGEN=967C, FC=84.25%, ox=4.6). The complete
voltage distribution data is not shown due to a non disclosure agreement with Siemens Power
Generation.
The average cell voltage is around 666 14 mV, and the estimated average area
specific resistance is around 1.04 0.08 cm , comparable with the data reported in
2
some literature papers [3,18]. The performance homogeneity (obtained by the estimated
standard deviation) of the single cell voltage is then 2%, which is a very good index of the
operation; during different experimental sessions, values of performance homogeneity of the
single cell voltage have been found up to 5%. It is also evident that the edge sectors show a
lower cell voltage.
82 M. Santarelli, P. Leone, M. Cal et al.
In the next graphs the distribution of the sensitivity of the voltage of the various sectors to the
overall fuel consumption factor is shown: in Figure 6 the voltage sensitivity map is reported,
in Figure 7 the value of the voltage sensitivity of the sectors is divided in Power Leads and
Ejectors side; in Figure 8 the value of the voltage sensitivity of the sectors is divided in North
and South side.
52
48
Bundle row
46
44
42
40
38
South
Pow er leads Ejector
side
Figure 6. Distribution map of the sectors voltage sensitivity to overall fuel consumption.
Figure 7. Sectors voltage sensitivity to overall fuel consumption divided in Power Leads and Ejectors
side.
Experimental Activity on a Large SOFC Generator 83
-44
-46
-48
-50
-52
-54
-56
Board Center
Sectors
Figure 8. Sectors voltage sensitivity to overall fuel consumption divided in North and South side.
In Table 2 the mean value and the standard deviation of the voltage sensitivity of the
sectors to the overall fuel consumption are reported.
Table 1. Mean and standard deviation for the voltage sensitivity of the sectors to overall
fuel consumption divided in total, Power Leads and Ejectors side, North and South side
The mean sensitivity value for the Power Leads side is similar to the one of the Ejectors
side, but in this side the values are more dispersed.
The most evident difference is between the sectors placed at the center and the two
boundaries of the generator: the sectors at the boundaries have a higher voltage sensitivity to
the overall fuel consumption. The different behaviour could be related to the fluid-dynamic
effects caused by the position of the two ejectors: the segments more far from the ejectors
seem to be more sensitive to variation of the overall FC factor, and this is probably linked to
lower local values of fuel utilization.
Another explanation could be linked to the fact that the hydrocarbon fractions (essentially
methane) are not totally converted in the reforming boards. Therefore, part of the reforming
reactions occurs at the beginning of the tubes, that is at the bottom plane. This causes a local
cooling effect. A variation of the overall FC factor determines a modification of the chemical
equilibrium inside the reformer boards, and as a consequence a modification of the cooling
effect at the bottom plane: when increasing the overall fuel consumption, there is a decrease
84 M. Santarelli, P. Leone, M. Cal et al.
of the residual reforming reaction at the bottom plane of the tube; because this reaction is
endothermic, this causes a increase of the local temperature. This local overheating causes a
current density gradient along the tube, with an increase of overpotentials and, as a final
consequence, this fact can reduce the voltage of the segment. This is emphasized at the
boundary rows, where probably the reforming reactions are reduced (the reformer boards
placed at the external boundaries of the generator receive the water vapour rich-fuel gas
through orifices with lower diameter, while the reformers placed between the internal bundle
rows receive the fuel gas through orifices with higher diameter): thus the voltage reduction
effect is more evident on this part of the generator.
The South side is more sensitive and the values are more dispersed than in the North side.
This is probably due to the fuel gas distribution in the generator rows and to the operation of
the two ejectors: probably in the South side a minor amount of fuel gas arrives (especially at
the boundary), and the operation is characterized by a local FC factor higher than the nominal
one. For this reason also lower voltages in these regions are detected.
As a final quantitative remark, the increase of a percentage point of the overall fuel
consumption produces a decrease of around 4.5% in the voltage of the sectors.
In the next graphs the distribution of the sensitivity of the voltage of the various sectors to the
air stoichs factor is shown: in Figure 9 the voltage sensitivity map; in Figure 10 the value of
the voltage sensitivity of the sectors divided in Power Leads and Ejectors side; in Figure 11
the value of the voltage sensitivity of the sectors divided in North and South side.
28
26
24
Voltage sensitivity [m V/0.1 stoichs]
22
20
Bundle row
18
16
14
12
10
South
Pow er leads Ejector
side
-20
-25
-30
-35
1 2 3 4 5 6 7 8 9 10 11 12
R ow s
Figure 10. Sectors voltage sensitivity to air stoichs divided in Power Leads and Ejectors side.
-10
Voltage sensitivity [m V/0.1 stoichs]
-15
-20
-25
-30
-35
Board C enter
Sectors
Figure 11. Sectors voltage sensitivity to air stoichs divided in North and South side.
In Table 3 the mean value and the standard deviation of the voltage sensitivity of the
sectors to the air stoichs are reported.
86 M. Santarelli, P. Leone, M. Cal et al.
Table 3. Mean and standard deviation for the voltage sensitivity of the sectors to air
stoichs divided in total, Power Leads and Ejectors side, North and South side
The Power Leads side is more sensitive and the values are less dispersed than at the
Ejectors side. The North side is more sensitive and the values are more dispersed than in the
South side.
From the graphs, the center rows result to be the more sensitive to a variation of the
factor; this can be linked to the fact that the air is introduced from the piping in the center of a
plenum over the stack, and next it is sent to the sectors: in this way, the central sectors receive
a greater flow whereas the edge sectors receive less air. This effect is less pronounced in the
Power Leads side.
-20
FC sensitivity m V/% FC
-30
-40
-50
-60
2 4 6 8 10 12 14 16 18 20 22 24
Sectors
Figure 12. Voltage sensitivity of the sectors to overall fuel consumption and air stoichs.
There are five sectors where the sensitivity is not significant: considering Figure 7, these
sectors are characterized by a high sensitivity to the overall fuel consumption, which tends to
predominate on the other experimental factors such as the air stoichs.
The average sector sensitivity is around 200 mV/stoichs. The increase of the air
stoichiometry ox leads to a significant decrease of voltage in all the sectors (except for the
ones placed at the generator edges: but we have evaluated that the effect is statistically not
significant). Note that the central zone of the generator shows a maximum of sensitivity of
Experimental Activity on a Large SOFC Generator 87
voltage to air stoichiometry: it is interesting to notice the parabolic shape of the sensitivity
distribution along the bundle rows and the complementary behaviour between the voltage
sensitivity to air stoichs and overall fuel consumption.
As the SOFC CHP-100 is an air-cooled system, the decrease of voltage, due to an
increase of air stoichiometry, is mainly due to the variation of the equilibrium temperature: an
increase of ox causes a reduction of the equilibrium temperature with consequent increase of
overvoltages (mainly ohmic contribution). This decrease of the temperature is almost uniform
in all the bundle rows of the generator, then also at the edges. Thus, it would be expected that
also the edge sectors will present the same behaviour, that is a reduction of voltage with
increasing of ox. The negligible effect of this operation on these sectors is explained with the
observation that the local fuel consumption is the dominant factor of the operation in the edge
sectors.
In Figure 12 the graph of the voltage sensitivity of the various sectors to overall fuel
consumption and air stoichs is shown. The values of the voltage sensitivity to air stoichs have
been divided by 10 to better show and compare the results.
45
40
35
B undle row
30
not significant
25
20
15
10
5
South South
Powerleads Ejector Powerleads Ejector
Side Side
Figure 13. Distribution map of the sectors voltage sensitivity to overall fuel consumption and air
stoichs.
that the air stoichs control factor has a thermal effect. Their thermal effect is contrary,
because an increase of the setpoint temperature means an increase of the local temperature
and consequently of the local voltage, while an increase of air stoichs means a decrease of the
local temperature and consequently of the local voltage. Concerning the distribution inside
the generator, it seems that the cooling effect is more effective in the center of the generator
compared to the boundaries, and this could be linked to the distribution of the cooling air, and
to the fact that the center is the zone with the highest average temperatures due to the lower
heat transfer to the outside.
The effect of the overall fuel consumption is not properly a thermal effect, but especially
a fluid-dynamic effect which influences the Nernst contribution to the polarization. It is very
evident at the boundaries of the generator, where we have observed (see [19]) that the fuel
distribution is not optimised and the reforming reactions in the reformer boards have a lower
kinetic.
It could be interesting to analyze the fact that the boundary rows (1 and 12) have the
reformer unit on the external side working with less fuel gas flow (the reformer boards placed
at the external boundaries of the generator receive the water vapour rich-fuel gas through
orifices with a lower diameter, while the reformers placed between the internal bundle rows
receive the fuel gas through orifices with a higher diameter); this could mean that, in
particular, the external sides of the rows 1 and 12 are not optimally cooled by the endothermic
reforming reaction. This probably could affect the values of temperature and voltage
associated to the sectors at the boundary zones.
Voltage sensitivity ofthe segm ents placed Voltage sensitivity ofthe segm ents placed
atthe PowerLeads side atthe Ejectors side
-15 0
not
-20 significant
-10
O sensitivity m V /0.1S toichs
-25
F C sensitivity m V /% F C
-20
-30
FC
FC reg
-35 -30
Stoichs/10
Stoichs reg
-40
-40
-45
-50
-50
-55 -60
2 4 6 8 10 12 2 4 6 8 10 12
R ows R ows
Figure 14. Sectors voltage sensitivity to overall fuel consumption and air stoichs divided in Power
Leads and Ejector side.
Where the voltage sensitivity to overall fuel consumption is higher the voltage sensitivity
to air stoichs is not significant: in these sectors (at the boundaries of the generator) the effect
of the overall fuel consumption is very high and the other factors do not have a significant
Experimental Activity on a Large SOFC Generator 89
influence on the voltage. In particular, these sectors are close to the ejectors, which have a big
influence on fuel distribution.
It is possible to notice the symmetry between the distribution of the voltage sensitivity to
overall fuel consumption and air stoichs: where the sensitivity to overall fuel consumption is
high the sensitivity to air stoichs is low and vice versa. Exploiting this sensitivity distribution,
it could be possible to think at a regulation strategy in order to uniform the sectors voltage
changing these two regulation factors: during a regulation of one factor, the other one could
be regulated symmetrically to maintain a uniform distribution of voltage.
In Figure 14 the sectors voltage sensitivity is divided in Power Leads and Ejectors side,
while in Figure 15 the sectors voltage sensitivity is divided in North and South side.
Voltage sensitivity ofthe segm ents placed Voltage sensitivity ofthe segm ents placed
atthe South side atthe North side
0 -10
FC
-15 FC reg
-10
Stoichs/10
O sensitivity m V /0.1S toichs
-20
-25
-30 -30
-35
-40
-40
-50
-45
-60 -50
2 4 6 8 10 12 2 4 6 8 10 12
R ows R ows
Figure 15. Sectors voltage sensitivity to overall fuel consumption and air stoichs divided in North and
South side.
The voltage sensitivity distribution to the factors for the North and South side is nearly
the same: every part of the generator seems to have the same trend and it seems that there
isnt any difference to address to fluid-dynamics process of fuel or air.
A slightly different behaviour can be seen considering the Power Leads and Ejectors side:
at the Ejectors side the trends seem emphasized compared to the Power Leads side, and this
could be linked to fluid-dynamic considerations, especially concerning the fuel distribution
coming from the ejectors. In particular, the Ejectors side is more sensitive to overall FC
variation. This could be linked, as already noticed, to the fact that the hydrocarbon fractions
(essentially methane) are not totally converted in the reforming boards. Therefore, part of the
reforming reactions occurs at the beginning of the tubes, that is at the bottom plane. This
causes a local cooling effect. A variation of the overall FC factor determines a modification
of the chemical equilibrium inside the reformer boards, and as a consequence a modification
of the cooling effect at the bottom plane: when increasing the overall fuel consumption, there
90 M. Santarelli, P. Leone, M. Cal et al.
is a decrease of the residual reforming reaction at the bottom plane of the tube; because this
reaction is endothermic, this causes a increase of the local temperature. This local overheating
causes a current density gradient along the tube, with an increase of overpotentials and, as a
final consequence, this fact can reduce the voltage of the segment. This is emphasized at the
boundary rows, where the reforming reactions are reduced (the reformer boards placed at the
external boundaries of the generator receive the water vapour rich-fuel gas through orifices
with lower diameter, while the reformers placed between the internal bundle rows receive the
fuel gas through orifices with higher diameter), and therefore the cooling effect is more
evident on this part of the generator.
Finally, another difference from Power Leads and Ejectors side is the sensitivity to air
stoichs: at the Ejectors side at the boundaries they show a not significant effect; the same
sectors are the ones with the highest sensitivity to overall FC, which could represent a
predominant effect. Moreover, the center rows result to be the most sensitive to a variation of
the air stoichs; this can be linked to the fact that the air is introduced from the piping in the
center of a plenum over the stack, and next it is sent to the sectors: in this way, the central
sectors receive a greater air flow whereas the edge sectors receive less air. This effect is less
pronounced in the Power Leads side.
Figure 16. Voltage sensitivity of the single cells to the factors, along the sectors.
Looking at the Figure 16, it is interesting to notice the distribution along the sectors of the
voltage sensitivity of the single cell to both overall fuel consumption and air stoichs, outlined
for cells placed at the Ejectors and Power Leads sides.
The estimation of the single cell sensitivity to overall fuel consumption is around -2.8
mVcell/%FC, comparable with values of some literature papers [16]; the homogeneity of the
overall FC sensitivity among the various cells is around 10%. The single cell sensitivity to air
Experimental Activity on a Large SOFC Generator 91
stoichs has been estimated in around -1.3 mVcell/0.1stoichs. Also from this figure it is clear
that where the overall FC sensitivity is higher the sensitivity to the air stoichs is lower.
In Figure 17 the sensitivity to air stoichs and overall fuel consumption are correlated,
while in Figure 18 and 19 we have considered, for a deeper analysis, also the correlations
with the generator setpoint temperature (Figure 18: correlation between the sensitivity to air
stoichs and generator setpoint temperature; Figure 19: correlation between the sensitivity to
overall fuel consumption and generator setpoint temperature).
Voltage sensitivity
-5
V oltage sensitivity to air stoichs [m V /0.1 stoichs]
3
not significant values
2 6
-10 23 22 of sensitivity to
stoichs
-15
8 4
7
-20 1 5 19
1211
20
-25 24 9 13
18
21
1015 16
-30 14
17
-35
-56 -54 -52 -50 -48 -46 -44 -42 -40 -38 -36
Voltage sensitivity to fuelconsum ption [m V/% FC ]
Figure 17. Correlation between voltage sensitivity to air stoichs and to overall fuel consumption.
Voltage sensitivity
-5
V oltage sensitivity to air stoichs [m V /0.1 stoichs]
3
not significant values
6 2 of sensitivity to
-10 23
22
stoichs
-15
not significant
8 4
values of 7
-20 sensitivity to 1 5 19
setpoint 11 12
temperature 20
-25 913 24
18
21
101615
-30 14
17
-35
-3 -2 -1 0 1 2 3 4 5 6
Voltage sensitivity to setpointtem perature [m V/C ]
Figure 18. Correlation between voltage sensitivity to air stoichs and to generator setpoint temperature.
92 M. Santarelli, P. Leone, M. Cal et al.
Voltage sensitivity
1 4 not significant
8
values of
0 sensitivity to
23 22 setpoint
-1
temperature
2
-2
6
3
-3
-56 -54 -52 -50 -48 -46 -44 -42 -40 -38 -36
Voltage sensitivity to fuelconsum ption [m V/% FC ]
Figure 19. Correlation between voltage sensitivity to overall fuel consumption and generator setpoint
temperature.
As expected, it seems to occur a linear correlation between the voltage sensitivity to air
stoichs and to generator setpoint temperature (Figure 18). It is a direct correlation (same
trend), because both cause a thermal effect on the generator.
Concerning the correlation between the other two factors, the correlation seems very
weak, and it appears as an inverse correlation.
Concerning the sectors 2, 3, 6, 22 and 23, where the voltage sensitivity to air stoichs is
not significant, they appears also to be the only segments with negative sensitivity to the
generator setpoint temperature. In particular, they have the highest value of sensitivity to
overall fuel consumption. As already commented, in these sectors (at the boundaries of the
generator) the effect of the overall fuel consumption is very high and the other factors do not
have a significant influence on the voltage. In particular, these sectors are close to the
ejectors, which have a big influence on the fuel distribution.
utilization of the sectors is estimated coupling the experimental voltage sensitivity to overall
fuel consumption tests with a model of cell voltage sensitivity to fuel utilization.
In order to deduce the local fuel utilization of the various sectors, a model of the single
cell voltage sensitivity to FU has been developed. Each term of the polarization model is
discussed and then some hypothesis are assumed for the particular case study of a tubular
cathode supported cell (used in the SOGC CHP-100 plant).
5.1. Analytical Expression of the Sensitivity of the Cell Voltage to the Fuel
Utilization
The sensitivity of the cell voltage to the fuel utilization depends on several contributions: (1)
the Nernst potential; (2) the activation overpotential; (3) the diffusion overpotential; (4)
eventually, the effect of leakages of air at the anode side. Conversely, a modification of the
fuel utilization does not cause any effect on the ohmic overpotential inside the cell layers, and
therefore the ohmic overpotential is not affected by the fuel utilization term.
The cell voltage of a tubular SOFC is expressed by the equation:
where Vc is the cell voltage, V Nernst is the Nernst potential, act ,a|c is the activation
overpotential, ohm is the ohmic overpotential, conc , a|c is the diffusion overpotential.
First, if the performance of a cell is mainly limited by the ohmic contribution and the
other terms of polarization play a marginal role, then the sensitivity of the cell voltage to FU
is well described by the variation of just the Nernst potential with fuel utilization [20].
The Nernst potential is expressed, with the hypothesis of ideal gas, by the following
equation:
R T ( )( )
y H yO 1/ 2
= E0 + ln
( )
2 2
V Nernst (6)
2F y H 2O
Under the assumption that the oxygen utilization is negligible, that is the oxidant is fed to
the cell far in excess of the stoichiometrically required amount, the Nernst potential is a
function of the operating temperature and of the axial location along the cell length (x), at
fixed pressure:
R T ( )( )
y H ( x) y O 1/ 2
V Nernst ( x, T ) = E 0 (T ) + ln
( )
2 2
(7)
2 F y ( x )
H 2O
The average Nernst potential is the x-averaged Nernst potential, that is:
94 M. Santarelli, P. Leone, M. Cal et al.
L
1
L 0
V Nernst = V Nernst ( x) dx (8)
VNernst ( x, T ) = E0 (T ) +
R T
ln
(
y H ( x) yO )( ) 1/ 2
= A x + B
( )
2 2
(9)
2 F y ( x )
H 2O
This refers to an operation with a variation of the reactants partial pressure along the cell
length according to the exponential equation:
1 y ( L)
log i x
L yi ( 0 )
y i ( x ) = y i ( 0) e
(10)
The boundary conditions are assumed by knowing the fuel composition at the cell inlet
(in terms of molar fractions), and the overall fuel utilization.
At x=0 (closed end), that is at the cell inlet, it can be written:
y H 2 (0) = y H0 2
(11)
y H 2O (0) = y 0H O
2
y H 2 ( L) = y H0 2 (1 FU )
(12)
y H 2O ( L) = y H0 2O + y H0 2 FU
The x-averaged Nernst potential at high current densities is thus written in the form:
R T (
y H0 (1 FU ) y O )( ) 1/ 2
V Nernst , HC (T ) = E 0 (T ) + ln
( )
2 2
(13)
4 F y H 2O + y H 2 FU
0 0
At low current densities, a significant length of the cell operates close to the exit Nernst
potential, that is most of the fuel is consumed very close to the fuel inlet (closed end of the
Experimental Activity on a Large SOFC Generator 95
cell). As a consequence, at low current densities the average Nernst potential is approximated
to the exit Nernst potential and maintained constant along the cell tube:
R T ( )( )
y H0 (1 FU ) y O 1/ 2
V Nernst , LC (T ) = E 0 (T ) + ln
( )
2 2
(14)
2 F y 0
+ y 0
FU
H 2O H2
As already introduced, when neglecting the diffusion overpotential and the air leakages,
the sensitivity of the cell voltage to the fuel utilization can be assumed to be equal to the
sensitivity of the Nernst potential to the same parameter. With this assumption it is possible to
write:
dVc dV
= Nernst (15)
dFU dFU
With this assumption, the first derivative of the cell voltage with respect to the fuel
utilization (that is, the voltage sensitivity to the fuel utilization) at low current densities is
twice the same term at high current densities: in fact, at high current density the voltage
sensitivity is given by the equation (16), while at low current density it is given by the
equation (17).
VNernst RT 1
= (16)
FU HC
4F (1 FU ) ( yH 2 O + yH 2 FU )
0 0
VNernst RT 1
= (17)
FU LC
2F (1 FU ) ( yH 2 O + yH 2 FU )
0 0
To refine the model, some considerations about the effect of the fuel utilization on the
activation overpotentials are now discussed.
The anode and cathode activation overpotentials are described by the Butler-Volmer
equation [22]:
a RT c act
F F
act
I = I 0 e e RT
(18)
The exchange current I0 is expressed as a function of the composition of the reactant
gases. In particular, the expression of the anode activation overpotential as a function of H2
and H2O concentration has been carefully analysed, as it could be of significance for the
96 M. Santarelli, P. Leone, M. Cal et al.
simulation of the characteristic curves at high fuel utilization. Due to contradictory data
discussed in literature, it is difficult to choose a model for the anode exchange current density.
The model proposed by Yamamura et al.[23] shows an apparent reaction order near to the
stoichiometry of the anode electrochemical reaction (the resulting polarization resistance is
inversely proportional to the partial pressure of the reactant, and directly proportional to the
partial pressure of the reaction product):
1
Eact , anode
pH p H 2O 2
RT
I 0,anode = anode 2 e
(19)
p p
ref ref
Mogensen et al. [24] investigated Ni/YSZ cermet electrodes at 1273 K and found the
electrode impedance to be formed by a low-frequency contribution, with resistance decreased
by increasing both the partial pressure of hydrogen and water, and a high-frequency
contribution with resistance and capacitance almost independent of the partial pressures of
hydrogen and water. The proposed model does not respect the stoichiometry of the
electrochemical reaction:
E act , anode
pH p H 2O
I 0, anode = anode 2 e RT
(20)
p p
ref ref
Anyway, from both these models it is possible to express the anode activation
overpotential as a function of the local partial pressures of fuel and products, expressing thus
the dependence of the activation overpotential on the fuel utilization factor.
Nevertheless, the activation contribution to the cell voltage sensitivity to FU is neglected
in this part of the analysis because of its small effect of the voltage drop [16,21]. In the next
Paragraph 6, dealing with parameter estimation of some terms of the cell polarization model,
the anode exchange current density will be estimated and a correlation with the local fuel
utilization will be outlined.
Concerning the diffusion overpotential, the main relation with the fuel utilization is present at
the anode side. The model equations and hypothesis are assumed by literature [21,25,26].
Transport of gaseous species usually occurs by binary diffusion, where the effective binary
diffusivity is a function of the fundamental binary diffusivity D H 2 D H 2O and the
microstructural parameters of the anode. In electrode microstructures with very small pore
size, the possible effects of Knudsen diffusion, adsorption/desorption and surface diffusion
may also be present.
The fundamental binary diffusivity is evaluated using the Chapman-Enskog model. This
model describes the diffusion in binary gas mixtures at low to moderate pressures and it
results from solving the Boltzmann equation. The derived working equation is:
Experimental Activity on a Large SOFC Generator 97
3 (4 k T / M AB )1 / 2
D AB = fD (21)
16 n AB 2
D
where, MA, MB are the molecular weights of the chemical species A an B, n is the number
density of molecules in the mixture, k is the Boltzmann constant, T the absolute temperature,
and the term MAB is expressed as:
M AB = 2 [(1 / M A ) + (1 / M B )]
1
(22)
Moreover, D is the collision integral for diffusion (depending on temperature and on the
intermolecular force law between colliding molecules), and AB is a characteristic length.
Finally there is a correction term (fD) which is in the order of unity and particularly lies
between 1 and 1.02 for molecules with comparable molecular weights, while it is between 1
and 1.1 if the molecular masses are very unequal. If fD is chosen as unity and n is expressed
by the ideal gas law, then:
0.00266 T 3 / 2
D AB = (23)
P M 1AB/ 2 AB
2
D
12 6
= 4 (24)
r r
where and are the characteristic Lennard-Jones energy and length, respectively. It is
possible to demonstrate that:
AB = ( A B )1 / 2
A + B (25)
AB =
2
and that the collision integral is only depending upon the term T*=kT/ AB. The relation of
Neufield is:
A C E G
D = * B
+ ( DT * ) + ( F T * ) + ( H T * ) (26)
(T ) e e e
98 M. Santarelli, P. Leone, M. Cal et al.
with:
k T
T* = (27)
AB
The equation (23) is derived for dilute gases consisting of non-polar, spherical,
monoatomic molecules.
In case of mixtures with one or both polar components, a modified Lennard-Jones
relation is used. In this case the collision integral can be evaluated according to the Brokaw
equation [27,28]:
A C E G 0.19 AB
2
D = + + + + (28)
(T * ) B e ( DT ) e ( F T ) e ( H T )
* * *
T*
2 F p Hb 2 ( FU ) Da ( eff )
ias = (29)
R T ta
R T ta
p Ha 2 = p Hb 2 ( FU ) ic (30)
2 F Da ( eff )
where the pressure of fuel at the bulk depends on the fuel utilization.
The effective diffusion coefficients have been evaluated according to the equation:
a/c T 1.5
Da / c ( eff ) = Di293
j
K
( ) (31)
a/c 293
The diffusion overpotential due to the anode diffusion is then given by:
Experimental Activity on a Large SOFC Generator 99
R T i
a
Vdiff = log1 (32)
2 F ias
The tubular cell used in the SOFC CHP-100 is a cathode-supported cell: then, the main
effect of diffusion is mainly addressed to the cathode layer. In terms of physical measurable
parameters, the cathode limiting current density can be evaluated in the form:
4 F pOb 2 Dc ( eff )
ics = (33)
p p Ob 2
R T tc
p
(This equation is an approximation for anode supported cells with very small cathode
thickness). In general, the limiting current density at the cathode is given by imposing to zero
the equation:
R T t C
ic
4 F p Dc ( eff )
p c
O2 = p ( p p ( , FU )) e
b
O2 (34)
where the partial pressure of the oxidant at the bulk depends on fuel utilization and the air
stoichiometry according to the following equation [5]:
FU y H0
y O2 = 2
(35)
FU y H 2
0
y O0 2
The diffusion overvoltage due to the cathode diffusion is then given by:
R T i
c
Vdiff = log1 (36)
4 F ics
In this model, the drop of voltage due to a variation of fuel utilization is modelled also
considering the effect on the cathode diffusion. The effect of diffusion on the cell voltage
sensitivity to fuel utilization has a significant effect when the limiting current density is
approached, or for low value of the air stoichiometry. It should also be noticed that in the
equations (34) and (35), as posed in [3], the variable has not the meaning of excess of
oxidant with respect to the stoichiometric requirement, and so it has a different meaning
compared to the air stoichiometry ox as defined and used in this Chapter.
As a result, it is possible to write, with respect to equations (32) and (36):
100 M. Santarelli, P. Leone, M. Cal et al.
i
Vdiffa ic as
R T 1 FU = R T ic i
= as (37)
FU 2 F i ias2 2 F (ias ic ) ias FU
1 c
ias
ics
V i
R T 1 FU = R T i
c c ic
= cs
diff
(38)
FU 4 F i ics2 4 F (ics ic ) ics FU
1 c
ics
Vdiffa / c 1
(39)
FU Da / C ( eff )
Vdiffa / c
ta / c (40)
FU
The effect of the mass diffusion on the sensitivity of voltage to the fuel utilization is
inversely proportional to the effective binary diffusivity on the electrodes layers and directly
proportional to the electrodes thickness.
The parameters used for solving the equations of the contribution of the diffusion
overpotential to the analytical expression of the sensitivity of the cell voltage to the fuel
utilization are been listed in Table 4:
An important role is played by eventual leakages of air at the anode side; in fact some fuel
can burn with the oxidant with the consequence that it does not react at the anode surface to
generate current. If it would occur, then the effective fuel utilization will be higher than the
expected and the voltage sensitivity to a FU perturbation will be very high, because the
Experimental Activity on a Large SOFC Generator 101
situation approaches the one with high values of fuel utilization. This is the case of the tubular
cathode-supported cells considered in this Chapter, as clearly explained in [16].
5.1.4. Behavior of the Sensitivity of the Cell Voltage to the Fuel Utilization
The results of the model are shown in Figure 20. Two models are considered: (1) the
simplified model (considering just the sensitivity of the Nernst potential to fuel utilization);
(2) the complete model (which accounts also for the effect of mass diffusion and 3% air
leakage).
M odelofvoltage sensitivity to FU
com plete m odel:low current com plete m odel:high current
nernstcontribution:low current nernstcontribution:high current
15
14
13
12
11
dV/dFU (m V/% FU )
10
9
8
7
6
5
4
3
2
1
0
70 75 80 85 90 95
FU (% )
Figure 20. Model of the cell voltage sensitivity to fuel utilization for a tubular cathode-supported SOFC
dV/dFU vs FU
100 m A/cm 2 400 m A/cm 2 m odelapprox m odellow currentapprox
14
12
dV/dFU, mV/%FU
10
0
70 72 74 76 78 80 82 84 86 88 90 92 94 96
FU, %
Figure 21. Experimental data of the cell voltage sensitivity to fuel utilization for a tubular cathode-
supported SOFC.
102 M. Santarelli, P. Leone, M. Cal et al.
The curves are drawn for an operating condition of 1000C and for a fuel with 89% H2
and 11% H2O of composition. In Figure 21 some experimental results (continuous lines) of
single cell voltage sensitivity are reported according to [16].
Figure 22. Experimental data of the cell voltage sensitivity to fuel utilization for a tubular anode-
supported SOFC.
Once the model is developed, it can be used in order to explain the different values of local
sensitivity to fuel utilization and to analyse the shape of the distribution of the local fuel
utilization along the various segments of the SOFC CHP-100 generator.
The values of the voltage sensitivity of each sector to the overall fuel consumption have
been estimated experimentally through the experiments described in the Paragraph 4.
Experimental Activity on a Large SOFC Generator 103
Now, the experimental values of the voltage sensitivity of the various sectors to the
overall fuel consumption can be used together with the analytical model of the sensitivity of
the cell voltage to the fuel utilization, in order to deduce the distribution of the local values of
fuel utilization pertaining to every sector of the generator.
According to the model equations, the voltage sensitivity to the fuel utilization depends
on the local temperature in the sectors and on the fuel inlet conditions at the anode in terms of
hydrogen and water vapour partial pressures. The effect of the local temperature is taken into
account in the analysis, because it has been directly measured on the generator. The effect of
the fuel inlet conditions is taken into account by direct measurement in the generator, but with
the strong hypothesis that the fuel composition is the same for all the sectors of the generator,
that is there is the same reformation degree in all the in-stack standard reformation boards.
The condition of the inlet fuel is: yH2=0.4 and yH2O=0.2.
Then, the local fuel utilization can be estimated by solving the voltage sensitivity model
in terms of fuel utilization.
-3
x 10
9
Theoreti
calcurve
8 Experim entaldata
% FC]
7
i
uelconsumpt mV/
on[
5
vi
i yt
t of
4
t
agesensi
3
t
vol
1
0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95
Fuelconsum pti
on
Figure 23. Graph of the theoretical curve (200 mA/cm2) and the experimental measurements of local
voltage sensitivity to the overall fuel consumption
A graph with the curve of the voltage sensitivity to the fuel utilization, obtained by the
model, and expressed as a function of the overall fuel consumption, can be developed. In this
graph, the measurements of the voltage sensitivity of every sector is inserted. In this way, the
local values of fuel utilization are deduced from the graph. In Figure 23 the graph of the
theoretical curve (for a current density of 200 mA/cm2, that is the experimental conditions of
the generator) and the experimental measurements of local voltage sensitivity to the overall
fuel consumption is shown. From the curve, the local fuel utilization values are deduced.
104 M. Santarelli, P. Leone, M. Cal et al.
82 0
80 -0.0005
78 -0.001
76
Utilization %
FUSensitivity/
-0.0015
Local Fuel
V/%FU
74
-0.002
72
-0.0025
70
68 -0.003
66 -0.0035
64 -0.004
1 2 3 4 5 6 7 8 9 101112131415161718192021222324
cellsector
Figure 24. Estimation of the distribution of the local values of fuel utilization inside the generator.
6. Polarization Analysis
In the experimental session, a polarization analysis was performed in order to investigate the
electrical behavior of the sectors of the generator.
The analysis was performed during real operating conditions of the generator, thus only
short-range V-I characteristics were measured. The current was changed between 435A and
475A, at fixed overall fuel consumption of 84.25% and setpoint temperature of 967C.
Experimental Activity on a Large SOFC Generator 105
ASR cm2
The polarization behavior has shown significant differences among the various sectors of
the generator: different slopes in the current range imposed for the polarization. As an
example, in Figure 25 the polarization curves of two sectors are shown. The polarization
curve has been detected for all the 24 sectors of the generator.
As a first approximation, these slopes can be associated to the concept of Area Specific
Resistance (ASR).
The polarization analysis consisted in the fitting of the experimental V-I curves with a
proposed polarization model. The model has been already presented in equation (5).
The Nernst potential is expressed by the equation (6) and evaluated according to equations (7)
and (10).
The anode exchange current density has been considered as a fitting parameter. For the
estimation of its first attempt value in the fitting analysis, the equation (19) has been
considered.
The parameters used in equations (18), (19) and (41) have been assumed by the current
literature: a=8.2 107 [mA/cm2] (first attempt value in the fitting analysis); c=1.8 107
[mA/cm2]; Eact,anode=120000 [J/mol]; Eact,cathode=120000 [J/mol]; a,anode=2, c,anode=1,
a,cathode=1.4, a,cathode=0.6.
Despite the ionic resistance of the electrolyte plays an important role in the cell ohmic
resistance, it is the contribution of the anode and cathode resistance which determines the
high ohmic resistance of the cell: this is due to the long path of the electrons in the electrodes
[29] (high in-plane resistance). The equivalent electrical circuit has been assumed by the
current literature [30,31] and the ohmic overpotential has been evaluated according to the
equations (42) and (43):
R j = j path j (43)
160
Resistivity/ohm * cm
140
120
100
80
60
40
20
0
700 750 800 850 900 950 1000
Temperature/C
At 1000C the ohmic area specific resistance of tubular cathode-supported cells is around
0.40 cm2; for an HPD5 cell is around 0.25 cm2; for an HPD10 cell is around 0.15 cm2;
for a Delta 9 cell is around 0.18 cm2 [32]. In Figure 26 the ionic resistance of 8YSZ is
drawn as a function of temperature considering the main models used in literature [22,30,31]
and a comparison with experimental values [33].
Equations for modelling the resistivities of the main cells layers are listed in the
following and have been assumed by the literature [3, 21,22,26,30,31,32,33,34]:
1392
T
cat = 0.00298 e cell
(44)
600
T
an = 0.044 e cell
(45)
10350
T
el = 0.00294 e cell
(46)
4690
T
int = 0.1256 e cell
(47)
The description of the materials used in these cells is largely available in literature
[32,35].
The diffusion overpotentials are described according to equations (32) and (36). The
fundamental binary diffusivities have been calculated using the previously described Brokaw
model and the values of microscopic parameters have been assumed by literature and are
listed in Table 4.
The parameter estimation procedure involved the anode exchange current density, considered
in the expression of the anode activation overpotential (equation (18)). As already noticed, in
literature there are some expressions which describe the anode exchange current (equations
(19) and (20): according to these equations, this parameter should be significantly affected by
the fuel utilization values, and the aim is to outline how the distribution of the local fuel
utilization (that is, the distribution of fuel) inside the generator affects the activation of the
reaction at the anode side in the various sectors.
The procedure of parameter estimation consisted in the evaluation of the values of the
anode exchange current density in each sector of the generator, in order to obtain its
distribution inside the stack. In the estimation, the local values of fuel utilization evaluated in
108 M. Santarelli, P. Leone, M. Cal et al.
the previous Paragraph 5, and the measured local values of temperature, have been associated
to each sector.
200
150
100
Powerlead
Ejector
50
1 2 3 4 5 6 7 8 9 10 11 12
Sector
Figure 27. Distribution of the estimated values of the anode exchange current density.
In Figure 27 the distribution of the anode exchange current density (i0a) of every sector is
shown:
95% confidence region for estimated value of anodic exchange current density
300
Lower limit
Anode axchange current density [m A/cm2 ]
Calculated value
250 Upper limit
200
150
100
50
0
0 5 10 15 20 25
Sector
Figure 28. 95% confidence region for the anode exchange current density estimation.
Experimental Activity on a Large SOFC Generator 109
It is possible to notice a symmetric trend centered on the row 6 for the Power Leads side
and on the row 7 for the Ejectors side. The behaviour for the two side is nearly the same:
lower value of anode exchange current density at the boundary rows of the generator, and
higher values for the center sectors.
This behaviour is the direct and logic consequence of the distribution of fuel utilization
inside the generator. There are high values of fuel utilization at the boundaries of the
generator while at the center of the generator the values of fuel utilization are lower.
Therefore, the activation of the anode reaction is favoured at the center rows, while the
kinetic is reduced at the boundary rows. This behaviour is expressed by the values of the
anode exchange current density.
In Figure 28 the 95% confidence region for the anode exchange current density
estimation is shown.
The 95% confidence region is very small and the error on the estimated value is very low
(below 2%). The regression produces a very good estimate of the anode exchange current
density.
In Figure 29 the correlation between the local anode exchange current density and the
local fuel utilization (as usual, divided in Power Leads and Ejectors side) is shown.
As shown, the distribution of the anode exchange current density is the direct and logic
consequence of the distribution of fuel utilization inside the generator. There are high values
of fuel utilization at the boundaries of the generator, and therefore the kinetic is reduced at the
boundary rows: this behaviour is expressed by the lower values of the anode exchange current
density.
C orrelation between anode exchange currentdensity C orrelation between anode exchange currentdensity
95%andconffueli
dence
ut region
ilization power for
leads estim ated value ofanodicand
side exchange cur
fuelutilization rent
eject densi
orsi
de ty
300 0.85 300 0.85
300
Lowerlim it
current density [m A /cm 2]
240 240
0.82 0.82
200
220 220
0.81 0.81
F uelutilization
F uelutilization
200 200
0.8 0.8
150
180 180
0.79 0.79
A node Aaxchange
160 160
100 0.78 0.78
140 140
0.77 0.77
120 120
50
100 0.76 100 0.76
80 0.75 80 0.75
0 2 4 6 8 10 12 2 4 6 8 10 12
0 5 or
Sect 10 15 20
Sect
or 25
Sector
Figure 29. Correlation between the local anode exchange current density and the local fuel utilization.
110 M. Santarelli, P. Leone, M. Cal et al.
C orrelation between anode exchange currentdensity C orrelation between anode exchange currentdensity
and fuelutilization powerleads side and fuelutilization ejectorside
300 300
280 280
260 260
A node exchange currentdensity [m A /cm2]
220 220
T em perature [C ]
T em perature [C ]
200 200
900 900
180 180
160 160
140 140
120 120
100 100
80 80
2 4 6 8 10 12 2 4 6 8 10 12
Sector Sector
Figure 30. Correlation between the local anode exchange current density and the local temperature.
The anode exchange current density depends not only on the local fuel utilization, but it
is also influenced by the local temperature. A higher temperature should have a increasing
effect on the anode exchange current density. Therefore, in Figure 30 the correlation between
the local anode exchange current density and the local temperature (as usual, divided in
Power Leads and Ejectors side) is shown. It seems that the correlation with the local
temperature is less evident than the correlation with the local fuel utilization, especially at the
Ejectors side (fuel inlet). Thus, it seems that the effect of the local fuel utilization is more
significant than the effect of the local temperature on the activation of the anodic reaction.
The distribution of temperature and fuel utilization is quite similar: where the temperature
is high (at the boundary rows of the generator) the fuel utilization is high too. This behaviour
can compensate the effect of temperature (positive) and fuel consumption (negative) on the
anode exchange current density. It can be observed that the temperature effect attenuates the
dependence of the anode exchange current density from the fuel utilization (in fact, the
distribution of temperature is symmetric compared to the fuel utilization distribution): the
negative activation effect at the boundary rows due to high fuel utilization is reduced by
higher local values of temperature, and the positive activation effect at the center (low fuel
utilization) is reduced too by lower local values of temperature.
As a conclusion, the effect of fuel utilization distribution on the anode exchange current
density is predominating, while the temperature distribution attenuates this behaviour.
Experimental Activity on a Large SOFC Generator 111
50 50
40 40
percentualloss [% ]
P ercentualloss
30 30
activation
concentration
ohm ic
20 20
10 10
0 0
5 10 15 20 activation concentration ohm ic
Sector
Figure 31. Values of the overpotentials for every sector at about 180 mA/cm2.
Based on the proposed electrochemical model, the percentage values of the overpotential
contributions of the considered electrochemical model (with the anode activation
overpotential calculated using the estimated values of the anode exchange current density) at
about 180 mA/cm2 are shown. In Figure 31 the percentage values of the overvoltages for
every sector at about 180 mA/cm2 are shown.
As expected, it is possible to notice that the activation overpotential and the ohmic
overpotential are the main losses factor (their sum is about the 90% of the losses), while the
concentration (diffusion) overpotential is less relevant in the low current range considered for
the experiment.
In particular, the activation overpotentials are quite high: this is a consequence of the
already highlighted not optimal fuel distribution inside the generator, with high values of
local fuel utilization especially at the boundary rows of the stack.
In Figure 32 the percentage weight of the overpotential contributions at 180 mA/cm2 are
shown. A comparison is made: (1) contribution evaluated with the value of anode exchange
current density find in literature; (2) contribution evaluated with the value of anode exchange
current density estimated by the regression procedure on the SOFC CHP-100 in two different
sectors of the generator (the highest and the lowest values of anode exchange current density
are considered).
112 M. Santarelli, P. Leone, M. Cal et al.
100
90
Acti
vation
80 Ohm ic
Concentrati
on
70
60
oss
uall
cent
50
Per
40
30
20
10
0
Li
terature I
_0_,_a=295.15 [m A/cm ^2] I
_0_,_a=91.75 [m A/cm ^2]
Figure 32. Percentage weight of the overvoltage contributions at 180 mA/cm2 (evaluated for the value
of anode exchange current density find in literature and the values find by the regression on the SOFC
CHP-100).
200 200
18 19
20 3
11
180 13 4 6 180
8
160 160
140 15 12 5 140
1621 7 9
10 2
120 14 120
17
1 23
100 100
80 80
24
60 60
900 0-10k h 10k-16k h over16k h
Tem perature [C ] R unning hours
Figure 33. Correlation between the cell anode exchange current density and its pedigree.
Experimental Activity on a Large SOFC Generator 113
It is possible to notice that the results of the regression procedure on the SOFC CHP-100
determine lower values of the activation overpotential and higher values of the ohmic and
diffusion overpotentials, compared to the results obtained with literature data.
Probably, it could be due to the fact that the literature equations are proposed for planar
cells, and their behaviour is towards an underestimation of the anode exchange current
density, with the consequence of overestimating the activation losses.
Finally, in Figure 33 a correlation between the anode exchange current density and the
local temperature (grouped by sector running hours), and the mean values for the running
hours (the cell pedigree), are shown.
As a general comment, it can be noticed that in the new sectors the anode exchange
current density is lower compared to the old sectors. Therefore, it is possible to conclude that
the effect of the position of the sector inside the generator predominates on its pedigree: in
fact, the boundary sectors (sectors 1 and 24) have a very low anode exchange current density
compared to other sectors with the same pedigree disposed in different positions in the
generator.
7. Conclusion
The Chapter dealt with the description of the experimental analysis of the CHP-100 SOFC
Field Unit built by Siemens Power Generation Stationary Fuel Cells (SPG-SFC).
The aim was to illustrate a methodology of experimental analysis of a real large SOFC
generator in operation, with the complexity linked to the large number of variables which
affect its operation, the limited number of measurements points in the generator volume, the
necessity to avoid malfunctions in the real operation.
As a consequence, the experimental analysis of the CHP-100 SOFC Field Unit has been
developed with methods of Design of Experiments, and with a statistical analysis of the
collected data.
The following conclusions can be outlined:
the design of experiments procedure is a powerful tool to study in a rigorous manner the
collected experimental data and also to relate, through analytical expressions, the
investigated dependent variables with the control factors (independent variables); the
design of experiments approach allows one to analyse the significance of the main and
interaction effects of the factors;
the experimental analysis of the SOFC generator is presented in terms of sensitivity
analysis of the local voltages with respect to the overall fuel consumption and the air
stoichiometry control factors;
the use of local voltage sensitivity tests to the overall fuel consumption, coupled with the
analytical model of the sensitivity of the cell voltage to the fuel utilization, is a useful
procedure to estimate the local fuel utilization of the various sectors of a real large
generator during its operation: the indication of the fuel distribution inside the generator
volume has therefore been found;
high local fuel utilization have been detected: they can lead to local overheating which
can limit the operation (power density, efficiency and degradation) of the generator;
114 M. Santarelli, P. Leone, M. Cal et al.
Nomenclature
BoP Balance of Plant
CHP Cogeneration Heat and Power
Di-j fundamental binary diffusivity in the electrode (cm2/s)
Da(eff) effective binary diffusivity in the anode layer (cm2/s)
Dc(eff) effective binary diffusivity in the cathode layer (cm2/s)
Ea activation energy of reaction (J mol-1)
E0 open circuit voltage (V)
F Faraday number (C mol-1)
FC Fuel Consumption (%)
FU Fuel Utilization (%)
Gair ,tot air mass flow (kg s-1)
Gair ,ech electrochemical air mass flow utilization (kg s-1)
G f ,ech
electrochemical fuel mass flow utilization (kg s-1)
G f ,tot fuel mass flow (kg s-1)
HC High Current
HPD High Power Density
I generator current (A)
I0 exchange current density (mA/cm2)
I 0,anode anode exchange current density (mA/cm2)
I 0,cathode anode exchange current density (mA/cm2)
ias anode limiting current (A cm-2)
ics cathode limiting current (A cm-2)
ic cell current density (A cm-2)
k Boltzmann constant (J K-1)
L cell length (m)
LC Low Current
Mi molecular weights of the chemical species (g mol-1)
n number density of molecules in the mixture
p stack pressure (Pa)
p Ha 2 hydrogen pressure at the anode/electrolyte interface (Pa)
pOa 2
oxygen pressure at the cathode/electrolyte interface (Pa)
Experimental Activity on a Large SOFC Generator 115
pOb 2
oxygen pressure at the cathode bulk (Pa)
R universal gas constant (J mol-1 K-1)
Rj cell ohmic resistivity ( cm2)
SOFC Solid Oxide Fuel Cell
T temperature (K)
TGEN setpoint temperature (K)
Tair air pre-heating temperature (K)
ta thickness of the anode layer (cm)
tc thickness of the cathode layer (cm)
Vc cell voltage (V)
Vdiff diffusion overpotential (V)
V Nernst Nernst potential (V)
V Nernst x-averaged Nernst potential (V)
V Nernst , HC
average Nernst potential, high current approximation (V)
V Nernst , LC average Nernst potential, low current approximation (V)
x axial position along the tubular cell (m)
xi regressor variables
Yi dependent variable
yi molar fraction of reactants at the electrodes bulk
y H0 2 hydrogen molar fraction at fuel inlet condition
Greek
i regression coefficient
pre-exponential coefficient in activation polarization equation (A cm-2)
electrode porosity
act ,a|c
activation overpotential at anode and cathode (V)
conc ,a|c concentration (diffusion) overpotential anode and cathode (V)
ohm ohmic overpotential (V)
transfer coefficient in activation polarization equation
air stoichiometric excess of oxidant
j ohmic resistivity of materials ( cm)
AB characteristic length of Chapman-Enskog model ()
D collision integral for diffusion
electrode tortuosity
116 M. Santarelli, P. Leone, M. Cal et al.
A1.1. Introduction
In many problems two or more variables are related and it is of interest to model and explore
this relationship.
Suppose that there is a single dependent variable or response Y that depends on k
independent or regressor variables, for example x1 , x2 ,..., xk . The relationship between
these variables is characterized by a mathematical model, called a regression model, that is fit
to a set of sample data. In some instances, the experimenter knows the exact form of the true
functional relationship between Y and x1 , x2 ,..., xk , say Y = ( x1 , x 2 ,..., xk ) .
However, in most cases, the true functional relationship is unknown and the experimenter
chooses an appropriate function to approximate . Low-order polynomial models are widely
used as approximating functions.
In general, the response variable Y may be related to k regressor variables. The model:
Y = 0 + 1 x1 + 2 x2 + ... + k xk + (A1)
is called a multiple linear regression model with k regressor variables. The parameters j
( j = 0 ,1,..., k ) are called the regression coefficients. This model describes a hyperplane in
the k-dimensional space of the regressor variables. The parameter j represents the expected
change in response Y per unit change in x j when all the remaining independent variables xi
(i j ) are held constant.
A1.2. Estimation of the Parameters in Linear Regression Models
The method of least squares is typically used to estimate the regression coefficients in a
multiple linear regression model. Suppose that n > k observations on the response variable
are available, say Y1 , Y2 ,..., Yn . Along with each observed response Yi , an observation on
each regressor variable is available; let xij denote the i-th observation or level of variable x j .
Y x1 x2 ... xk
Y1 x11 x12 x1k
Y2 x21 x22 x2 k
Yn xn1 xn 2 xnk
Experimental Activity on a Large SOFC Generator 117
The model equation (A1) can be written in terms of the observations as:
Yi = 0 + 1 x i1 + 2 x i 2 + ... + k x ik + i
k (A2)
Yi = 0 + j x ij + i i = 1, 2,..., n
j =1
Assume that the error term i in the model has E ( i ) = 0 and V ( i ) = 2 . The method
of least squares chooses the coefficients j in equation (A2) so that the sum of the squares of
the errors i is minimized. The least squares function is:
2
n k n
L = = Yi 0 j xij
i
2
(A3)
i =1 i =1 j =1
n
L k
= 2 Yi 0 j xij = 0 (A4a)
0 0 , 1 ,..., k i =1 j =1
n
L k
= 2 Yi 0 j xij xij = 0 j = 1,2,...k (A4b)
j i =1 j =1
0 , 1 ,..., k
n n n n
n 0 + 1 x i1 + 2 x i 2 + ... + k x ik = Yi
i =1 i =1 i =1 i =1
n n n n n
0 x i1 + 1 x i21 + 2 x i1 x i 2 + ... + k x i1 x ik = x i1Yi
i =1 i =1 i =1 i =1 i =1
The equations (A5) are called the least squares normal equations. Note that there are
p = k + 1 normal equations, one for each of the unknown regression coefficients. The
118 M. Santarelli, P. Leone, M. Cal et al.
solution to the normal equations will be the least squares estimators of the regression
coefficients 0 , 1 ,..., k .
The normal equations are solved in matrix notation. The model in terms of the
observations (A2) may be written in matrix notation as:
Y = X +
where:
L = i2 = ' = (Y X )'(Y X )
n
i =1
L = Y 'Y ' X 'Y Y ' X + ' X ' X = Y 'Y 2 ' X 'Y + ' X ' X (A6)
because ' X 'Y is a (1 1) matrix, or a scalar, and its transpose ( ' X 'Y )' = ' X 'Y is the
same scalar.
The least squares estimators must satisfy:
L
= 2 X 'Y + 2 X ' X = 0
Equation (A7) is the matrix form of the least squares normal equations. To solve the
normal equations, multiply both sides of equation (A7) by the inverse of X ' X . Thus, the
least squares estimator of is:
= (X ' X ) X 'Y
1
(A8)
It is easy to see that the matrix form of the normal equations is identical to the scalar
form. Writing out equation (A7) in detail, the following form is obtained:
n n n
n
n x x ... x yi
0 ni =1
i1 i2 ik
n i =1
n n
i =1
n
i =1
x ... xi1 xik . 1 = xi1 yi
x x
2
x
i =1
i1
i =1
i1
i =1
i1 i 2
i =1
i =1
#
# # # ... # #
n n
... xik2 k xik yi
n n n
xik x x i1 x xi 2
i =1 i =1
ik ik
i =1 i =1 i =1
If the indicated matrix multiplication is performed, the scalar form of the normal
equations (A5) will result. In this form it is easy to see that X ' X is a ( p p ) symmetric
matrix and X 'Y is a ( p 1) column vector. Note the special structure of the X ' X matrix:
the diagonal elements of X ' X are the sums of squares of the elements in the columns of X ,
and the off-diagonal elements are the sums of cross-products of the elements in the columns
of X .
Furthermore, note that the elements of X 'Y are the sums of cross-products of the
columns of X and the observations Y . The fitted regression model is:
Y = X (A9)
k
Yi = 0 + j xij i = 1, 2,..., n
i =1
The difference between the actual observation Yi and the corresponding fitted value Yi
e = Y Y (A10)
120 M. Santarelli, P. Leone, M. Cal et al.
( ) = e
n n
SS E = Yi Yi
2 2
i = e'e
i =1 i =1
Substituting e = Y Y = Y X :
Equation (A11) is called the error or residual sum of squares and it has n p degrees of
freedom associated with it, where n is the total number of observations and p = k + 1 .
It can be shown that:
E (SS E ) = 2 (n p )
SS E
2 = (A12)
n p
H 0 : 1 = 2 = ... = k = 0
(A13)
H1 : j 0 for at least one j
Experimental Activity on a Large SOFC Generator 121
SS T = SS R + SS E (A14)
n
SS T = Yi 2
( Y )n
i =1 i
2
= Y 'Y
( Y ) n
i =1 i
2
(A15)
i =1 n n
SS E = Y 'Y
( Y ) n
i =1 i
2
' X 'Y
( Y ) = SS
n
i =1 i
2
SS R
n n T
SS R = ' X ' Y
( Y ) n
i =1 i
2
(A16)
n
SS R k MS R
F0 = = (A18)
SS E (n k 1) MS E
and to reject H 0 if F0 > F , k , n k 1 . From the expected mean squares it could be shown that,
of the numerator of the test statistic is greater than the expected value of the denominator, and
122 M. Santarelli, P. Leone, M. Cal et al.
H 0 should be rejected for values of the test statistic that are too large. H 0 would be rejected
and concluded that there are differences in the treatment means if F0 > F , k , n k 1 .
Error or residual SS E n k 1 MS E
Total SS T n 1
SS R SS
R2 = = 1 E (A19)
SS T SS T
may be qualitative, such as two machines, two operators, the high and low levels of a
factor, or perhaps the presence and absence of a factor. A complete replicate of such a design
requires 2 2 2 ... 2 = 2 observations and is called 2 factorial design.
k k
k
The 2 design is particularly useful in the early stages of experimental work, when many
factors are likely to be investigated. It provides the smallest number of runs with which k
factors can be studied in a complete factorial design. Consequently, these designs are widely
used in factor screening experiments.
Factor Replicate
Treatment combination Total
A B I II III
- - A low, B low y A B I y A B II y A B III (1)
+ - A high, B low y A+ B I y A+ B II y A+ B III a
- + A low, B high y A B + I y A B + II y A B + III b
+ + A high, B high y A+ B + I y A+ B + II y A+ B + III ab
By convention, the effect of a factor is denoted by a capital Latin letter. Thus A refers
to the effect of factor A, B refers to the effect of factor B, and AB refers to the AB
interaction. In the 22 design the low and high levels of A and B are denoted by - and +,
respectively, on the A and B axes.
b ab
+
Fattore B
_
(1) a
_
+
Fattore A
The four treatment combinations in the design are also represented by lowercase letters,
as shown in Figure A1. In the figure it could be noted that the high level of any factor in the
treatment combination is denoted by the corresponding lowercase letter and that the low level
of a factor in the treatment combination is denoted by the absence of the corresponding letter.
Thus, a represents the treatment combination of A at the high level and B at the low
level, b represents A at the low level and B at the high level, and ab represents both factors at
the high level. By convention, (1) is used to denote both factors at the low level.
In a two-level factorial design, it is possible to define the average effect of a factor as the
change in response produced by a change in the level of that factor averages over the levels of
the other factor. Also, the symbols (1), a, b, ab now represent the total of all n replicates taken
at the treatment combination.
The effect of A at the low level of B is [a (1)] n and the effect of A at the high level of
B is [ab b ] n . Averaging these two quantities yields the main effect of A:
A=
1
{[ab b] + [a (1)]} = 1 [ab + a b (1)] (A20)
2n 2n
The average main effect of B is found from the effect of B at the low level of A,
[b (1)] n , and at the high level of A, [ab a ] n , as:
B=
1
{[ab a] + [b (1)]} = 1 [ab + b a (1)] (A21)
2n 2n
The interaction effect AB is defined as the average difference between the effect of A at
the high level of B and the effect of A at the low level of B. Thus:
AB =
1
{[ab b] [a (1)]} = 1 [ab + (1) a b] (A22)
2n 2n
Alternatively, AB may be defined as the average difference between the effect of B at the
high level of A and the effect of B at the low level of A, also leading to Equation (A22).
The formulas for the effects of A, B and AB may be derived by another method. The
effect of A can be found as the difference in the average response of the two treatment
combinations in which A is at high level ( y A+ ) and the two treatment combinations in which
A is at low level ( y A ), say:
ab + a b + (1) 1
A = y A+ y A = = [ab + a b (1)]
2n 2n 2n
Experimental Activity on a Large SOFC Generator 125
The effect of B is found as the difference between the average of the two treatment
combinations in which B is at high level ( y B + ) and the average of the two treatment
combinations in which B is at low level ( y B ), say:
ab + b a + (1) 1
B = yB+ yB = = [ab + b a (1)]
2n 2n 2n
Finally, the interaction effect AB is the average of the right-to-left diagonal treatment
combinations in the square in Figure A1, minus the average of the left-to-right diagonal
treatment combinations (a e b), or:
ab + (1) a + b 1
AB = = [ab + (1) a b]
2n 2n 2n
Note that the previous equations are identical to equations (A20), (A21) and (A22).
The magnitude and the direction of the factor effects are often examined in order to
determine which variables are likely to be important. The analysis of variance can generally
be used to confirm this interpretation.
Consider the sums of squares for A, B and AB. We define the total effect or contrast of A
as:
The contrast sum of squares is equal to the contrast squared divided by the total number
of observations in the design:
SS A =
[ab + a b (1)]2 (A24)
4n
SS B =
[ab + b a (1)]2 (A25)
4n
SS AB =
[ab + (1) a b]2 (A26)
4n
The total sum of squares is found in the usual way, that is:
2 2 n
y...2
SS T = y ijk
2
(A27)
i =1 j =1 k =1 4n
126 M. Santarelli, P. Leone, M. Cal et al.
where SS T has 4n 1 degrees of freedom. The error sum of squares, with 4(n 1) degrees
of freedom, is usually computed by subtraction as:
SS E = SS T SS A SS B SS AB (A28)
Furthermore, based on the p-values, it is possible to conclude that the main effects are
statistically significant and that there is no interaction between these factors.
It is often convenient to write down the treatment combinations in the order
(1) , a , b , ab , known as standard order or Yatess order.
The contrast coefficients used in estimating the effects are in Table A4.
Effects (1) a b ab
A -1 +1 -1 +1
B -1 -1 +1 +1
AB +1 -1 -1 +1
Note that the contrast coefficients for estimating the interaction effect are just the product
of the corresponding coefficients for the two main effects. The contrast coefficient is always
either +1 or -1, and a table of plus and minus signs such as in Table A5 can be used to
determine the proper sign for each treatment combination.
The column headings in Table A5 are the main effects (A and B), the AB interaction, and
I, which represents the total or average of the entire experiment. To find the contrast for
estimating any effect, simply multiply the signs in the appropriate column of the table by the
corresponding treatment combination and add.
Y = 0 + 1 x1 + 2 x 2 + (A29)
1 and 2 are one-half the corresponding factor effect estimates. The regression coefficient
is one-half the effect estimate because a regression coefficient measures the effect of a unit
change in x on the mean of y , and the effect estimate is based on a two-unit change (from -
1 to +1).
In order to check the model adequacy it is necessary to calculate the residuals, make the
plots of normal probability and residuals versus fitted values. Furthermore, response surface
plots can be realized using the regression model expression.
The treatment combinations can be written in standard order by introducing the factors
one at a time, with each new factor being successively combined with those that precede it.
For example, the standard order for a 24 design is: (1), a, b, ab, c, ac, bc, abc, d, ad, bd, abd,
cd, acd, bcd and abcd.
The general approach to the statistical analysis of the 2k design is summarized in Table
A6.
The first step is to estimate factor effects and examine their signs and magnitudes. This
gives the experimenter preliminary information regarding which factors and interactions may
be important, and in which directions these factors should be adjusted to improve the
response.
In forming the initial model for the experiment, the full model is usually chosen, that is,
all main effects and interactions, provided that at least one of the design points has been
replicated. The analysis of variance is used to formally test for significance of main effects
and interaction. Table A7 shows the general form of an analysis of variance for a 2k factorial
design with n replicates. Step 4, refine the model, usually consists of removing any not
significant variables from the full model. Step 5 is the usual residual analysis to check for
model adequacy and to check assumptions. Sometimes model refinement will occur after
residual analysis, if it is found that the model is inadequate or assumptions are badly violated.
The final step usually consists of graphical analysis, either main effect or interaction plots, or
response surface and contour plots.
In general, the contrast for an effect are determined by expanding the right-hand side of
equation (A30):
The sign in each set of parentheses is negative if the factor is included in the effect and
positive if the factor is not included. For example, consider a 23 factorial design and the
contrast for AB would be:
Once the contrasts for the effects have been computed, the effects can be estimated and
the sums of squares can be computed according to:
AB " K =
2
(Contrast AB"K )
n2k
SS AB"K =
1
(Contrast AB"K )2
n 2k
k
Table A7. Analysis of variance for a 2 design
k two-factor interactions
2
AB SS AB 1
AC SS AC 1
# # #
JK SS JK 1
k three-factor interactions
3
ABC SS ABC 1
ABD 1
# # #
IJK SS IJK 1
# # #
k
= 1 k-factor interaction
k
ABC " K SS ABC"K 1
Error SS E 2 (n 1)
k
Total SS T n2 k 1
k
Y = 0 + j x j + ij xi x j + (A31)
j =1 i< j
then there is a model capable of representing some curvature in the response function. This
curvature, of course, results from the twisting of the plane induced by the interaction
terms ij xi x j . In some situations the curvature in the response function will not be
adequately modelled by Equation (A31). In such cases, a logical model to consider is:
k k
Y = 0 + j x j + ij xi x j + jj x 2j + (A32)
j =1 i< j j =1
130 M. Santarelli, P. Leone, M. Cal et al.
n F nC ( y F y C )
2
where n F is the number of factorial design points. This sum of squares may be compared to
the error mean square to test for pure quadratic curvature.
k
When points are added to the center of the 2 design, then the test for curvature actually
tests the following hypotheses:
k
H 0 : jj = 0
j =1
k
H 1 : jj 0
j =1
Furthermore, if the factorial points in the design are unreplicated, one may use the nC
center points to construct an estimate of error with nC 1 degrees of freedom.
The mean square for pure error is calculated from the center points as follows:
(y y)
2
i
SS E
MS E = =
central po int s
(A34)
nC 1 nc 1
Experimental Activity on a Large SOFC Generator 131
Y = 0 + 1 x1 + 2 x2 + 12 x1 x2 + 11 x12 + 22 x22 +
it is not possible to estimate the six parameters because there are only 5 points in the
design, considering also the central point.
An effective solution consists in adding 4 axial runs in 22 design, as shown in Figure A2,
obtaining the Central Composite Design (CCD) for fitting a second-order model.
x2 x3
y +
z x2
y +
bc
+
abc
b ab
c ac
x x
x cc x +
x1
+
x1
cc
b ab
(1) a
z (1) a
y
y
The face-centered central composite design is not rotatable, but it does not require as
many center points as the spherical CCD: nC = 2 3 is sufficient to provide a good variance
of prediction throughout the experimental region.
References
[1] Kabs H., Operational Experience with Siemens-Westinghouse SOFC Cogeneration
Systems, Proceedings of Lucerne Fuel Cell Forum 2001, Lucerne (Switzerland), 2001.
[2] George R.A., Status of tubular SOFC field unit demonstrations, J. Power Sources, V.
73, pp. 251-256, 1998.
[3] Singhal S.C., Kendall K., High Temperature Solid Oxide Fuel Cells: Fundamentals,
Design and Applications, Elsevier, 2004.
[4] Saroglia S., Leone P., Santarelli M., Design and Development of a cogenerative system
for a SOFC CHP100, Eco-efficiency 2005 HYSYDays 1st World Congress of Young
Scientists on Hydrogen, 18-20 May 2005, Torino, Italy.
[5] Leone P., Santarelli M., Cal M., Computer experimental analysis of the CHP
performance of a SOFC stack by a factorial design, J. Power Sources, 156 (2006),
400-413.
[6] Cal M., Santarelli M.G., Leone P., Design of experiments for fitting regression models
on the tubular SOFC CHP100 kWe: screening test, response surface analysis and
optimization, accepted for publication on International Journal of Hydrogen Energy,
2006.
[7] Leone P., Santarelli M., Computer Experimental analysis of a tubular SOFC CHP to
evaluate factors effects on performances and S/C ratio, Eco-efficiency 2005
HYSYDays 1st World Congress of Young Scientists on Hydrogen, 18-20 May 2005,
Torino, Italy.
[8] Cal M., Santarelli M., Leone P., Comparison of the behavior of the CHP-100 SOFC
Field Unit fed by natural gas or hydrogen through a computer experimental analysis,
World Hydrogen Technology Convention, Singapore, 2005.
[9] Cal M., Fontana E., Giaretto V., Orsello G., Santarelli M., The EOS Project: a SOFC
pilot plant in Italy- safety aspects, HySafe - International Conference on Hydrogen
Safety, Pisa (Italy), 2005.
Experimental Activity on a Large SOFC Generator 133
[10] Cali M., Debenedictis F., Leone P., Santarelli M., Verda V., Benchmark characterization
of a tubular SOFC CHP generator: model description and experimental validation,
ASME 1st European Fuel Cell Technology and Applications Conference, Rome (Italy),
December 2005.
[11] Cal M., Leone P., Santarelli M., Orsello G., Disegna G., Operation of the tubular SOFC
CHP100 kWe Field Unit in Italy. General topics and operation description by means of
regression models, World Hydrogen Energy Conference WHEC 2006, Lyon (France),
June 2006.
[12] Cal M., Orsello G., Santarelli M., Leone P., Experimental activity on the tubular SOFC
CHP100 kWe Field Unit in Italy: factor significance, effects and regression model
analysis, Proceedings of ESDA2006, 8th Biennial ASME Conference on Engineering
Systems Design and Analysis, Torino (Italy), July 2006.
[13] Santarelli M., Leone P., Cal M., De Benedictis F., Experimental activity on the tubular
SOFC CHP100 generator in Italy: regression models analysis and optimization, Lucerne
Fuel Cell Forum 2006, Lucerne (Switzerland), July 2006.
[14] Montgomery D.C., Design and analysis of experiments, John Wiley & Sons, INC,
(2005).
[15] Walpole R.E., Myers R.H., Probability and Statistics for Engineers and Scientists,
Prentice-Hall International Inc., New Jersey, (1993).
[16] Gopalan S., DiGiuseppe G., Fuel sensitivity tests in tubular solid oxide fuel cells, J.
Power Sources, 125 (2004) 183188.
[17] Santarelli M., Leone P., Gariglio M., Cal M., Spinelli P., Orsello G., De Benedictis F.,
Experimental analysis of the polarization effects at variable local temperature and fuel
consumption in a 100 kWe SOFC stack, Proceedings of Fuel cell Seminar 2006,
Honolulu, Hawaii (USA), 2006.
[18] Singhal S.C., in Proceedings of the 17th Riso International Symposium on Materials
Science: High Temperature Electrochemistry: Ceramics and Metals. W. Poulsen, N.
Bonanos, S. Linderoth, M. Mogensen and al., Riso National Laboratory, Roskilde.
Denmark, 1996, p. 123.
[19] Cal M., Santarelli M., Leone P., Experimental analysis of the behavior of some
performance index of the primary generator and Balance of Plant of the SOFC CHP-100,
as function of the Set Up Temperature and the Fuel Consumption control factors,
Politecnico di Torino - TurboCare Siemens Confidential Report, 2006.
[20] Bessette N., Schmidt D.S., Rawson J., Foster R., Litka A., Technical Progress Report
Semi Annual, Acumentrics Advanced Power & Energy Technologies, February 2006.
[21] Draper R., DiGiuseppe G., Optimal design of current take off bus bars for tubular solid
oxide fuel cells, Proceedings of ASME European Fuel Cell Conference 2005, Rome
(Italy), 2005.
[22] Costamagna P., Honneger K., Modeling of Solid Oxide Heat Exchanger Integrated
Stacks and Simulation at High Fuel Utilization, Journal of Electrochemical Society, Vol.
145, No.11, pp. 3995-4007, 1998.
[23] Yamamura T. et al., Preparation of Nickel Pattern Electrodes on YSZ and Their
Electrochemical Properties in H2-H2O Atmospheres, Journal of Electrochemical
Society, Vol. 141, No. 8, pp. 2129-2134, 1994.
[24] Mogensen M. et al., H2-H2O-Ni-YSZ Electrode Performance, Journal of
Electrochemical Society, Vol. 151, No. 9, pp. A1436-A1444, 2004.
134 M. Santarelli, P. Leone, M. Cal et al.
[25] Kim J., Virkar A.V., Fung K.Z., Metha K., Singhal S.C., Polarization effects in
intermediate temperature, Anode-Supported Solid Oxide Fuel Cells, J. Electrochemical
Soc., Vol. 146 (1), pp. 69-78, 1999.
[26] Zhao F., Anil Virkar V., Dependence of polarization in anode-supported solid oxide fuel
cells on various cell parameters, Journal of Power Sources, Vol. 141, pp. 7995, 2005.
[27] Reid R.C., Prausnitz J.M., Poling B., The properties of gases & liquids, Fourth Edition,
McGraw-Hill, Inc., 1987.
[28] Todd B., Young J.B., Thermodynamic and transport properties of gases for use in solid
oxide fuel cell modeling, Journal of Power Sources, Vol. 110, pp. 186200, 2002.
[29] Vora S.D., Advances in Solid Oxide Fuel Cell Technology (SOFC) at Siemens
Westinghouse, Proceedings of Fuel cell Seminar 2005, Palm Springs, California (USA),
2005.
[30] Bessette N.F., Wepfer W.J., Winnick J., A mathematical model of a solid oxide fuel
cell, Journal of Electrochemical Society, Vol. 142 (11), pp. 3792-3800, 1995.
[31] Campanari S., Iora P., Definition and sensitivity analysis of a finite volume SOFC
model for a tubular cell geometry, J. of Power Sources, Vol. 132, pp.113-126, 2004.
[32] Huang K., Cell Power Enhancement via Materials Selection, Proceedings of the 7th
European SOFC Forum, Lucerne (Switzerland), 2006.
[33] US Patent n6,207,311 B1, Inventors: Baozhen L., Ruka R., Singhal S.C., Solid Oxide
Fuel Cell operable over wide temperature range, Assigne: Siemens Westinghouse Power
Corporation, 27/3/2001.
[34] Pratihar S.K., Dassharma A., Maiti H.S., Materials Res. Bull., Vol 40, pp. 1936, 2005.
[35] George R.A., Status of tubular SOFC field unit demonstrations, J. Power Sources, Vol.
73, pp. 251-256, 1998.
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 135-160 2007 Nova Science Publishers, Inc.
Chapter 3
Abstract
Commonly, polymer electrolyte membrane fuel cells operate at temperature <100C
because above this temperature the electrochemical performances of the Nafion drop down.
Therefore, many scientists have studied different types of no-fluorinated polymers as
alternative to the Nafion. According to the literature, comparable performances to the Nafion
in terms of proton conductivity and thermo-chemical properties, lower crossover and cost are
the characteristics that can be obtained by using treated polyetheretherketone (PEEK).
Different methods are used for producing electrolyte membranes from PEEK: a) PEEK
electrophilic sulfonation (S-PEEK); b) S-PEEK and no-functional polymers blending; c) S-
PEEK, heteropolycompounds and polyetherimide doping with inorganic acids; and so on. In
particular, the sulfonation of polyetheretherketone/cardo-group (PEEK-WC) by using
sulphuric acid is also presented in this chapter.
Introduction
Fuel cell technology is considered as a promising alternative for growing energy request and
cleaner environment. Actually, polymer electrolyte membrane fuel cells (PEMFCs) and direct
methanol fuel cells (DMFCs) are known to utilize the proton conducting membranes. DMFC
and PEMFC systems are conceptually the same, except for using a different feeding fuel
(methanol and hydrogen, respectively). Commonly, PEMFCs operate at temperature <100 C
because above this temperature the electrochemical performances of the Nafion (the most
important commercial electrolyte membrane) drop down. However, due to the poisoning
effect of CO on the catalyst (generally platinum), the ideal working temperature of the
*
E-mail address: a.basile@itm.cnr.it, Phone: (+39) 0984 492013, Fax: (+39) 0984 402103
136 Angelo Basile and Adolfo Iulianelli
PEMFCs is above 100 C. In fact, taking into account that the adduct Pt-CO is thermolabile,
operative temperatures at about 120-130 C could drastically reduce or eliminate the CO-
poisoning.
In general, the performances of the DMFCs are lower than the PEMFC ones but,
nevertheless, they show several advantages over PEMFCs:
Vice versa, some disadvantages are present to prevent DMFC from commercialization:
Currently, Nafion is the most widely proton exchange membrane used for both PEMFC
and DMFC systems. Nafion is based on sulfonated fluorocarbon polymer and shows good
thermal stability and high proton conductivity as advantages, while high methanol
permeability (methanol crossover), high cost (about 900-1000 US $/m2) and proton
conductivity loss above 100 C represent the disadvantages. Therefore, several studies have
been carried out to identify different types of non-fluorinated polymers as alternative to the
Nafion and one of the most promising is represented by the polyetheretherketone (PEEK)
polymer, which shows, in sulfonated form (S-PEEK) and in some case, comparable
performances to the Nafion in terms of proton conductivity and thermo-chemical properties,
as well as lower crossover and costs.
Different method can be utilized for producing electrolyte membranes from PEEK:
Among other studies, the sulfonation of a new PEEK based polymer, the
polyetheretherketone/cardo-group (PEEK-WC), represents at moment a way to obtain a new
sulfonated polymer (S-PEEK-WC) showing interesting electrochemical performances and
very good thermo-chemical properties with respect to the Nafion.
Actually, Victrex Company is the main producer of PEEK polymer and its sulfonation is
commonly carried out by introducing directly the sulfonic acid group (SO3H) onto the
polymer back-bone by modification or by polymerizing sulfonated monomers. The choice of
PEEK instead of perfluorinated polymer backbones is mainly due to cost and stability
considerations. Sulfonated PEK, PEEK and PEEKK show interesting behaviours in terms of
wettability, water flux, antifouling capacity, permselectivity and increased solubility in
solvents for fuel cell processing. In fact, the sulfonic acid functional groups aggregate to form
a hydrophilic domain and, when the sulfonated polymers are hydrated, the protonic charge
carriers form, within inner space, charge layers by the dissociation of the acidic functional
groups and the proton conductance (), assisted by water dynamics, occurs.
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 137
Several studies confirmed that S-PEEK membranes may be durable enough under fuel
cell operation conditions and with a life-time more than 3000 h. At the same time, the
mechanical properties of PEEK tend to progressively deteriorate by increasing the sulfonation
degree (DS), which affects the long term stability of highly sulfonated PEEK membranes, due
to the hydroxyl radical initiated degradation. Differently, low sulfonated PEEK membranes
show good thermal stability but not sufficient proton conductivity to be used in PEM fuel
cell.
To overcome this problem, S-PEEK is used as a major component in blend membranes
production. The range of S-PEEK content in the blends from 50 to 80% seems the most
suitable and several S-PEEK based composite membranes exhibit high proton conductivity
and show at the same time good thermo-mechanical properties.
However, many variables affect the performances of the pure S-PEEK and the S-PEEK
based membranes and, only by a good combination within them, it is possible to realize good
alternative membranes to the Nafion for PEMFC and DMFC applications.
Figure 1. Schematic representation of the microstructures of the NAFION and S-PEEK illustrating a
comparison in terms of hydrophobic/hydrophilic separation. Reprinted from [4] with permission from
Elsevier.
reactions, allowing to reach better thermal stability and mechanical properties. Moreover, the
concentration as well as the positions of the sulfonate groups (e.g., meta-, para-, and ortho-)
within the directly-synthesized S-PEEKs can be readily controlled [10]. Vice versa, the post-
sulfonating procedure can not only degrade the mechanical and thermal stabilities of the S-
PEEK polymers but also lack the control of the sulfonating process [11]. However, in both
cases the sulfonation modifies the chemical character of the PEEK, reducing the cristallinity
and consequently affecting the polymer solubility. These effects are more evident depending
on the sulfonation degree (DS) of the polymer, which represents the sulfonic groups
concentration in the PEEK matrix polymer. When the DS is over 30%, the S-PEEK polymers
are soluble in dimethylformamide (DMF), dimethylsulfoxide (DMSO) or in N-
methylpyrrolidone (NMP); for DS above 70%, they are soluble in methanol and at DS =
100% in hot water [5,6]. Moreover, if the DS is above 60%, the S-PEEKs are highly swollen
in methanol/water solution at 8090C and for these reasons, they are not suitable for the
DMFC applications [14].
Figure 3. Nomenclature of the aromatic protons for the S-PEEK repeat unit.
Carrying out the PEEK sulfonation procedure via concentrated sulphuric acid, the
sulfonation is limited and occurs only on the four chemically equivalent positions of the
hydroquinone segment [15,16] (Figure 2) and the DS does not exceed 100% due to the
electron-withdrawing deactivating effect of the SO3H group once it is introduced exactly in
that ring. The other two phenyl rings connected through the ether linkages are therefore
deactivated for electrophilic sulfonation by the electron-withdrawing effect of the carbonyl
group [17]. When this PEEK sulfonation procedure is followed, the DS can be determined
quantitatively by 1H NMR analysis [18]. Taking into account the nomenclature of the
aromatic protons for the S-PEEK repeating unit of Figure 3, the presence of a sulfonic acid
group in the 1H NMR spectra causes a significant down-field shift of the hydrogen HE pick
compared with HC, HD picks in the hydroquinone ring, Figure 4. The intensity of the HE
signal allows to estimate the HE content which is equivalent to the SO3H group. The ratio
between the peak area of the distinct HE signal (AHE) and the integrated peak area of the
signals, corresponding to all the other aromatic hydrogens (AHA,AB,B,C,D), represents the DS
140 Angelo Basile and Adolfo Iulianelli
calculation, which is limited ranging between values of 0 100%. The non-sulfonated PEEK
is insoluble in any solvent except in strong acids and its 1H NMR spectrum could not be
recordable. Taking into account the example of 1H NMR for a S-PEEK reported in Figure 4,
in low sulfonated S-PEEK (low DS), a significant proportion of repeating units are non-
sulfonated and the HC and HD of the unsubstituted hydroquinone ring appear as a
characteristic singlet.
Figure 4. 1H NMR spectra of a S-PEEK sample. Reprinted from [18] with permission from Elsevier.
If the S-PEEK is synthesized via two steps, monomer synthesis and polymer preparation,
the DS indicates the ratio between the sulfonate monomer group number and the repeating
unit in the PEEK polymer chain and it can overcome 100% [10]. In Figure 5, the formation of
S-PEEKs by means of the nucleophilic aromatic substitution reaction is shown. The first step
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 141
is the sulfonated monomer synthesis by using fuming sulphuric acid, represented in Figure 5
(a) for example by the 2-fluorobenzene sulfonate (Monomer M), which reacts with a 4,4-
difluorobenzophenone (Monomer K) and a 3,3,5,5-tetramethyl-4,4-biphenol, Figure 5 (b).
The S-PEEKs formation is realized by co-condensation reaction and the DS is controlled by
the ratio between the Monomer M and Monomer K, Figure 5 (b). By increasing the ratio
between the sulfonated monomer and the unsolfonated monomer, the DS of the S-PEEKs can
overcome 100%, Table 1.
(a)
(b)
Figure 5. (a) Scheme of synthesizing the sulfonated monomer, (b) Formation of the S-PEEKs through
the co-condensation reaction.
However, not depending on which way is realized the PEEK sulfonation, the DS can
affect several important parameters, describing the thermo-mechanical and electrochemical
behaviour of the S-PEEKs, such as:
the glass transition temperature (Tg),
the water uptake (W.U.),
the weight loss,
the ion exchange capacity (IEC),
the proton conductivity.
Figure 6. Glass transition temperature depending on DS for a S-PEEK sample. Reprinted from [18]
with permission from Elsevier.
SO3H groups per repeating unit). The water absorption of S-PEEK membranes increases
linearly up to a DS of 65% and thereafter very rapidly above 70%. In the highly sulfonated S-
PEEK the density of SO3H groups is high and it may involve clustering or agglomeration.
Clustered ionomers absorb more water and, therefore, a large W.U. may be suggestive of the
presence of ion-rich regions where proton transfer is particularly fast [18,14].
Figure 7. Water absorption for a S-PEEK sample. Reprinted from [18] with permission from Elsevier.
Figure 8. Weight loss versus DS calculated by TGA due to sulfonic acid decomposition. Reprinted
from [18] with permission from Elsevier.
144 Angelo Basile and Adolfo Iulianelli
c) For what concerns the weight loss, it is to consider that the PEEK polymer is
temperature resistant and the temperature weight loss starts at about 520C. The pyrolytic
degradation of PEEK results in the formation of phenols and benzene. The major pyrolysis
product of PEEK is phenol, which is produced in the early chain scission reaction involving
ether rather than carbonyl linkages. Two weight loss steps are observed for S-PEEK, which
are reflected by two broad peaks in the DTG curve in two separate temperature ranges. The
first weight loss peak in S-PEEK probably is due to the splitting-off of sulfonic acid groups.
From the first peak, the weight loss corresponding to the sulfonic acid decomposition can be
plotted against DS, as shown in Figure 8, in which the theoretical weight loss is calculated
assuming that the splitting-off of a sulfonic acid group releases one SO3 molecule. As shown
by the figure, there is a rather close agreement between the weight loss calculated
theoretically and the one determined from TGA experiments for the sulfonic acid
decomposition. Therefore, it can be said that the S-PEEK membranes are thermally stable up
to approximately 300C and that this temperature is only marginally affected by an increasing
in the degree of sulfonation up to 80% [18].
d) Sulfonated cation-exchange membranes such as S-PEEKs have to show ion-exchange
capacities (IEC) of minimum 1.4 to 1.7 meq/g to permit low ionic resistances, which make
them suitable for the application in electromembrane processes. The IEC, in fact, indicates the
density of ionisable hydrophilic groups in the membrane matrix, which are responsible for the
ionic conductivity of the membranes. The uncrosslinked sulfonated membranes show very
high swelling, which leads to poor mechanical properties and low ion permselectivity and,
thus, disqualifies them for the application in PEM fuel cell, especially at temperature > 80C.
Therefore, it is necessary to reduce the S-PEEK swelling to improve the mechanical and
conductive properties and different strategies are available for the swelling reduction of
cation-exchange membranes and few of them can be resumed as follows:
The IEC of the S-PEEKs depends on the DS and on the reaction time and, in some case,
they can reach values close to 2.4 meq/g with DS equal to 85% after 120 hours of sulfonation,
carried out via concentrated sulphuric acid [19]. Also when the PEEK sulfonation proceeds
via sulfonated monomer copolymerization without any side reactions, the IEC increases with
increasing DS and shows values ranging between 0.7 and 1.5 meq/g, corresponding to DS of
40 and 120%, respectively [11].
e) One of the most important parameter to be taken into account to evaluate the
electrochemical performances of the S-PEEKs is the proton conductivity, which depends
mainly on the DS, but also on other important parameters such as temperature, relative
humidity (R.H.), etc.
Generally, the S-PEEK sulfonic concentration, indicated by the DS, plays the main role
for what concerns the proton migration through an electrolyte polymer. With increasing DS,
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 145
the S-PEEKs show a solubility increase in organic solvents (e.g. DMAc), due to a cristallinity
decrease [5]. As a consequence, the polymer becomes more hydrophilic and absorbs more
water, which facilitates the proton transport. Hence, the sulfonation raises the conductivity of
the PEEK not only by increasing the number of protonated sites (SO3H), but also through the
formation of water mediated pathways for protons.
The DS is able to affect also the temperature influence on the S-PEEK conductivity. In
fact, a S-PEEKs membrane with a 40% DS does not show any improvement in conductivity
with increasing the temperature, but, on the contrary, it decreases. The conductivity of a S-
PEEK with 48% DS increases up to 85C and then dropped sharply to very low values, while
DS between 70 and 74% allow a continuous conductivity increase, showing a gradual
decrease only above 100C. S-PEEKs with a DS of 80% show a different conductivity
behaviour with respect to lower ones: the conductivity increases slowly in the range of 20
50C, more rapidly between 50 and 100C and then gradually again up to 145C [18].
These results can be explained if it is taken into account that there are two different
competing trends depending on the temperature influence on the conductivity: the first one
enhances and the second one reduces the conductivity. At low temperature, the effect of
dehydration on proton conductivity may be negligible, but at high temperature the effect of
dehydration on proton conductivity can not be neglected and may be assumed as dominant
effect. This dehydration apparently starts at room temperature for S-PEEKs having a 40% DS
and, then, shifts towards higher temperatures when the DS increases. For instance, the S-
PEEK samples with 40% DS or less lose water so fast that the dehydration suppresses any
conductivity rise, while for S-PEEKs with DS around the 60%, the waters lost velocity
induces the conductivity decrease at temperature of almost 50C. S-PEEK membranes with
higher DS start to lose conductivity at higher temperature due to their stronger ability of
retaining water. In fact, for S-PEEKs with DS equal to 80% the water retention is so high that
the dehydration induces only a decrease in the conductivity increment rate up to 145C.
Therefore, not only the capacity of S-PEEKs water uptake is important to improve the proton
conductivity but also the capacity to retain water [20].
As previously said, another important parameter affecting the conductivity is the relative
humidity, which indicates the hydration state of polymer. The acid strength, for example, has
an appreciable influence on proton conduction only at low R.H. values. At high humidity
values, the increasing material hydration tends to level the acid strength of the sulfonic group,
determining its full dissociation. Accordingly, the differences in the S-PEEKs protonic
conductivity become progressively smaller at high R.H. values. The acid strength of the SO3H
groups in perfluorinated polymers is considerably higher than that in nonperfluorinated
polymers. At the moment, Nafion is the best protonic conductor at low R.H. but, in
comparison with the other less superacids polymers such as S-PEEKs, the differences in
conductivity become small or vanish for R.H. close to 100%.
As already seen, the S-PEEK conductivity increases appreciably with increasing DS and,
in contrast with Nafion behaviour, the temperature has a strong effect on the conductivity.
Furthermore, at the medium temperatures and high R.H., the conductivity essentially depends
on the concentration of the sulfonic groups while the effect of the acid strength is evident only
at low R.H.. A reasonable explanation could be the following: the interaction forces between
the polymeric chains gradually decreases with the increasing temperature, favouring a greater
hydration of the polymer for a given constant value of R.H.[21].
Table 2. Properties of pure S-PEEK and S-PEEK based polymers in comparison with Nafion
In Table 2, several literature data concerning proton conductivity for different pure S-
PEEKs are reported at 25 and 80C and, moreover, other significant information are added.
As can be easily noted, several discrepancies exist within the reported proton conductivities,
as well as they exist in literature. For example, the S-PEEK conductivity obtained by
Kaliaguine et al. [14] at 25C is 3.8 10-3 S/cm while that of Xing et al. [17] is 1.6 10-2
S/cm at the same temperature. Kobayashi et al. [12] found a conductivity 1.0 105 S/cm at
room temperature and a DS equal to 65% and at 60C it is ranged between 7 104 and 8
104 S/cm, while D.J. Jones et al. [46] found a conductivity of 3.5 103 S/cm with a 65%
DS at 80C. It should be pointed out that similar discrepancies were observed in many other
cases. A possible explanation can be related to the preparation and treatment conditions of S-
PEEKs.
The SO3H groups of the S-PEEK, responsible for charge transfer in PEMs, are able to
form strong hydrogen-bonding with some solvents such as DMF or DMAc, affecting
significantly the conductivity of the membranes and reducing the charge carrier number
and/or mobility. In fact, by using various casting solvent it is possible to find different
conductivity values depending on the different interactions between the casting solvent and
the S-PEEK samples.
It is highly plausible that S-PEEKs with high DS contain some amount of residual
sulphuric acid, which is rather difficult to wash out. When this acid reacts with DMF or
DMAc, a decomposition into DMAm sulphate and formic or acetic acids is produced. The
reaction scheme between DMF or DMAc with sulphuric acid is shown in Figure 9 and the
result of this interaction is a conductivity decrease due to the decrease of the sulfonic acid
concentration.
In the presence of water, the SO3H groups inevitably form hydrogen bonds with oxygen.
The intensities of the H-bonding absorption is exactly two and six times the intensity of the
aromatic hydrogen in ortho-position with respect to the sulfonic acid group, Figure 10. This
indicates that, after interaction with DMF, one molecule having exactly two types of
hydrogen atoms in the ratio of 1:6 is present for each SO3H group. As shown in the figure,
DMF has exactly these hydrogen atoms numbers, six in the two methyl groups and one on the
aldehyde carbon. When another solvent such as DMAc or NMP is used to prepare the S-
PEEK membranes, no evidence of hydrogen bonded is recorded. This suggests that the DMF
molecule is particularly prone to hydrogen bonding with SO3H groups, while DMAc is not.
148 Angelo Basile and Adolfo Iulianelli
Therefore, the S-PEEK proton conductivity differences existing between the membranes
prepared by using different casting solvents are obviously explicable on the basis of strong
bonding, which blocks the protons of the SO3H groups excluding them from charge transfer
and causing a conductivity decrease [22,16].
Figure 10. Structures of DMF and DMAc and a possible configuration of hydrogen-bonding between
SO3H groups of S-PEEK and DMF molecules.
The proton conductivity performances of the S-PEEKs, as well as for other polymers,
describe the membrane capacity to process the feeding fuel (hydrogen in PEMFC and
methanol in DMFC) in order to produce electricity and to form water as secondary product by
reaction with oxygen at the anode electrode. Part of the feed can also goes through the
polymeric membrane without processing, producing a fuel efficiency lost well known as fuel
crossover. It is also well known that the main disadvantage related to the Nafion use in
DMFC is properly the methanol crossover and its alternative membranes such as the S-
PEEKs have to be able to constitute a barrier for the methanol permeation.
For what concerning the S-PEEKs in DMFC, the main parameter to take into account is
that DS, which, if is higher than 70%, leads the membranes to become methanol soluble and,
consequently, not applicable for this fuel cell application.
The different methanol permeation of the S-PEEKs and the Nafion can be explained by
the difference in their microstructures. In the Nafion membrane, because of its high
hydrophobicity of the perfluorinated backbone and its high hydrophilicity of the SO3H
groups, it would give rise to form hydrophobic/hydrophilic domains, especially in the
presence of water. As previously discussed, the sulfonic acid groups aggregate to form
hydrophilic domains. These hydrophilic domains are interconnected in Nafion membrane and,
not only proton and water but also a smaller polar molecule such as methanol can go through
them, leading to the methanol crossover. The S-PEEKs microstructure is distinctly different
with respect to the Nafion due to the smaller hydrophobic/hydrophilic domains and to the
lower flexibility of the polymer backbone. In other words, for the S-PEEKs the separation
into hydrophilic and a hydrophobic domains is less pronounced and for this reason their
methanol permeability are lower than the Nafion one [14,23,20].
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 149
Figure 11. The selectivity of a S-PEEK sample and Nafion117. Reprinted from [20] with permission
from Elsevier.
High proton conductivity () and low methanol permeability (PCH3OH) are two of the
essential characteristics that a polymer electrolyte membrane has to possess in order to be
validly proposed for use in DMFC. Combining these two parameters into a factor defined
selectivity (=log(/P)) [24], a comparison can be more easily carried out within different S-
PEEK membranes. In Figure 11, the factor comparison between a S-PEEK sample and
Nafion117 membranes is shown. Increasing the DS, the S-PEEKs selectivity decreases
because, overcome DS = 70%, the sulfonated polymer becomes more soluble in methanol and
then methanol permeability increases [20].
Although generally the Nafions proton conductivities are higher than those of the S-
PEEKs, the Nafion methanol permeability increases with increasing temperature and the
factor decreases making the S-PEEKs more attractive than Nafion for DMFC applications.
expected if one of the components causes a decrease of the average separation between acidic
groups. The proximity of acidic functions might even lead to significant conductivity in
anhydrous conditions, where a non-vehicular mechanism is considered to be responsible of
proton transfer from a donor to an acceptor site, without the assistance of water molecules
acting as carriers. Several anhydrous electrolytes have been indeed investigated for
applications in fuel cells operating at intermediate temperature (80200C) and most of them,
being based on composite materials or acidbase interactions of two components, are
summarized in Table 2. For each type of modified S-PEEK based membrane, the main
parameters affecting the proton conductivity and other important variables are following
illustrated.
The SiS-PPSU/S-PEEK
A physical blend between S-PEEK with elevated DS and polyphenylsulfone (PPSU), a fully
aromatic polymer with good thermal stability and oxidation resistance, can improve its
mechanical properties. The PPSU are structurally affine to S-PEEKs, its solubility in organic
solvents allows to easily carry out functionalization reactions in homogeneous conditions,
giving also the possibility to introduce sulfonic acid groups. As for the S-PEEKs, the
conductivity of the sulfonated PPSU (S-PPSU) is a function of the DS: conductivity values of
about 103 S/cm have been reported for S-PPSU with DS = 0.7, while larger DS lead to water
soluble species, preventing possible application in fuel cells.
A silylated telechelic hybrid polymer can allow to obtain proton-conducting membranes
with proper mechanical and proton conductivity properties by introducing an inorganic
network, covalently linked to the organic backbone of highly sulfonated PPSU, (SiSPPSU) in
a blend with highly sulfonated PEEK (DS = 90%). The SiSPPSU shows very low solubility
and poor plastic properties in the solvents conventionally used for membrane casting (DMSO,
DMF, DMA, NMP, etc). In anhydrous conditions, proton transfer occurs via H+ hopping from
a sulfonic group to another, so that all the SO3H substituents appear to contribute to
conductivity. Up to the limit of 5 wt.% of SiSPPSU in blends with S-PEEKs, the large
number of SO3H groups positively contributes to S-PEEK conductivity (see Table 2) making
available at close distance extra sites for proton hopping. Above this SiSPPSU concentration
in the S-PEEK blend, the interactions between SO3H groups become predominant and
negatively affect electrochemical performances [25].
The SiS-PEEK
Hybrid S-PEEK based membranes can be prepared also carrying out a reaction between S-
PEEK with a DS up to 90% and a sufficient amount of inorganic component such as SiCl4,
obtaining an improvement of S-PEEK mechanical properties without altering significantly the
proton conductivity performances. The first step in the preparation of this hybrid membranes
type is the reaction within S-PEEK and butyllithium. The subsequent reaction with SiCl4,
followed by alcoholysis, leads to the formation of a product with the Si presence in the S-
PEEK (SiS-PEEK). While the S-PEEK with elevated DS is reported to be soluble or form gel
when immersed in water [17], SiS-PEEK membranes shows an elevated water uptake
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 151
(240%), remaining flexible and easy to handle. The formation of covalently bound inorganic
clusters increases the level of water uptake without excessive swelling or solubility in water
of the membrane. With the hybrid SiS-PEEK membranes is possible to achieve stable
conductivity values in the range of 210-2 110-2 S/cm and, at same time, to improve their
solubility properties [26].
The S-PEEK/HPAs
Another approach to improve the S-PEEKs proton conductivity consists on the composite
membranes synthesis by incorporation of solid heteropolyacids (HPA) into partially
sulfonated PEEK polymer matrix. The HPAs such as tungstophosphoric acid (TPA), its
sodium salt (Na-TPA) and molybdophosphoric acid (MPA) are known as the most conductive
solids among the inorganic solid electrolytes at ambient temperature [27-29]. For instance,
hydrated tungstophosphoric acid exhibits a room temperature conductivity of 1.9101 S/cm
and its sodium form has a conductivity of about 102 S/cm. The HPAs are soluble in polar
solvents and their strong acidity is attributed to the large size of the polyanion yielding low
delocalized charge density. When appropriately embedded in a hydrophilic polymer matrix,
the hydrated HPAs are expected to endow the composite membrane with their high proton
conductivity, retaining the desirable mechanical properties of the polymer film. The binder
matrix is constituted by partially sulfonated PEEK, which provides the polymer matrix with
some hydrophilicity. The composite S-PEEK/HPA membranes are prepared by incorporation
of HPA (TPA, MPA and Na-TPA) into the matrix S-PEEK polymer (in Table 2 three
different S-PEEK DS, respectively 70, 74 and 80%, are reported blended with HPAs). They
contain HPAs and display higher Tg than the pure S-PEEK membranes and show a thermal
stability up to temperatures above 250C. This Tg increase may be due to a reduction in chain
mobility, probably caused by the interaction of the solid acid with polar groups of the S-
PEEK polymer chain [30].
Incorporating 60 wt.% of HPAs in S-PEEKs with a DS ranging within 70-80%, the
proton conductivity is increased with respect to the pure S-PEEK at temperatures higher than
80C, improving also the conductivitys stability. The conductivity increasing due to the TPA
incorporation in the S-PEEK based membranes is more relevant with respect to the
membranes in which Na-TPA and MPA are incorporated. Being TPA a stronger acid,
systematically it yields a higher proton conductivity increasing as well as a better water
retention at high temperature. In fact, the W.U. of the composite membranes increases by
means of the incorporation of HPAs into S-PEEK matrices. The increasing association of the
PEEK repeat unit with the anionic counter charge (SO3), immobilized on the polymer
backbone of a neighbouring chain, suggests a less effective separation of the aqueous phase
compared to the Nafion. This increased association is not however sufficient to upset the
effect of the increased water content, which brings about an increase in conductivity of S-
PEEK membranes.
The S-PEEK/HPA composite membranes absorb more water than the pure S-PEEK
membranes. The W.U. increase associated with the HPAs incorporation into the S-PEEK
matrix is only one of the factors affecting the membrane conductivity. The other factors,
including: the polymer intrinsic conductivity, the strength, density and softness of the solid
acid sites, the solid phase loading, the particle size and spatial distribution, the aqueous phase
152 Angelo Basile and Adolfo Iulianelli
dispersion, affect the dependence of conductivity on the water content. In the case of Nafion
membranes, which are described as nanoporous inert sponges for water of hydration, the
water shows a little interaction with the polymer chain and forms a hydration shell around the
SO3H acid groups [31]. In HPA/S-PEEK composite membranes, the aqueous phase is more
continuous. This leads to higher W.U. and contributes to a greater extent to proton
conductivity, achieving values of around 10-1 S/cm (see Table 2) [18].
The BPO4/S-PEEK
Other composite membranes can be prepared by fine powder BPO4 incorporation into
partially sulfonated PEEK polymer. In several cases, this type of BPO4/S-PEEK composite
membranes can show higher conductivity than that of the HPAs/S-PEEK even if the intrinsic
conductivity of HPAs is about twice that of BPO4 under the same conditions (room
temperature, fully hydration). DS and BPO4 content contributes to the swelling increase of the
composite membranes and the W.U. due to the BPO4 substantially exceeds the sorption
capacity of the unsupported solid electrolyte. The reason of this effect can only be due to the
BPO4/S-PEEK composite membranes porosity similarly to the HPAs/S-PEEK, while the pure
S-PEEK is essentially nonporous. The influence of solid electrolyte incorporation into the S-
PEEK matrix produces two effects: the porosity rise, which does not improve the membrane
electric properties, and the conductivity increase due to the higher intrinsic conductivity of the
BPO4 solid electrolyte, which allows to reach BPO4/S-PEEK conductivity of around 7.5 10
2
S/cm (see Table 2) [32].
S-PEEK/PANIs
Sulfonated PEEK/polyaniline composite membranes can be synthesized by chemical
polymerization of a thin layer of polyaniline (PANI), in the presence of a high oxidant
concentration on a single face modification, in order to prevent the methanol crossover and
conserving good proton conductivity performances. The IEC, the W.U. and the proton
conductivities of the S-PEEK/PANI composite membranes depend on the coating density of
the PANI in the S-PEEK. In some case, the methanol permeability of these membranes show
to be four times lower than Nafion117, while the high proton conductivity of the S-
PEEK/PANIs can be attributed to the sulfonic acid and amine groups present in the
membrane matrix, which give rise to hydrophilic regions in the polymer because of their
strong affinity toward water. These hydrophilic areas formed around the cluster of chains lead
to absorption of water, enabling a easy proton transfer [33]. Protons in the form of hydronium
ions pass through the hydrophilic regions of S-PEEK/PANIs and this fact is responsible of the
proton conductivity increase with respect to the pure S-PEEK membrane. The S-PEEK/PANI
membranes with high PANI coating density exhibit slightly lower conductivity in comparison
to the Nafion (see Table 2), while the increasing and decreasing trends of the methanol
permeability, as a function of polymerization time, arise because of the presence of
hydrophilic groups ( SO3H) with amine groups. In this acidbase composite material, the
methanol permeability may be influenced either by cross-linking between these charged
moieties or by significant phase segregation of the PANI. In fact, induced by the PANI
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 153
The S-PEK/ZP-ZRO2
By means of the in situ generation of inorganic oxides like zirconia by hydrolysis of the
alkoxides in the S-PEK/Zp polymer solution, it is possible to obtain a decreasing of water and
methanol flux through the membranes. The addition of well-dispersed zirconium phosphate
into the polymer solution (S-PEK/ZP-ZRO2) increases the membranes conductivity (see
Table 2). Both the effects can be explained by the swelling behaviour of the composite
membranes. For what concerns the cross-over, it is to be taken into account that the
water/methanol flux increases linearly with the degree of swelling. For the Nafion, which
forms channels terminated by SO3H groups inside the matrix of the hydrophobic backbone,
the hydrophilic channels are responsible for a high flux, the hydrophobic matrix hinders an
excessive swelling. Vice versa, SO3H groups in S-PEKs are attached to the polymer main
chain and they are more statistically distributed, leading to an overall swelling of the
membrane, but with a lower flux due to the absence of distinct channels like in Nafion
membranes. The inorganic modification with zirconia reduces the S-PEK swelling, but it can
be present some region of lower resistance to water/methanol transport, which may be
generated especially in the interface between polymer and zirconium phosphate [35].
The PEEK-WC
PEEK, PEEKK and PEK are polymers with a high cristallinity degree, insoluble in water and
in almost all common organic solvents. The PEEKs insolubility is overcome by a new type
of polymer, the PEEK-WC, which is no recordable as a blend, a composite or a hybrid
membrane but only as a chemically modified PEEK based polymer. This polymer is
characterized by the presence of the cumbersome lattonic group, able to reduce the
cristallinity degree, showing consequently to be amorphous. PEEK-WC polymer shows very
good thermal, chemical and mechanical stability [36] and it is used in PEM fuel cell in
sulfonated form. Two ways can be followed to sulfonate PEEK-WC: the first one by using
chlorosulfuric acid [37,38] and the second one by using concentrated sulphuric acid [39-41].
Following the chlorosulfuric acid sulfonation procedure, at temperature around 70 C it is
possible to obtain a PEEK-WC having a high DS, the capability to be a polymer water soluble
and able to be used as polyelectrolyte. At the same time, the disadvantage due to this
sulfonation procedure is represented by the difficulty to control the substitution degree. By
working at lower temperatures (around 0 C), this problem can be avoided, but the sulfonated
PEEK-WC (S-PEEK-WC) show lower sulfonation degree, around 30%, with a decrease of
the electrochemical performances, but also with DS 82%, the conductivity showed by S-
PEEK-WC (see Table 2) is however three times lower than Nafion117 membrane at 115 C
and relative humidity 100% [38].
When the S-PEEK-WC membranes are obtained by using concentrated sulphuric acid, a
DS 40% after 3 hours of polymer sulfonation is achieved [39,40]. The S-PEEK-WC,
154 Angelo Basile and Adolfo Iulianelli
synthesized via sulphuric acid, shows a proton conductivity of 1.7 x 10-2 S/cm at temperature
higher than 100 C (see Table 2). This proton conductivitys value is comparable with respect
to S-PEEK-WC obtained by chlorosulfuric acid sulfonation (DS = 82%). The comparison
within them indicates that the direct polymer sulfonation with concentrated sulphuric acid is
more performing because almost the same conductivity value is achieved, but with a DS of
40% with respect to a DS of 82% obtained by using the chlorosulfuric acid sulfonation
procedure.
Moreover, composite membranes based on S-PEEK-WC can be obtained by
incorporating an amorphous zirconium phosphate sulfophenylenphosphonate gel with a fixed
20 wt% content. They, indicated as Zr(SPP)/S-PEEK-WC, absorb less water and methanol
than the pure S-PEEK-WC and the methanol absorption does not increase significantly with
temperature on the contrary of the pure S-PEEK-WC, especially in the case of high DS. The
Zr(SPP)/S-PEEK-WC methanol permeability is lower than that of Nafion due to its
microstructure with narrow and highly branched channels, which are responsible for the
lower methanol crossover.
Generally, the proton conductivity increases with DS, but for what concerns the
Zr(SPP)/S-PEEK-WC composite membrane the increase is not significant and the
introduction of Zr(SPP) into S-PEEK-WC improves the proton conductivity only for low DS
values [42] (see Table 2).
where G is a change of Gibbs free energy of formation in the system, n the number of
electrons transferred during the reaction (2 in the case of hydrogen), N Avogadros number,
and e the charge of an electron. The Gibbs free formation energy depends on the temperature
(T), pressure (P) and the phase state of the reactants and product (liquid or gas, fn). For all the
materials in a fuel cell in the gaseous state, at atmospheric pressure and temperature around
80C, it can be easily calculated that E = 1.17 V. This derived voltage can be treated as the
maximum that it can be obtained under no-loss conditions. In an ideal case scenario, this
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 155
Figure 12. Schematic dependence of output voltage and power density on electrical current density
from a proton exchange membrane fuel cell. Reprinted from [44] with permission from Elsevier.
Table 3. Open circuit voltage and resistance of few literature data regarding S-PEEK, S-
PEEK-WC and Nafion membranes
voltage should be independent of the electrical current drawn. In reality, it can not be avoided
losses and, moreover, other limitations are present. Therefore, the process is irreversible and
results in lower output voltages. A typical voltage and current characteristic of a cell is given
in Figure 12. Four different processes are at least responsible in a fuel cell for the observed
behaviour. The ohmic resistance to both currents (electrons and protons) generates heat and
results in the slow and linear drop of voltage when the current increases (middle part of the
curve). This voltage drop can be presented as
Vr = i (Re + Rp)
and
Re = Rcont + Rbackpl
where i is the current, Re the resistance to electron current, Rp the resistance to proton current
through the PEM, Rcont the contact resistance between catalyst and backplate, Rbackpl the
resistance of backplates. In Table 3 few literature data concerning the S-PEEK ohmic
resistance are reported in comparison with the Nafion. As illustrated in the table, the R
values, which mainly cause the linear variation of the cell potential with current density, are
156 Angelo Basile and Adolfo Iulianelli
very different within them and that of Nafion results to be the lowest one. The voltage and
current relationship during the electrochemical reaction itself on each catalyst layer are linked
by a simple dependence named Tafel equation, which is used as electrochemistry law for
irreversible reactions. The voltage drop occurring on the electrode is known as the over-
voltage (overpotential) and reflects the fact that some energy is needed to generate a reaction
product. The dependence has the simple form
2.3RT i
VOV = ln( )
nF i0 ( M , T , S )
where R is the gas constant, T temperature, a symmetry coefficient (usually around 0.5), n
the number of exchanged electrons, i0 a exchange current which is dependent on the materials
involved, temperature and material active area S. The chemical meaning of i0 can be
understood from the fact that, even without any applied external current, the electrochemical
reaction takes place with a certain probability, but the reaction products at the same time
collapse back to form the initial reactants. Therefore, i0 is a balance current of the fully
reversible reaction. When an external current is applied, the balance is shifted and this is
reflected by generating reaction products. Two different electrodes in a fuel cell with two
different reactions on them lead to two different current/voltage dependences. Some hydrogen
in molecular form always passes through a PEM and ends at the cathode area. The catalyst on
the cathode effectively splits this hydrogen to protons, which react with oxygen making a
small electrical load on the cathode side even without external current. Since the activity of
the cathode is not particularly high, this leads to a noticeable over-voltage even without
external current. For this reason, a real open circuit voltage (OCV) will be less than the ideal
EMF by this over-voltage amount [43-45]. An example of this assert is illustrated in Table 3,
where all few data regarding the OCV are lower than 1.17 V and the Nafion shows the
highest value with respect to the S-PEEK and to the S-PEEK-WC.
Conclusion
The different mechanical and transport properties of hydrated perfluorosulfonic polymers
such as Nafion and low-cost sulfonated polyaryls such as sulfonated PEEK appear due to the
differences in the acidity of the sulfonic groups and of the microstructures. The less
pronounced hydrophobic/hydrophilic separation of the pure S-PEEKs compared to the Nafion
corresponds to narrower, less connected hydrophilic channels and to larger separations
between the less acidic sulfonic acid functional groups. This leads to a disadvantageous
swelling behaviour and a stronger decrease of water and proton transport with decreasing
water content. On the other hand, the hydrodynamic flow of water is reduced compared to
that Nafion, which is an essential advantage, especially for DMFC applications.
Blending of sulfonated PEEK with other polymers such as PPSU in sulfonated form or
with inorganic component such as SiCl4 and inorganic oxides like zirconia, as well as with
solid heteropolyacids as HPA, TPA and Na-TPA or with BPO4, it is possible significantly to
improve the membrane swelling behaviours, the mechanical properties and the proton
conductivity. Other efforts have been carried to reach the same results following different
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 157
Aknowledgements
Special thanks to Dr. Isabella Nicotera for reviewing the manuscript and for her precious
suggestions.
References
[1] K.D. Kreuer, On the development of proton conducting materials for technological
applications, Sol. State Ion., 97 (1997) 1-15
[2] G. Hubner, E. Roduner, EPR investigation of HO radical initiated degradation of
sulfonated aromatics as model compounds for fuel cell proton conducting membranes, J.
Mater. Chem., 9 (1999) 409-418
[3] K.D. Kreuer, Th. Dippel, J. Maier, Membrane materials for PEM fuel cells: a
microstructural approach, Proc. Electrochem. Soc., 95-23 (1995) 241-246
[4] K.D. Kreuer, On the development of proton conducting polymer membranes for
hydrogen and methanol fuel cells, J. Membrane Sci., 185 (2001) 29-39
[5] X. Jin, M.T. Bishop, T.S. Ellis, F.E. Karasz, A sulphonated poly(arylether ether ketone),
Br. Polym. J., 17 (1985) 4-10
[6] M.T. Bishop, F.E. Karasz, P.S. Russo, K.H. Langle, Solubility properties of a poly(aryl
ether ether ketone) in strong acids, Macromolecules, 18 (1985) 86-93
[7] C. Bailly, D.J. Williams, F.E. Karasz, W.J. MacKnight, The sodium salts of sulphonated
poly(aryl-ether-ether-ketone) (PEEK): preparation characterization, Polymer, 28 (1987)
1009-1016
[8] F. Wang, M. Hickner, Y.S. Kim, T.A. Zawodzinski, J.E. McGrath, Direct
polymerization of sulfonated poly(arylene ether sulfone) random (statistical)
copolymers: candidates for new proton exchange membranes, J. Membrane Sci. 197
(2002) 231-242
[9] F. Wang, J. Li, T. Chen, J. Xu, Synthesis of poly(ether ether ketone) with high content
of sodium sulfonate groups and its membrane characteristics, Polymer, 40 (1999)
795-799
158 Angelo Basile and Adolfo Iulianelli
[10] M. Gil, X. Ji, X. Li, H. Na, J.E. Hampsey, Y. Lu, Direct synthesys of sulfonated
aromatic poly(ether ether ketone) proton exchange membranes for fuel cell applications,
J. Membrane Sci., 234 (2004) 75-81
[11] X. Li, Z. Wang, H. Lu, C. Zhao, H. Na, C. Zhao, Electrochenical properties of
sulfonated PEEK used for ion exchange membranes, J. Membrane Sci., 254 (2005)
147-155
[12] T. Kobayashi, M. Rikukawa, K. Sanui, N. Ogata, Proton-conducting polymers derived
from poly(ether-etherketone) and poly(4-phenoxybenzoyl-1,4-phenylene, Sol. State Ion.,
106 (1998) 219-225
[13] P. Xing, G.P. Robertson, M.D. Guiver, S.D. Mikhalenko, S. Kaliaguine, Sulfonated
Poly(aryl ether ketone)s Containing the Hexafluoroisopropylidene Diphenyl Moiety
prepared by Direct Copolymerization, as Proton Exchange Membranes for Fuel Cell
Application, Macromol., 37 (2004) 7960-7967
[14] L. Li, J. Zhang, Y. Wang, Sulfonated poly(ether ether ketone) membranes for direct
methanol fuel cell, J. Membrane Sci., 226 (2003) 159-167
[15] D. Daoust, J. Devaux, P. Godard, A. Jones, R. Legras, in H. Kausch (Ed.), Advanced
thermoplastic composites characterization and processing, Carl Hanser, Munich,
1992, Chapter 1
[16] G.P. Robertson, S.D. Mikhailenko, K. Wang, P. Xing, M.D. Guiver, S. Kaliaguine,
Casting solvent interactions with sulfonated poly(ether ether ketone) during proton
exchange membrane fabrication, J. Membrane Sci., 219 (2003) 113-121
[17] P. Xing, G.P. Robertson, M.D. Guiver, S.D. Mikhailenko, K. Wang, S. Kaliaguine,
Synthesis and characterization of sulfonated poly(ether ether ketone) for proton
exchange membranes, J. Membrane Sci., 229 (2004) 95-106
[18] S.M.J. Zaidi, S.D. Mikhailenko, G.P. Robertson, M.D. Guiver, S. Kaliaguine, Proton
conducting composite membranes from polyether ether ketone and heteropolyacids for
fuel cell applications, J. Membrane Sci., 173 (2000) 1734
[19] W. Cui, J. Kerres, G. Eigenberg, Development and characterization of ion-exchange
polymer blend membranes, Sep. Purif. Tech., 14 (1998) 145-154
[20] S. Xue, G. Yin, Methanol permeability in sulfonated poly(ether ether ketone)
membranes: A comparison with Nafion membranes, European Pol. J., 42 (2006)
776-785
[21] G. Alberti, M. Casciola, L. Massinelli, B. Bauer, Polymeric proton conducting
membranes for medium temperature fuel cells (110-160C), J. Membrane Sci., 185
(2001) 73-81
[22] S. Kaliaguine, S.D. Mikhailenko, K.P. Wang, P. Xing, G. Robertson, M. Guiver,
Properties of SPEEK based PEMs for fuel cell application, Catal. Tod., 82 (2003)
213-222
[23] S. Ren, C. Li, X. Zhao, Z. Wu, S. Wang, G. Sun, Q.Xin, X. Yang, Surface modification
of sulfonated poly(ether ether ketone) membranes using Nafion solution for direct
methanol fuel cells, J. Membrane Sci., 247 (2005) 59-63
[24] B.S. Pivovar, Y. Wang, E.L. Cussler, Pervaporation membranes in direct methanol fuel
cells, J. Membrane Sci., 154 (1999) 155-162
[25] M.L. Di Vona, A. DEpifanio, D. Marani, M. Trombetta, E. Traversa, S. Licoccia,
SPEEK/PPSU-based organicinorganic membranes: proton conducting electrolytes in
anhydrous and wet environments, J. Membrane Sci., 279 (2006) 186191
Alternative Sulfonated Polymers to Nafion for PEM Fuel Cell 159
[44] V.M. Vishnyakov, Proton exchange membrane fuel cells, Vacuum, 80 (2006) 1053-1065
[45] L. Kim, S-M Lee, S. Srinivasan, CE Chamberlin, Modeling of proton exchange
membrane fuel cell performance with an empirical equation, J. Electroch. Soc., 142
(1995) 2670-2674
[46] D.J. Jones, J. Rozire, Recent advances in the functionalisation of polybenzimidazole
and polyethrketone for fuel cell applications, J Membrane Sci., 185 (2001) 41-58
[47] M. Wakizoe, O.A. Velev, S. Srinivasan, Analysis of proton exchange membrane fuel
cell performance with alternate membranes, Electr. Acta, 40 (1995) 335-344
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 161-209 2007 Nova Science Publishers, Inc.
Chapter 4
Francisco Jurado*
University of Jan, Department of Electrical Engineering, 23700 Alfonso X,
n 28, EPS Linares (Jan), Spain
Abstract
A fully mature fuel cell industry constitutes a potential opportunity to electric utilities. It
could meliorate the technical and financial performance of existing distribution lines by
improving service quality and reliability.
When fuel cells connect to the power system, both the owner of the energy resource and
the central power system benefit. Reliability increases for both because they can support each
other.
However, the interconnection of fuel cell plants to the grid is still hindered by restrictive
conditions and procedures for grid connection. Problems arise with regard to determination of
the point of connection, safety and stability issues. Most important it is needed to establish a
standardized technical interface for allocation of connection that take into account possible
positive effects of distributed generation on transmission and distribution losses.
At last, the aim would be the substitution of conventional coal-fired generation with
Integrated Gasification Combined Cycle plants that integrate fuel cells and gas turbines. This
chapter studies the use of distributed resources for ancillary services and simulating the
impact of distributed resources on utility distribution networks.
Keywords: Biomass, distributed generation, electric system stability, fuel cells, hybrid
systems.
1. Introduction
The interaction realized by fuel cellmicroturbine hybrids derive primarily from using the
rejected thermal energy and combustion of residual fuel from a fuel cell in driving the gas
*
E-mail address: fjurado@ujaen.es, Telephone: +34-953-648518, Fax: +34-953-648586
162 Francisco Jurado
turbine. This leveraging of thermal energy makes the high temperature Molten Carbonate
Fuel Cells (MCFCs) ideal candidates for hybrid systems. Use of a recuperator contributes to
thermal efficiency by transferring heat from the gas turbine exhaust to the fuel and air used in
the system.
Traditional control design approaches, consider a fixed operating point in the hope that
the resulting controller is robust enough to stabilize the system for different operating
conditions. On the other hand, adaptive control incorporates the time-varying dynamical
properties of the model (a new value of gas composition) and considers the disturbances
acting at the plant (load power variation).
A system showing great promise is the integration of gasification with a fuel cell. A high-
temperature fuel cell-microturbine combination has the potential to achieve up to 60 percent
efficiency and near-zero emissions. Fuel flexibility enables the use of low-cost indigenous
fuels, renewables, and waste materials. However, the characteristics of gas from biomass
gasification may vary significantly. In many practical cases, the designer has a model of the
system, but the system parameters are subject to uncertainty. In this case robust control can be
used to design a control system with a guaranteed level of performance, as long as the system
uncertainties remain the assumed bounds.
To determine the potential impacts of fuel cells on future distribution system, dynamic
models of fuel cells should be created, reduced in order, and scattered throughout test feeders.
This chapter presents the implementation of an efficient method for computing low order
linear system models of Solid Oxide Fuel Cells (SOFCs) from time domain simulations. The
method is the Box-Jenkins algorithm for calculating the transfer function of a linear system
from samples of its input and output.
When connected in small amounts, the impact of distributed generation on distribution
system stability will be negligible. However, if its penetration level becomes higher,
distributed generation may start to influence the dynamic behavior of the system as a whole.
This chapter presents a mathematical representation of a SOFC plant that is suitable for use in
distribution system stability studies. The model is applied to a distributed utility grid that uses
a solid oxide fuel cell plant as distributed resource. Examinations include transient stability
and voltage stability of the system.
The chapter is organized as follows. In Section 2, general principles of MCFC-
microturbine hybrid power cycles are explained. In Section 3, robust control for fuel cell-
microturbine hybrid power plant using biomass is introduced. Section 4 presents the modeling
of SOFC plants on the distribution system using identification algorithms. Effect of a SOFC
plant on distribution system stability is discussed in Section 5, and finally, conclusions are
provided in Section 6.
2002). However, it is first necessary to reform fuels such as natural gas, converting it into a
gas with a high hydrogen content, so that it can be electrocatalytically oxidized with air.
The direct carbonate fuel cell is a variant of molten carbonate fuel cells (MCFC) in that it
internally reforms methane-containing fuels within the anode compartment of the fuel cell
(Matsumoto et al., 1990; Farooque, 1991; Sasaki et al., 1993; Shinoki et al., 1995). The
largest demonstration of MCFC technology has been Californias 2-MW Santa Clara
Demonstration Project (Fuel Cell Engineering Corporation, 1997).
Microturbines, which are typically fueled with natural gas, generate between 25 and 200
kW of electricity. Their small size and relatively low cost allow them to be located near
where they are needed. They can operate at very low emission levels and reduce the
efficiency losses and environmental impact of large transmission and distribution systems. In
this paper, a MCFC is associated with a gas microturbine to produce electric power.
Hybrid systems offer a solution to two important problems, the low efficiency and
relatively high emissions of small gas turbines, and the high cost of small fuel cell power
plants (Steinfeld et al., 1996; Layne and Holcombe, 2000; Ghezel-Ayagh et al., 2002).
The Section is structured as follows. Subsection 2.1 presents a review of the MCFC.
Some basic concepts of the gas turbine theory are presented in Subsection 2.2. Subsection 2.3
describes the gas turbine control configuration. Subsection 2.4 briefly discusses the fuel cell-
microturbine hybrid power cycles. Subsection 2.5 outlines the adaptive control. At last,
Subsection 2.6 depicts some simulation results and discussion.
MCFC can reach fuel-to-electricity efficiencies approaching 60%, considerably higher than
the 37-42% efficiencies of a phosphoric acid fuel cell (PAFC) plant. When the waste heat is
captured and used, overall fuel efficiencies can be as high as 85%.
Improved efficiencies are one reason why MCFC offers significant cost reductions over
PAFC technology. Another is that the electrodes of a MCFC be made of nickel catalysts
rather than the more costly platinum of PAFC systems.
Natural gas is internally reformed, partially in an internal reforming unit and partially at
the cells, eliminating the need for a large external reforming unit to produce hydrogen fuel.
The approach, Fig. 1, is a combination of indirect internal reforming (IIR) and direct internal
reforming (DIR) which provides for better thermal management. At the cathode, oxygen
reacts with carbon dioxide and electrons to form carbonate ions:
The carbonate ions flow through the electrolyte matrix from cathode to anode. At the
anode, the carbonate ions are consumed by the oxidation of hydrogen to form steam and
carbon dioxide, releasing electrons to the external circuit:
Cell voltage under load current (A/cm) is expressed as (Hirschenhofer et al., 1998):
1/ 2
RT p H 2 ,a pO2 ,c pCO2 ,c
Vo = Eo + ln (4)
2F p H 2O ,a pCO2 ,a
A thorough introduction to gas turbine theory is provided in (Cohen et al., 1987). There also
exists a large literature on the modeling of gas turbines. Model complexity varies according to
A Potential Alternative in the Electric Utility 165
the intended application. Detailed first principles modeling based upon fundamental mass,
momentum and energy balances is reported in (Fawke et al., 1987) and (Shobeiri, 1987).
These models describe the spatially distributed nature of the gas flow dynamics by dividing
the gas turbine into a number of sections. Throughout each section, the thermodynamic state
is assumed to be constant with respect to location, but varying with respect to time.
Mathematically, the full partial differential equations model description is reduced to a set of
ordinary differential equations, which facilitate easier application within a computer
simulation program. For a detailed model, a section might consist of a single compressor or
turbine stage. Much simpler models result if the gas turbine is decomposed into just three
sections corresponding to the main turbine components i.e. compressor, combustor and
turbine, as in (Hussain and Seifi, 1992).
Instead of applying the fundamental conservation equations, as described above, another
modeling approach is to characterize gas turbine performance by utilizing real steady state
engine performance data, as in (Hung, 1991). It is assumed that transient thermodynamic and
flow processes are characterized by a continuous progression along the steady state
performance curves. This is known as the quasi-static assumption. The dynamics of the gas
turbine, e.g. combustion delay, motor inertia, fuel pump lag etc. are then represented as
lumped quantities separate from the steady-state performance curves. Very simple models
result if it is further assumed that the gas turbine is operated at all times close to rated speed
(Rowen, 1983).
Air at atmospheric pressure enters the gas turbine at the compressor inlet. After
compression of the air to achieve the most favorable conditions for combustion, fuel gas is
mixed with the air in the combustion chamber. Combustion takes place and the hot exhaust
gases are expanded through the turbine to produce mechanical power. In terms of energy
conversion, the chemical energy present in the combustion reactants is transferred to the gas
stream during combustion. This energy - measured in terms of gas enthalpy- is then converted
into mechanical work by expanding the gas through the turbine. Thus, the excess mechanical
power available for application elsewhere, after accounting for the power required to drive the
compressor, is derived ultimately from the combustion process.
Compressor power consumption equation:
wa hIC
Pc = (5)
ctrans
PT = T wg hIT (7)
166 Francisco Jurado
Pmec = PT Pc (8)
The concept of the gas turbine control system, which is applied in this chapter, is based on the
Speedtronic Mark 4 description as presented in (Rowen, 1988). Some considerations
concerning the subject may be also found in (Hannett and Afzal Khan, 1993; Hannett et al.,
1995; Jurado et al., 2002). The simplified gas turbine model is divided into two
interconnected subsystems. The subsystems are the fuel system (fuel valve with actuator), and
the turbine. The fuel flow out from the fuel systems results from the inertia of the fuel system
actuator and of the valve positioner.
The fuel system actuator equation is:
kf
wf = e1 (9)
f s +1
a
e1 = Fd (10)
bs + c
where the input variable to the fuel system is Fd. The output variable from the fuel system
model is wf . A single gas turbine does not require the digital setpoint feature. The kLHV factor
depends on the LHV. The kLHV and 0.23 factors cater for the typical turbine power/fuel rate
A Potential Alternative in the Electric Utility 167
characteristic, which rises linearly from zero power at 23 % fuel rate to rated output at 100 %
fuel rate. The turbine torque function is given by:
Equation (11) allows the turbine torque to be calculated algebraically. This torque is used
in the equations which model the mechanical system:
Pmec = T (12)
Input variables to the turbine are wf , and . Output variable from the turbine is Pmec .
For the purpose of this chapter only modulating control of mechanical side of the gas turbine
is of interest. The simplified model of the gas turbine controller in this work consists of two
inputs and one output. Inputs to the controller are Pmec and . The output from the controllers
is Fd .
The block diagram of the gas turbine control system is presented in Fig. 3 and described
by the data in Table 1. The diagram consists of two Proportional Integral Derivative (PID)
controllers. LVG stands for Least Value Gate that transmitting the minimum of two incoming
signals.
a b c kf b
1 0.05 1 1 0.4
The coupling of high temperature fuel cells with gas turbines is by no means new, and has
been discussed for several years. Recent technical papers have covered a wide range of topics,
these including the following: (1) systems involving midsized and large gas turbines (Veyo
168 Francisco Jurado
and Lundberg, 1999); (2) dynamic modeling, performance, technical issues, and integration
considerations (Liese et al., 1999); (3) microturbine specific studies (Campanari, 1999); and
(4) details of experimental programs to demonstrate the coupling of a microturbine with a
solid-oxide fuel cell SOFC (Leeper, 1999).
When used in combination with turbines, fuel cells can produce from 55 90% of the
electricity of the system while turbines produce the remainder. Several cycle configurations
have been proposed and the terminology is still evolving, but one useful way of looking at the
differences is to divide the configurations into those that include directly fired turbines and
those with indirectly fired turbines. A hybrid power system with a directly fired turbine
normally uses a pressurized fuel cell to provide input to the turbine, thus acting as a
combustor. In the indirectly fired turbine system, an atmospheric fuel cell is used and a
pressurized heat exchanger provides input to the turbine. When utilizing the fuel cell/gas
turbine combination the combined efficiency of the system is raised to greater than 60 % and
criteria pollutant emissions are essentially eliminated. An example of a hybrid system with an
indirectly fired turbine is shown in Fig. 4.
This type of integration could utilize any gas turbine system. The indirect-fired approach
places the fuel cell in the exhaust of an indirect fired gas turbine. The air and residual
products of the fuel cell are then fed to an atmospheric combustor. This combustor heats the
air leaving the compressor via a heat recovery unit (HRU) and delivers it to the turbine. The
combination of the atmospheric combustor and heat exchanger, replaces the normal internal
pressurized combustor.
The turbine exhaust flows to the fuel cell anode exhaust oxidizer. Exhaust from the anode
exhaust oxidizer flows to the heat exchanger, which provides the heat for the compressor air.
The exit from the heat exchanger flows through the fuel cell cathode providing the oxygen
and carbon dioxide needed in the carbonate fuel cell process.
In the fuel cell/turbine hybrid power plant the fuel cell does not need to operate at the
turbine pressure, instead it operates at the preferred ambient pressure and its independent of
gas turbine cycle pressure ratio. The system works efficiently with a wide range of turbine
compression ratio. This allows taking a system developed for integration at the multi-MW
A Potential Alternative in the Electric Utility 169
scale with industrial size turbines, and configuring a small MW class system using a
microturbine at a lower pressure ratio.
y (i ) = z k
( )
B z 1
u (i ) + e(i )
( )
(13)
A z 1
Equation (13) is one of the most typical in the field of adaptive control and non-standard
discrete-time control algorithms in general. If only stable factors exist in the B polynomial,
the plant will be called minimum phase. The part e(i) is the stochastic part of the disturbance.
In this chapter, this disturbance is the load power variation PL.
Generally, most control algorithms can be described by the structure and parameters of
the difference equation:
( ) ( ) ( )
R z 1 u (i ) + S z 1 y (i ) T z 1 (i ) + h = 0 (14)
The coefficients of the R(z-1), S(z-1) and T(z-1) polynomials and the h term are chosen
before the simulation experiment and stay constant during the experiment.
The aim of the minimum-variance control algorithm is the minimization of the following
performance index (Astrm and Wittenmark, 1995; Astrm and Wittenmark, 1990).
2
( ) (( )) [( ) ]
1
B z
B z
( )
J = E P z 1 r 1 y(i + K ) V z 1 (i ) + q Q z 1 u (i )
2
(15)
170 Francisco Jurado
where Br-(z-1) is the reciprocal polynomial of the B-(z-1) polynomial. Minimization of this
performance index leads to a control algorithm of the same structure as Equation (14).
The estimation scheme used in this work is the recursive least squares (RLS). The
adaptive control is shown in Fig. 5. Stack current Idc is measured and used in calculation of
fuel flow setpoint. This measurement of the stack current determines the plant output signal
y(i). Stack current is continuously adjusted by the inverter control to maintain power. This
current is the setpoint of the plant output signal (i).
A plant consisting of a load is fed from the fuel cell/ turbine hybrid system. The selected
system comprises a 250 kW fuel cell and a 30 kW gas microturbine. The plant and the fuel
cell/ microturbine system are modeled using MATLAB.
All parameters correspond to a 2-stack equivalent. The fuel cell stacks used in this
chapter are rated at 125 kW. Stack voltage is taken across a parallel connection of 2 stacks,
each stack consisting of 258 cells.
This chapter develops the control system with an adaptive minimum variance controller.
The plant and the controller are simulated as discrete in time (Phillips and Nagle, 1995). The
plant parameters are: polynomials A(z-1), B(z-1) and standard deviation sd of white noise of
normal distribution. Parameters of RLS are: initial values of diagonal of covariance matrix
P_init and forgetting factor .
The disturbance considered is a load power variation PL. This is zero-mean, white noise
and has constant variance 2.
The controller has the form (Astrm and Wittenmark, 1990):
u (i ) =
( )
S c z 1
y (i )
( )
(16)
Rc z 1
where Sc of order 2 and Rc of order 2 . The structure of the controller is calculated according
to the plant structure. To show the ability of the controller to adapt to varying operating
conditions, a new value of gas composition is introduced at 100 s what provides time-varying
plant. Controller parameters are identified on-line and the simulation time is 200 s.
The minimum-variance controller serves as an example of a self-tuner. The optimal value
of the output signal standard deviation is obtained in the steady state. The plant is assumed to
be of ARX type (AutoRegressive with eXogenous input).
The results with the derived model are summarized. Fig. 6 a presents the control signal
U. Referring to the results shown in Fig. 6 b, it can be seen the stack current deviation Idc.
Stack voltage starts at its initial value, corresponding to regulated power, and drops quickly to
a sudden increase in stack current, in turn, controlled by the inverter to maintain the new
power setpoint. Stack voltage varies according to the temperature dependence, while stack
current is continuously adjusted by the inverter control to maintain power.
Figure 7. Output standard deviation sd and the optimal output standard deviation.
Fig. 7 shows the output standard deviation sd and the optimal output standard deviation.
Fig. 8 a depicts the polynomial R parameters where the varying parameters considered are r1
and r2. Fig. 8 b displays the polynomial S parameters, s1 and s2 being the varying parameters.
This controller guarantees some additional robustness margins in the case that the model does
not cover the entire plant uncertainty.
A power plant can generate electric power using biomass from the trees. A gasifier is capable
of converting tons of wood chips per day into a gaseous fuel that is fed into a fuel cell. The
composition of gas depend on biomass gasification technology.
174 Francisco Jurado
Gasification is the term often used to describe the thermochemical processes that take
place between a solid fuel and a gaseous reacting media, like air, oxygen or steam in order to
produce a combustible gas or synthesis gas rich in carbon monoxide and hydrogen. The
difference compared to combustion is that the amount of oxidant, i.e. oxygen in air or pure
oxygen or steam, is not high enough to enable complete oxidation, i.e. combustion. Instead, a
partial oxidation, i.e. gasification, will take place. The overall gasification process can be
divided into four steps, resulting in a product mixture of gaseous, liquid and solid products.
The first step is drying, where the moisture of the fuel fed to the gasifier, coal or biomass,
evaporates. The second step where volatile compounds in the fuel evaporate is called
devolatilization. This is followed by pyrolysis, the step where the major part of the carbon
content of the fuel is converted into gaseous compounds in the reduction zone of the gasifier.
The result of the pyrolysis is, apart from gases, a carbon rich solid residue called char. In the
last step, called gasification, the char is partly oxidized into gaseous products and partly burnt
to generate the heat needed for the overall gasification process. Even though the reactions
taking place during gasification are not very complicated and complex themselves, the
difficult part is to establish the kinetic relations between the different species.
Hydrogen products are defined as any hydrogen (H2) and other gases, whose usefulness
or value is defined by its hydrogen or hydrogen-plus-methane (CH4) content.
Biomass has a low fixed carbon/volatile matter ratio, favoring processes that maximize
pyrolysis and cracking, producing higher CH4/H2 ratios, desirable for use in the most nearly
commercial fuel cells.
The reactions taking place during the different steps of gasification are:
C (s ) + H 2O CO + H 2 (17)
CO + H 2O CO2 + H 2 (18)
C (s ) + 2 H 2 CH 4 (19)
CH 4 + H 2O CO + 3H 2 (20)
In this work, the fuel cell model uses a constant average cell voltage and a constant
average fuel utilization to calculate the production of the fuel cells according to Faraday's
law, Equation (21):
reactions. If the tar is allowed to condense when the gas is cooled (condensation temperature
of tar from woody biomass is 200 to 500C) considerable problems with equipment
contamination (such as filter clogging) result. Fuel nitrogen is largely converted to ammonia,
which on firing of the gas will form nitrogen oxides. Apart from tars, gas turbines require a
minimum LHV of the gas and are sensitive to dust. In addition, alkali compounds are
detrimental to gas turbines, causing corrosion and deposits at the high working temperatures
of turbines.
The requirement for a higher gas pressure has led to the development of two distinct
schemes for integrating the gasification system to the gas turbine; pressurized gasification-hot
gas filtering and cleaning and; atmospheric gasification-cold gas filtering and cleaning.
A variety of relatively large-scale biomass gasification technologies are at various
advanced stages of development. Three gasifier/gas cleanup designs are considered here: (i)
atmospheric-pressure air-blown fluidized-bed gasification with wet scrubbing (Waldheim and
Carpentieri, 1998), its higher heating value (HHV) is 6.47 MJ/kg, % H2 = 16.6, and % CH4 =
3.4; (ii) pressurized air-blown fluidized-bed gasification with hot-gas cleanup (Salo et al.,
1998), the HHV is 5.48 MJ/kg, % H2 = 15.5, and % CH4 = 0.05; and (iii) atmospheric-pressure
indirectly-heated gasification with wet scrubbing (Paisley and Anson, 1997) the HHV is 18.1
MJ/kg, % H2 = 29.5, and % CH4 =13.05.
Gumz is the earliest reference found describing the concept of combining a pressurized
gasifier with a gas turbine engine (Gumz, 1950).
There are many ways to design a control system. Optimal control is a method that has been
around since the 1950s and has seen a lot of successful applications, especially in the
aerospace field. Adaptive control is a way of designing a control system to adjust to changes
in the system that is being controlled. Fuzzy and neural control allow a designer to control a
nonlinear system without having a mathematical model of the system.
In many practical cases, the designer has a model of the system, but the system
parameters are subject to uncertainty. In this case robust control can be used to design a
control system with a guaranteed level of performance, as long as the system uncertainties
remain the assumed bounds.
Traditional control design approaches, consider a fixed operating point in the hope that
the resulting controller is robust enough to stabilize the plant for different operating
conditions. These approaches definitely yield good results if the parameter variations are
small or the system dynamics is not too sensitive with respect to these parameters
(Ackermann et al., 2002). For significant parameter variations these control design methods
reach their performance limits. New design approaches, which already incorporate the plant
uncertainty in the design step are then required.
Eigenvalue assignment plays an important role in control system theory because of its
many applications, such as stabilization and optimization (Lee et al., 2001). Schemes to
assign the closed-loop eigenvalues of the controllable linear time-varying system have been
proposed in (Wolovich, 1968; Nguyen, 1987; Valasek and Olgac, 1995 a; Valasek and Olgac,
1995 b).
176 Francisco Jurado
To stabilize this plant it would suffice to shift the poles only a little into the left complex
half -plane. In view of a practical realization this is certainly not sufficient since the damping
still could be arbitrarily small. The notion of -stability is therefore introduced to assure
sufficient stability margins. Considering the operating domain it cannot be the aim of the
design to specify precisely the location of the poles in dependency of the operating point. The
design goal is reached if the damping is sufficiently large. This is the case if the roots of the
closed-loop system for all possible operating points have a degree of damping larger then a
value D0. Furthermore, a maximal settling time might be required. This corresponds to a
maximal real part of the roots not larger than a certain value 0. Both conditions can be
visualized graphically in the complex s-plane.
The admissible region for closed-loop eigenvalues is denoted as , a system is called -
stable if all its eigenvalues are located in this region and an uncertain system is called robustly
-stable if all eigenvalues for all operating conditions are contained in . The definition of -
stability permits arbitrary regions in the complex s-plane and does not underlie any
restrictions. It also includes the special cases of the left half-plane for Hurwitz stability and
the unit circle for Schur stability.
The region is specified by its boundary . Besides lines, adequate elements to
describe the boundary of a -region are circles, hyperbolas, and ellipses. Lines parallel to the
imaginary axis restrict the settling time, circles the bandwidth. Circles can be used to limit the
bandwidth, or in its inverted version to guarantee a minimal bandwidth.
The field of application of this method is not only robustness analysis but also controller
design. In this case the set of stabilizing controller parameters is determined. All controllers
from this set stabilize the plant, thus, allowing to incorporate further design criteria to select
the final controller. The task is to determine a controller which robustly -stabilizes the
system for the entire operating domain.
3.3. Results
A plant consisting of a load is fed from the fuel cell/ turbine hybrid system. The selected
system comprises a 250 kW fuel cell and a 30 kW gas microturbine. The plant and the fuel
cell/ microturbine system are modeled using MATLAB (Math Works, 2006).
All parameters correspond to a 2-stack equivalent. The fuel cell stacks used in this
chapter are rated at 125 kW. Stack voltage is taken across a parallel connection of 2 stacks,
each stack consisting of 258 cells.
To investigate transient behavior, the plant is assumed to be at steady state corresponding
to rated power and subjected to a sudden variation in power demand; this is zero-mean, white
noise, and has constant variance.
The robust control is shown in Fig. 5, where the HRU, power conditioning system (PCS),
and plant control system are included in the simulation (Working Group on Prime Mover and
Energy Supply Models, 1994; Hannett and Feltes, 2001; Tolbert et al., 1999; Lukas et al.,
2000). The inverter is assumed to regulate load voltage perfectly, and simply draws stack
current proportional to load current and inversely proportional to stack voltage using a power
demand setpoint. Stack current is measured and used in calculation of fuel flow setpoint.
A Potential Alternative in the Electric Utility 177
Fig. 9 presents the eigenvalues. Referring to the results shown in Fig. 10, it can be seen
the -region, which is a hyperbola with a = -1, b = 1.73205, that corresponds to a maximal
real part of =-1 (maximal settling time), that guarantees a minimal degree of damping
equals to 0.5.
Figure 9. Eigenvalues.
Many utilities prefer to use the PI controller for better system dynamic response and in
the present study, the PI controller is used. The parameters of the controller are KP and KI.
Fig. 11 shows parameter space where the varying parameters considered are KP and KI. This
figure displays the set of simultaneously -stabilizing controller parameters for the vertices of
the operating domain ( % H2 = 15, % CH4 =0; % H2 = 30, % CH4 = 0; % H2 = 15, % CH4
=13; and % H2 = 30, % CH4 = 13). The stability boundary is computed and if the point is -
stable the point is marked (Paradise, 2004). The controller is selected from the set with
178 Francisco Jurado
maximal distance from the stability boundaries. This guarantees some additional robustness
margins in the case that the model does not cover the entire plant uncertainty.
Fig. 12 depicts the parameter space where the varying parameters considered are %
hydrogen and % methane. The controller is robustly -stable since the vertices are -stable
and the operating domain is not intersected by stability boundaries.
Fig. 13 displays the stack current deviation Idc where the simulation time is 200 s, this
figure demonstrates that the response of the PI controller with varying parameters is better
than that of PI controller with fixed parameters.
Stack voltage starts at its initial value, corresponding to regulated power, and drops
quickly to a sudden increase in stack current, in turn, controlled by the inverter to maintain
the new power setpoint. Stack voltage varies according to the temperature dependence, while
stack current is continuously adjusted by the inverter control to maintain power.
terms to a given signal (Smith et al, 1993), methods based on spectral analyses (Bounou et al,
1992), and the eigensystem realization algorithm (Sanchez-Gasca and Chow 1997). In
addition, methods designed to estimate the frequency response of a system have also been
proposed (Trudnowski et al, 1994) The Prony method is perhaps the method for which more
extensive results and applications have been documented. Its application to the analysis and
control of electromechanical oscillations has shown the value of deriving linear models from
time domain simulations and measured data (Sanchez-Gasca and Chow, 1999).
This Section presents the application of the Autoregression with exogenous signal (ARX)
identification algorithm to compute low-order system models, suitable for analysis and
control design (Ljung and Glad, 1994; Ljung, 1999; Sderstrm and Stoica, 1989). This
algorithm consists of a simple procedure for calculating the transfer function of a linear
system from samples of its input and output.
Using MATLAB/Simulink (Mathworks, 2006), a dynamic model of a SOFC-penetrated
distribution system is created.
This Section is structured as follows. Subsection 4.1 presents a review of the SOFC.
Subsection 4.2 introduces the utility-connected inverter control. Some basic concepts of ARX
models are described in Subsection 4.3. Subsection 4.4 compares the response of identified
system versus the response of the actual system. In the end, Subsection 4.5 depicts some
simulation results.
There are several types of fuel cells being developed for a variety of applications and these
have been extensively discussed in the open literature. Unlike other variants, the SOFC is
entirely solid state with no liquid components. Operation at elevated temperature is needed to
achieve the necessary level of conductivity in the cells solid electrolyte for it to operate
efficiently. With an outlet temperature in the range of 9001000C, the efficiency of the cell
alone is about 50 percent.
Typically the fuel cell system consists of SOFC generator modules in a parallel flow
arrangement, with the number of standard modules being determined by the plant power
requirement. The SOFC generator module embodies a number of tubular cells, which are
combined to form cell bundle rows, several of which are arranged side by side to make up the
complete assembly.
Most likely, fuel cell will be a dominant distributed energy resource. The SOFCs are
dynamic devices and when connected to the distribution system they will affect its dynamic
behavior. Hence, researchers have developed dynamic models for these components
(Massardo and Lubelli, 2000; Padulls et al., 2000; Campanari, 2001; Rao and Samuelsen,
2002; Zhu and Tomsovic, 2002).
The chemical response in the fuel processor is usually slow. It is associated with the time
to change the chemical reaction parameters after a change in the flow of reactants. This
dynamic response function is modeled as a first-order transfer function with a 5-s time
constant.
The electrical response time in the fuel cells is generally fast and mainly associated with
the speed at which the chemical reaction is capable of restoring the charge that has been
A Potential Alternative in the Electric Utility 181
drained by the load. This dynamic response function is also modeled as a first-order transfer
function but with a 0.8-s time constant.
With aid of the inverter, the fuel cell system can supply not only real power but also
reactive power. Usually, power factor can be in the range of 0.8-1.0. The SOFC system
dynamic model is given in Fig. 14.
The fuel utilization (U) is the ratio between the fuel flow that reacts and the input fuel
flow, such as the follows:
q Hin2 q Ho 2 q Hr 2
U= = (22)
q Hin2 q Hin2
p H 2 V an = n H 2 RT (23)
It is possible to isolate the pressure and to take the time derivative of the previous
expression, obtaining:
182 Francisco Jurado
dp H 2 RT
= qH (24)
dt Van 2
There are three relevant contributions to the hydrogen molar flow: the input flow, the
flow that takes part in the reaction and the output flow, thus:
dpH 2
dt
=
RT in
Van
(
qH 2 qHo 2 qHr 2 ) (25)
According to the basic electrochemical relationships, the molar flow of hydrogen that
reacts can be calculated as:
N0I
qHr 2 = = 2 K r I rfc (26)
2F
dp H 2
dt
=
RT in
Van
(
qH 2 qHo 2 2 K r I rfc ) (27)
If it could be considered that the molar flow of any gas through the valve is proportional
to its partial pressure inside the channel, according to the expressions:
qH 2 K an
= = KH2 (28)
pH 2 M H2
Replacing the output flow by (28), taking the Laplace transform of both sides and
isolating the hydrogen partial pressure, yields the following expression:
pH 2 =
1 / KH2
1 + H2 s
(q in
H2 2 K r I rfc ) (29)
(
where H 2 = Van / K H 2 RT . )
A similar operation can be made for all the reactants and products. Applying Nernsts
equation and Ohms law (to consider ohmic losses), the stack output voltage is represented by
the following expression:
RT p H 2 pO2
0 .5
Vc = N 0 E0 + ln rI fc
r (30)
2 F p H 2O
A Potential Alternative in the Electric Utility 183
Control of the flux vector has been shown to have good dynamic and steady-state
performance (Chandorkar, 2001). It also provides a convenient means to define the power
angle since the inverter voltage vector switches position in the d-q plane, whereas there is no
discontinuity in the inverter flux vector.
For a six-pulse Voltage Source Inverter (VSI), the inverter output voltage space vector
can take any of seven positions in the plane specified by the d-q coordinates. The time
integral of the inverter output voltage space vector is called the inverter flux vector for
short. The d and q axis components of the inverter flux vector v are defined as
t t
dv = v d d
dq = v
q d (31)
The magnitude and the angle of v with respect to the q axis are determined as
dv
v = qv2 + dv2 v = tan 1 (32)
qv
The d and q axis components of the ac system voltage flux vector e , its magnitude, and
angle are defined in a similar manner. The angle between v and e is defined as
p = v e (33)
It is useful to develop the power transfer relationships in terms of the flux vectors. The
basic real power transfer relationship for the control system of Fig. 15 in the d-q reference
frame is
P=
3
(eqiq + ed id ) (34)
2
In (34), eq and ed are the q- and d-axis components, respectively, of the ac system voltage
vector E. In addition, iq and id are the components of the current vector I. When iq and id are
expressed in terms of the fluxes, taking into account the spatial relationships between the two
flux vectors and assuming the ac system voltage to be sinusoidal, (34) can be expressed as
3
P= e v sin p (35)
2 LT
In this expression, e and v, are the magnitudes of the ac system and the inverter flux
vectors, respectively, and p is the spatial angle between the two flux vectors. is the
184 Francisco Jurado
frequency of rotation of the two flux vectors. The expression for reactive power transfer can
be derived in a similar manner. This is
3
Q=
2LT
[
e v cos p e2 ] (36)
Equations (35) and (36) indicate that P can be controlled by controlling p, which can be
defined as the power angle, and Q can be controlled by controlling v.
The two variables that are controlled directly by the inverter are v and p. The vector v
is controlled to have a specified magnitude and a specified position relative to the ac system
flux vector e.
The errors between actual and desired amounts activate the remainder of the firing
scheme only if they exceed a threshold value. If the error is larger than the hysteresis band
(whose widths are p and v ) then a decision towards a new switching sequence is made.
If the errors are within their hysteresis band, the switches will hold their current status.
Therefore, the SOFC plant has two major control loops:
1. Power control: done by adjusting the set point P* of the inverter for fast transient
variations and fuel flow input control for slow variations.
2. Voltage control: done by adjusting the set point E* of the inverter, which effects the
magnitude of the converter output voltage.
A Potential Alternative in the Electric Utility 185
Using the fast response of the inverter, the maximum power can be used to alleviate
transients when fault occurs. However, the inverters for fuel cells have the following
limitations:
Overused fuel, U > 90% (fuel starvation and permanent damages to cells).
Underused fuel, U < 70%, (the cell voltage would rise rapidly).
Undervoltage, stack voltage < certain point ( loss of synchronism with the network).
The most used model structure is the simple linear difference equation
y (t ) + a 1 y (t 1) + ... + a na y (t na ) = b1 u (t nk ) + ... +
bnb u (t nk nb + 1) + e(t )
(37)
which relates the current output y(t) to a finite number of past outputs y(t-k) and inputs u(t-k).
The structure is thus entirely defined by the three integers na, nb, and nk. na is equal to
the number of poles and nb1 is the number of zeros, while nk is the pure time-delay (the
dead-time) in the system. For a system under sampled-data control, typically nk is equal to 1
if there is no dead-time.
For multi-input systems nb and nk are row vectors, where the i-th element gives the
order/delay associated with the i-th input.
There are two methods to estimate the coefficients a and b in the ARX model structure:
Least Squares: Minimizes the sum of squares of the right-hand side minus the left-hand
side of the expression above, with respect to a and b.
Instrumental Variables: Determines a and b so that the error between the right- and left-
hand sides becomes uncorrelated with certain linear combinations of the inputs.
There are several elaborations of the basic ARX model, where different disturbance models
are introduced. These include well known model types, such as ARMAX, Output-Error, and
Box-Jenkins (Knudsen, 1994; Stoica et al., 1985; Van Overschee and DeMoor, 1996).
A general input-output linear model for a single-output system with input u and output y
can be written:
nu
A(q ) y (t ) = [Bi (q )Fi (q )]ui (t nk i ) + [C (q ) / D (q )]e(t ) (38)
i =1
186 Francisco Jurado
Here ui denotes input #i, and A, Bi , C, D, and Fi , are polynomials in the shift operator (z
or q). It is just a compact way of writing difference equations.
The general structure is defined by giving the time-delays nk and the orders of these
polynomials (i.e., the number of poles and zeros of the dynamic model from u to y, as well as
of the disturbance model from e to y).
Most often the choices are confined to one of the following special cases:
Note that A(q) corresponds to poles that are common between the dynamic model and the
disturbance model. Likewise Fi(q) determines the poles that are unique for the dynamics from
input # i, and D(q) the poles that are unique for the disturbances.
Although each method has a somewhat different set of parameters that a system analyst
can adjust, one requirement an identified system must meet is that its response to a given
input should match the response of the actual system. This practical criterion was used as a
guideline to adjust the order of the identified systems.
4.4. Performance
SOFC models used in the distribution system analysis were constructed as shown in the
following :
In this Section SOFC modeled is denoted as the actual system. Once an identified
system was obtained, its time domain response and transfer function were compared against
the corresponding quantities of the actual system. To accomplish this, the actual system was
linearized around an operating point.
The results presented here correspond to a 0.02 p.u. by 0.1 s probing pulse into the real
power block of SOFC in Fig.14. The sampling time was 0.01 s and 600 points were used to
perform the system identification.
Assume a SOFC is operating with constant rated voltage and power demand 0.6 p.u.
There is 0.3 p.u. of step increase in the total load at t = 10 s. Fig. 16 compares the time
response of identified system versus the response of the actual system. The identified system
was obtained using Box-Jenkins algorithm, and is of 4th order. This method estimates
parameters of the Box-Jenkins model structure using a prediction error method. The order of
the identified system is the minimum order required to obtain a good time domain match.
A Potential Alternative in the Electric Utility 187
Fig.17 compares the magnitude and phase of the transfer function Vfcr(s)/P(s) of the
identified and the linearized actual systems. These plots show a very good match in the
frequency range.
Figure 17. Transfer function magnitude and phase comparison. Actual system and identified system.
188 Francisco Jurado
The IEEE test feeders (Kersting, 2001) are used as the test system to investigate the dynamic
characteristics of the distribution system with fuel cells. Figs. 18, 21 and 23 show the test
systems with the fuel cells. In this chapter, all loads are balanced, and characterized by
constant power.
The majority of data for the fuel cell model has been extracted from (Kuipers, 1998;
Singhal, 1999) and a commercial leaflet describing a SOFC 100 kW plant.
Here, all fuel cells in the test feeders have the same dynamic response and share the
generation equally. The IEEE 13 node test feeder is very small, short and relatively highly
loaded for a 4.16 kV feeder. Penetration means the proportion of the distribution feeder load
being supplied by SOFCs associated with the distribution feeder (Donnelly et al., 1996). In
this model, an initial load of Pi is assumed and the penetration is thus,
P
Penetration = (43)
P + Pi
in this Section, the penetration level of the IEEE test feeders is set at 10 %.
Figure 18. One line diagram of IEEE 13 node feeder with fuel cells.
The first controlled transient was a 0.1 p.u. step in voltage at the point of connection of
the distribution system at 0.5 s, while frequency was relatively stable during each transient.
The second controlled transient was a 0.1 p.u. step in frequency at the point of connection
of the distribution system at 0.5 s, while voltage was held constant
A Potential Alternative in the Electric Utility 189
Figure 19. Fuel cell response to a frequency step transient at node 634. IEEE 13 node feeder.
Figure 20. Fuel cell response to a frequency step transient at node 634. IEEE 13 node feeder.
The IEEE 34 node test feeder is an actual feeder. The feeders nominal voltage is 24.9
kV. It is characterized by very long and lightly loaded.
190 Francisco Jurado
Figure 21. One line diagram of IEEE 34 node feeder with fuel cells.
Figure 22. Fuel cell response to a voltage step transient at node 848. IEEE 34 node feeder.
The IEEE 123 node test feeder operates at a nominal voltage of 4.16 kV. It does provide
voltage drop problems that must be solved with the application of voltage regulators and
shunt capacitors.
A Potential Alternative in the Electric Utility 191
Figure 23. One line diagram of IEEE 123 node feeder with fuel cells.
Figure 24. Fuel cell response to a response for a voltage step transient at node 46. IEEE 123 node
feeder.
192 Francisco Jurado
Figure 25. Fuel cell response to a voltage step transient at node 634 for the IEEE 13 node feeder.
tap changer, supplementary line regulators installed on distribution feeders, shunt capacitor
switched on distribution feeders (Barker and De Mello, 2000) and shifting transformers
towards the load center (Ijumba et al. 1999). A multiple line drop compensation voltage
regulation method that determines tap positions of under-load tap changer transformers is
proposed in (Joon-Ho Choi, and Jae-Chul Kim, 2001) to maintain the customers voltages
within the permissible limits.
The model derived is based on the main equations. It is developed in the Laplace domain
and transient simulation is done using a software developed based on the MATLAB package.
The Section is structured as follows. Subsection 5.2 presents a review of transient
stability. Some basic concepts of voltage stability are introduced in Subsection 5.3. and
Subsection 4.4 depicts some simulation results.
Transient stability is a term applied to alternating current (ac) electric power systems,
denoting a condition in which the various synchronous machines of the system remain in
synchronism, or in step each other. Conversely, instability denotes a condition involving loss
of synchronism, or falling out of step (Kimbark, 1995).
Hence, for a simplified intuitive description of transient stability, a power system may be
regarded as a set of synchronous machines and of loads interconnected through the
transmission network. Under normal operating condition, all the system machines run at the
synchronous speed. If a large disturbance occurs the machines start swinging with respect to
each other, their motion being governed by differential equations. Depending upon the power
system modeling, the number of such first-order differential equations is lower bounded by
twice the number of system machines, but may be orders of magnitude larger.
DG units normally supply power to the local load centers but the excess power could also
be exported to the regional power grid, adding to the capacity and stability of the grid system.
The key to interconnection is the safety of the people who have to clear faults on the line,
and protecting the DG generator from feeding into a low-impedance fault. A fault will knock
DG off the system, requiring it to be resynchronized with the grid. There are various means to
enhance the transient stability performance of the distribution systems (Nagrath and Kothari,
1994; Kundur, 1994; Sauer and Pai, 1998).
Fast valving is one of the most effective and economic means of improving the stability
of a power system under large and sudden disturbances. Fast valving schemes involve rapid
closing and opening of thermal turbine valves in a prescribed manner to reduce the generator
acceleration following a severe fault (Patel et al., 2001).
During steady state operation of a power system, there is equilibrium between the
mechanical input power of each unit and the sum of losses and electrical power output of that
unit. The problem arises when there is a sudden change in the electrical power output due to a
severe and sudden disturbance. The severity is measured by drop of this power to a very low
or to zero value and a consequential sudden acceleration of the machines governed by the
swing equation:
194 Francisco Jurado
2
2H d
= Pm Pe (44)
0 dt 2
Where,
= rotor angle, in electrical radian
Pm = mechanical power, in p.u.
Pe = electrical power output, in p.u.
H = inertia constant, in MW-s/MVA
0 = nominal speed, in electrical radian/s
From Equation (44), it is apparent that the decrease in the mechanical power has the same
impact on the rotor angle swings as that of increase in the electrical power output. Fast
valving has a function of reducing the mechanical power input to the turbine and so the
generated power.
A change in load demand (PL, QL) causes corresponding change in both bus voltage
magnitude and the power output of the generator. The voltage magnitude value is calculated
in the power flow analysis. The change in the generators active power output results in
system frequency variations through the swing equations.
Voltage instability spans a range in time from a fraction of second to tens of minutes. Time
response charts have been used to describe dynamic phenomena (Taylor, 1994; Caizares,
2002). Many electric system components and controls play a role in voltage stability. Only
some, however, will significantly participate in a particular incident or scenario. The system
characteristics and the disturbance will determine which phenomena are important.
Mid-term voltage stability.The time frame is several minutes, typically two-three
minutes. Operator intervention is often not possible. This scenario involves high loads, and a
sudden large disturbance. The system is transiently stable because of the voltage sensitivity of
loads. The disturbance (loss of generators in a load area or loss of major distribution lines)
causes high reactive power losses and voltage sags in load areas.
Longer-term voltage stability. The instability evolves over a still longer time period and
is driven by a very large load buildup, or a large rapid power transfer increase. The load
buildup, measured in MW/min, may be quite rapid.
Several papers have shown the direct relation between saddle-node bifurcations and
voltage collapse problems, e.g., (Caizares and Alvarado, 1993; Caizares, 1995). Saddle-
node bifurcations, also known as turning points, are generic codimension one local
bifurcations of nonlinear dynamical systems of the form:
= f ( x, )
dx
(45)
dt
A Potential Alternative in the Electric Utility 195
where x Rn are the state variables, R is a particular scalar parameter that drives the
system to bifurcation in a quasi-static manner, and f: Rn x R Rn is a nonlinear function.
System exhibits a saddle-node bifurcation at the equilibrium point (x0, 0), i.e., f (x0, 0) = 0,
if the corresponding system Jacobian Dxf|0=Dxf (x0, 0) has a unique zero eigenvalue, and
some particular transversality conditions hold at that equilibrium point.
The state variables x and nonlinear function f (.) are typically defined in terms of the
quasi-steady state phasor models used in transient stability studies. Thus, generator angular
speed deviations and phasor bus voltages, magnitudes and angles, are usually an important
part of vector x. The parameter is typically used to represent changes in the system loading,
regardless of the load model used.
Therefore, typically one starts from an initial stable operating point and increases the
constant power loads by a factor , until the singular point of the linearization of the power
flow equations is reached. The loads can be defined as,
PL = P0 (1 + PnL ) (46)
where P0 is the active power base load, PnL represents the load distribution factor and PL is the
active load at a bus L for the current operating point.
The introduction of a micro-source within an existing power system assumes that there may
be a cluster of small generators located in electrical proximity, interconnected through on-site
feeders and operation can be isolated or connected to a distribution grid.
SOFCs have a slow response to changes in commands and also do not provide any kind
of internal form of energy storage. These inertia-less systems are not well suited to handle
step changes in the requested output power. It must be remembered that the current power
systems have storage in the mechanical energy of the inertia of the generators. When a new
load is applied the initial energy balance is satisfied by the inertia of the system, this results in
a slight reduction in system frequency.
The step sized power demand in the distribution system should be instantaneously
matched by an identical supply of power from the micro source power source.
Load coming on-line with grid connection. SOFCs have a problem in instantaneous
power tracking. SOFC shows a poor load tracking, as a load is applied and the micro-source
ramps up to pick up the whole quota of extra power request. The missing transient power that
the micro-source is not fast enough to provide is taken from the connection with the grid that
supplies only a part of the local requested power during steady state. Voltage is maintained by
the power grid.
Load coming on-line without grid connection. The requested power from the load coming
on-line is a step function, while the inertia-less micro source always takes a finite amount of
time to ramp up to the newly requested value. Since the unit cannot change its output power
instantaneously, the power is balanced by voltage reduction. As the power injected from the
micro source increases to supply the needed load the voltage is restored.
196 Francisco Jurado
There is a missing quota of energy from the difference of the power that the load is
requesting and the power that the unit is able to provide. If the connection to the main grid is
active, then the missing quota can be provided by the utility. It will be seen from the grid
terminals as a temporary, pulse-like power request.
5.4. Results
SOFC plant G1
4 1
2
5
6 9 10
7 13
8 11 12
A power measurement is fed in the fuel flow regulator, whose output determines the position
of a fuel valve. A step change in the input reactant element will not be noticed as a sudden
increase in charges. Typically, there is a sudden but contained rise in output power that takes
place in 1-3 s, while the newly desired output power level is reached limited until the SOFC
reaches its thermal equilibrium, which may tens of seconds. This characteristic makes the
SOFC a poor candidate for isolated systems, where loads require instantaneous power.
Figure 27. SOFC plant response to a step change in the power demand.
198 Francisco Jurado
Assume a SOFC is operating with constant rated voltage and power demand (PL) 0.6 p.u.
There is 0.3 p.u. of step increase in the total load at t = 10 s. In the first 2 or 3 s after PL is
increased, P has a rapid increase due to the fast electrical response time in the fuel cell.
Subsequently, P increases slowly and continuously until reaching the required power. This is
due to the slow chemical response time in the fuel processor. The total response time of P
from 0.6 to 0.9 p.u. is about 30 s. Fig. 27 shows the response of the SOFC to a step change in
the fuel valve.
Normally the fuel cell supports base load thus, the fuel flow regulator in Fig. 14 is set to
provide a constant input flow-rate for the cell fuel. The response of the fuel flow regulator is
slow. The plant, however, can respond continuously to small instantaneous power adjustment
by the control of the angle p* . Any steady state power change must be followed by a
corresponding change in the fuel flow-rate. If the fuel cell power increases while the fuel flow
rate is maintained constant, the steady state cell utilization will increase and the cell voltage
will drop.
This Section discusses the nature of the response of system to a linear increase in load
demand. In the load flow calculation, SOFC plant is considered as a PV bus where P and E
(shown in Fig. 15) are specified.
Figure 30. Voltage profiles for the IEEE 13 bus test system.
200 Francisco Jurado
The power load PL increases by a factor until the voltage collapse is reached. The analysis
is tested on the distribution system. Continuation method is used to identify the location of
bifurcation point ( = 2.45), as presented in Fig. 30.
Fig. 31 displays the output power for the SOFC. Simulation times are 180 s for mid-
term voltage stability and 1000, 10000 s for longer-term voltage stability, using a ramp
function equal to the linear increase in load demand.
The plant can contribute to the utility reactive power supply by continuously adjusting
the set points v* of the inverter to regulate the magnitude of the interface bus voltage.
However, according to IEEE P1547, a SOFC system should trip offline if the magnitude
of the local electric power system voltage falls below 88% for 2 s or 50% for 10 cycles.
Figure 31. Dynamic response of SOFC plant.Voltage instability at 180 s, 1000 s, and 10000 s.
6. Conclusions
The emerging fuel cells produce very high-temperature exhaust gases that can be used to
drive a gas turbine. A dynamic model for this MCFC-microturbine system has been
elaborated.
A MCFC stack interacts with other system components and causes an interdependency
between them. To investigate transient behavior, the plant is assumed to be at steady state
corresponding to rated power and subjected to a sudden variation in power demand. This
chapter has developed the control system with an adaptive minimum variance controller and
based on the simulation study, the resulting controller is robust enough to stabilize the system
for different operating conditions.
A Potential Alternative in the Electric Utility 201
A simple, biogas-fed, hybrid MCFC power-generation system has been developed. The
characteristics of gas from biomass gasification may vary significantly. Energy, measured in
terms of HHV, and the composition of gas depend on biomass gasification technology.
Traditional control design approaches, consider a fixed operating point in the hope that
the resulting controller is robust enough to stabilize the system for different operating
conditions. On the other hand, robust control incorporates the varying parameters of the
model.
This chapter establishes the possibilities for biomass-based fuel cell and gas microturbine
integration. Referring to the results obtained, it can be seen that the derived model is quite
valid for robust control studies.
In this work it was felt to be premature, based on the current state of technology
readiness, to identify an optimum hybrid plant in terms of parameter selection, efficiency, and
major feature selection. A better understanding of these will be realized with the tests of a
coupled system, and this will facilitate more accurate system modeling in the future.
The capability to calculate low-order equivalent linear systems from time domain
simulations of SOFC models using the ARX algorithm has been established. After the SOFC
model was created, it was reduced to transfer functions using the ARX algorithm; thus, the
transfer function (reduced-order model) exhibited the same dynamic response as the original
SOFC model.
A significant reduction in the model order was achieved. The time domain response of
the identified system matches the response of the actual system.
Therefore, each SOFC reduced-order model influences the grid in the same manner as
SOFC and loads would, modulating real and reactive power in response to voltage and
frequency changes on the grid.
The introduction of SOFC within the context of an already existing grid can provide
control of local bus voltage, control of base power flow thus reducing the power demand from
the main grid feeder, and ultimately frequency control associated with load sharing within
units located in the micro-grid.
The SOFC is capable of providing effective load-following service in the distribution
system. However, the results also show that the SOFC system is not an uninterruptible power
supply and does not protect the load from voltage stability while in grid-connect mode. When
SOFC plant is connected to a point where it gives support to a load in fault conditions,
transient stability is enhanced with aid of the SOFC inverter.
Nomenclature
Adaptive Control
J performance index
K prediction horizon in the performance index
P_init covariance matrix
q weighting coefficient
r1, r2 varying parameters of the polynomial Rc(z-1)
s1, s2 varying parameters of the polynomial Sc(z-1)
sd standard deviation
u(i) plant control signal
y(i) plant output signal
Idc stack current deviation (A)
PL load power variation (W)
forgetting factor
2 constant variance
(i) setpoint of the plant output signal
Fuel cell
O 2
response time for oxygen flow (s)
Gas turbine
a, b, c valve parameters
cpa specific heat of air at constant pressure (J/(kg K))
cpg specific heat of combustion gases (J/(kg K))
cps specific heat of steam (J/(kg K))
e1 valve position
Fd fuel demand signal
HHV higher heating value (MJ/kg)
KI PID parameter
KP PID parameter
kf fuel system gain constant
204 Francisco Jurado
Inverter
References
Ackermann J., et al., 2002. Robust Control: The Parameter Space Approach (2nd Edition).
Springer, London.
Akai M., Nomura N.,Waku H., Inoue M., 1997. Life-cycle analysis of a fossil-fuel power
plant with CO2 recovery and a sequestering system. Energy Vol. 22, No. (2-3), pp.
249-255.
Astrm K.J., Wittenmark B., 1990. Computer-controlled systems: theory and design. 2nd ed.
Englewood Cliffs: Prentice Hall, cop.
Astrm K.J., Wittenmark B., 1995. Adaptive control. 2nd ed. Massachusetts: Addison-Wesley.
Barker P.P., De Mello R.W., 2000. Determining the impact of distributed generation on
power systems. I. Radial distribution systems, IEEE/ PES Summer Meeting, Vol. 3, pp.
1645 -1656.
Bessette N.F., 1994. Modeling and Simulation for Solid Oxide Fuel Cell Power Systems, Ph.
D. thesis, Georgia Institute of Technology.
Bounou M., Lefebvre S., Malhame R.P., 1992. A spectral algorithm for extracting power
system modes from time recordings, IEEE Trans. Power Systems, Vol. 7, No. 2, pp.
665-672.
Campanari S., 1999. Full load and part load performance prediction for integrated SOFC and
microturbine systems. ASME paper GT-65.
Campanari S., 2001. Thermodynamic model and parametric analysis of a tubular SOFC
module, Journal of Power Sources, Vol. 92, No. (1-2), pp. 26-34.
Caizares C.A., 1995. On bifurcations, voltage collapse and load modelling, IEEE Trans.
Power Systems, Vol. 10, No.1, pp. 512-522.
Caizares C.A., editor, 2002. Voltage Stability Assessment: Concepts, Practices and Tools,
Final report, IEEE/PES Power Systems Stability Subcommittee Special Publication.
Caizares C.A., Alvarado F.L., 1993. Point of collapse and continuation methods for large
ac/dc systems, IEEE Trans. Power Systems, Vol. 8, No.1, pp. 1-8.
Chandorkar M.C., 2001. New techniques for inverter flux control, IEEE Trans. on Industry
Applications, Vol. 37, No. 3, pp. 880887.
Cohen H., Rogers G.F.C., Saravanamuttoo H.I.H, 1998. Gas turbine theory. 4th ed., Pearson-
Longman, England.
Ding J., Patel P.S., Farooque M., Maru H.C., 1997. A computer model for direct carbonate
fuel cells. In: Proceedings of the Fourth International Symposium on Carbonate Fuel
Cell Technology, Montreal, Quebec.
Donnelly M.K., Dagle J.E., Trudnowski D.J., Rogers G.J., 1996. Impacts of the distributed
utility on transmission system stability, IEEE Trans. on Power Systems, Vol. 11, No. 2,
pp. 741746.
Farooque M., 1991. Development of internal reforming carbonate fuel cell technology, final
report, prepared for U.S. DOE/METC, DOE/MC/23 274-2941, pp. 3-6 to 3-11.
Fawke A.J., Saravanamuttoo H.I.H., Holmes M., 1972. Experimental verification of a digital
computer simulation method for predicting gas turbine dynamic behaviour. Institution of
Mechanical Engineers Proc. Vol. 186, No. 27.
Fuel Cell Engineering Corporation, Santa Clara, 1997. 2 MW Fuel cell demonstration power
plant test, report TR-108 252. Electric Power Research Institute.
206 Francisco Jurado
Ghezel-Ayagh H., Leo A.J., Maru H.C., 2002. Direct Fuelcell/turbine hybrid system for
ultra high efficiency power generation. FuelCell Energy, Inc.
Gumz W., 1950. Gas Producers and Blast Furnaces, John Wiley and Sons, New York, pp.
166-167.
Hannett L.N., Afzal Khan, 1993, Combustion turbine dynamic model validation from tests.
IEEE Trans. Power Systems Vol. 8, No. 1, pp. 152-158.
Hannett L.N., Jee G., Fardanesh B., 1995. A governor/turbine model for a twin-shaft
combustion turbine. IEEE Trans. on Power Systems Vol. 10, No. 1, pp. 133-140.
Hannett L.N., Feltes J.W., 2001. Testing and model validation for combined-cycle power
plants, in Power Engineering Society Winter Meeting, Columbus, OH, USA, pp. 664-670.
Haynes C.L., 1999. Simulation of Tubular Solid Oxide Fuel Cell Behavior for Integration
Into Gas Turbine Cycles, Ph.D. thesis, Georgia Institute of Technolog.
Hirschenhofer J.H., Stauffer D.B., Engleman R.R., Klett M.G., 1998. Fuel Cell Handbook. U.
S. Department of Energy.
Hung W.W., 1991. Dynamic simulation of gas-turbine generating unit. IEE Proc.-C Vol.
138, No. 4.
Hussain A., Seifi H., 1992. Dynamic modeling of a single shaft gas turbine. In: Welfonder, E,
Lausterer G.K., Weber H., editors. Proc. of the IFAC Symposium on Control of Power
Plants and Power Systems, Munich, Germany, Pergamon Press, pp. 43-48.
Ijumba N.M., Jimoh A.A., Nkabinde M., 1999. Influence of distribution generation on
distribution network performance, Africon IEEE, Vol. 2, pp. 961964.
Joon-Ho Choi, Jae-Chul Kim, 2001. Advanced voltage regulation method of power
distribution systems interconnected with dispersed storage and generation systems, IEEE
Trans. on Power Delivery Vol. 16, No. 2, pp. 329334.
Jurado F., Ortega M., Cano A., Carpio J., 2002. Neuro-fuzzy controller for gas turbine in
biomass-based electric power plant. Electric Power Systems Research Journal, Vol. 60,
No. 3, pp. 123-135.
Jurado F., 2003. Power supply quality improvement with a SOFC plant by neural-network-
based control, Journal of Power Sources, Vol. 117, No. 1-2, pp. 75-83.
Kersting W.H., 2001. Radial distribution test feeders, in Proc. IEEE/PES Summer Meeting,
Vol. 2, pp. 908 912.
Kimbark E.W., 1995. Power System Stability, IEEE Press, New York.
Knudsen T., 1994. A new method for estimating ARMAX models, in Proc. IFAC Symposium
on System Identification, Copenhagen, Denmark, pp. 611-617.
Kuipers J.A., 1998. in Proc. Fuel Cell Seminar, pp. 450.
Kundur P., 1994. Power System Stability and Control, EPRI Power System Engineering
Series, Mc Graw-Hill, New York.
Lasseter R., 2001. Dynamic models for micro-turbines and fuel cells. PES Summer Meeting,
Vol. 2, pp. 761-766.
Layne A., Holcombe N., National Energy Technology Laboratory, 2000. Fuel cell / gas
turbine hybrid power systems for distributed generation: a status report, Global Gas
Turbine News, No.2.
Leal E.M., Silveira J.L., 2002. Study of fuel cell cogeneration systems applied to a dairy
industry, Journal of Power Sources Vol. 106, No. (1-2), pp. 102-108.
A Potential Alternative in the Electric Utility 207
Lee H.C., Choi J.W., Kim C.D., 2001. Efficient eigenvalue assignment for linear time-
varying systems. In: Proceedings of the IEEE International Symposium on Industrial
Electronics, Vol. 3, pp. 1696-1701.
Leeper J.D., 1999. The hybrid cycle: integration of a fuel cell with a gas turbine. ASME Paper
GT-430.
Liese E.A., et al., 1999. Technical development issues and dynamic modeling of gas turbine
and fuel cell hybrid systems. ASME paper GT-360.
Lukas M,D,, Lee K,Y,, Ghezel-Ayagh H., 1999. Development of a stack simulation model for
control study on direct reforming molten carbonate fuel cell power plant. IEEE Trans.
Energy Conversion Vol. 14, No. 4, pp. 16511657.
Lukas M.D., Lee K.Y., Ghezel-Ayagh H., 2000. Operation and control of direct reforming
fuel cell power plant. in Proceedings of the IEEE Power Engineering Society Winter
Meeting.
Lukas M,D,, Lee K,Y,, Ghezel-Ayagh H., 2001. An explicit dynamic model for direct
reforming carbonate fuel cell stack. IEEE Trans. Energy Conversion Vol. 16, No.3, pp.
289-295.
Lasseter R., 2001. Dynamic models for micro-turbines and fuel cells, IEEE/PES Summer
Meeting, Vol. 2, pp. 761-766.
Ljung L., Glad T., 1994. Modeling of Dynamic Systems, Englewood Cliffs, N.J., Prentice
Hall.
Ljung L., 1999. System Identification - Theory for the User, 2nd edition. Upper Saddle River,
N.J., Prentice Hall.
Massardo A.F., Lubelli F., 2000. Internal reforming solid oxide fuel cell-gas turbine
combined cycles (IRSOFC-GT): Part A-Cell model and cycle thermodynamic analysis, J.
Eng. Gas Turbines and Power, Vol. 122, No. 27, pp. 27-35.
MATLAB, 2006. High-performance numeric computation and visualization software, The
Mathworks Inc., Natick, MA.
Matsumoto S., Sasaki A., Urushibata H., Tanaka T., 1990. Performance model of molten
carbonate fuel cell. IEEE Trans. Energy Conversion Vol. 5, No. 2, pp. 252-258.
Monn M., 1995. Project brings fuel cell closer to utility use. Power Vol. 139, No. 4,
pp. 7679.
Moore T., 1997. Market potential high for fuel cells, EPRI J., Vol. 22, No. 3, pp. 6-17.
Moscinski J., Ogonowski Z., 1995. Advanced control with Matlab and Simulink. London:
Ellis Horwood.
Nagrath I.J., Kothari D.P., 1994. Power System Engineering, Tata McGraw-Hill, New Delhi.
Nguyen C.C., 1987. Arbitrary eigenvalue assignment for linear time-varying multivariable
control systems. Internationl Journal of Control, Vol. 45, No. 3, pp. 1051-1057.
Padulls J., Ault G.W., McDonald J.R., 2000. An integrated SOFC plant dynamic model for
power systems simulation, Journal of Power Sources, Vol. 86, No. (1-2), pp. 495-500.
Paisley M., Anson D., 1997. Gasification for gas turbine based power generation, ASME
paper GT-5.
Paradise, Parametric Robustness Analysis and Design Interactive Software Environment,
2004. DLR Institute of Robotics and Mechatronics, Robust Control, Germany.
Patel R., Bhatti T.S., Kothari D.P., 2001. Improvement of power system transient stability
using fast valving a review, Electric Power Components and Systems, Vol. 29, No.10, pp.
927-938.
208 Francisco Jurado
Phillips C.L., Nagle. H.T., 1995. Digital control system: analysis and design 3rd ed. New
Jersey: Prentice Hall, cop.
Rao A.D., Samuelsen G.S., 2002. Analysis strategies for tubular solid oxide fuel cell based
hybrid systems, J. Eng. Gas Turbines and Power, Vol. 124, No.3, pp. 503-509.
Rowen W.J., 1983. Simplified mathematical representations of heavy duty gas turbines.
ASME Journal of Engineering for Power Vol. 105, No. 4, pp. 865-869.
Rowen W.J., 1988. Speedtronic Mark IV control system. Alsthom Gas Turbine Reference
Library, AGTR 880.
Salo K., Horvath A., Patel J., 1998. Pressurized gasification of biomass, ASME paper
GT-349.
Sanchez-Gasca J.J., Chow J.H., 1997. Computation of power system low-order models from
time domain simulation using a Hankel matrix, IEEE Trans. Power Systems, Vol. 12, No.
4, pp. 1461-1467.
Sanchez-Gasca J.J., Chow J.H., 1999. Performance comparison of three identification
methods for the analysis of electromechanical oscillations, IEEE Trans. Power Systems,
Vol. 14, No. 3, pp. 9951002.
Sasaki A., Matsumoto S., Fujitsuka M., Shinoki T., Tanaka T., Ohtsuki J., 1993. CO2
recovery in molten carbonate fuel cell system by pressure swing adsorption, IEEE Trans.
Energy Conversion Vol. 8, No. 1, pp. 26-32.
Sauer P.W., Pai M.A., 1998. Power System Dynamics and Stability, Prentice Hall, New
Jersey.
Scientific American, 1999. The future of fuel cells, Vol. July, pp. 72-83.
Shinoki T., Matsumura M., Sasaki A., 1995. Development of an internal reforming molten
carbonate fuel cell stack, IEEE Trans. Energy Conversion Vol. 10, No. 4, pp. 722-729.
Shobeiri T., 1987. Digital computer simulation of the dynamic operating behaviour of gas
turbines. Brown Boveri Review, Vol. 3.
Silveira J.L., Leal E.M., Ragonha L.F., 2001. Analysis of a molten carbonate fuel cell:
cogeneration to produce electricity and cold water. Energy Vol. 26, No.10, pp. 891-904.
Singhal S.C., 1999. Progress in tubular solid oxide fuel cell technology, in Proc. Solid Oxide
Fuel Cells, Honolulu, HI, USA, Oct. 17-22, 1999, pp. 39-51.
Smith J.R., Hauer J.F., Trudnowski D.J., 1993. Transfer function identification in power
system applications, IEEE Trans. Power Systems, Vol. 8, No. 3, pp. 1282-1290.
Sderstrm T., Stoica P., 1989. System Identification, London, Prentice Hall International.
Steinfeld G., Maru H.C., Sanderson R.A., 1996. High efficiency carbonate fuel cell/turbine
hybrid power cycles, in Second Workshop on very high efficiency fuel cell/advanced
turbine power cycles. In: Fuel Cells Conference.
Stephenson D., Ritchey I., 1997. Parametric study of fuel cell and gas turbine combined cycle
performance. ASME paper GT-340.
Stoica P., Soderstrom T., Friedlander B., 1985. Optimal instrumental variable estimates of the
AR-parameters of an ARMA process, IEEE Trans. Automatic Control, Vol. 30, No. 11,
pp. 1066-1074.
Taylor C.W., 1994. Power System Voltage Stability. New York: McGraw-Hill, Inc, Chap.2.
Tolbert, L.M., Peng, F.Z., Habetler, T.G., 1999. Multilevel converters for large electric
drives. IEEE Trans. Industry Applications, Vol. 35, No. 1, pp. 36-44.
Trudnowski D.J., Donnelly M.K., Hauer J.F., 1994. A Procedure for Oscillatory Parameter
Identification, IEEE Trans. Power Systems, Vol. 9, No. 4, pp. 2049-2055.
A Potential Alternative in the Electric Utility 209
Valasek M., Olgac N., 1995 a. Efficient eigenvalue assignments for general linear mimo
systems. Automatica, Vol. 31 b, No. 11, pp. 1605-1617.
Valasek M., Olgac N., 1995 b. Efficient pole placement technique for linear time-variant siso
systems. IEE Proceedings - Control Theory and Applications, Vol. 142, No. 5, pp.
451-458.
Van Overschee P., DeMoor B., 1996. Subspace Identification of Linear Systems: Theory,
Implementation, Applications, Kluwer Academic Publishers.
Veyo S.O., Lundberg W.L., 1999. Solid oxide fuel cell power system cycles. ASME paper
GT-419.
Waldheim L., Carpentieri E., 1998. Update on the progress of the Brazilian wood BIG-GT
demonstration project, ASME paper GT-472.
Wolovich W.A., 1968, On the stabilization of controllable systems. IEEE Trans. on
Automatic Control, Vol. 13, No. 5, pp. 569-572.
Working Group on Prime Mover and Energy Supply Models, 1994. Dynamic models for
combined cycle plants in power system studies. IEEE Trans. Power Systems, Vol. 9, No.
3, pp.1698-1708.
Zhu Y., Tomsovic K., 2002. Development of models for analyzing the load-following
performance of microturbines and fuel cells, Electric Power Systems Research, Vol. 62,
No. 1, pp. 1-11.
In: Fuel Cell Research Trends ISBN 1-60021-669-2
Editor: L.O. Vasquez, pp. 211-245 2007 Nova Science Publishers, Inc.
Chapter 5
Abstract
This chapter will introduce the reader to the reasons for miniaturizing fuel cells
and to the specifications required by this miniaturization. It will then show what kinds
of fuel cells can fit to these specifications and which fuels can be employed to supply
them. The techniques presently used for the realization of miniature fuel cells will
be described, underlining particularly the growing part of the microfabrication tech-
niques inherited from microelectronics. It will present an overview on the applications
of these latter techniques on miniature fuel cells by presenting several solutions devel-
oped throughout the world. It will finally detail, as an example, the complete fabrica-
tion process of a particular microfabricated fuel cell based on a silane-grafted porous
silicon membrane as the proton-exchange membrane instead of a common ionomer
such as Nafion .
1. Introduction
Numerous electronic devices such as computers, cellular phones, camcorders, have become
portable with the miniaturization of electronic components. Such equipments require ap-
propriate power supply, which have to be more and more effective especially in terms of
power density and lifetime, in order to be able to provide for the increasing number of
functionalities of these portable applications (like Internet on cellular phones, for example).
In this strong growth market, fuel cells represent promising power sources for these
applications. Indeed even if they are generally associated with much larger applications
such as vehicles and power plant, they can be of any size, given their modularity. It enables
E-mail address: tristan.pichonat@isen.iemn.univ-lille1.fr. Tl. +33 (0)320 197 947.
Fax. +33 (0)320 197 884.
212 Tristan Pichonat
Figure 1. Comparison between the specific energies of the fuel cells and the competing
batteries, after [2, 4].
to conceive very small size fuel cells, involving only a few thin cells with a small effective
area.
30 years ago, only the Nickel Cadmium (NiCd) technology was available for the supply
of portable electronics. The last twenty years have seen the emergence of three new tech-
nologies on the market: Nickel Metal Hidride (NiMH) since the late 1980s, Lithium-Ion
(Li-Ion) since 1991, and Lithium-polymer (LiPo) since 1996. The main battery manufac-
turers have been searching since to increase the lifetime of batteries but despite their effort
batteries remain heavy, bulky, of limited charge and above all polluting. Actually, the major
problem for batteries is that their energy is stored in heavy and voluminous sources such as
metal oxides and graphitic materials [13].
Compared with batteries fuel cells can produce electrical energy from much lighter ma-
terials such as gases, alcohols, and hydrides by catalytic electrochemical processes. They
theoretically provide energy densities 3 to 5 times higher than their competing batteries
(fig. 1). Autonomies of 30 days in stand-by mode and more than 20 hours in communica-
tion are so forecasted for a cellular phone within the same volume. Moreover, contrary to
present batteries which suffer from their low autonomy combined with frequent and long-
time recharging, the autonomy of fuel cells only depends on the size of the fuel tank and
the refilling is immediate by changing the fuel tank. This technology is also theoretically
non-polluting. As many arguments that presently put research in effervescence.
Yet miniaturizing fuel cells is still a challenge, with many issues to consider such as
the fabrication techniques, the cost and the customer safety [5]. Actually fuel cells are
two-part systems, the fuel as the energy source and the fuel cell as the energy converter,
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 213
Figure 2. Fuel cells from the french CEA microfabricated on a silicon wafer.
both which have to work as one. If the ability to store and manage efficiently the fuel will
mostly influence the complete volume and the autonomy of the fuel cell, the capability to
transform efficiently the specific energy of the fuels into electrical energy will depend on
the effectiveness of catalytic reactions on the electrodes of the fuel cell. The cost is then
bound, among other factors, to the catalyst itself, mostly platinum, even if in very small
amounts. Moreover using fuel cells in portable applications implies methods to provide a
high level of safety. This means, given that fuel cells are not generally sealed since they
require oxygen from air, that the fuel has to be isolated to avoid any risk of accidental direct
oxidation with air or user contact if toxic [2]. Other issues such as the management of the
heat and the water produced are also challenges that actors of the domain have to deal with.
Focusing on the miniaturization, two main research axes are possible for the global
conception of the small fuel cells: either the reduction in the size of existing fuel cells,
or the search for new materials and structures usable for their fabrication. Although the
first solution seems to be the easiest, it involves the miniaturization of all the auxiliary
components and the difficulty to adapt "macrofabrication" techniques to miniature sizes. So
research rather tends to apply itself to develop size-adapted methods with suitable materials.
One of the technological ways to miniaturize fuel cells is to have recourse to standard
microfabrication techniques mainly used in microelectronics and more especially the fabri-
cation of micro- and nano-electro-mechanical systems (MEMS/NEMS). Actually more and
more papers show the interest in developing MEMS-based fuel cells, either directly with
silicon substrates (fig. 2), or adapting the methods to other substrates such as metals or
polymers. These techniques enable notably mass fabrication at low cost (very large number
of devices on a very small area) and then could lead to the reduction of the global cost of
the miniature fuel cells.
The significant issues concerning the miniaturization of fuel cells will be developed,
showing notably which kinds of fuel cells can fit to the requirements of portable electronics
and which kinds of fuels may be appropriate for the supply of small fuel cells. Major mi-
crofabrication techniques will be briefly introduced, in order to make more understandable
the review on microfabricated fuel cells. Finally the complete fabrication process of a mi-
214 Tristan Pichonat
cromachined fuel cell based on a silane-grafted porous silicon membrane will be described
as an example.
2. Miniaturization
2.1. Microelectronics and MEMS
Historically microfabrication techniques have first been developed to meet the requirements
of microelectronics, but they also have allowed the emergence and the development of a new
research field where mechanical elementscan be manufactured and actuated with electrical
signals at a micro- and even nanometer scale. To describe this emerging research field, R.
T. Howe and others proposed the expression Micro Electro Mechanical Systems or MEMS
in the late 1980s [14]. MEMS is not the only term used to describe this field which is also
called as micromachines, for instance in Japan, or more broadly referenced as microsystem
technology (MST) in Europe.
These submillimeter devices are machined using specific techniques globally called
microfabrication technology. This definition also includes microelectronic devices, but in
addition to electronic parts, MEMS also features mechanical parts like holes, cavity, chan-
nels, cantilevers, or membranes. This particularity has a direct impact on their manufac-
turing processes which need to be adapted for thick layer deposition, deep etching and to
introduce special steps to free the mechanical structures. Moreover, many MEMS are now
not only based on silicon but are also manufactured with polymer, glass, quartz or thin metal
films.
These MEMS are now part of our daily life with many applications such as inkjet heads,
accelerometers for crash air-bag deployment systems in automobiles, micromirrors for dig-
ital projectors (best example with the DLP (for Digital Light Processing), presently the
only all-digital display chip by Texas Instruments) but also pressure sensors for automotive
and medical applications. Further applications to come could be lab-on-chip for medical
analysis and radiofrequency MEMS for telecommunications (switches, filters, etc.). Two
examples of MEMS fabricated at the IEMN (Villeneuve dAscq, France), a microgripper
and a micromotor, are shown in figures 3 and 4.
Besides the many applications previously cited, there are three main reasons that prompt to
the use of MEMS technology today:
- Miniaturization of existing devices, like for example the production of silicon based
gyroscope reducing previously existing voluminous devices to a microchip;
- Development of new devices based on principles that do not work at larger scale, like
for instance the electro-osmotic effect in microchannels;
The first two aspects rejoin the motivations for the conception of miniature fuel cells.
The key idea is also to reduce the size to gain volume and so to enable low cost mass
fabrication, not denigrating the performances of the system. Furthermore the materials and
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 215
techniques used for fabricating MEMS are compatible with the fabrication of fuel cells, and
provide, for instance, easy ways to conceive microchannels to bring efficiently the fuel and
the oxidant to the membrane-electrode assembly, but mostly can offer to the fuel cell an
appropriate and dedicated micromachined support which may eventually be integrated with
other microelectronic or MEMS devices.
3. Microfabrication
3.1. Introduction
The designation microfabrication techniques gathers the collection of technologies used for
making microdevices [17]. To fabricate a microdevice, many processes must be performed,
one after the other, many times repeatedly. For instance the fabrication of memory chip
requires some 30 lithography steps, 10 oxidation steps, 20 etching steps, 10 doping steps,
and so on.
Typical microfabrication processes include the following steps: wafer cleaning, pho-
tolithography, etching, thin film deposition, doping (by either thermal diffusion or ion
implantation), bonding (wire and wafer bonding), and chemical-mechanical planarization
(CMP).The complexity of microfabrication processes can be described by their mask count.
This one represents the number of different pattern layers that will constitute the final de-
vice. Modern microprocessors are made with 30 masks while a few masks suffice for a
microfluidic device or a laser diode.
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 217
This section will briefly introduce the reader to the microfabrication techniques that
will be more generally used for applications in miniature fuel cells, i.e. photolithography,
etching, deposition, and sealing. These techniques mainly concern silicon as the basic
substrate, but are also fitted to other materials such as some polymers or foil materials.
For these materials, the specific techniques will be described in the following overview on
microfabricated fuel cells (section 4.). More informations on techniques and MEMS in
general can be found in [1821].
3.2. Photolithography
Photolithography or optical lithography is the transfer of a pattern by selective exposure to a
radiation source such as UV light from a photomask to a photosensitive material deposited
on a substrate [22].
This process involves a combination of several main steps: substrate preparation, pho-
toresist application, soft-baking, exposure, developing, and possible hard-baking. Crys-
talline silicon in the form of a wafer is the most employed material as the substrate but
glass or metal are also used as substrates. The photosensitive material is generally a light-
sensitive resist (photoresist) that experiences a change in its physical properties when ex-
posed to a radiation source. It is deposited on the substrate by spin-coating to produce a thin
and uniform film at the surface of the substrate. It is then soft-baked to harden the photore-
sist. A transparent plate with patterned chromium areas printed on it, called a photomask
(or shadowmask), is placed between the source of radiation and the wafer, selectively ex-
posing parts of the substrate to light. An exposed positive photoresist will thus become
soluble in a developing solution while a negative photoresist will become insoluble upon
illumination. Exposed areas (when using a positive photoresist) or unexposed areas (when
using a negative photoresist) are removed with the adapted developing solution, leaving the
substrate with the pattern of the original mask (or its opposite if negative resist is used).
The remaining photoresist provides a protecting layer for the etching of the underlying
substrate (fig. 5).
In this process, the substrate is placed inside a reactor supplied by different gases [21]. The
principle of the process is that a chemical reaction takes place between the source gases pro-
ducing a solid material which condenses on all surfaces inside the reactor. CVD is widely
used in the semiconductor industry to deposit various materials such as polycrystalline,
amorphous, and epitaxial silicon, carbon fiber, filaments, carbon nanotubes, SiO2 , silicon-
germanium, tungsten, silicon nitride, silicon oxynitride, titanium nitride, and various high-k
dielectrics.
The two most important CVD technologies in MEMS are the Low Pressure CVD
(LPCVD) and Plasma Enhanced CVD (PECVD). The LPCVD process produces layers
with excellent uniformity of thickness and material characteristics. The main problems
with the process are the high deposition temperature (higher than 600 C ) and the relatively
slow deposition rate. The PECVD process can operate at lower temperatures (down to 300
C ) thanks to the extra energy supplied to the gas molecules by the plasma in the reactor.
However the quality of the films tends to be inferior to processes running at higher temper-
atures. Moreover most PECVD deposition systems can only deposit the material on one
side of the wafers on 1 to 4 wafers at a time whereas LPCVD systems deposit films on both
sides of at least 25 wafers at a time.
Electrodeposition
It involves the coating of an electrically conductive sample with a layer of metal using
electrical current. The part to be plated is the cathode of the circuit while anode is made of
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 219
the metal to be plated. Both of these components are immersed into a solution containing
one or more metal salts as well as other ions that permit the flow of electricity. A rectifier
supplies a direct current to the cathode causing the metal ions in solution to lose their charge
and plate out on the cathode. As the electrical current flows through the circuit, the anode
slowly dissolves and replenishes the ions in the bath. Other electroplating processes use a
non-consumable anode such as lead.
Thermal Oxidation
This process allows the deposition of silicon dioxide on silicon. Silicon oxidation is made
at high temperature (typically between 800 and 1100 C ) in an oven. It is performed with
the reaction of an oxidizing gas (generally O2 ) on the silicon surface. It can be made more
effective by the use of water in the form of vapour. This process is particular because the
material for the deposition (atoms of Si) is taken directly from the substrate. For instance
when a thickness of 1 m is grown on the Si substrate, the thickness of the substrate is
decreased of a few hundreds of nm.
Epitaxy
This technology is quite similar to what happens in CVD processes, though here with this
process if the substrate is an ordered semiconductor crystal (i.e. silicon, gallium arsenide),
it is possible to keep on building on the substrate with the same crystallographic orientation
with the substrate acting as a seed for the deposition. If an amorphous/polycrystalline
substrate surface is used, the film will also be amorphous or polycrystalline.
There are several technologies for creating the conditions inside a reactor needed to
support epitaxial growth, of which the most important is Vapor Phase Epitaxy (VPE). In
this process, a number of gases are introduced in an induction heated reactor where only
the substrate is heated. The temperature of the substrate typically must be at least 50 % of
the melting point of the material to be deposited.
An advantage of epitaxy is the high growth rate of material, which allows the forma-
tion of films with considerable thickness (>100 m). Epitaxy is a widely used technology
for producing silicon on insulator (SOI) substrates. The technology is primarily used for
deposition of silicon.
a film. A schematic diagram illustrates the principle in figure 6. The ions (generally Ar+ )
for the sputtering process are supplied through a plasma induced in a sputtering reactor.
Sputtering processes are generally applicable to all materials: metals, alloys, semiconduc-
tors, and insulators.
3.4. Etching
There are three categories of etching processes for the silicon and silicon compounds (sili-
con dioxide, silicon nitride): wet etching, electroetching, and dry etching. In the former the
material is dissolved when immersed in a chemical solution. The same thing happens with
electroetching but with the help of electrical current. In dry etching the material is sput-
tered or dissolved using reactive ions or a vapor phase etchant. Concerning other materials
to etch, either metal of polymer, wet etching processes are generally used.
Electroetching of Silicon
Discovered in the late 1950s by Ulhir [23] and recently put into light again with new appli-
cations for porous silicon such as photoluminescence [24, 25], this process can be consid-
ered as a particular wet etching [26]. The silicon substrate is immersed in an hydrofluoric
acid (HF) bath between two platinum electrodes. When a current is established between the
electrodes, the silicon is etched on the anode side. According to the current, it can either
lead to porous silicon with pores of controlled diameter (from the nm to several m, still
according to the applied current but also to the type of silicon and the concentration of the
HF bath used) or electropolishing (complete etching of the silicon). Masking of the material
can be achieved with a precious metal layer (gold, platinum) or with silicon nitride but for
a limited time since HF also etches silicon nitride. More informations can be found in [27].
Dry Etching
Reactive ion etching (RIE) is a dry etching process. The substrate is placed inside a reactor
in which several gases are introduced. A plasma is struck in the gas mixture using a radio
frequency (RF) power source, breaking the gas molecules into ions. The ions are acceler-
ated towards, and react chemically at the surface of the material, etching the material and
forming another gaseous material. The ions can also physically etch the material due to
their high kinetic energy. By changing the balance between the chemical and physical parts
it is possible to influence the anisotropy of the etching, since the chemical part is isotropic
and the physical part is highly anisotropic. Then the combination can form sidewalls that
have shapes from rounded to vertical. Etch conditions in an RIE system are very much
dependent on the many process parameters, such as pressure, gas flows, and RF power.
Deep reactive ion etching (DRIE) is a highly anisotropic etch process developed espe-
cially for MEMS and used to create deep and high aspect ratio holes and trenches in silicon
(and other materials). In this process, etch depths of hundreds of m can be achieved with
almost vertical sidewalls. The main technology for etching silicon is based on the Bosch
process where two different gases are alternated in the reactor: C4 F8 for creating a Teflon-
like passivation and protective layer on the surface of the substrate, and SF6 for etching the
substrate. The passivation layer is immediately sputtered away by the physical part of the
etching, but only on the horizontal surfaces and not on the sidewalls. Since the passivation
layer only dissolves very slowly in the chemical part of the etching, it builds up on the
sidewalls and protects them from etching. As a result, etching aspect ratios of 50 to 1 can
be achieved. The process can easily be used to etch completely through a silicon substrate,
and etch rates are 3 to 4 times higher than wet etching.
3.5. Micromolding
Also referred as soft lithography because they rely on the use of an elastomer, usually
poly(dimethylsiloxane)(PDMS), these types of processes consist in the replication of pat-
terns either identical or complementary to those of a master. Soft lithography allows to
achieve conformal and high resolution (nm scale) patterning with flat or curval substrates
and enables the use of materials incompatible with optical lithography or other conven-
tional microfabrication techniques. Different replication techniques are possible such as
222 Tristan Pichonat
3.6. Sealing
MEMS devices have movable mechanical parts which require free space to move. Given
their size, these devices are particularly sensible to surrounding environment (moisture,
temperature for instance) and contaminants. In order to maintain the stability of perfor-
mances and to protect the inside of the cavities from air, moisture, and any contaminants,
MEMS devices have to be sealed. Several techniques of MEMS sealing exist, like for
instance anodic bonding. It allows the bonding of glass and silicon by applying a high
temperature between 320 and 400 C and a high voltage (400 V up to 1 kV). The silicon
devices are encapsulated by the glass cap and thus remain protected from the atmosphere
at a low vacuum.
In the following section, a review of MEMS-based fuel cells is going to show that all these
previous techniques are used to manufacture miniature fuel cells and to make them more
effective. Actually, photolithography / etching and micromolding will be useful to transfer
patterns such as microchannels for fuel flows or membranes on a substrate and deposition
techniques will be helpful to make surfaces conducting (for current collectors) or insulat-
ing. Sealing techniques will be convenient to ensure the isolation of fuel and oxidant, and
likewise the safety of consumers.
The real micromachined parts of the micro fuel cell are generally the electrode plates
and the fuel delivery system. The latter is often achieved by microchannels. Nguyen et
al made a review of the different designs for the fuel delivery systems [29], illustrated in
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 223
figure 9.
Concerning the usable fuels for the supply of miniature fuel cells, the more investigated
are presently hydrogen and methanol. Methanol is a low cost liquide fuel with a high
energy density and can be directly oxidized in the fuel cell. Though due to the poor
kinetics of the charge transfer reaction at the heart of the fuel cell, direct methanol fuel
cells need substantial amounts of noble metals as catalyst, increasing the cost. Moreover
concentrated methanol can not be used as it leaks through the membranes of fuel cells, so
diluted methanol has to be used which drastically reduced energy density. On the other
hand hydrogen is the basic fuel but is far too dangerous to be stored directly as compressed
gas and too complicated to be stored as a liquid. Intermediate storages such as hydrides
(chemical or metal) and methanol are under consideration though it adds complexity to the
overall fuel cell [2]. Fuels such as ethanol [11] or formic acid [12, 13] are also currently
considered as interesting alternatives.
In the following sections the indexation of the different FCs will be made according to
the materials used as basic substrates. The sections will not focus on the different fuels used
(hydrogen for Proton-Exchange-Membrane FC or PEMFC, methanol for Direct Methanol
224 Tristan Pichonat
Figure 9. Typical fuel delivery designs: (a) direct supply, (b) with distribution pillars, (c)
parallel microchannels, (d) serpentine microchannel, (e) parallel/serpentine microchannel,
(f) spiral microchannel, (g) interdigitated microchannel, (h) spiral/interdigitated microchan-
nel, after [29].
FC or DMFC, ethanol, formic acid) but the fuel supply conditions will be specified for each
FC described.
A variant of this design was reported by Min et al (Tohoky Univ, Japan) who pro-
posed two structures of -PEFC using microfabrication techniques [33], the "alternatively
inverted structure" and the "coplanar structure". These structures used Si substrates with
porous SiO2 layers with platinum (Pt)-based catalytic electrodes and gas feed holes, glass
substrates with micro-gas channels, and a polymer membrane (Flemion S). In spite of a
reported enhancement [34], the FC reached poor results (only 0.8 mW cm2 for the alter-
natively inverted structure).
Starting from a classical miniature FC consisting in a MEA between two microma-
chined Si substrates but sandwiching Cu between layers of gold, Yu et al (Hong Kong
Univ.) were able to decrease the internal resistance of the thin-film current collectors. It
involved an increase in the performance of the FC, achieving a peak power density of 193
mW cm2 with H2 and O2 [35].
Yen et al [36] (Univ. of California / Pennsylvania State Univ.) presented a bipolar
Si-based micro DMFC with a MEA (Nafion -112 membrane) integrated and micro chan-
nels 750 m wide and 400 m deep fabricated using Si micromachining (Deep RIE). This
DMFC with an active area of 1.625 cm2 was characterized at near room temperature (RT),
showing a maximum power density of 47.2 mW cm2 at 60 C then 1M methanol was fed
but only 14.3 mW cm2 at RT. Since then, Lu et al [37] have enhanced the performance of
the DMFC to a maximum power density of 16 mW cm2 at RT and 50 mW cm2 at 60
C with 2 M and 4 M methanol supply with a modified anode backing structure enabling
to reduce methanol crossover (fig. 11). A further evolution of their technology is shown in
the paragraph 4.2..
Yeom et al (Univ. of Illinois) reported the fabrication of a monolithic Si-based mi-
croscale MEA consisting of two Si electrodes, with catalyst deposited directly on them,
supporting a Nafion -112 membrane in-between [38]. The electrodes were identically gold-
covered for current collecting, and were also covered with electrodeposited Pt black. The
electrodes and the Nafion membrane were sandwiched and hot-pressed to form the MEA.
The complete fuel cells were tested with various fuels: H2 , methanol and formic acid and
reached 35 mW cm2 , 0.38 mW cm2 and 17 mW cm2 respectively at RT with forced
226 Tristan Pichonat
Figure 11. Picture of a silicon wafer with flow channels, after [37].
Figure 13. Photograph and schematic views of the final DMFC, after [48].
Figure 14. From the wafer to the phone, the fuel cells by the CEA, after [50].
six-cell pack. An average power density of 144 mW cm2 was measured with H2 and air
breathing, which was one of the highest performance currently reported.
Since many years now, Morse et al (Lawrence Livermore National Laboratory) have
been working on SOFCs and PEMFCs [51]. They combined thin film and MEMS tech-
nologies to fabricate miniature FCs. A silicon modular design served as the platform for
FCs based on either PEM or SO membrane. In 2002, the PEM cell yielded a computed
peak power of 37 mW cm2 at 0.45 V and 40 C whereas the SOFC reached 145 mW
cm2 at 0.35 V and 600 C . Since 2002, the UltraCell company has an exclusive license
with Lawrence Livermore National Laboratory to exploit their technology to develop com-
mercial FCs.
230 Tristan Pichonat
Figure 15. Franhofer IZM prototype of miniature fuel cell powering a LED, after [53].
Figure 16. Structure of the passive air breathing direct formic acid fuel cell, after [13].
one had a peak power density of 20 mW cm2 and the 5-cell compressed stack achieved
250 mW cm2 .
Though they do not clearly refer to microfabrication techniques, another interesting
metallic miniature FC structure can also be reported, once again coming from the University
of Illinois, and proposed by Ha et al [12]. They described the design and performance
of a passive air breathing direct formic acid FC (DFAFC). The MEA was fabricated in
house with catalyst inks (Pt black powder for cathode and Pt-Ru black for anode, Millipore
water and 5 % Nafion solution) directly painted on a Nafion -117 membrane. The current
collectors were fabricated from titanium (Ti) foils electrochemically coated with gold. The
miniature cell at 8.8 M formic acid produced a maximum power density of 33 mW cm2
with pieces of gold mesh inserted between the current collector and the MEA on both sides
of the MEA. With Pd black used as the catalyst at the anode side [13], they recently obtained
a maximum power density of 177 mW cm2 at 0.53 V for their passive DFAFC with 10 M
formic acid.
Figure 17. Assembled polymeric micro fuel cell, after [55]. (a) open view and (b) assem-
bled view.
collector and corrosion protection layer. The Gaussian shape allows gold to cover all sides
of the channel. In this -FC, water generated by the reaction was utilized for gas humidify-
ing. The flow channel has a spiral shape which enables the dry gas in the outer spiral line
to become hydrated by acquiring some of the moisture from the adjacent inner spiral line.
The MEA consists in a Nafion 1135 membrane with a hydrophobic carbon paper and a
diffusion layer (carbon powder and PTFE) on one side and coated with a catalyst layer (ink
with Nafion -112 solution and Pt carbon) on the other side. Silver conductive paint was
printed on the other side of the carbon paper to increase its conductivity and to contribute
to collect current. In the final step, the two PMMA substrates and the MEA were bonded
together using an adhesive gasket. H2 was supplied by hydride storage, the air by an air
pump with a constant flow rate of 50 mL min1 , O2 flow with a constant flow rate of 20
mL min1 , all tests were performed at RT. A high power density of 315 mW cm2 at 0.35
V has been achieved when O2 is supplied at the cathode side, 82 mw cm2 with air.
Another PMMA-based micro PEMFC has been developed by Hsieh et al. The technol-
ogy included excimer laser lithography for patterning the flow field in the PMMA struc-
ture [56]. The channels were about 400 m wide, and 200 m deep with a rib spacing of 50
m for both anode and cathode. Electrically conductive regions acting as current collector
for electrodes were defined by sputtering 200 nm-thick copper film on the flow-field plates.
The MEA was based on Nafion -117 membrane with thin sputtered Pt film on both sides.
The assembly was then completed by 2 mm PMMA gaskets on both sides and sealed by
surface mount technology or hot-pressing. A power density of 25 mW cm2 at 0.65 V was
obtained at room temperature, when supplied by H2 and O2 . Further developments using
high aspect ratio UV-curable epoxy SU-8 instead of PMMA and silver as current collector
were reported with similar performances (30 mW cm2 with H2 and forced air) [57].
SU-8 epoxy was also the solution adopted by Cha et al to create microchannels by
simple UV photolithography and lift-off from a glass substrate [60]. Nafion-115 was used
as the PEM and sandwiched between two polymer chips with microchannels and two layers
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 233
of photosensitive polymers. Microchannels were 300 m wide and 100 m deep. Current
collectors were formed by sputtering Pt on polymer chips. Carbon black, Pt black, and Pt-
Ru catalyst were sprayed in the microchannels before assembling with the membrane. The
maximum power density observed with this all-polymer -DMFC was 8 mW cm2 with 2
M methanol as fuel and oxygen feeding.
A complete PEMFC system consisting in a PDMS substrate with micro flow channels
upon which the MEA was vertically stacked has been developped by Shah et al [58, 59].
PDMS microreactors were fabricated by employing micromolding with a dry-etched silicon
master. The PDMS spincoated on micromachined Si was then cured and peeled off from the
master. The MEA consisted in a Nafion -112 membrane where Pt was sputtered through a
Mylar mask. Despite an interesting method, this FC gave poor results (0.8 mW cm2 ).
Figure 18. Carbonized fluidic plates shown to the left of 1 cm2 Cirlex square, after [61].
Ref. First author Year Substrate PEM Pmax OCV Fuel / oxidant T
[11] Aravamudhan 2005 silicon Nafion -115 8.1 NR 8.5M ethanol/air RT
[60] Cha 2004 SU-8 Nafion -115 8 0.45 2M methanol/O2 RT
[55] Chan 2005 PMMA Nafion -1135 315 0.35 H2 /O2 RT
82 NR H2 /air RT
[49] Chen 2006 silicon Nafion -1135 144 0.93 H2 /air RT
[42, 43] Chu 2006 silicon porous silicon 94 NR 5M FA + 0.5 SA/air RT
[12, 13] Ha 2006 metal Nafion -117 177 0.53 10M FA/air RT
[53] Hahn 2004 stainless steel 80 NR H2 /air 60
[56, 57] Hsieh 2004 PMMA Nafion -117 25 0.78 H2 /O2 RT
2005 SU-8 Nafion -117 30 NR H2 /air RT
[51] Jankowski 2002 silicon NR 37 0.9 humidified H2 /air 40
[39] Jayashree 2005 silicon Nafion -112 28 NR FA/O2 RT
[32] Lee 2002 silicon Nafion -115 40 NR H2 /O2 RT
[47, 48] Liu 2006 silicon modified Nafion 4.9 NR methanol/air RT
[36, 37] Lu 2004 silicon Nafion -112 16/50 NR 2M methanol/forced air RT/60
[52] Lu 2005 stainless steel Nafion -112 34/100 NR 2M methanol/forced air RT/60
[50] Marsacq 2005 silicon Nafion 300 NR H2 /air RT
[30, 31] Myers 2002 silicon Nafion 60 NR H2 /NR RT
[33, 34] Min 2006 silicon Flemion 0.8 NR H2 /air RT
[54] Mller 2003 metal Gore MEA 50 0.87 H2 /O2 RT
[61] Park 2006 pyrolyzed Cirlex Nafion -115 1.21 NR H2 /O2 RT
[75, 76] Pichonat 2006 silicon porous silicon 58 0.34 H2 /air RT
[58, 59] Shah 2004 PDMS Nafion -112 0.8 0.79 H2 /air 60
[62] Wan 2006 PTFE DE 2021 80 0.985 H2 /air RT
[45, 46] Xiao 2006 silicon Nafion -117 13.7 NR H2 /O2 RT
[38, 40] Yeom 2005 silicon Nafion -112 35 NR H2 /O2 RT
2006 Silicon Nafion 12.3 0.3 10M passive FA/air RT
[35] Yu 2003 silicon Nafion -112 193 NR H2 /O2 / RT
236 Tristan Pichonat
One can also notice that besides the use of microfabrication techniques, all the differ-
ent presented technologies have a major point in common which is the use of Nafion or
Nafion -like ionomer as the proton exchange membrane. However these proton-conducting
polymers are not really suitable for microfabrication: they can not be patterned with stan-
dard photolithography and their volumetric changes with humidification are a major
problem to deal with at a micrometer scale. Yet alternate solutions are currently under
active development. Actually, there have been several papers on membraneless laminar
flow-based microfluidic fuel cells [6365], which do not require any proton-exchange
membrane. These micro fuel cells use the liquid-liquid interface between the fuel and the
oxidant. An aqueous stream containing a liquid fuel, such as formic acid, methanol, or
dissolved hydrogen, and an aqueous stream containing a liquid oxidant, such as dissolved
hydrogen, permanganate, or hydrogen peroxide, are introduced into a single microfluidic
channel in which the opposite sidewalls are the anode and cathode [68]. Thanks to the
dominant laminar convective transport, fuel crossover can be avoided, and so the anode
dry-out and cathode flooding. Furthermore these fuel cells are media flexible, i.e. the
composition of the fuel and oxidant streams (e.g. pH) can be chosen independantly,
allowing to improve reaction kinetics and open cell potentials. For instance, using a new
dual electrolyte H2 /O2 fuel cell system, Cohen et al exhibited thus an OCV of 1.4 V and
1.5 mW was generated from a single planar device with Si microchannels [66]. With the
integration of a gas-diffusion electrode enabling oxygen delivery directly from air, the
laminar flow-based fuel cell proposed by Jayashree et al achieved performances as high
as 26 mW cm2 with 1 M formic acid as fuel [68]. This novel technology is thus already
competing with proton-exchange membrane fuel cells and allows to consider further
miniaturization of fuel cells.
In the following section, another approach consisting in using standard MEMS tech-
niques and chemical grafting to create a proton-conducting membrane will be described as
an example of a miniature fuel cell manufactured thanks to microfabrication technology.
The complete process for the fabrication of the proton-conducting membranes, previously
reported in [69, 73], can be described as follows. A 4-inch 520 m thick n+ (100)-oriented
double-side polished silicon wafer is first thermally oxidized in an oven at 1000 C under
O2 and water steam flows to obtain a 2 m thick SiO2 layer on both sides of the substrate.
These layers will allow the electrical insulation between the two parts of the future fuel cell.
Then these previous layers are covered with sputtered Cr-Au layers on both sides. The Cr
layers are used as adherence layers for the Au layers and are relatively thin (30 nm). The Au
layers are 800 nm thick and will serve as current collector layers for the fuel cells. They are
also useful as masking layers for the next different etchings since Au is not etched neither
by KOH solution nor by HF solution, the two wet etchants used in the next steps.
This is followed by classical photolithography (with adapted etch for the metal layers)
and chemical wet etching with a wet KOH solution to produce 50 m thick double-sided
membranes. These membranes are then made porous by anodization in an ethanoic HF bath
with an average pore diameter of 10 nm (size controlled with the input current density). In
order to be sure that all the pores are open on the rear side of the membranes, a short reactive
ion etching (RIE) is added to the process with a classical silicon etching process using SF6
and O2 gases. The figure 20 sums up the different steps of the microfabrication process and
the figures 21 and 22 show the result of the process.
The previous manufactured membranes are not usable as they are. Actually, PS is not
naturally proton-conducting. In order to make the membranes proton-conducting, a solu-
tion consists in grafting silane molecules containing proton-conducting groups on the pores
walls. A commercially available silane salt from United Chemical Technologies Inc (UCT)
containing three carboxylate groups have been used for the first investigations. The first step
consists in creating silanol functions (Si-OH) at the surface of PS. A soft process involving
UV-ozone cleaner has been successfully implemented. The grafting of silane molecules is
then realized by immersing the hydrophilic porous membranes into a 1 % silane solution in
ethanol for 1 h at RT and ambient air. In order to replace -Na endings from the silane salt
by -H endings to get the real carboxylic behaviour for the grafted function, membranes are
immersed for 12 hours in a 20 % solution of sulfuric acid, then carefully rinsed in deionized
water. To complete the FC assembly, E-tek carbon conducting cloth electrodes filled with Pt
(20 % Pt on Vulcan XC-72) are added on both sides of the membrane as a H2 / O2 catalyst.
Figure 23 shows the final single cell.
5.2. Results
Figure 23 shows (top view) a typical 8 mm by 8 mm FC realized with an active area (in
black on the figure) of 7 mm2 .
Measurements were carried out at RT. H2 feeding was provided by a 20 % NaOH so-
lution electrolysis while passive ambient air was used at the cathode. In order to bring the
238 Tristan Pichonat
Figure 20. Sketch of a complete process for manufacturing porous silicon membranes for
miniature fuel cells.
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 239
Figure 21. Photograph of the silicon wafer after the membrane etching process. 69 mem-
branes of 7 mm2 are micromachined on a 4-inch wafer.
Figure 22. SEM (scanning electron microscope) cross section view of a typical porous
silicon membrane made with the process in figure 20.
240 Tristan Pichonat
Figure 23. Membrane-electrode assembly (8 x 8 mm), scale comparison with a 1 cent euro
coin.
gas to the membrane, a home-made test cell was used in which up to 4 single cells can be
tested separately or together with a serial or parallel connexion. A maximum power density
of 58 mW cm2 at 0.34 V was reached with an open circuit voltage (OCV) of 0.68 V -fig.
24). The relatively low voltage is mostly due to the crossover of H2 through the pores.
In order to reduce this crossover and then to increase the OCV, the pore diameter should
be reduced at least as small as that of assumed pores in Nafion (3 to 5 nm). Moreover, to
increase the power density, the grafting density should be controlled to be sure that all the
internal surface of the pores is grafted. The use of a catalyst ink instead of a Pt-loaded
carbon cloth should create a more intimate link between the membrane and the catalyst
and further increase the performances. A study on the thickness of current collecting metal
layers [49] had also recently shown that the thickness of the Au layer was probably too thin
to provide the lowest electrical resistance.
6. Conclusion
Though no commercial product is available yet, microfabrication techniques have now
proved themselves to be very useful tools for the development of miniature fuel cells. Ac-
tually, these techniques enable the miniaturization and the mass fabrication of almost every
component of a fuel cell: flow channels for the proper circulation of fuel and waste with
photolithography and reactive ion etching or micromolding, catalyst support by etching and
deposition, current collector by metal deposition and proton-conducting membrane with
insulator deposition, photolithography, wet etching, electroetching and post-fabrication
chemical treatment.
Several examples of miniature fuel cells using microfabrication techniques have been
presented. Among the numerous solutions developed today, the basic structure of fuel cells
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 241
Figure 24. Performances of the miniature fuel cell obtained with a mesoporous silicon
membrane grafted with proton-conducting silane (voltage vs current density in continuous
line, power density vs current density in dashed line).
remains the same: thin film planar stack (generally silicon, foils, polymer or glass) with
commercial ionomer, most often Nafion , the reported layers being micromachined (mi-
crochannels or porous media) for gas/liquid management and coated with gold for current
collecting. Performances range from the tenth of W cm2 to several hundreds of mW
cm2 according principally to the fluids management and the sealing. As the basic mate-
rial for MEMS technologies, silicon remains the most employed material for MEMS-based
fuel cells, but foils and polymers have shown interesting perspectives for future commercial
developments. Thus in most of the cases, this is often the simple application of microtech-
nology to conventional fuel cell structures. Yet new ways of conceiving fuel cells using
real microscale effects such as the membraneless laminar flow-based fuel cells begin to ap-
pear. They show that new technological breakthroughs have to be expected, especially to
get rid of ionomer membranes that have notably the disadvantages to change in size with
the humidification and to be inappropriate with microtechnology.
Acknowledgement
All previously published figures were reproduced with kind permission from the editors and
/or the authors.
References
[1] Dyer, C.K. Scientific American 1999, 281 (1), 88-93.
[2] Dyer, C.K. In: 2004 IEEE Symposium on VLSI Circuits, Digest of Technical Papers
2004, 124-127.
[6] Morse, J.D.; Graff, R.T.; Hayes, J.P.; Jankowski, A.F. In: Proceedings of the Materials
Research Society Symposium, 2002; Vol. 575, pp 321-324.
[7] Srikar, V.T.; Turner, K.T.; Ie, T.Y.A.; Spearing, S.M. J. Power Sources 2004, 125,
62-69.
[8] Baertsch, C.D.; Jensen, K.F.; Hertz, J.L.; Tuller, H.L.; Vengallatore, S.K.; Spearing,
S.M.; Schmidt, M.A. J. Mat. Res. 2004, 19 (9), 2604-2615.
[9] Tang, Y.; Stanley, K.; Wu, J.; Ghosh, D.; Zhang, J. J. Micromech. Microeng. 2005, 15,
S185-S192.
[10] Chen, X.; Wu, N.J.; Smith, L.; Ignatiev, A. App. Phys. Lett. 2004, 84 (14), 2700-2702.
[11] Aravamudhan, S.; Rahman, A.R.A.; Bhansali, S. Sens. Actuators A 2005, 123-124,
497-504.
[12] Ha, S.; Adams, B.; Masel, R.I. J. Power Sources 2004, 128, 119-124.
[13] Ha, S.; Dunbar, Z.; Masel, R.I. J. Power Sources 2006, 158, 129-136.
[14] Chollet, F.; Liu, H.B. "A (not so) short Introduction to Micro Electromechanical Sys-
tems", version 2.0, Aug. 2006, <http://memscyclopedia.org/IntroMEMS.html>.
[15] Millet, O.; Bernardoni, P.; Rgnier, S.; Bidaud, P.; Tsitsiris, E.; Collard, D.; Buchail-
lot, L. Sens. Actuators A 2004, 114, 371-378.
[16] Legrand, B.; Collard, D.; Buchaillot, L. In: Proceedings of the 13th IEEE Inter-
national Conference On Solid-state Sensors, Actuators and Microsystems Transduc-
ers05, Soul, Korea, 2005; pp.57-60.
[17] http://en.wikipedia.org/wiki/Microfabrication
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 243
[18] Sze, S.M. VLSI technology, 2nd edition, McGraw-Hill, New York, 1988.
[19] Senturia, S.D. Microsystem Design, Springer Science, New York , 2001.
[21] http://www.memsnet.org/mems/
[22] http://en.wikibooks.org/wiki/Microwiki
[25] Lehmann, V.; Gsele, U. Appl. Phys. Lett. 1991, 58, 856-858.
[26] Halimaoui, A. In: Properties of Porous Silicon; Canham, L. (Ed.), Emis Datareviews
Series, INSPEC, 1997; pp 12-22.
[29] Nguyen, N. -T.; Chan, S.H. J. Micromech. Microeng. 2006, 16, R1-R12.
[30] Maynard, H.L.; Meyers, J.P. In: Proceedings of the 2nd Annual Small Fuel Cells and
Battery Technologies for Portable Power Applications International Symposium, New
Orleans, LA, U.S.A., 2000.
[31] Meyers, J.P.; Maynard, H.L. J. Power Sources 2002, 109, 76-88.
[32] Lee, S.J.; Chang-Chien, A.; Cha, S.W.; OHayre,R.; Park, Y.I.; Saito, Y.; Prinz, F.B.
J. Power Sources 2002, 112, 410-418.
[33] Min, K.B.; Tanaka, S.; Esashi, M. In: Proceedings of the 16th International Confer-
ence on MEMS; Kyoto, Japan, 2003, pp 379-382.
[34] Min, K.B.; Tanaka, S.; Esashi, M. J. Micromech. Microeng. 2006, 16, 505-511.
[35] Yu, J.; Cheng, P.; Ma, Z.; Yi, B. J. Power Sources 2003, 124, 40-46.
[36] Yen, T.J.; Fang, N.; Zhang, X.; Lu, G.Q.; Wang, C.Y. App. Phys. Lett. 2003, 83 (19),
4056-4058.
[37] Lu, G.Q.; Wang, C.Y.; Yen, T.J.; Zhang, X. Electrochim. Acta 2004, 49, 821-828.
[38] Yeom, J. ; Mozsgai, G.Z.; Flachsbart, B.R.; Choban, E.R.; Asthana, A.; Shannon,
M.A.; Kenis, P.J.A. Sens. Actuators B 2005, 107 (2), 882-891.
[39] Jayashree, R.S.; Spendelow, J.S.; Yeom, J.; Rastogi, C.; Shannon, M.A.; Kenis, P.J.A.
Electrochim. Acta 2005, 50 (24), 4674-4682.
244 Tristan Pichonat
[40] Yeom, J.; Jayashree, R.S.; Rastogi, C.; Shannon, M.A.; Kenis, P.J.A. J. Power Sources
2006, 160 (2), 1058-1064.
[41] Gold, S.; Chu, K.-L.; Lu, C.; Shannon, M.A.; Masel, R.I. J. Power Sources 2004, 135,
198-203.
[42] Chu, K.-L.; Gold, S.; Subramanian, V.; Lu, C.; Shannon, M.A.; Masel, R.I. J. Micro-
electromech. Syst. 2006, 15 (3), 671-677.
[43] Chu, K.-L.; Shannon, M.A.; Masel, R.I. J. Electrochem. Soc. 2006, 153 (8), A1562-
A1567.
[44] Yao, S.C.; Tang, X.; Hsieh, C.-C.; Alyousef, Y.; Vladimer, M.; Fedder, G.K.; Amon,
C.H. Energy 2006, 31, 636-649.
[45] Xiao, Z.; Yan, G.; Feng, C.; Chan, P.C.H.; Hsing, I-M. In: Proceedings of the 13th
International Conference on Solid-State Sensors, Actuators and Microsystems (Trans-
ducers 05), Seoul, Korea, 2005; pp 1856-1859.
[46] Xiao, Z.; Yan, G.; Feng, C.; Chan, P.C.H.; Hsing, I-M. J. Micromech. Microeng. 2006,
16, 2014-2020.
[47] Liu, X.; Suo, C.; Zhang, Y.; Zhang, H.; Chen, W.; Lu, W. In: Proceedings of the
8th Symposium on Design, Test, Integration and Packaging of MEMS/MOEMS (DTIP
06), Stresa, Italy, 2006; pp 360-364.
[48] Liu, X.; Suo, C.; Zhang, Z.; Zhang, H.; Wang, X.; Sun, C.; Li, L.; Zhang, L. J.
Micromech. Microeng. 2006, 16, S226-S232.
[49] Chen, C.; Li, X.; Wang, T.; Zhang, X.; Li, J.; Dong, P.; Zheng, D.; Xia, B. J. Micro-
electromech. Syst. 2006, 15 (5), 1088-1097.
[50] Marsacq, D. Clefs CEA 2004, 50-51.
[51] Jankowski, A.F.; Hayes, J.P.; Graff, R.T.; Morse, J.D. In: Proceedings of the Materials
Research Society Symposium, San Francisco, CA, U.S.A., 2002; Vol. 730, pp 93-98.
[52] Lu, G.Q.; Wang, C.Y. J. Power Sources 2005, 144, 141-145.
[53] Hahn, R.; Wagner, S.; Schmitz, A.; Reichl, H. J. Power Sources 2004, 131, 73-78.
[54] Mller, M.; Mller, C.; Gromball, F.; Wlfe, M.; Menz, W. Microsyst.Technol. 2003,
9, 159-162.
[55] Chan, S.H.; Nguyen, N.-T.; Xia, Z.; Wu, Z. J. Micromech. Microeng 2005, 15,
231-236.
[56] Hsieh, S.-S.; Kuo, J.-K.; Hwang, C.-F.; Tsai, H.-H. Microsyst. Technol. 2004, 10,
121-126.
[57] Hsieh, S.-S.; Hwang, C.-F.; Kuo, J.-K.; Tsai, H.-H. J. Solid State Electrochem. 2005,
9 121-131.
Microfabrication Techniques: Useful Tools for Miniaturizing Fuel Cells 245
[58] Shah, K.; Shin, W.C.; Besser, R.S. J. Power Sources 2003, 123, 172-181.
[59] Shah, K.; Shin, W.C.; Besser, R.S. Sens. Actuators B 2004, 97, 157-167.
[60] Cha, H.Y; Choi, H.G.; Nam, J.D.; Lee, Y.; Cho, S.M.; Lee, E.S.; Lee, J.K.; Chung,
C.H. Electrochim. Acta 2004, 50, 795-799.
[61] Park, B.Y.; Madou, M.J. J. Power Sources 2006, 162 (1), 369-379.
[63] Ferrigno, R.; Stroock, A.D.; Clark,T.D.; Mayer, M.; Whitesides, G.M. J. Am. Chem.
Soc. 2002, 124, 12930-1.
[64] Choban, E.R.; Markoski, L.J.; Wieckowski, A.; Kenis P.J.A. J. Power Sources 2004,
128, 54-60.
[65] Choban, E.R.; Spendelow, J.S.; Gancs, L.; Wieckowski, A.; Kenis P.J.A. Electrochim.
Acta 2005, 50, 5390-8.
[66] Cohen, J.L.; Westly, D.A.; Pechenik, A.; Abruna, H.D. J. Power Sources 2005, 139,
96-105.
[67] Cohen, J.L.; Volpe, D.J.; Westly, D.A.; Pechenik, A.; Abruna, H.D. Langmuir 2005,
21, 3544-50.
[68] Jayashree, R.S.; Gancs, L.; Choban, E.R.; Primak, A.; Natarajian, D.; Markoski, L.J.;
Kenis, P.J.A. J. Am. Chem. Soc. 2005, 127, 16758-9.
[69] Pichonat, T.; Gauthier-Manuel, B.; Hauden, D. Chem. Eng. J. 2004, 101 (1-3),
107-111.
[70] Pichonat, T. Study, conception and realization of a miniature fuel cell for portable
applications, Ph.D. Thesis, 2004; available at http://tel.ccsd.cnrs.fr/tel-00010710.
[75] Pichonat, T.; Gauthier-Manuel, B. In: Proceedings of the 8th Symposium on Design,
Test, Integration and Packaging of MEMS/MOEMS (DTIP06), Stresa, Italy, 2006; pp
354-359.
Chapter 6
Abstract
Hydrogen electrode reaction has been widely studied regarding to its applications in fuel
cell technology. In the last years, a lot of studies concerning kinetic and mechanistic aspects
about smooth and well defined surfaces have been recently published, but less is known about
the electrochemical behavior of technical electrodes, in half-cells and complete cells. In this
paper, the main manufacture methods of technical hydrogen electrodes reported in the
literature for low temperature fuel cells such as alkaline fuel cells, phosphoric acid fuel cells
(PAFCs), and polymer electrolyte fuel cells (PEFCs) is examined. The kinetics of hydrogen
electrode oxidation reaction both, in liquid and solid electrolytes is also reviewed. Previous
work constitutes a significant background that can help to develop technical hydrogen
diffusion anodes for application in practical fuel cells. In particular, the electrochemical
behavior of such anodes is correlated with that observed on well characterized surfaces.
1. Introduction
The potential ability of fuel cells (FCs) to become an alternative to the conventional energy
sources has been well demonstrated. However, high cost, durability, high system complexity
and lack of fuel infrastructure are drawbaks that still keep in standby their large-scale
commercialization [1, 2]. Table 1 summarizes some technical characteristics of the main
*
E-mail address: falcaide@cidetec.es
248 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
types of FCs, classified according to the electrolyte used and the operation temperature,
which, are:
Cathode catalyst Pt, Pt alloys Pt/Au, Pt, Ag Pt/Cr/Co, Pt/Ni Li/NiO LaSrMnO3
Polymer electrolytre fuel cells (PEFCs). The electrolyte is usually an ion exchange
membrane, which has to be hydrated for a proper operation. The working temperature is
between 60-120 C, limited by the polymer membrane.
Alkaline Fuel Cells (AFCs). The electrolyte is 35-50 or 85 wt. % KOH in aqueous
solution, depending on the operating temperature, smaller than 120 or about 240 C,
respectively. The electrolyte is circulated or retained in a matrix (usually of asbestos).
Phosphoric Acid Fuel Cells (PAFCs). The electrolyte used is a 85-95 vol. % H3PO4
solution retained in a matrix of silicon carbide. The working temperature is normally in the
range 150-220 C, but it can be increased to about 300 C, because of the stability of the
concentrated acid solution.
Molten Carbonate Fuel Cell (MCFCs). The electrolyte is a mixture of sodium, potassium
and lithium carbonates, retained in a ceramic matrix of LiAlO2. The working temperature is
around 600-700 C.
Solid Oxide Fuel Cells (SOFCs). Usually, Y2O3-stabilized ZrO2 (yttria-stabilized zirconia
YSZ) is used as electrolyte. The cell operates at 600-1000 C.
Nowadays, such FCs are in different stages of development and have different
applications. The only practical application of low temperature fuel cells considered in this
chapter are AFCs in the American space shuttles. On the other hand, PAFCs have been used
in stationary power generation plants since the 70s. Finally, PEFCs have experimented a
resurgence in the 90s due to their performance improvement, as a consequence of the use of a
new proton exchange membrane (Nafion) and new techniques that enhanced the efficiency of
platinum catalyst in the electrodes. This resurgence has been mainly directed towards portable
and transport applications rather than stationary applications [3].
Technical H2 Electrodes for Low Temperature Fuel Cells 249
Most of practical applications of low temperature fuel cells require high-purity hydrogen
as fuel, because platinum is used as electrocatalyst in the anode. Pure hydrogen is expensive
and hence, alternative sources of this product are required in a hydrogen-based economy.
Usually, it is produced by reforming processes that lead to a mixture that contains some
impurities such as CO2 and CO or diluents (e.g. N2). The consequence of the use of this fuel
is a loss in the fuel cell performance because the anode becomes poisoned. One way to avoid
this is to develop new electrocatalysts that could maintain or improve fuel cell performance.
This requires the knowledge of the anode behavior, that is, the mechanism of hydrogen
oxidation reaction (HOR) taking place in the anode.
Hydrogen diffusion anodes utilized in practical fuel cells are technical electrodes. In this
kind of electrodes, complications arise from the undefined structure of the catalyst.
Furthermore, most of the surface analytical methods used with smooth surfaces can not be
applied with highly dispersed technical electrodes. That is the reason because, traditionally,
HOR has been studied in model electrodes, used to reduce complexity and to simplify
interpretation of electrochemical experiments. For example, in model electrodes there is no
influence of gas flow nor effect of the ionomer, as would be the case for a technical electrode
measurement. Besides, another motivation to work with model electrodes lies in the
preparation of defined structures, like well-defined surfaces. In this scenario, the relationship
between electronic property and electrochemical reactivity may be evident in the
experimental measurements.
It seems that still, the results obtained with model electrodes can not be applied in a
straightforward way to technical electrodes. However, they are very interesting studies carried
out with technical electrodes in experimental conditions that reproduce as faithfully as
possible those found in real FCs.
In this chapter, we will describe first the main methods utilized to fabricate technical
electrodes, because it is well known that they can influence their electrochemical behavior.
Next, we will give some insights on the electrocatalysis and mechanistics aspects of HOR,
necessary to understand properly the studies made with technical electrodes, which we will
discuss later. Finally, we will try to correlate the electrochemical behavior of model
electrodes with those shown for technical electrodes on the basis of their chemical
composition.
The development of hydrogen electrodes for fuel cells (FCs) is directly related to the
development of the different kinds of FCs. In this sense, it has been established that a good
electrode structure has to verify the following conditions: (i) to allow a thin film of electrolyte
over the electrode with easy access of reactants and avoid drying and drowning, (ii) to have a
high surface area for reaction and (iii) to keep the electrolyte with constant thickness and
composition.
Hydrogen technical electrodes for low temperature fuel cells such as AFCs, PAFCs and
PEFCs are porous gas diffusion electrodes (GDEs) [4]. These electrodes have a large area
reaction zone with minimum mass transport hindrances, thus allowing the easy access of
250 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
reactants and removal of products. Then, the goal of GDEs fabrication is to provide a
structure that maximizes the catalyst utilization and avoids large mass transport losses in the
electrode.
GDEs typically consist, at least, of two layers: a gas diffusion layer (GDL) and an active
layer (AL). The GDL should provide mechanical support and electrical contact (current
collector), optimal distribution of reactant gases and a pore structure suitable for the remova
of liquid or vapor phase water (water management). The AL contains the catalyst, where the
electrochemical reaction takes place, but only in those sites where reactant, electrolyte and
catalyst meet, that is, in the three-phase zone or boundary (TPB).
There are a lot of articles and patents that describe the preparation of several types of GDEs.
These electrodes can be used as anodes or cathodes, depending on the electrocatalyst present
in them. We will report below general procedures for their preparation, mainly focused to the
manufacture of hydrogen diffusion electrodes [5]. The main steps in the fabrication of GDEs
are the preparation of the gas diffusion medium and the preparation and application of the
catalyst layer.
The preparation of a GDL involves the use of a substrate, carbon cloth or paper [6-8], which
are in general commercially available. They are usually treated to have
hydrophobic/hydrophilic properties, typically using polytetrafluoroethylene (PTFE) [9]. A
microporous carbon layer, made with carbon and PTFE with controlled porosity is applied to
the substrate in the catalyst layer side or to both sides [10]. This improves the gas and water
transport properties.
The PTFE-bonded electrode was introduced by Neidrach and Alford [11]. It consists of metal
blacks or carbon supported metal catalysts that are hydrophilic, blended with fine particles of
hydrophobic PTFE, that flow and bind the structure as a result of heat treatment during
fabrication. Thanks to its physical properties, PTFE flows and penetrates the pores, thus
allowing a good interfacial contact between catalyst and carbon and providing hydrophobic
gas pores for reactants.
During the electrode operation, some pores are filled with the electrolyte while others are
empty. This produces a large vapor/liquid interfacial area and increases the interfacial
reaction rate due to better catalyst utilization and electrode performance. The interface is quite
stable by the existence of high capillary forces in small pores. The pressure control system of
this electrode is simpler than that of porous sintered metal electrodes.
Nowadays, the main type of H2 electrodes used in AFCs is a PTFE-bonded electrode with
a Pt load of about 0.3 mg cm-2 [12-14]. PAFCs employ H2-diffusion PTFE-bonded electrodes
with Pt supported on carbon as catalyst for low loadings of 0.1-1.0 mg cm-2 [15]. In contrast,
PEFCs utilize H2 electrodes in which the catalyst (Pt/C) and the ionomer (Nafion) are
Technical H2 Electrodes for Low Temperature Fuel Cells 251
premixed before the catalyst layer is deposited (thin film method), being the catalyst loading
between 0.1-0.35 mg cm-2 [16-18].
The most important methods for catalyst layer preparation and application are:
Figure 1. Preparation of H2 technical electrodes for PEFCs employing the thin-film method.
Figure 2. Flow-sheet for fabrication of porous carbon H2 technical electrodes used in PAFCs.
Figs. 1-4 show flow chart diagrams for the manufacturing of H2 technical electrodes used
in PEFCs, PAFCs and AFCs, respectively. Nowadays, hybrid techniques involving
Technical H2 Electrodes for Low Temperature Fuel Cells 253
combination of two processes explained above, for example, spray and screen-printing, are
also tested to fabricate PEFCs electrodes [28]. On the other hand, vacuum deposition methods
such as sputtering and electrodeposition [18, 29], in which platinum is deposited onto a non-
catalyzed porous carbon support from a plating bath are used in a lesser extent to obtain
catalyst layers because its scaling-up is difficult.
Figure 3. Production steps for the manufacturing of dual-layer H2 technical electrodes using PTFE as a
powder for AFCs.
Figure 4. Production steps for the manufacturing of dual-layer H2 technical electrodes using PTFE as a
powder and in emulsion for AFCs.
254 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
The use of a particular catalyst for HOR in alkaline media depends on the operating
conditions and cost. For example, non-noble metals such as nickel [37] have been used in H2
technical electrodes for AFCs. It is known that the catalytic activity of Raney Ni (porous Ni
doped with Co or alloyed with Fe, Ti or Mo to improve its performance) approaches that of
noble metals in these systems. Another catalyst that has been used is nickel boride (Ni2B).
Sodium tungsten bronzes [38] and carbon supported Pt and Pt-Pd [39] have also been
employed to catalyze HOR in alkaline solutions.
Technical H2 Electrodes for Low Temperature Fuel Cells 255
HOR has a very high exchange current density and platinum is generally considered to be the
most effective electrocatalyst amongst the stable noble metals [40]. When pure hydrogen is
fed, the reaction is highly reversible and proceeds at high current densities with low
polarization loss, i.e. the overpotential associated with the hydrogen electrode is very low
even at very high current densities. However, when hydrogen contains impurities such as CO
or CO2, the anode overpotential increases and the cell performance is reduced. Tungsten
trioxide (WO3) has been found to catalyze HOR in acid solution and, mixed with platinum
(WO3-Pt) appears to have a synergistic effect with respect to its activity [41]. The practical
challenge in the HOR research is to find electrocatalysts with high exchange currents per unit
cost and resistance to poisoning, i.e., low cost and stable catalysts able to oxidize hydrogen at
a very high rate in the presence of impurities. Of course, carbon supported electrocatalysts are
nowadays generally used, since their performance depends on the overall active surface area.
Commercial application of H2 is generally based on the gas obtained by hydrocarbons
reforming and therefore, it is necessary to dispose of an anode electrocatalyst able to tolerate
a limited amount of CO, either under steady state operation or under transient conditions of
high CO content. In addition, CO2 adversely affects the catalyst performance, because a
reverse gas-shift reaction yielding CO could take place depending on the fuel composition
[42]. It seems that Pt-Ru better tolerates CO2 than Pt [43]. The preferred electrocatalysts for
HOR in the presence of CO include Pt and several Pt alloys or Pt mixtures with other noble or
non-noble metals [44]. The most used include ruthenium, molybdenum and tin.
At temperatures below about 125C, CO adsorption on platinum is very strong. Even few
ppms in the H2 stream cause substantial performance losses on the anode. Therefore, the use
of platinum alone is not viable for HOR in the presence of CO in low temperature fuel cells.
Thus, platinum-ruthenium, platinum-molybdenum and platinum-tin are being used as anode
electrocatalysts for hydrogen oxidation in the presence of CO because they tolerate low ppms
of CO without excessive polarization losses. Tungsten carbide (WC) also shows high CO-
tolerance [38, 44].
H2 2 H+ + 2 e- (1)
whereas in alkaline solution, the overall anodic process can be written as:
These overall electrode reactions can be splitted into the following sequences:
a) Transport of molecular hydrogen from the gas phase to the electrolyte and from the
electrolyte to the electrode surface:
256 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
b) Hydration and ionization of molecular hydrogen, which may take place through one or
both of the two following parallel reaction paths:
Hads H+ + e- (5)
or
Hads + OH- H2O + e- (6)
Reaction steps (4) and (5) or (6), the two latter in acid and alkaline media, respectively,
correspond to the Tafel-Volmer (T-V) mechanism, whereas reaction steps (7), in acid, and
(8), in alkaline media, are known as Heyrovsky-Volmer (H-V) mechanism.
In acid media, the rate equations for reactions (4), (5) and (7) can be formulated assuming
that adsorption of hydrogen atoms on the electrode surface follows a Langmuir isotherm, with
insignificant adsorption of molecular H2. It is assumed that such reactions occur on a
homogeneous surface. Then, current densities for Tafel, Volmer and Heyrovsky reactions can
be written as [45]:
j T = k T (1 - ) a H 2 - k -T 2
2
(9)
F - (1 - V ) F
j V = k V exp V - k -V (1 - ) a H + exp (10)
RT RT
F - (1 - H ) F
j H = k H (1 - ) a H 2 exp H - k -H a H + exp (11)
RT RT
where is the hydrogen overpotential, is the surface coverage by Had, kis and k-is (i = T,
V, H) are rate constants of the forward and backward reactions, respectively; i are the
Technical H2 Electrodes for Low Temperature Fuel Cells 257
transfer coefficients and the as with subcripts mean the thermodynamic activities of the
indicated species. The subcripts T, V and H, refer to Tafel, Volmer and Heyrovsky steps,
respectively.
When equilibrium is attained, the partial anodic current density is equal to the partial
cathodic current density of the same reaction. Tafel, Volmer and Heyrovsky exchange current
densities are given by equations (12), (13) and (14), respectively:
j 0,T ( ) = k T (1 - ) a H 2 = k -T 2
2
(12)
F - (1 - V ) F
j 0,V ( ) = k V exp V = k -V (1 - ) a H + exp =
RT RT
= k V1- k -V 1- (1 - ) a H +
(13)
F - (1 - H ) F
j 0,H ( ) = k H (1 - ) a H 2 exp H = k - H a + exp
RT RT
H
= k H1- k -H (1 - )
1
a 1H-2 a H + (14)
where j 0,i ( ) (i = T, V, H) denote the exchange current densities for reactions (2)-(4).
Taking into account these relations, equations (9)-(11) can be expressed as follows:
1 - 2 p H
2
j T = j 0,T 2 - (15)
1 - 0 p H 2 0
F 1 - aH+ - (1 - V ) F
j V = j 0,V exp V - exp (16)
0 RT 1 - 0 a H + RT
1 - p H 2 F aH+ - (1 - H ) F
j H = j 0,H exp H - exp (17)
1 - 0 p H 2 RT 0 a H + RT
j 0,V , j 0,H 10 j R , j R being the total residual current density due to oxidation and
258 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
reduction of impurities, to anodic metal dissolution and oxide formation, and to cathodic
reduction of oxygen.
The T-V mechanism is involved when j 0,T > 10 j 0,H and the steady-state condition of
hydrogen oxidation (or evolution) is given by
j = jT = jV (18)
The general steady-state current-potential curves can be determined from equation (18),
which can be simplified using certain approximations to be verified under particular
experimental conditions. The following cases can be considered:
RT j RT j
=- ln 1 - + ln1 - (19)
2 F j d,H 2 F j +
d, H
In equation (19) j d,H 2 and j d,H + are the limiting current densities due to convective
diffusion:
2 DH 2 F
j d,H 2 = c H 2 (20a)
H 2
DH + F
j d,H + = - c H + (20b)
H +
where Dis are the diffusion coefficients and is are the thicknesses of the diffusion layers of
the indicated species.
If the hydrogen diffusion alone determines the j curve, then j 0,T , j 0,V >> j d,H 2 , and
taking into account that at 25 C the second term in equation (19) is negligible for solutions
with values of c H + 0.1 M, one obtains j j d,H 2 at = 0.06 V.
Technical H2 Electrodes for Low Temperature Fuel Cells 259
ii) In the case j 0,T 10 j 0,V ; 0.1 j 0,T > j > 10 j 0,V , the Volmer reaction is the rate-
determining step, so that no limiting current density is expected, and the Tafel reaction is at
quasi-equilibrium. Then, equation (16) for anodic current density yields:
F
j V = j 0,V exp V (21)
0 RT
The following Tafel relation is obtained when the ratio of coverages in equation (21)
remains close to the unity in the studied range:
2 RT 2 RT
= ln j0,V + ln jV = a + b log jV (22)
F F
iii) In the case that j 0,V 10 j 0,T ; 0.1 j 0,V > j > 10 j 0,T , the Tafel reaction is a pure
chemical reaction. Considering that all partial reactions that occur before and after the Tafel
reaction are fast, the following expression for the hydrogen overpotential is established [46]:
RT j
=- ln 1 - (23)
2 F j l,T
where j l,T is a limiting anodic current density, which corresponds to the dissociation rate of
molecular hydrogen into adsorbed atoms.
According to equation (19), if the reaction is rapid, only a diffusion overpotential ( d )
appears. When the Tafel reaction is also the rate-determining step, the overall potential will
have then another contribution, T , coming from equation (23) and the following equation
can be derived:
RT j
d + T = - ln 1 - (24)
2 F jl
where 1 j l = 1 j d, H 2 + 1 j l,T .
The limiting current density for H2 oxidation, determined by the Tafel reaction, j l,T , is
proportional to the concentration of hydrogen. Since, j d, H 2 c H 2 (see equation (20)), j l is
260 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
RT 1 - 0 a H +
= ln (25a)
F 1 - 0 a H +
1 aH+ F
= exp - (25b)
0 1 0 aH+
RT
Combining equation (15) with equation (25a), and assuming that polarization due to
protons concentration is insignificant, a H + a H + 1 , which is equivalent to consider that
only the concentration of molecular hydrogen decreases in the plane of reaction with respect
to the bulk concentration, one can deduce that:
RT p H 2 j
=- ln - 2
(26)
2F p
H 2 j 1 -
0,T
1 - 0
From equation (26) a Tafel line with slope b = RT / 2F equal to 0.0128 V or 0.0296 V
-1
dec at 25 C is expected for small values of .
On the other hand, if equation (25b) is put into equation (15) and the overpotential is
assumed to be a large positive number (tending towards the limiting current density), the
following equation can be obtained:
1 - 2 p H
j T = j 0,T 2 (27)
1 - 0 pH2
pH2
j l,T = j 0,T (28)
p H 2
In contrast, when << 1, the maximum anodic limiting current density will be:
Technical H2 Electrodes for Low Temperature Fuel Cells 261
j 0,T pH2
j l,T = (29)
(1 - 0 ) 2
p H 2
Equations (28) and (29) hold for j d,H 2 , being expected that p H 2 p H 2 = 1 in this
case [47].
If the Tafel reaction alone is the rate-controlling step, then, j d,H 2 >> j l,T for 0 0.1
and j l = j l,T is reached at 0.05 V , whereas for 0 = 0.9, j l,an is attained at
0.120 V [46]. Only in these conditions, the comparison of theory and experiment is
simple, because of the existence of a limiting current in the anodic region [48].
When the Tafel and the Volmer reactions are both slow steps, the overall overpotential is
given by = T + V , that is, it contains reaction and charge-transfer contributions. The
overall overpotential can be expressed through the following generalized equation, which
takes into account the effect of coverage, [49]:
-1
4
RT j 2 RT j 1- 2
-1
j
=- ln 1 - 2
+ sinh -1
1 - (30)
2F F 2 j 0,V 1 - 0
2
j 1 - 1-
0,T j 0,T
1 - 0 1 - 0
Figure 5. Theoretical anodic hydrogen overpotential for the Tafel-Volmer mechanism, with <<1 and
= 1/2, according to equation (30) at different j 0,T j 0,V ratios: (a) 0.1, (b) 1 and (c) 10.
Temperature 298 K.
262 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
Fig. 5 shows theoretical j j 0,T vs. curves for several j0,T j 0,V ratios according to
equation (30), with 0 and = 0.5. As can be seen, curve a represents the contribution
of the reaction overpotential ( T ), whereas the remaining part corresponds to the charge
transfer overpotential ( V ).
On the other hand, if one considers the diffusion of molecular hydrogen coupled to the
reaction and charge transfer, the anodic hydrogen overpotential is given by
RT F
1/2
j
ln ( j ) - ln 1 - - exp -
RT
=- ln ( j 0 ) + (31)
F F jl RT
Equation (31) is valid for << 1 and represents a Tafel equation with a correction term,
which takes into account the lowering of concentration due to the rate control by diffusion
and reaction, and the reverse reaction [44]. The slope and the extrapolation to = 0 of the
straight lines obtained from equation (31) allows the determination of the charge transfer
coefficient, , and the exchange current density, j0 , respectively, without diffusion and
reaction rate control.
iv) When j 0,T , j 0,V < j R , the potential difference metal/solution is determined by two
different electrode reactions, which occur simultaneously. The anodic current density of one
reaction is equal to the absolute value of the cathodic current density of the other reaction.
Hydrogen oxidation or evolution may be one of the two electrode reactions.
The V-H mechanism is involved if j 0,H > 10 j 0,T and the steady-state condition for
hydrogen oxidation is given by
j = jV = jH (32)
v) In the case that j 0,V , j 0,H > 0.1 j , the situation is identical to case i) of Tafel-
Volmer mechanism.
vi) When j 0,H 10 j 0,V ; 0.1 j 0,H > j >10 j 0,V , the Volmer reaction is the rate-
determinig step and the following equation is obtained:
Technical H2 Electrodes for Low Temperature Fuel Cells 263
F
j V = 2 j 0,V exp V (33)
0 RT
vii) For j 0,V 10 j 0,H ; 0.1 j 0,V > j >10 j 0,H , the Heyrovsky reaction is the rate-
determining step and the coverage is given by the steady-state condition from equation (32).
Then, the anodic current density is:
1 - p H 2 F
j H = j 0,H exp H (34)
1 - 0 p H 2 RT
Under these conditions, the anodic current becomes negligible at overpotentials where the
ratio of pressures for molecular hydrogen decreases up tpclose to 0.
viii) When j 0,V , j 0,H < j R , the situation is the same as the case (iv) in the T-V
mechanism. The theoretically calculated current-potential curves for both T-V and H-V
mechanisms can only be discriminated in the case that Tafel reaction is the rate-determining
step [49, 50].
Table 2. Kinetic parameters for HOR on Pt(hkl) in 0.05 mol dm-3 H2SO4
and 0.1 mol dm-3 KOH (data taken from ref. [54]).
j0a/ mA cm-2
Pt(hkl) 274 K 333 K
acid alkaline acid alkaline
Pt(111) 0.21 0.01 0.83 0.3
Pt(110) 0.65 0.125 1.35 0.675
Pt(100) 0.36 0.05 0.76 -
a
Obtained from the micropolarization region
[59]. When using a solid electrolyte such as an ionic exchange membrane is used, the study is
then performed in a fuel cell environment.
The hydrogen oxidation kinetics on technical carbon-supported Pt-Ru catalysts was
investigated using the thin-film rotating electrode technique [60]. The electrode was prepared
placing a certain volume of catalyst suspension onto a glassy carbon rotating electrode. In this
case, the kinetic parameters were determined by nonlinear fitting of the polarisation curve in
the mixed kinetic- and diffusion-controlled region, assuming a Tafel slope of RT / 2F, which
corresponds to a TafelVolmer mechanism with the dissociative adsorption of hydrogen as
the r.d.s.
Vogel et al. [61] studied the electrochemical oxidation of hydrogen on carbon-supported
Pt electrodes in acidic electrolyte. The electrodes were made by pressing the catalyst with a
minimum amount of PTFE into a small gold disk.
Experimental current density vs. overpotential curves in hydrogen-saturated H3PO4 were
analysed in terms of theoretical equations derived from the model of porous flooded
electrodes considering a TafelVolmer mechanism, in which the Tafel reaction was the r.d.s.
These theoretical relationships can be reduced to simpler ones for two limiting cases: i)
equation (35) corresponds to a mixture of diffusion and reaction control and applies to
electrodes with relatively thick catalyst layers, high concentration of active catalyst, and high
exchange current density and ii) equation (36) related to a pure reaction control ocurring in
the case of a very thin catalyst layer, low concentration of active catalyst or a slow reaction
rate:
2 F
1 - exp -
j RT
= (1 - 0 ) (35)
jl 2F
1 - 0 1 - exp -
RT
and
2 F
1 - exp -
j RT
= (1 - 0 ) (36)
jl 2F
2
1 - 0 1 - exp -
RT
Figure 6 shows the comparative shapes for the theoretically normalised j - curves
decribed by equations (35) and (36). As can be seen, the shape of such curves normalised by
the limiting current density, jl, is function only of the hydrogen equilibrium coverage, 0 . For
the carbon-supported Pt electrodes used in the work of Vogel et al. [61], the experimental
data fitted better to equation (35), suggesting that the electrode is in a mixed region of
activation and diffusion control. The poisoning effect of CO on HOR in the Pt electrodes was
also studied by the same authors using similar equations to (35) and (36), but replacing j 0 2 ,
H
266 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
the exchange current density for hydrogen oxidation, by j 0 , the exchange current density
CO
j 0CO = j 0H 2 (1 - CO ) (37)
where CO is the coverage of platinum by CO. Results obtained from anodic polarisation for
CO-poisoned Pt electrodes confirm again the Tafel (r.d.s.)Volmer mechanism for H2
oxidation on platinum. The proportionality shown in equation (37) is consistent only with the
sequence proposed.
Figure 6. Comparison between the shapes of the normalised polarisation curves in the two limiting
cases: (a) and (c) diffusion-reaction control; (b) and (d) reaction control. 0 = 0.1 (curves a and b) and
0.95 (curves c and d). Temperature: 298 K.
The equations derived by Vogel et al. were further used to model the performance of a
Pt/C anode when it is fed with H2 containing a certain amount of CO and CO2 in H3PO4
electrolyte. The electrode was made with Pt supported on a carbon black, attached to a carbon
paper, being tested as a floating electrode in a half-cell assembly [62]. The porous platinum
anode was mechanically floated on top of the acid electrolyte, while pure or mixed hydrogen
were supplied through the electrode surface opposite to the active layer This configuration
minimizes diffusion resistances [63]. From the current density vs. overpotential plots for H2
and H2/CO mixtures, CO , j 0H and j0CO , can be calculated and used to measure the
2
catalytic activity at different overpotentials from equations (35) and (36) [64].
As pointed out in section 3 of this chapter, even though platinum is the most active
catalyst for hydrogen oxidation in acidic solution, it is quickly deactivated by CO. This
represents a trouble for PEFCs and PAFCs running with industrial hydrogen or reformate
feed streams. To solve this problem, some binary platinum alloys tolerant to CO were
developed [51].
An electrode kinetics for the evaluation of hydrogen oxidation on carbon-supported
binary Pt-M alloys (M = Cr, Mn, Fe, Co and Ni) was conducted at a solid polymer electrolyte
membrane interface [65]. The gas diffusion electrodes consisted of a carbon cloth substrate
Technical H2 Electrodes for Low Temperature Fuel Cells 267
and an active layer. The membrane electrode assembly (MEA) was fabricated by the hot-
pressing method and was incorporated in a single-cell test fixture with capabilities for half-
cell measurements. The current vs. potential curves thus obtained were analysed in terms of
equation (35), indicating a similar Tafel (r.d.s.)Volmer mechanism for Pt and Pt alloys.
Pozzio et al. [66] studied the anodic oxidation of pure hydrogen on gas diffusion
electrodes catalysed by Pt/C, Pt-Ru/C and Pt-Mo/C in aqueous H2SO4. The substrate was a
carbon paper and the GDL was composed of a mixture of carbon and PTFE. Electrochemical
measurements were carried out using a conventional three-electrode cell, where the GDE was
mounted on a holder furnished with a metal ring current collector and hydrogen back feeding.
The following non-linear semi-empirical equation was proposed to fit the anodic
galvanostatic polarisation data, obtained with pure H2 at room temperature:
where Eocp is the anode open circuit potential (OCP); R is the ohmic resistance; Rp is the
0
polarisation resistance at OCP, and f is a mass transfer related factor ( f 1 j limiting ). R and
catalytic sites, revealed to the authors some characteristics of the mechanism of hydrogen
oxidation on Pt-Ru/C and Pt-Mo/C catalysts. In particular, the decay of limiting current
density in these catalysts was partially related to the change of the HOR rate constant. Then,
they proposed the HeyrovskyVolmer mechanism given by equations (5) and (7), was
proposed for HOR in Pt-Ru and Pt-Mo supported electrodes. This proposal was reinforced by
the same authors investigating the H2 oxidation mechanism on Pt and Pt-Ru electrocatalysts
supported on a high surface area carbon powder, by electrochemical impedance spectroscopy
measurements at OCP, using the same experimental setup described above. In the light of the
analyisis of impedance data with the help of equivalent circuits it was confirmed the Tafel
Volmer mechanism for HOR on Pt/C catalysed electrodes, whereas a Heyrovsky-Volmer
mechanism was proposed for Pt-Ru/C [67].
Janssen et al. [68] investigated the mechanism of hydrogen oxidation on a commercial
fuel cell grade electrode on Toray paper with a platinum loading of 0.50 mg cm-2 [68]. The
solution used was 0.5 mol dm-3 H2SO4(aq) at 25 C. Experimental j- curves in the low
overpotential region up to 200 mV were obtained for pure and diluted H2. By comparing
experimental and theoretical relations, they concluded that HOR occurs according to the
TafelVolmer mechanism. To obtain the kinetic parameters of HOR, they further applied an
equation similar to (31), the corrected Tafel slope depending on 0, V 0,T , d,l 0,T and
0 , assuming that the gas compartment behaves like a continously fed and stirred tank
reactor [69]. They found that the reactivity of the electrode, which depends on the
electrochemical pretreatment applied, had a large effect on the kinetic parameters of H2
oxidation. The transfer coefficient of the Volmer reaction, V , was , with 0, V 0,T 1 ,
and the limiting current was determined by hydrogen diffusion for a very reactive gas
268 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
diffusion electrode. In contrast, the Tafel reaction was rate determining for a low reactive gas
diffusion electrode, whereas V increased and 0, V 0,T decreased with lowering
reactivity of the gas diffusion electrode. The effect of hydrogen concentration (partial
pressure of H2) on the charge transfer resistance or exchange current density for the Tafel
Volmer mechanism supported the results obtained.
HOR on Pt-Ru/C anodes has also been studied in a symmetrical H2|H2 polymer
electrolyte membrane fuel cell, because the polarisations involved in this reaction are small
[70]. This cell contained a MEA in which a proton exchange membrane is sandwiched
between a Pt-Ru catalysed anode and a Pt catalysed cathode. The anode (working electrode)
was then fed with H2, as well as the cathode, which was used as counter and reference
electrode. The analysis of the anodic polarisation scans together with the dependence of the
exchange current density on the partial pressure of hydrogen allowed concluding that the Pt-
Ru catalysed anode follows the TafelVolmer mechanism.
HOR on a Pt/C electrode in alkaline solutions has also been studied by us in a three-
electrode arrangement [71]. A semi-empirical equation in agreement with the TafelVolmer
mechanism was proposed to quantitatively explain the current vs. potential curves obtained in
a variety of experimental conditions:
j
E = OCP + A + B ln ( j an ) - C ln1 - an + R j an (39)
jl
where OCP is the open circuit potential; A and B are related to the charge transfer; C is
related to diffusion or reaction overpotential, and R is an ohmic-type resistance. Equation
(39) accounts for diffusion, adsorption, inhibition, charge transfer and ohmic-type drop in the
wetted electrode pores.
In addition, the effect of H2 partial pressure was modelized modifying equation (35) for
high overpotentials as follows:
RT 1 - 0 RT j an
+ R' j an
= - ln + ln
F 0 F j l - j an
(40)
complications due to water management and mass transport limitations for fuel transport, a
situation very far from the realistic operating conditions in a fuel cell, and it may lead to some
loss of information.
The transition to more realistic conditions, near fuel cell environment, is achieved using a
thin-film rotating disk electrode with a practical supported catalyst. It allows obtaining kinetic
parameters such as the the exchange current density and Tafel slopes, avoiding the use of
complex models. This is of fundamental importance, because small catalyst particles having a
significant fraction of surface atoms could behave differently compared to bulk materials.
The effects cited above are found when studying electrode kinetics at GDEs, in liquid or
solid electrolyte, but in this case, the complexity is related to the evaluation of diffusion and
ohmic overpotentials, closely connected to the charge transfer kinetics. However, choosing
appropiate models, it is possible to calculate different contributions from the simulation of
current vs. potential data.
HOR is so fast that it becomes very difficult to determine experimentally the kinetic
parameters of the reaction due to the interference of mass transport resistances. Nevertheless,
it appears to be general agreement regarding the mechanism of HOR on Pt and Pt alloys in
model and technical electrodes, the r.d.s. being the dissociative adsorption of H2 (Tafel
reaction). On the other hand, significant differences in activity are reported because of the
structure, composition and the electroactive area of the electrodes used in the measurements.
For example, the exchange current densiy, j0, of HOR on smooth Pt in acid media is reported
to be equal to 1.9 mA cm-2 [51]. On highly dispersed platinum supported on carbon, it could
be one order of magnitude greater due to its bigger specific area and particular surface
structure of the catalytic particles. In alkaline media, j0 reported for smooth and highly
dispersed Pt electrodes are one order of magnitude lower than those given in acid media
[51,71]. The same trend as in acid media is observed when comparing j0 between smooth and
highly dispersed Pt in alkaline electrolyte.
From the above considerations, it seems that the extrapolation of the results obtained with
model electrodes to technical ones is still open to discussion. In this way, Gonzalez et al. [72]
remarked that it is not always possible to extrapolate the results obtained in porous RDE to
technical diffusion electrodes in fuel cells.
8. Conclusion
The main preparation methods for H2 technical electrodes for low temperature fuel cells have
been examined. It has been demonstrated that the electrochemical behavior of the electrodes
depends on their fabrication, thus affecting the fuel cell operation. The preparation of the
catalyst of the active layer also influences its physical properties and electrochemical
performance. Different electrochemical approaches to study HOR on model, as a first
approximation, and technical electrodes, are exhaustively analyzed and their kinetic
parameters are discussed to evaluate their performance and system modelling. The existence
of a gap between the knowledge obtained from studies on model electrodes and technical
electrodes is emphasized. To optimize the performance of practical fuel cell electrodes, the
preparation of high surface area catalysts with the same characteristics as those shown at the
atomic level then seems necessary. In this sense, mechanistic studies are fundamental to
270 Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas
provide kinetics parameters, which can help us to design catalysts to operate in a real fuel cell
environment.
References
[1] Carrete, L.; Friedrich, K. A.; Stimming, U. Fuel Cells. 2001, 1, 5-39.
[2] Bagotzky, V. S.; Osetrova, N. V.; Skundin, A. M. Russ. J. Electrochem. 2003, 39,
919-934.
[3] Cropper, M. Fuel Cells. 2004, 4, 233-240.
[4] Kocha, S. S. In Handbook of Fuel Cells. Fundamentals Technology and Applications;
Vielstich, W.; Lamm, A.; Gasteiger, H. A.; Eds.; John Wiley & Sons: Chichester (UK),
2003; Vol. 3, pp 539-552.
[5] Gascoyne J. M.; Hards, G. A.; Ralph, T. H. International (WIPO) Publication N.
WO0205365, 2002.
[6] Narsavage, S. T.; Vine, R.; Emanuelson, R. U.S. Patent 3,859,138. 1975.
[7] Froberg, R. U.S. Patent 3,944,686. 1976.
[8] Christner, L.; Nagel, D.; Watson, P. U.S. Patent 4,115,528. 1978.
[9] Schulz, D. U.S. Patent 3,960,601. 1974.
[10] Bevers D.; Wagner, N.; Bradke, M. Int. J. Hydrogen Energy. 1998, 23, 57-63.
[11] Neidrach, L. W.; Alford, H. R. U.S. Patent 3,432,355. 1969.
[12] Gourec, P.; Poletto, L.; Denizot, J.; Sanchez-Cortezon, E.; Miners, J. H. J. Power
Sources. 2004, 129, 193-204.
[13] Han, E.; Eroglu, I.; Turker, L. Int. J. Hydrogen Energy. 2000, 25, 157-165.
[14] Lin, B. Y. S.; Kirk, D. W.; Thorpe, S. J. J. Power Sources. 2006, 161, 474-483.
[15] Sammes, N.; Bove, R.; Stahl, K. Curr. Opin. Solid State Mat. Sci. 2004, 8, 372-378.
[16] Mehta, V.; Cooper, J. S. J. Power Sources. 2003, 114, 32-53.
[17] Antolini, E. J. Appl. Electrochem. 2004, 34, 563-576.
[18] Litster, S.; McLean, G. J. Power Sources. 2003, 130, 61-76.
[19] Zelenay, P.; Ren, X. International (WIPO) Publication N. WO0245188, 2002.
[20] Hampden-Smith, M. J.; Kodas, T. T.; Atanassova, P. et al. U.S. Patent 7,098,163. 2006.
[21] Goller, G. J., U.S. Patent 4,185,131. 1980.
[22] Datz, A.; Schricker, B. U.S. Patent 6,645,660. 2003.
[23] Kawakara, T. U.S. Patent 6,653,252. 2003.
[24] M. Neergat, A. K. Shukla. J. Power Sources. 2001, 102, 317-321.
[25] Gervais, W. U.S. Patent 6,679,979. 2004.
[26] Srinivasan, S. Fuel Cells. From Fundamentals to Applications; Springer: New York
(NY), 2006; pp 276-278.
[27] Solomon, F.; Grun, C. U.S. Patent 4,379,772. 1983.
[28] Abaoud, H. A.; Ghouse, M.; Lovell, K. V.; Al-Motairy, G. N. Int. J. Hydrogen Energy.
2005, 30, 385-391.
[29] Alcaide, F.; Crespo, O.; Grande, H. Catal. Today. 2006, 116, 408-414.
[30] Wendt, H.; Spinac E. V.; Neto, A. O.; Linardi, M. Quim. Nova. 2005, 28, 1066-1075.
[31] Brandon, N. P.; Skinner, S.; Steele, B. C. H. Annu. Rev. Mater. Res. 2003, 33, 183-213.
[32] Martz, N.; Roth, C.; Fue, H. J. Appl. Electrochem. 2005, 35, 85-90.
[33] Dicks, A. L.; J. Power Sources. 2006, 156, 128-141.
Technical H2 Electrodes for Low Temperature Fuel Cells 271
[60] Wang, J. X.; Brankovic, S. R., Zhu, Y.; Hanson, J. C.; Adi, R. R. J. Electrochem. Soc.
2003, 150, A1108-A1117.
[61] Vogel, W.; Lundquist, J.; Ross, P.; Stonehart, P. Electrochim. Acta. 1975, 20, 79-93.
[62] Dahr, H. P.; Christner, L.G.; Kush, A.K.; Maru, H.C. J. Electrochem. Soc. 1986, 133,
1574-1582.
[63] Tarasevich, M. R.; Bogdanovskaya, V. A.; Grafov, B. M.; Zagudaeva, N. M.; Rybalka,
K. V.; Kapustin, A. V.; Kolbanovskii, Y. A. Russ. Journal Electrochem. 2005, 41,
840-851.
[64] Dahr, H. P.; Christner, L.G.; Kush, A.K. J. Electrochem. Soc. 1987, 134, 3021-3026.
[65] Mukerjee, S.; McBreen, J. J. Electrochem. Soc. 1996, 143, 2285-2294.
[66] Pozio, A.; Giorgi, L., Antolini, E.; Passalacqua, E. Electrochim. Acta. 2000, 46,555-561.
[67] Giorgi, L.; Pozio, A.; Bracchini, C.; Giorgi, R.; Turt, S. J. Appl. Electrochem. 2001,
31, 325-334.
[68] Vermeijlen, J. J. T. T.; Janssen, L. J. J.; Visser, G. J. J. Appl. Electrochem. 1997, 27,
497-506.
[69] Vermeijlen, J. J. T. T.; Janssen, L. J. J. J. Appl. Electrochem. 1993, 23, 1237-1243.
[70] Halseid, R.; Tunold, R. J. Electrochem. Soc. 2006, 153, A2319-A2325.
[71] Alcaide, F.; Brillas, E.; Cabot, P.-L., J. Electrochem. Soc. 2005, 152, E319-E327.
[72] Lizcano-Valbuena, W. H.; Paganin, V. A.; Leite, C. A. P.; Galembeck, F.; Gonzalez, E.
R. Electrochim. Acta. 2003, 48, 3869-3878.
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 273-379 2007 Nova Science Publishers, Inc.
Chapter 7
Abstract
Fuel Cells are growing in importance as sources of sustainable energy and will doubtless
form part of the changing program of energy resources in the future. Two key issues limiting
the widespread commercialization of fuel cell technology are better performance and lower
cost. The development of physically representative models that allow reliable simulation of
the processes under realistic conditions is essential to the development and optimization of
fuel cells, the introduction of cheaper materials and fabrication techniques, and the design and
development of novel architectures. The difficult experimental environment of fuel cell
systems has stimulated efforts to develop models that could simulate and predict multi-
dimensional coupled transport of reactants, heat and charged species using computational
fluid dynamic (CFD) methods. The strength of the CFD numerical approach is in providing
detailed insight into the various transport mechanisms and their interaction, and in the
possibility of performing parameters sensitivity analyses.
Among all kinds of fuel cells, proton exchange membrane (PEM) fuel cells are compact
and lightweight, work at low temperatures with a high output power density and low
environmental impact, and offer superior system startup and shutdown performance. These
advantages have sparked development efforts in various quarters of industry to open up new
field of applications for PEM fuel cells, including transportation power supplies, compact
cogeneration stationary power supplies, portable power supplies, and emergency and disaster
backup power supplies.
This chapter of "PEM Fuel Cell Modeling" looks at how engineers can model PEM fuel
cells to get optimal results for any application. A review of recent literature on PEM fuel cell
modeling was presented. A full three-dimensional, non-isothermal CFD model of a PEM fuel
cell with straight flow field channels has been developed in this chapter. The model was
developed to improve fundamental understanding of transport phenomena in PEM fuel cells
and to investigate the impact of various operation parameters on performance. This
comprehensive model accounts for the major transport phenomena in a PEM fuel cell:
convective and diffusive heat and mass transfer, electrode kinetics, transport and phase change
*
E-mail address: maherars@hotmail.com. Present address: Currently at the Higher Institute of Mechanical
Engineering, Head of Mechanical & Energy Engineering Department, Yefren, P.O. Box 65943, Libya.
274 Maher A.R. Sadiq Al-Baghdadi
mechanism of water, and potential fields. In addition, the hygro and thermal stresses in
membrane, which developed during the cell operation, were modeled and investigated. The
new feature of the algorithm developed in this model is its capability for accurate calculation
of the local activation overpotentials, which in turn results in improved prediction of the local
current density distribution. Fully three-dimensional results of the velocity flow field, species
profiles, liquid water saturation, temperature distribution, potential distribution, water content
in the membrane, stresses distribution in the membrane, and local current density distribution
are presented and analyzed with a focus on the physical insight and fundamental
understanding. The model is shown to be able to understand the many interacting, complex
electrochemical, and transport phenomena that cannot be studied experimentally. This chapter
is a practical summary of how to create CFD models, and how to interpret results.
Acknowledgments
I would like to express my deepest thanks and sincere gratitude to Prof. Dr. Haroun A.K.
Shahad Al-Janabi for his support, valuable advice, and encouragement which he provided
during the research.
My great appreciation is expressed to International Technological University (ITU),
London, UK for providing available facilities.
I am thankful to Prof. Dr. Frank Columbus, the president and editor-in-chief of the Nova
Science Publishers, for his invitation to the publishing program "Fuel Cell Research and
Development."
My gratitude and appreciation is due to my wife for her patience, care, and support
during the period of preparing this work.
Most of all, I want to thank my parents for their unconditional support throughout all the
years of my education. This study would not have been possible without them. Many thanks
also are due to my brother and sister.
Thanks to all who have helped me in carrying out this research.
Nomenclature
yi Mass fraction -
zf Fixed-site charge -
O 2
Oxygen concentration parameter -
Abbreviations Definition
Subscript Definition
a Anode
c Cathode
g Gas phase
l Liquid phase
w Water
1. Introduction
1.1. Background
Fuel Cells are electrochemical devices that directly convert the chemical energy of a fuel into
electricity. In general, fuel cells offer many advantages over conventional energy conversion
devices. They have higher energy efficiencies at both design and off-design [1].
A comparison of fuel cell systems versus other energy conversion systems is shown in
Figure 1.1. The figure provides a comparison between the exergy efficiencies for electricity
generation by the principal energy conversion system in use today [2]. The solid bars
represent what currently is possible. The figure demonstrates that the exergy efficiency of fuel
cell systems compares favorably to all other energy conversion systems using hydrocarbons
for fuel. In fact, fuel cell systems have the highest overall average efficiency of all systems
except hydroelectric plants. The non-solid extensions above each efficiency column indicate
the theoretical improvements that are predicted for each system. With these predicted
improvements in fuel cell system efficiency, fuel cell systems would gain an even larger
efficiency advantage over other systems.
Figure 1.1 also indicates fuel cell systems can be used to generate power over a large
range of power requirements. This scalability will be addressed shortly.
These higher efficiencies allow for a better use of natural resources such as hydrocarbons.
Being more efficient at off-design allows a fuel cell system to deliver peak power while still
being efficient at lower power requirements. This offers a distinct advantage over internal
combustion (IC) engines, whose efficiencies drop off drastically the further they operate from
their peak (i.e. design) power point [3]. In fact, the vast majority of time, a vehicle, whose
primary power source is an IC engine, operates far below peak power. This is also true of
most other energy conversion systems, depending of course, on the application.
PEM Fuel Cell Modeling 279
Fuel cell systems also are easily scaled, allowing them to be used for small applications
such as the power source for a personal computer as well as large applications like a
stationary power plant. Large or small, all these applications harness the fuel cells inherent
efficiency advantages to make better use of a fuel's chemical energy.
As the world becomes more industrialized, a second key advantage of fuel cells, lower
emissions, becomes essential for controlling global pollution. With fuel cells, the conversion
of chemical energy to electrical energy is accomplished electrochemically only water is
created. Although some emissions do occur during the fuel reforming process, temperatures
are not high enough for NOx, sulfur components are removed prior to reforming so that SOx
formation does not occur, and much less CO2 is released due to the higher average
efficiencies of fuel cell systems. In addition, CO concentrations are too small by necessity and
un-reacted hydrocarbons are re-circulated in the system [2, 4]. Furthermore since there are no
moving parts in a fuel cell system, such systems operate much more quietly, resulting in less
noise pollution. While fuel cell systems offer an excellent alternative for powering vehicles,
at this stage in their development they might be most economically feasible in stationary
residential power applications [5].
Figure 1.1. Exergy efficiencies of the principal types of energy conversion systems [2].
According to their electrolytes, fuel cells are classified into four types; PEMFC (Proton
Exchange Membrane Fuel Cell), PAFC (Phosphoric Acid Fuel Cell), MCFC (Molten
280 Maher A.R. Sadiq Al-Baghdadi
Carbonate Fuel Cell), and SOFC (Solid Oxide Fuel Cell). An exception to this classification
is the DMFC (Direct Methanol Fuel Cell) which is a fuel cell in which methanol is directly
fed to the anode. The electrolyte of this cell is not determining for the class. Table 1.1
compares the different types of fuel cell systems [2, 5-8]. A schematic representation of a fuel
cell with reactant and product, and ions flow directions for these types of fuel cells are shown
in Figure 1.2 [6].
From the Table 1.1 can conclude that the PEMFC is the only fuel cell that excels in all
the characteristics essential for private vehicle applications. Additional advantages that the
PEMFC offers over some of the other fuel cells are that the PEMFC is a less complicated
system to implement and has a longer expected lifetime [7, 8]. Among all kinds of fuel cells,
proton exchange membrane (PEM) fuel cells are compact and lightweight, work at low
temperatures with a high output power density, and offer superior system startup and
shutdown performance [9, 10]. These advantages have sparked development efforts in various
quarters of industry to open up new field of applications for PEM fuel cells, including
transportation power supplies, compact cogeneration stationary power supplies, portable
power supplies, and emergency and disaster backup power supplies [10, 11].
It is believed that there will be a time in the future when global energy demands will be
met by some source other than fossil fuels. It is believed that hydrogen will play a major role
in such a future [11]. The concept of a hydrogen economy describes an economy where the
principal source of energy is hydrogen related. Fuel cells, in particular proton exchange
membrane fuel cells (PEMFC), are expected to play a major role in a future hydrogen
economy [12]. Fuel cells are particularly attractive for use in vehicles as a replacement to the
combustion engine [13]. The low operating temperatures of a PEMFC allows for easy start up
and quick response to changes in load and operating conditions.
Power Density
0.7 8.1 3.8 6.5 > 1.5 0.8 1.9 0.1 1.5 1.5 2.6
(kW/m2)
Practical
60 60 60 55 55-65 60-65
Efficiency (%)
Combine heat and Combine heat and power
power for for stationary
Transportation, Space, Military,
Applications decentralized decentralized systems
Energy storage systems
stationary power and for transportation
systems (trains, boats, )
Start-Up Time min sec sec-min hrs hrs hrs
Figure 1.3 shows the operation principle of a PEM Fuel Cell. Humidified air enters the
cathode channel, and a hydrogen gas enters the anode channel. The hydrogen diffuses through
the anode diffusion layer towards the catalyst layer, where each hydrogen molecule splits up
into two hydrogen protons and two electrons on catalyst surface according to [14]:
2 H 2 4 H + + 4e (1.1)
The protons migrate through the membrane and the electrons travel through the
conductive diffusion layer and an external circuit where they produce electric work. On the
cathode side the oxygen diffuses through the diffusion layer, splits up at the catalyst layer
surface and reacts with the protons and the electrons to form water:
282 Maher A.R. Sadiq Al-Baghdadi
O 2 + 4 H + + 4e 2 H 2 O (1.2)
2H 2 + O2 2H 2O (1.3)
Based on its physical dimensions, a single cell produces a total amount of current, which
is related to the geometrical cell area by the current density of the cell in [A/cm2]. The cell
current density is related to the cell voltage via the polarization curve, and the product of the
current density and the cell voltage gives the power density in [W/cm2] of a single cell.
The PEM fuel cell consists of a current collector (including gas channels), gas diffusion layer,
and catalyst layer on the anode and cathode sides as well as an ion conducting polymer
membrane (Figure 1.4).
PEM Fuel Cell Modeling 283
Reactants enter the cell through gas channels, which are embedded in the current
collectors (bipolar plate). The gas diffusion layers (GDL) are used to uniformly distribute the
reactants across the surface of the catalyst layers (CL), as well as to provide an electrical
connection between the catalyst layers and the current collectors.
The electrochemical reactions that drive a fuel cell occur in the catalyst layers which are
attached to both sides of the membrane. The catalyst layers must be designed in such a
manner as to facilitate the transport of protons, electrons, and gaseous reactants. Protons,
produced by the oxidation of hydrogen on the anode, are transported through ion conducting
polymer within the catalyst layers and the membrane. Electrons produced at the anode are
transported through the electrically conductive portion of the catalyst layers to the gas
diffusion layers, then to the collector plates and through the load, and finally to the cathode.
Gaseous reactants are transported by both diffusion and advection through pores in the
catalyst layers. Protons are conducted across the polymer membrane from the anode, where
they are produced, to the cathode where they combine with oxygen and electrons to form
water, which may be in vapor or liquid form, depending on the local conditions (Figure 1.5).
Liquid water is transported through the pores in the catalyst and gas diffusion layers through a
284 Maher A.R. Sadiq Al-Baghdadi
mechanism that may be similar to capillary flow. Upon reaching the gas channels, liquid
water is transported out of the cell along with the bulk gas flow [15]. Water may also be
transported, in dissolved form, through the polymer portion of the catalyst layers and through
the membrane [16]. The mechanisms of dissolved water transport are diffusion, due to a
concentration gradient between anode and cathode, and electro-osmotic drag (Figure 1.6)
[17]. Heat produced in the cell is removed principally by conduction through the cell and
convection by a coolant in contact with the collector plates.
Figure 1.5. Transport of gases, protons, and electrons in a PEM fuel cell electrode.
plates is the high cost of machining the flow fields. Metallic plates are cheap and easy to
manufacture, but they have high contact resistance due to the metal oxide layer forming
between the plate and the gas diffusion layer. Metallic plates also suffer high degradation
from the corrosive fuel cell environment that leads to short life cycles. However, some coated
metallic plates have been shown to produce performance comparable to graphite plates.
Finally, composite plates can offer the combined advantages of high electrical and thermal
conductivity of graphite plates and low manufacturing cost of metallic plates [23].
g = h Ts (1.4)
The basic thermodynamic functions are internal energy U, enthalpy H, entropy S, and
Gibbs free energy G. These are extensive properties of a thermodynamic system and they are
first order homogenous functions of the components of the system. Pressure and temperature
are intensive properties of the system and they are zero-order homogenous functions of the
components of the system. Electrochemical potentials are the driving force in an
electrochemical system. The electrochemical potential comprises chemical potential and
electrostatic potential in the following relation.
Chemical potential:
G
i = (1.5)
N i T , P , N j
Electrochemical potential:
i = i + zF (1.6)
Where i is the chemical potential, z is the charge number of the ion, is the potential
at the location of the particles i.
a ivi
RT
E = Eo ln i vj (1.7)
ne F a j
j
o
Where E is the reversible potential at standard state, a i and a j are activity
coefficients of the products and reactants respectively, v i and v j are the stoichiometric
coefficients respectively. The standard state reversible electrode potential can be calculated
from the thermodynamic property of standard Gibbs free energy change of the reaction as
[24];
PEM Fuel Cell Modeling 289
G o
Eo = (1.8)
ne F
It can be seen from these relations that the reversible potential is dependent on
temperature and pressure since the Gibbs free energy is a function of temperature and the
activity coefficients are dependent on temperature, pressure for gases and ionic strengths for
ionic electrolytes. The Nernst equation (1.7) is used to derive a formula for calculating the
reversible cell potential as follows: Anode electrode potential;
RT a H +
2
E a = E ao + ln (1.9)
2 F a H 2
RT a H 2O ( g )
E c = E co + ln (1.10)
2 F a H2 + a O1 22
The absolute electrode potential of the fuel cell is difficult to measure. However, only the
electrode potential difference between the cathode and anode is important in fuel cells. The
reversible cell potential can be obtained from the difference between the reversible electrode
potentials at the cathode and anode.
RT a H 2O
E = Eo + ln (1.11)
2 F a H 2 a O1 22
The reversible potential obtained from the Nernst equation corresponds to the thermodynamic
equilibrium state of the electrochemical system. However, when current starts flowing
through the cell, the cell potential drops below the reversible potential due to several types of
overpotential including activation overpotential, ohmic overpotential, and concentration
overpotential. In addition to these typical electrochemical overpotentials, PEM fuel cells also
suffer from other losses such as internal currents and fuel crossover even at open circuit [26].
Figure 1.10 shows a typical polarization curve of a PEM fuel cell. The curve can be divided
into four regions, which are governed by different losses. Fuel crossover loss occurs when the
outer circuit is disconnected in region 1. Activation loss dominates at low current densities in
region 2. Region 3 is governed by the ohmic loss and the bending down of the polarization
curve in region 4 is due to the diffusion overpotential.
Figure 1.10. Typical polarization curve of a PEM fuel cell and predominant loss mechanisms in various
current density regions.
small compared to other types of losses. This loss can be reduced by increasing the thickness
of the electrolyte at the cost of a higher ohmic loss.
2- Activation Overpotentials
The activation overpotential is the potential loss to drive the electrochemical reactions
from equilibrium state. Therefore, it is the potential loss when there is a net current
production from the electrode, i.e. a net reaction rate. In PEM fuel cell, the activation
overpotential at the anode is negligible compared to that of the cathode. Activation
polarization depends on factors such as the properties of the electrode material, ion-ion
interactions, ion-solvent interactions and characteristics of the electric double layer at the
electrode-electrolyte interface. Activation polarization may be reduced by increasing
operating temperature and by increasing the active surface area of the catalyst.
3- Ohmic Overpotentials
Two types of ohmic losses occur in fuel cells. These are potential losses due to electron
transport through electrodes, bipolar plates, and collector plates; and potential loss due to
proton transport through the membrane. The magnitudes of these potential losses depend on
the materials used in the construction of the fuel cells and its operating conditions [27].
Membrane conductivity increases with membrane water content. Reduction in the thickness
of the membrane between anode and cathode may be thought of as an expedient way to
eliminate ohmic overpotential. However, thin membrane may cause the problem of
crossover or intermixing of anodic and cathodic reactants [27].
2. Literature Review
2.1. Introduction
Fuel cell modeling has received much attention over the past 20 years in an attempt to better
understand the phenomena occurring within the cell. Parametric models allow engineers and
designers to predict the performance of the fuel cell given geometric parameters, material
properties and operating conditions. Such models are advantageous because experimentation
is costly and time consuming. Furthermore, experimentation is limited to designs, which
already exist, thus does not facilitate innovative design. Given the highly reactive
environment within the fuel cell, it is often impossible to measure critical parameters, such as
temperature, pressure and potential gradients, or species concentration within the cell. Thus,
292 Maher A.R. Sadiq Al-Baghdadi
detailed transport models, which accurately predict the flux and concentration of multiple
species, are required. Such information is very useful to develop the fuel cells.
This section reviews the important works done in PEMFC modeling over the past 20
years, discusses contemporary trends and compares various approaches to modeling in recent
times.
A fuel cell model may fall into one of three categories: analytical, semi-empirical, and
mechanistic (theoretical) [28]. Table 2.1 categorizes the models reviewed in this chapter
according to their areas of investigation and dimension of study.
Maggio et al. [34] studied the water transport in a fuel cell using a semi-empirical
approach. They modeled the concentration overpotential effect using an empirical function
between the cathode gas porosity and current density (since current density is related to water
production). The effective gas porosity was assumed to decrease linearly with increasing
current density. This is due to the increasing percentage of gas pores occupied by liquid
water. Their results indicate that dehydration of the membrane is likely to occur on the anode
side rather than the cathode side.
Semi-empirical modeling has also been used to model fuel cell stacks.
Maxoulis et al. [35] used such an approach to model a fuel cell stack during automobile
driving cycles. They combined the model of Amphlett et al. [32] with the commercial
software ADVISOR, which was used to simulate vehicular driving conditions. They studied
the effects of the number of cells per stack, electrode kinetics and water concentration in the
membrane on the fuel consumption. They concluded that a larger number of cells per stack
result in greater stack efficiency resulting in better fuel economy.
Lee et al. [36] developed an artificial neural network model for use to design and analysis
PEM fuel cell power systems. The artificial neural network model can simulate the
experimental data for different operating conditions and hence can be used to investigate the
influence of process variables.
Al-Baghdadi [37] developed a semi-empirical model to provide a tool for the design and
analysis of fuel cell total systems. The model take into account the process variations, such as
the gas pressure, temperature, humidity, and used to cover operating processes, which are
important factors in determining the real performance of fuel cell.
Semi-empirical modeling has also been used to optimize fuel cell performance. Al-
Baghdadi and Al-Janabi [38] developed a semi-empirical parametric model for investigating
the performance optimization of a PEM fuel cell. Their results indicate that operating
temperature and pressure can be optimized, based on cell performance, for given design and
operating point (cell voltage and related current density).
Semi-empirical models are, however, limited to a narrow corridor of operating
conditions. They cannot accurately predict performance outside of that range. They are very
useful for making quick predictions for designs that already exists. They cannot be used to
predict the performance of innovative designs, or the response of the fuel cell to parameter
changes outside of the conditions under which the empirical relationships were developed.
Empirical relationships also do not provide an adequate physical understanding of the
phenomena inside the cell. They only correlate output with input.
may include a large portion of the relevant fuel cell physics while at the same time having
relatively short solution times. However, these narrowly focused models neglect important
parts of the fuel cell making it impossible to get a complete picture of the phenomena
governing fuel cell behavior.
One of the early mechanistic models for a PEM fuel cell was the pioneering work of
Bernardi and Verbrugge [45, 46]. They developed a one-dimensional, steady state, isothermal
model which described water transport, reactant species transport, as well as ohmic and
activation overpotentials. Their model assumed a fully hydrated membrane at all times, and
thus calculated the water input and removal requirements to maintain full hydration of the
membrane. The model was based on the Stefan Maxwell equations to describe gas phase
diffusion in the electrode regions, the NernstPlanck equation to describe dissolved species
fluxes in the membrane and catalyst layers, the Butler Volmer equation to describe electrode
rate kinetics and Schlogls equation for liquid water transport.
This model was used primarily to predict the polarization effects (due to ohmic and
activation overpotentials) and the water management requirements. The model computed the
required water input at the anode side and required water removal rate at the cathode side
necessary to maintain full hydration of the Nafion membrane at all times.
Their model also predicted the dissolved hydrogen and oxygen concentrations within the
catalyst layers and it was found that at sufficiently high current densities most of the
electrochemical reactions occurred on the outer surface of the catalyst layer. This information
is vital to designers of fuel cells. It allows them to economically distribute the catalyst where
it is most needed. Considering that the platinum catalyst is one of the largest expenses in a
fuel cell, this could help reduce the cost. Although their model was basic and many
improvements have been made since, this work served as a foundation for PEM fuel cell
modeling.
Baschuk and Li [47] developed a one-dimensional model, which accounted for cathode
mass limitation effects by allowing variable degrees of flooding at the cathode catalyst
layer/backing region. They account for concentration overpotential as a result of the
decreased concentration of dissolved oxygen in the catalyst region due to the excessive water
content. Darcys law is used to obtain the drop in partial pressure of the oxygen at the cathode
catalyst layer, and Henrys law is used to determine the dissolved oxygen concentration.
Their results showed excellent agreement with experimental results. The model also predicted
that increasing the cell pressure lowers the limiting current density. High pressures result in
maximum flooding occurring at lower current densities, and this effect is more significant
than the increase in partial pressure of the oxygen. The results also showed, predictably, that
increasing the temperature increases the limiting current density.
Models that include all parts of a fuel cell are typically two- or three-dimensional and
reflect many of the physical processes occurring within the fuel cell. In a real PEM fuel cell
geometry, the gas diffusion layers are used to enhance the reaction area accessible by the
reactants. The effect of using these diffusion layers is to allow a spatial distribution in the
current density on the membrane in both the direction of bulk flow and the direction
orthogonal to the flow but parallel to the membrane. This two-dimensional distribution cannot
be modeled with the well-used two-dimensional models, (like models that developed by the
researchers in Ref. [48-52]), where the mass-transport limitation is absent in the third
direction.
296 Maher A.R. Sadiq Al-Baghdadi
Large
Computational 67
Domains
Heat Generation
66
Mechanism
Multi Phase
65
Interdigitated
64
Flow Field
Interdigitated
63
Flow Field
Serpentine Flow
54 62
Field
Multi Phase 60 51
3D non 58
55 56
isothermal 57
65
3D
59 64
isothermal
61
Catalyst Layer
52
Mode
Gas Channel
50
Flow Mode
Anode Mass
49
Limitation
CFD methods 53
Cathode Mass
47
Limitation
2D 48 71
1D 45 46
91 92 93 94 95 96 97 98 99 00 01 02 03 04 05
Year
Fuller and Newman [48] developed a pseudo two-dimensional model, which predicted
water and thermal management as well as fuel utilization for a fuel cell operating with
reformed methanol as the fuel.
Um et al. [49] developed a two-dimensional transient model based on the single domain
approach, which studied the effects of hydrogen dilution along the anode gas channel. The
two-dimensional model considered flow perpendicular to the membrane electrode assembly
(MEA) cross-section, as well as in the direction of flow in the gas channels. As hydrogen
diffuses from the gas channel into the gas diffusion region, its concentration along the gas
channel decreases resulting in a two-dimensional concentration gradient in the gas diffusion
electrodes. The result is that mass transport limitations are seen on the anode side especially
at high current densities, and when reformed fuel is used instead of pure hydrogen. At high
current densities, hydrogen is extracted from the flow channels at a much fast rate than at low
current densities. With reformed fuel, the partial pressure of the hydrogen is already lowered
by the presence of carbon dioxide in the gas feed, so as it is used up toward the end of the gas
channel, the partial pressure of the hydrogen may be too low and it may not be able to diffuse
fast enough to the anode catalyst layer. The result is anode side mass transport limitations.
Such phenomena cannot be studied using one-dimensional models.
Ge and Yi [50] developed a two-dimensional model to study the effects of flow mode in
straight gas channels, i.e., counter flow versus co-flow. It was found that the flow mode only
made a difference when dry or low humidity inlet gases were used. For such cases, counter
flow operation produced better results since by so doing the reactant gases were sufficiently
humidified internally. If the inlet gases are already humidified, the flow mode makes little
difference. The reason for this is that for high humidity gases, the increase in membrane
conductivity due to the high humidity is counteracted by the increase in cathode concentration
overpotential due to the presence of liquid water. This is the case whatever the flow mode.
However, for low humidity gases, counter flow operation allows for internal humidification
of the gas streams. For co-flow low humidity gases, the membrane dehydrates. This
information gives the designers of fuel cells an alternative to humidifying the gas streams.
Seigel et al. [51] developed a multi-phase, two-dimensional model. The model was a two-
dimensional steady state model, which studied transport limitations due to water build up in
the cathode catalyst region. They considered water in three phases: liquid, gas and dissolved
(membrane phase). They found that treating the catalyst layer as a very thin interface
underestimates the transport limitations due to water build-up. Hence, they modeled the
catalyst layer as a finite region. Their model showed that 2040% of the water building up at
the cathode catalyst layer comes from water which is transported across the membrane. This
problem may be counteracted by applying a pressure differential to force back diffusion of
water, i.e., from cathode to anode.
Mechanistic models can also be characterized as single domain (unified) models or multi-
domain models. The single-domain approach consists of equations governing the entire
domain of interest, with source and sink terms accounting for species consumption and
generation within the cell. The single-domain is less cumbersome in that no internal boundary
conditions and conditions of continuity need to be specified. It is also easier to incorporate
into commercial CFD codes. As a result, the time for model development is shortened. Multi-
domain models involve the derivation of different sets of equations for each region of the fuel
cell, namely the anode and cathode gas diffusion regions, anode and cathode gas flow
298 Maher A.R. Sadiq Al-Baghdadi
channels, membrane and catalyst layers. These equations are solved separately and
simultaneously.
Siegel et al. [52] modeled fuel cell with an interdigitated flow field using single domain.
Their model accounts for specific transport, electrochemical kinetics, energy transport,
current distribution, and water uptake and release in the catalyst layer. The governing
differential equations are solved over a single computational domain, which consists of a gas
channel, gas diffusion layer, and catalyst layer for both the anode and cathode sides of the
fuel cell as well as the solid polymer membrane. The model for the catalyst regions is based
on an agglomerate geometry, which requires water species to exist in both dissolved and
gaseous forms simultaneously.
Zhou and Liu [53] modeled a PEM fuel cell with a single-domain approach. The single-
domain approach combines all the regions of interest into one domain. The computational
effect is that all regions can be considered as one domain where no internal boundary
conditions or statements of continuity need be defined. The only difference is that material
properties and source terms assume different values for the all regions. Conservation
equations which govern the entire domain of interest, typically the entire fuel cell (gas flow
regions and the membrane electrode assembly) are defined. In each region, the differences are
accounted for by source and sink terms. All equations are written in the form of a generic
convectiondiffusion equation, and all terms, which do not fit that format are dumped into the
source or sink term. This formulation allows for solution using known computational fluid
dynamics (CFD) methods. The strength of the CFD numerical approach is in providing
detailed insight into the various transport mechanisms and their interaction, and in the
possibility of performing parameters sensitivity analyses.
Two-dimensional models can accurately predict spatial variations in species
concentration and fluxes for simple flow regimes, e.g., straight flow channels (co-flow and
counter-flow arrangements). However, for more complex flow regimes, such as interdigitated
flow and serpentine flow channels, or for modeling, studying, and analyzing the mass-
transport limitation in the third direction (e.g. land area), three-dimensional modeling is
required. Comprehensive three-dimensional models can study the effect of a large number of
properties and operating parameters and therefore much more computationally intensive,
leading to longer solution times. However, these disadvantages are typically outweighed by
the benefit of being able to assess the influence of a greater number of design parameters and
their associated physical processes.
Dutta et al. [54] used the unified approach to study mass transport between the channels
of a PEM fuel cell with a serpentine flow field. Their model is three-dimensional and allows
for multi-species transport. They studied the effect of flow channel width in the serpentine
flow field on velocity distribution, gas mixture distribution and reactant consumption.
Serpentine flow fields allow for a greater area for diffusion of the supply gases. Their results
showed that for low humidity conditions, water transport is dominated by electro-osmotic
effects, i.e., water flows from anode to cathode at the side of the cell closer to the gas channel
inlet. At the outlet side of the cell, water transport is dominated by back diffusion, and it
flows in the opposite direction. Thus the serpentine flow field allows for circulation of the
water within the cell.
Berning et al. [55] used the unified approach to develop a three-dimensional non-
isothermal fuel cell model. The model studied reactant concentrations, current density
distributions and temperature gradients within the cell as well as water flux and species
PEM Fuel Cell Modeling 299
transport. For gas flow fields separated by current collecting plates, three-dimensional effects
were observed due to the unevenness of the hydrogen and oxygen supply. These effects were
pronounced under the collector plate land areas. These effects may result in transport limiting
conditions at high current densities. The development of a non-isothermal model was
intended to study the heat transfers within the cell. It was observed that a temperature
difference of 23K existed within the cell. Yan et al. [56] performed a similar study and
found a temperature variation within the cell of the same magnitude. However, the magnitude
of the heat transfer was not reported so it is difficult to compare the magnitude of the
conductive heat transfer relative to the total heat transfer. This information would have helped
to justify the need for non-isothermal modeling. If the heat transfer by conduction were small
compared to other heat transfers, then a temperature difference of 23K could hardly be
significant.
Wang et al. [57] developed a single-phase, three-dimensional parametric model,
considering the effects of temperature, humidity and pressure. It was found that the
performance of the fuel cell is improved with increasing temperature if the inlet gases are
fully humidified. If the gases are not fully humidified, dehydration of the membrane is likely
to occur resulting in reduced conductivity values, hence reduced cell performance. They also
found that at low current densities anode humidification is required, but not at higher current
densities. This is because at high current densities, sufficient water is produced at the cathode
to keep the membrane hydrated. Their results further showed that cathode humidification is
not significant at all, especially at high current densities. This is because dehydration is likely
to occur on the anode side and flooding on the cathode side. Therefore, humidifying the
cathode gas stream adds no benefit. Finally, increasing the pressure of the inlet gases was
seen to improve performance by increasing the activation currents and the partial pressures of
reactant gases. The authors report that at higher current densities, their model overestimates
the cell current density compared to experimental results. The reason for this is that the model
did not take into account mass transport effects. The impact of liquid water on transport in the
gas-diffusion electrode was, however, not account for.
Berning et al. [58] performed a parametric study using their previously described single-
phase, three-dimensional model [55]. The effect of various operational parameters such as the
temperature and pressure on the fuel cell performance was investigated. In addition,
geometrical and material parameters such as the gas diffusion electrode thickness and
porosity as well as the ratio between the channel width and the land area were investigated. It
was found that in order to obtain physically realistic results experimental measurements of
various modeling parameters were needed. In addition, the contact resistance inside the cell
was found to play an important role for the evaluation of impact of such parameters on the
fuel cell performance. The impact of liquid water on transport in the gas-diffusion electrode
was, however, not account for.
Using the unified approach, Kumar and Reddy [59] studied the effects of having metal
foam in the flow field of the bipolar plates. Their three-dimensional steady state model shows
that decreasing the permeability of the gas flow field improves performance. This is because
at low flow field permeability reactant gases are transported by forced convection rather than
diffusion. Having many tiny gas channels results in a lower permeability than having few
large channels. However, due to limitations in machining processes, the flow channels cannot
be made too small. Placing metal foam in the flow field allows the flow field permeability to
be lowered without resorting to precise machining processes. They found that decreasing the
300 Maher A.R. Sadiq Al-Baghdadi
permeability from 106 to 1012 m2 increases the average current density of their system
from 5943 to 8425Am2.
Berning and Djilali [60] developed a computational fluid dynamics multiphase model of
a PEM fuel cell. Their model accounts for three-dimensional transport processes including
phase change and heat transfer, and including the gas diffusion layers and gas flow channels
for both anode and cathode, as well as a cooling channel. Transfer of liquid water inside the
gas diffusion layers was modeled using viscous force and capillary pressure terms. The
physics of phase change was accounted for by prescribing local evaporation as a function of
the undersaturation and liquid water concentration. Simulation has been performed for fully
humidified gases entering the cell. The results showed that different competing mechanisms
lead to phase change at both anode and cathode sides of the fuel cell. The predicted amount of
liquid water depends strongly on the prescribed material properties, particularly the hydraulic
permeability of the gas diffusion layer. Analysis of the simulations at a current density of 1.2
Acm-2 showed that both condensation and evaporation take place within the cathode gas
diffusion layer, whereas condensation prevails throughout the anode, except near the inlet.
However, the liquid water saturation does not exceed 10% at either anode or cathode side.
Cha et al. [61] studied the effect of flow channel scaling on fuel cell performance. In
particular, the impact of dimensional scales on the order of 100 micrometers and below has
been investigated. A model based on isothermal three-dimensional computational fluid
dynamics has been developed which predicts that very small channels result in significantly
higher peak power densities compared to their larger counterparts. The experimental results
confirm the predicted outcome at relatively large scale. At especially small scale ( 100m ),
the model (which dose not consider two-phase flow) disagree with the measured data. Liquid
water flooding at the small channel scale is hypothesized as a primary cause for this
discrepancy.
Nguyen et al. [62] developed a three-dimensional model which accounts for mass and
heat transfer, current and potential distribution within a cell using a serpentine flow field.
Their results show that oxygen concentration along the gas channels decrease in the direction
of flow. Also, in the gas diffusion layer, the oxygen concentration is a minimum under the
land area. At high current densities the oxygen is almost completely depleted under the land
areas. The result is an uneven distribution of oxygen concentration along the catalyst layer
resulting in local overpotentials, which vary spatially. A unique feature of this model is a
voltage-to-current (VTC) algorithm, which allows for the solution of the potential field and
the local activation overpotential. Since the reactant concentration is not constant across the
catalyst layer, the activation overpotential will not be constant. Their simulations show a
variation in local activation overpotential from 0.31 to 0.37 V at a current density of 1.2
Acm2. This VTC algorithm however, comes with a computational cost. It slows down the
solution requiring 60008000 iterations for convergence. In addition, the predicted
distribution of current densities show profiles which are fundamentally different from the
distribution obtained in all other previous models that used current-to-voltage algorithm.
Um and Wang [63] used a three-dimensional model to study the effects an interdigitated
flow field. The model accounted for mass transport, electrochemical kinetics, species profiles
and current density distribution within the cell. Interdigitated flow fields result in forced
convection of gases, which aids in liquid water removal at the cathode. This would help
improve performance at high current densities when transport limitations due to excessive
PEM Fuel Cell Modeling 301
water production are expected. The model shows that there is little to no difference at low to
medium current densities between an interdigitated flow field and a conventional flow field.
However, at higher current densities, a fuel cell with an interdigitated flow field has a limiting
current, which is nearly 50% greater than an equivalent cell with a conventional flow field.
Because of the flow field, three-dimensional effects under the current collector land area,
known as rib effects, are prominent.
Using the multi-domain approach Hu et al. [64, 65] developed an isothermal, three-
dimensional, two-phase model for a fuel cell. They gave boundary conditions, which could be
used for straight flow channels as well as interdigitated flow fields. Unlike previous models,
which assume separate flow channels for gases and liquids, this model assumes a two-phase
mixture. Water properties such as specific volume change depending on the degree of
mixture. They used a CFD algorithm to solve for the flow field in the gas flow channels and
diffusion regions, and the fourth order RungeKutta method together with a shooting
technique to solve for the flow field in the catalyst layers and the membrane. For the
interdigitated flow field, results show that the oxygen concentration is higher and liquid water
saturation is lower than those for a conventional straight channel flow field. The higher
oxygen concentration results in fast reaction rates and the lower liquid water saturation results
in less concentration overpotential. It is also shown that the local current densities are much
more uniform with an interdigitated flow field than with a conventional flow field. However,
the performance of a fuel cell with an interdigitated flow field is only shown to be better than
that with a conventional flow field if the inlet gases are well humidified. This is because the
interdigitated field aids in water removal, but does not aid in hydration of an already de-
hydrated membrane. So, the internal gases need to be humidified.
Hyunchul et al. [66] developed a single-phase, three-dimensional, non-isothermal model
to account rigorously for various heat generation mechanisms, including irreversible heat due
to electrochemical reactions, entropic heat, and Joule heating arising from the electrolyte
ionic resistance. Their results show that the irreversible reaction heat and entropic heat in the
cathode catalyst layer are major contributors to heat generation in PEM fuel cells, accounting
for about 90% of total waste heat released. Further, it was revealed that the gas diffusion layer
thermal conductivity strongly impacts on the membrane temperature rise. There exist vastly
different thermal behaviors under different humidification conditions. In the medium range of
current density (i.e. the ohmic control regime of IV curve), the low inlet humidity case show a
significant decrease in fuel cell performance as the membrane temperature rise, indicating
that the fuel cell performance is primarily controlled by membrane hydration. Thus, efficient
cooling through the current collector ribs becomes critical in low-humidity operation in order
to maintain good membrane proton conductivity. On the other hand, for fully humidified
cases, the performance is dominated by oxygen reduction reaction kinetics and therefore
become higher with larger membrane temperature rises. In addition, severe flooding of
electrodes is more likely in the fully humidified operation, particularly in the cold region near
ribs. Therefore relatively higher membrane temperature rise enabled by low gas diffusion
layer conductivity should be helpful to alleviate electrode flooding and enhance the oxygen
reduction reaction kinetics.
Sivertsen and Djilali [67] developed a single-phase, non-isothermal 3D model which is
implemented into a computational fluid dynamic code. The model allows parallel computing,
thus making it practical to perform well-resolved simulations for large computational
domains. The parallel solver allows them to use a large computational grid (total of 546000
302 Maher A.R. Sadiq Al-Baghdadi
cells). The maximum number of processors used was eight. The model solves for the electric
and ionic potentials in the electrodes and membrane. The local activation overpotential
distribution is resolved, rather than assumed uniform, using a voltage-to-current (VTC)
algorithm. Their results showed that the predicted distribution of current densities show
profiles which are fundamentally different from the distribution obtained in all other previous
models that used current-to-voltage algorithm. The maximum current density occurs under
the land areas as a result of the dominant influence of ohmic losses over concentration losses
on the activity at the catalyst layer.
The need for improved lifetime of PEM fuel cells necessitates that the failure
mechanisms be clearly understood and life prediction models be developed, so that new
designs can be introduced to improve long-term performance.
Weber and Newman [68] developed one-dimensional model to study the stresses
development in the fuel cell. They showed that hygro-thermal stresses might be an important
reason for membrane failure, and the mechanical stresses might be particularly important in
systems that are non-isothermal. However, their model is one-dimensional and does not
include the effects of material property mismatch among PEM, GDL, and bipolar plates.
Tang et al. [69] studied the hygro and thermal stresses in the fuel cell caused by step-
changes of temperature and relative humidity. Influence of membrane thickness was also
studied, which shows a less significant effect. However, their model is two-dimensional,
where the hygro-thermal stresses are absent in the third direction (flow direction). In addition,
a simplified temperature and humidity profile was assumed, (constant temperature for each
upper and lower surfaces of the membrane was assumed), with no internal heat generation.
An operating fuel cell has varying local conditions of temperature, humidity, and power
generation (and thereby heat generation) across the active area of the fuel cell in three-
dimensions. Nevertheless, no models have yet been published to incorporate the effect of
hygro-thermal stresses into actual fuel cell models to study the effect of these conditions on
the stresses developed in the membrane. In addition, the transport phenomena in a fuel cell
are inherently three-dimensional, but no models have yet been published to address the hygro-
thermal stresses in PEM with the three-dimensional effect.
2.3. Summary
Fuel cell technology is expected to play an important role in meeting the growing demand for
distributed power generation. In an ongoing effort to meet increasing energy demand and to
preserve the global environment, the development of energy systems with readily available
fuels, high efficiency and minimal environmental impact is urgently required. A fuel cell
system is expected to meet such demands because it is a chemical power generation device,
which converts the chemical energy of a clean fuel (e.g. Hydrogen) directly into electrical
energy. Still a maturing technology, fuel cell technology has already indicated its advantages,
such as its high-energy conversion efficiency, modular design and very low environmental
intrusion, over conventional power generation equipment.
Two key issues limiting the widespread commercialization of fuel cell technology which
are better performance and lower cost. Recent years have seen significant increase in power
densities, reliability and overall performance of PEM fuel cells but the underlying physics of
the transport processes in a fuel cell, (which involve coupled fluid flow, heat and mass
PEM Fuel Cell Modeling 303
A computational model of an entire cell would require very large computing resources and
excessively long simulation times. The computational domain in this chapter is therefore
limited to one straight flow channel with the land areas. The full computational domain
consists of cathode and anode gas flow channels, and the membrane electrode assembly as
shown in Figure 3.1.
3.3. Assumption
(u ) = 0 (3.1)
[
(u u u ) = P + u + (u )
2 T
] (3.2)
3
The mass balance is described by the divergence of the mass flux through diffusion and
convection. The steady state mass transport equation can be written in the following
expression for species i;
M M P
+ (x j y j )
N
y i Dij y j + y j + y i u
M j M P =0
j =1
(3.3)
T
+ Di
T
T
306 Maher A.R. Sadiq Al-Baghdadi
Where the subscript i denotes oxygen at the cathode side and hydrogen at the anode side,
and j is water vapor in both case. Nitrogen is the third species at the cathode side.
The temperature field is obtained by solving the convective energy equation;
(CpuT kT ) = 0 (3.4)
12
8 T 1.75 1 1
Dij = 3.16 10 + (3.5)
13 2
13
M i M j
P Vki + Vkj
k k
The values of the molar diffusion volumes, (V ) , are given by Fuller et al. [70] as
ki
In order to determine the inlet gas composition to the fuel cell, the following equation of
the saturation pressure of water vapor has been used [49];
(3.6)
+ 1.4454 10 7 (T 273.15)
3
The above calculated saturated pressure is in bars. The molar fraction of water vapor in
the fully humidified inlet gas stream at the cathode gas flow channel is simply the ratio of the
saturation pressure and the total pressure;
Psat
x H 2O ,in ,c = (3.7)
PC
PEM Fuel Cell Modeling 307
Since the ratio of nitrogen and oxygen in dry air is known to be 79:21, the inlet oxygen
fraction can be found via;
1 x H 2O ,in ,c
x O2 ,in = (3.8)
79
1+
21
Since the sum of all molar fractions is unity, the molar fraction of nitrogen can be
obtained;
The gas composition of the anode inlet stream is much simpler since there are only two
gas species, hydrogen and water vapor. The molar fraction of water vapor in the anode
incoming gas stream can be determined as follow;
Psat
x H 2O ,in ,a = (3.10)
Pa
The molar fraction of hydrogen is determined from the unity relation as follow;
x i .M i
yi = n
(3.12)
xj =1
j .M j
P
= n
(3.13)
y
R.T i
i =1 M i
The molar fraction of each species is related to the mass fraction by [55];
308 Maher A.R. Sadiq Al-Baghdadi
yi
M
xi = n i (3.14)
yj
M j =1 j
n
M = x i .M i (3.15)
i =1
The correlation for the gas viscosity of each species can be expressed as [71];
where A, B, and C are correlation constants of species i as shown in Table 3.2 [71].
Species A B C
2
O2 18.11 66.32 10 187.9 10 6
N2 30.43 49.89 10 2 109.3 10 6
H2 21.87 22.20 10 2 37.51 10 6
H 2O 31.89 41.45 10 2 8.272 10 6
The Herning-Zipperer correlation [71] for the calculation of gas mixture viscosity is used
in this work;
. x .(M )i i i
0.5
= i =1
n
(3.17)
x .(M )
0.5
i i
i =1
The correlation for the gas thermal conductivity of each species can be expressed as [71];
where A, B, C, and D are correlation constants of species i as shown in Table 3.3 [71].
PEM Fuel Cell Modeling 309
Species A B C D
O2 0.7816 23.80 10 2
0.8939 10 4
2.324 10 8
N2 0.9359 23.44 10 2 1.21 10 4 3.591 10 8
H2 19.34 159.74 10 2 9.93 10 4 37.29 10 8
H 2O 17.53 2.42 10 2 4.3 10 4 21.73 10 8
n
k = ki . yi (3.19)
i =1
The correlation for the gas specific heat capacity of each species can be expressed as
[71];
4.1868 10 3
Cp i =
Mi
( A + B.T + C.T 2 + D.T 3 ) (3.20)
where A, B, C, and D are correlation constant of species i as shown in Table 3.4 [71];
Species A B C D
3 6
O2 6.22 2.710 10 0.37 10 0.22 10 9
N2 7.07 1.320 10 3 3.31 10 6 1.26 10 9
H2 6.88 0.022 10 3 0.21 10 6 0.13 10 9
H 2O 8.10 0.720 10 3 3.63 10 6 1.16 10 9
n
Cp = Cp i . y i (3.21)
i =1
(u ) = 0 (3.22)
310 Maher A.R. Sadiq Al-Baghdadi
Kp
u= P (3.23)
M M P T
+ (x j y j )
N
y i Dij y j + y j + y i u + DiT =0 (3.24)
j =1 M j M P T
In order to account for geometric constraints of the porous media, the diffusivities are
corrected using the Bruggemann correction formula [72];
Where the term on the right hand side accounts for the heat exchange to and from the
solid matrix of the GDL.
is a modified heat transfer coefficient that accounts for the convective heat transfer in
[W/m2] and the specific surface area [m2/m3] of the porous medium [55]. Hence, the unit of
is [W/m3]. The heat transfer coefficient between the gas phase and the solid matrix of the
electrodes, , has been found by trial-and error. It has been adjusted so that the temperature
difference between the solid and the gas-phase is minimal, i.e. below 0.1 K throughout the
whole domain. This is equivalent to assuming thermal equilibrium between the phases. The
low velocity of the gas-phase inside the porous medium and the high specific surface area
which accommodates the heat transfer justify this assumption.
The effective thermal conductivity can be calculated from the thermal conductivities of
gas and graphite by an expression given by Gurau et al. [73];
1
k eff = 2k gr + (3.27)
1
+
2k gr + k 3k gr
However, the effective thermal conductivity of Ballard Arvcarb was used in this model.
The potential distribution in the gas diffusion layers is governed by;
( e GDL ) = 0 (3.28)
PEM Fuel Cell Modeling 311
M O2
S O2 = ic (3.29)
4F
M H2
S H2 = ia (3.30)
2F
The production of water is modeled as a source terms, and hence can be written as;
M H 2O
S H 2O = ic (3.31)
2F
The generation of heat in the cell is due to entropy changes as well as irreversibilities due
to the activation overpotential;
T ( s )
q = + act ,c i c (3.32)
ne F
The local current density distribution in the catalyst layers is modeled by the Butler-
Volmer equation;
O2
C O2 a F cF
ic = i exp RT act ,c + exp RT act ,c
ref
(3.33)
o ,c
C ref
O2
H2
CH a F cF
i a = ioref,a ref2
C exp RT act ,a + exp RT act ,a (3.34)
H2
The variation of the reference exchange current density with the operating cell pressure
and temperature was computed using the procedure given by Parthasarathy et al. [74];
ref Pc x O2 ,in
io ,c = 6.028 10 exp 2.06
4
(3.35)
101325 T = 353.15 K
312 Maher A.R. Sadiq Al-Baghdadi
ref Pa x H 2 ,in
io ,a = 822 exp 2.06
(3.36)
101325 T = 353.15 K
ref 21 T
io ,c = 3.05 10 exp 8.6295 (3.37)
PC = 3 atm
ref T
io ,a = 10.1815 exp 64.3306 (3.38)
Pa = 3 atm
3.4.4. Membrane
The balance between the electro-osmotic drag of water from anode to cathode and back
diffusion from cathode to anode yields the net water flux through the membrane [44];
(DW yW )
i
N W = n d M H 2O (3.39)
F
1 1
DW = 1.3 10 10 exp 2416 (3.40)
303 T
(k mem T ) = 0 (3.41)
The potential loss in the membrane is due to resistance to proton transport across
membrane, and is governed by;
( m mem ) = 0 (3.42)
The membrane type is Nafion 117. Bernardi and Verbrugge [46] developed the
following theoretical expression for the electric conductivity of the fully humidified
membrane;
F2
m = .Z f .D H+ .C f (3.43)
RT
PEM Fuel Cell Modeling 313
As a result of in the changes in temperature and moisture, the membrane, GDL and bipolar
plates will all experience expansion and contraction. Because of the different thermal
expansion and swelling coefficients between these materials, hygrothermal stresses are
expected to be introduced into the unit cell during operation. In addition, the non-uniform
current and reactant flow distributions in the cell result in non-uniform temperature and
moisture content of the cell which could in turn, potentially causing localized increases in the
stress magnitudes.
Using hygrothermoelasticity theory, the effects of temperature and moisture as well as
the mechanical forces on the behavior of elastic bodies have been addressed.
The total strain tensor is determined using Tang et al. [69] formula;
= M + T + S (3.44)
where, is the contribution from the mechanical forces and , are the thermal and
M T S
Similarly, the swelling strains caused by moisture uptake are given by;
S = mem ( Re f ) (3.46)
= (3.47)
The initial conditions corresponding to zero stress-state are defined; all components of
the cell stack are set to reference temperature 20 C, and relative humidity 35% (corresponding
to the assembly conditions) [69, 75]. In addition, a constant pressure of (1 MPa) is applied on
the surface of lower graphite plate, corresponding to a case where the fuel cell stack is
equipped with springs to control the clamping force.
( )
( )
E = 1.229 0.83 10 3 (T 298.15) + 4.3085 10 5 T ln PH 2 + ln PO2 (3.49)
1
2
is usually run at high pressure. The anode and cathode diffusion overpotentials are calculated
from the following equations [26];
RT i
Diff ,c = ln1 c
(3.50)
2 F i L,c
RT i
Diff ,a = ln1 a
(3.51)
2 F i L,a
2 FDO2 CO2
i L,c = (3.52)
GDL
2 FD H 2 C H 2
i L ,a = (3.53)
GDL
The diffusivity of oxygen and hydrogen are calculated from the following equations [63];
32
T 5 101325
DO 2 = 3.2 10 (3.54)
353 P
32
T
4 101325
DH2 = 1.1 10 (3.55)
353 P
2 E cell F
E fc = (3.57)
M H 2 .LHV H 2
Boundary conditions are specified at all external boundaries of the computational domain as
well as boundaries for various mass and scalar equations inside the computational domain.
316 Maher A.R. Sadiq Al-Baghdadi
3.7.1. Inlets
The inlet values at the anode and cathode are prescribed for the velocity, temperature and
species concentrations (Dirichlet boundary conditions). The inlet velocity is a function of the
desired current density, the geometrical area of the membrane, the channel cross-section area,
and stoichiometric flow ratio. The inlet velocities of air and fuel are calculated according to
[66]:
I 1 RTin ,c 1
u in ,c = c AMEA (3.58)
4F x O2 ,in Pc Ach
I 1 RTin ,a 1
u in ,a = a AMEA (3.59)
2F x H 2 ,in Pa Ach
3.7.2. Outlets
At the outlets of the gas-flow channels, only the pressure is being prescribed as the desired
electrode pressure; for all other variables, the gradient in the flow direction (x) is assumed to
be zero (Neumann boundary conditions).
solutions were independent of the grid size. The coupled set of equations was solved
iteratively, and the solution was considered to be convergent when the relative error in each
field between two consecutive iterations was less than 10-6. The number of iterations required
to obtain converged solutions dependent on the nominal current density of the cell; the higher
the load the slower the convergence.
Once the flow field is obtained, the mass fraction equations are solved for the mass fractions
of oxygen, hydrogen, water vapor, and nitrogen. Scalar equations are solved last in the
sequence of the transport equations for the temperature field in the cell and potential fields in
the gas diffusion layers and the membrane. The local current densities are solved based on the
Butler-Volmer equation. After the local current densities are obtained, the local activation
overpotentials can be readily calculated from the Butler-Volmer equation. The local
activation overpotentials are updated after each global iterative loop. Convergence criteria are
then performed on each variable and the procedure is repeated until convergence. The
properties and then source terms are updated after each global iterative loop based on the new
local gas composition and temperature. The strength of the current model is clearly to
perform parametric studies and explore the impact of various parameters on the transport
mechanisms and on fuel cell performance. The new feature of the algorithm developed in this
work is its capability for accurate calculation of the local activation overpotentials, which in
turn results in improved prediction of the local current density distribution. The flow diagram
of the algorithm is shown in Figure 3.2.
Choosing the right modeling parameters is important in establishing the base case validation
of the model against experimental results. Since the fuel cell model that is presented in this
chapter accounts for all basic transport phenomena simply by virtue of its three-
dimensionality, a proper choice of the modeling parameters will make it possible to obtain
good agreement with experimental results obtained from a real fuel cell. It is important to
note that because this model accounts for all major transport processes and the modeling
domain comprises all the elements of a complete cell, no parameters needed to be adjusted in
order to obtain physical results.
Table 3.5 shows the dimensions of the computational domain. All parameters listed in
Table 3.5 refer to both side, anode and cathode. The membrane refers to fully wetted Nafion
117 membrane.
The operational parameters are based on the experimental operating conditions used by
Wang et al. [57]. These values are listed in Table 3.6.
PEM Fuel Cell Modeling 319
Electrode material properties have important impact on fuel cell performance. The
important electrode material properties are thickness, porosity, and thermal and electronic
conductivities. The base case values are listed in Table 3.7.
The membrane properties are required to model various transport phenomena across the
membrane. Table 3.8 lists the membrane properties taken for the base case. The membrane
type is fully humidified Nafion 117.
impracticality of performing a large number of parametric simulations with the finer grid, the
base case grid was employed for all simulations presented in the this chapter.
Figure 3.3. Polarization curves comparing base case mesh and 50% refined mesh results.
3.11. Results
In addition to the polarization curve, the comprehensive three-dimensional model also allows
for the assessment of important information about the detail of transport phenomena inside
the fuel cell. These transport phenomena are the velocity flow field, variation of local
concentration of gas reactants, temperature field, local current densities, and potential field.
The velocity profiles in the cathode GDL at nominal current densities of 1.2 A/cm2 are
shown in Figure 3.6. The figure shows the 2D velocity vectors (cone) in the cross plane of
GDL (y-z) at x=25 mm (lower figure) and 3D velocity vectors arrows (upper figure). The
vectors are pointed upward from the cathode GDL, spreading into the catalyst layer, and
pointed up to the membrane surface. In the cathode, oxygen is consumed on the cathode
catalyst layer surface due to the electrochemical reactions while the water vapor is produced
on the surface.
Figure 3.5. Velocity profiles in the mid-plane of the cathode and anode gas flow channels for three
different nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
324 Maher A.R. Sadiq Al-Baghdadi
Figure 3.6. Velocity profiles in the cathode GDL: 3D profile (upper); on the y-z plane at x=25 mm
(lower). The nominal current density is 1.2 A/cm2.
The velocity profiles in the anode GDL at nominal current densities of 1.2 A/cm2 are
shown in Figure 3.7. The figure shows the 2D velocity vectors (cone) in the cross plane of
GDL (y-z) at x=25 mm (lower figure) and 3D velocity vectors arrows (upper figure). The
vectors are pointed downward from the anode GDL, spreading into the catalyst layer, and
pointed down to the membrane surface because both species (H2 and water vapor) are
PEM Fuel Cell Modeling 325
consumed on the anode catalyst layer surface due to the electrochemical reactions and the
water transport across the membrane.
Figure 3.7. Velocity profiles in the anode GDL: 3D profile (upper); on the y-z plane at x=25 mm
(lower). The nominal current density is 1.2 A/cm2.
channel area. The concentration of oxygen at the catalyst layer is balanced by the oxygen that
is being consumed and the amount of oxygen that diffuses towards the catalyst layer, driven
by the concentration gradient. The lower diffusivity of the oxygen along with the low
concentration of oxygen in ambient air results in noticeable oxygen depletion under the land
areas. At a low current density, the oxygen consumption rate is low enough not to cause
diffusive limitations, whereas at a high current density the concentration of oxygen under the
land areas has already reached near-zero values. Since, in addition, the local current density
of the cathode side reaction depends directly on the oxygen concentration; this means that the
local current density distribution under the land areas is much smaller than under the channel
areas, especially near the outlet. It becomes clear that the diffusion of the oxygen towards the
catalyst layer is the main impediment for reaching high current densities.
Due to the relatively low diffusivity of the oxygen compared to the hydrogen, the cathode
operation conditions usually determine the limiting current density when the fuel cell is run
on humidified air. This is because an increase in current density corresponds to require higher
oxygen consumption, which is limited by diffusivity of oxygen.
Figure 3.8. Oxygen molar fraction distribution in the cathode side for three different nominal current
densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
328 Maher A.R. Sadiq Al-Baghdadi
Figure 3.9. Hydrogen molar fraction distribution in the anode side for three different nominal current
densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
Figure 3.10. Water molar fraction distribution in the cell for three different nominal current densities:
0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
the catalyst layer and this leads to a significantly larger amount of heat being generated here.
The maximum temperature is more than 7 K above the gas inlet temperature and it occurs
inside the cathode catalyst layer.
Figure 3.11. Temperature distribution in the cell for three different nominal current densities: 0.3 A/cm2
(upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
Figure 3.12. Temperature distribution at the cathode side catalyst layer for three different nominal
current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
PEM Fuel Cell Modeling 335
Figure 3.13. Temperature profiles along through-plane direction for three different nominal current
densities of 0.3 A/cm2, 0.7 A/cm2, and 1.2 A/cm2.
Figure 3.14. Dimensionless local current density distribution ( i c I ), at the cathode side catalyst layer
for three different nominal current densities: 0.3 A/cm (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
2
the non-uniform local current production and show that ohmic losses are larger in the area of
the catalyst layer under the flow channels.
Figure 3.15. Activation overpotential distribution at the cathode side catalyst layer for three different
nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
340 Maher A.R. Sadiq Al-Baghdadi
Figure 3.16. Activation overpotential distribution at the anode side catalyst layer for three different
nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
Figure 3.17. Ohmic overpotential distribution in the anode and cathode gas diffusion layers for three
different nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
PEM Fuel Cell Modeling 343
Figure 3.18. Membrane overpotential distribution across the membrane due to proton transport for three
different nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
Figure 3.19. Diffusion overpotential distribution at the cathode side catalyst layer for three different
nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
346 Maher A.R. Sadiq Al-Baghdadi
Figure 3.20. Diffusion overpotential distribution at the anode side catalyst layer for three different
nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
displacement (contour plots) and deformation shape (scale enlarged 300 times) for GDLs,
membrane, and MEA at a nominal current density of 1.2 A/cm2 on the y-z plane at x=10 mm.
The figure illustrates the effect of stresses on the MEA. Because of the different thermal
expansion and swelling coefficients between gas diffusion layers and membrane materials
with non-uniform temperature distributions in the cell during operation, hygrothermal stresses
and deformation are introduced. The non-uniform distribution of stresses, caused by the
temperature gradient in the MEA, induces localized bending stresses, which can contribute to
delaminating between the membrane and the GDLs.
3.12. Summary
A three-dimensional, single-phase CFD model of a PEM fuel cell has been presented in this
section. The complete set of equations was given, and the computational procedure was
outlined. The results of the base case show good agreement with experimentally obtained
data, taken from the literature. A detailed distribution of the reactants and the temperature
field inside the fuel cell for different current densities were presented. In addition, potential
distribution in the membrane and gas diffusion layers, activation overpotential distribution,
diffusion overpotential distribution, and local current density distribution were presented.
Water management issues for the polymer membrane were addressed. Furthermore, the hygro
and thermal stresses in membrane, which developed during the cell operation, were modeled
and the behavior of the membrane was investigated. A grid refinement study revealed that
already for the base-case grid that was used the solution proofed to be grid-independent.
This model can be used to provide fundamental understanding of the transport
phenomena that occur in a fuel cell, and furthermore provide guidelines for fuel cell design
and prototyping.
One of the simplifications of the current model is the assumption that the volume of the
liquid water inside the gas diffusion layer is negligible (single-phase). In order to eliminate
PEM Fuel Cell Modeling 349
this shortcoming, a multi-phase model has been developed, which will be presented in the
following section.
Figure 3.21. von Mises stress distribution in the membrane for three different nominal current densities:
0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).
Figure 3.22. Total displacement (slice contour plots) and deformed shape plots (scale enlarged 180
times) for the membrane for three different nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2
(middle); 1.2 A/cm2 (lower).
352 Maher A.R. Sadiq Al-Baghdadi
Figure 3.23. Total displacement (slice contour plots) and deformed shape plots (scale enlarged 300
times) for the GDLs (upper), membrane (middle), and MEA (lower) at a nominal current density of 1.2
A/cm2 on the y-z plane at x=10 mm.
Figure 3.24. The break-up of different loss mechanisms at base case conditions.
354 Maher A.R. Sadiq Al-Baghdadi
Figure 3.25. Efficiency and power density of the cell at base case conditions.
Water management is one of the critical operation issues in proton exchange membrane fuel
cells. Spatially varying concentrations of water in both vapor and liquid form are expected
throughout the cell because of varying rates of production and transport. Water emanates
from two sources: the product water from the oxygen-reduction reaction in the cathode
catalyst layer and the humidification water carried by the inlet streams or injected into the
fuel cell.
One of the main difficulties in managing water in a PEM fuel cell is the conflicting
requirements of the membrane and of the catalyst gas diffusion layer. On the cathode side,
excessive liquid water may block or flood the pores of the catalyst layer, the gas diffusion
layer or even the gas channel, thereby inhibiting or even completely blocking oxygen mass
transfer. On the anode side, as water is dragged toward the cathode via electro-osmotic
transport, dehumidification of the membrane may occur, resulting in deterioration of protonic
conductivity. In the extreme case of complete drying, local burnout of the membrane may
result.
PEM Fuel Cell Modeling 355
Devising better water management is therefore a key issue in PEMFC design, and this
requires improved understanding of the parameters affecting water transport in the membrane.
Using as a basis the single-phase model presented in section 3, a multi-phase model has
been developed that accounts for both the gas and liquid phase in the same computational
domain and thus allows for the implementation of phase change inside the gas diffusion
layers. The model includes the transport of liquid water within the porous electrodes as well
as the transport of gaseous species, protons, energy, and water dissolved in the ion conducting
polymer.
Water is assumed to be exchanged among three phases; liquid, vapor, and dissolved, and
equilibrium among these phases is assumed. Water transport inside the porous gas diffusion
layer is described by two physical mechanisms: viscous drag and capillary pressure forces
and is described by advection within the gas channels. Liquid water, created by the
electrochemical reaction and condensation, is dragged along with the gas phase. At the
cathode side, the humidity level of the incoming air determines whether this drag is directed
into or out of the gas diffusion layer, whereas at the anode side this drag is always directed
into the GDL. The capillary pressure gradient drives the liquid water out of the gas diffusion
layers into the flow channels.
Water transport across the membrane is also described by two physical mechanisms:
electro-osmotic drag and diffusion. The balance between the electro-osmotic drag of water
from anode to cathode and back diffusion from cathode to anode yields the net water content
through the membrane. The present multi-phase model is capable of identifying important
parameters for the wetting behavior of the gas diffusion layers and can be used to identify
conditions that might lead to the onset of pore plugging, which has a detrimental effect of the
fuel cell performance.
4.2. Assumption
The assumptions made in this multi-phase model are basically identical to the ones stated in
section 3. In order to implement the phase change of water, the following additional
assumptions were made:
1- Liquid water exists in the form of small droplets of specified diameter only.
2- Inside the channels the liquid phase and the gas phase share the same pressure field.
3- Equilibrium prevails at the interface of the water vapor and liquid water.
4- Heat transfer between the gas-phase and the liquid water is idealized, i.e. both phases
share the same temperature field.
(rg g u g ) = 0 (4.1)
and for the liquid water (liquid phase) inside the channel becomes;
(rl l u l ) = 0 (4.2)
Two sets of momentum equations are solved in the channels, and they share the same
pressure field;
Pg = Pl = P (4.3)
Under these conditions, it can be shown that the momentum equations becomes;
2
[
( g u g u g g u g ) = rg P + g u g + g (u g )
T
] (4.4)
3
and
[
( l u l u l l u l ) = rl P + l u l + l (u l )
2 T
] (4.5)
3
Multiple species are considered in the gas phase only, and the species conservation
equation in multi-component, multi-phase flow is the same as in the single-phase
computations in section 3, except for the consideration of the volume fraction of the gas
phase;
M M P
+ (x j y j )
N
g g i Dij
r y y j + y j +
j =1 M j M P =0 (4.6)
T
rg g y i u g + DiT
T
(rg ( g Cp g u g T k g T )) = 0 (4.7)
The gas phase and the liquid phase are assumed to be in thermodynamic equilibrium;
hence the temperature of the liquid water is the same as the gas phase temperature.
The constitutive equations are the same as in the single-phase case. For the liquid phase
the following properties are considered;
The liquid water density is calculated as follow [71];
PEM Fuel Cell Modeling 357
B 2
A + + CT + DT
l = 1 10 3.10 T
(4.9)
and
The momentum equation for the gas phase reduces to Darcys law, which is, however,
based on the relative permeability for the gas phase (KP ) . The relative permeability accounts
for the reduction in pore space available for one phase due to the existence of the second
phase. The relative permeability for the gas phase is given by;
The momentum equation for the gas phase inside the gas diffusion layer becomes;
u g = (1 sat )
Kp
P (4.14)
g
358 Maher A.R. Sadiq Al-Baghdadi
Two liquid water transport mechanisms are considered; shear, which drags the liquid
phase along with the gas phase in the direction of the pressure gradient, and capillary forces,
which drive liquid water from high to low saturation regions [60]. Starting from Darcys law,
the following equation can write;
Kp l
ul = Pl (4.15)
l
where the liquid water pressure stems from the gas-phase pressure and the capillary pressure
according to [60];
Pc
Pl = P Pc = P sat (4.16)
sat
Introducing this expression into Equation (4.15) yields a liquid water velocity field
equation;
KPl KPl Pc
ul = P + sat (4.17)
l l sat
12
Pc = f (sat ) (4.18)
KP
Z
T T
= 1 c (1 10 3 ) (4.20)
Tc T1
M M P
+ (x j y j )
N
(1 sat ) y i Dij y j + y j +
P
g
j =1 M j M
= m phase (4.21)
T T
(1 sat ) g y i u g + Di
T
The heat transfer in the gas diffusion layers is governed by the energy equation as
follows;
The gas phase and the liquid phase are assumed to be in thermodynamic equilibrium, i.e.,
the liquid water and the gas phase are at the same temperature. The enthalpy of evaporation
for water is calculated as follow [71];
Z
T T
H evap = H evap ,l c 4186.80 (4.23)
c
T T1
x w 0 x w
m evap = M H 2O N D k xm Ddrop (4.24)
1 x w0
condition, the surface concentration can be calculated based on the saturation pressure and is
only a function of temperature.
The heat-transfer coefficient for convection around a sphere is well established, and by
invoking the analogy between convective heat and mass transfer, the following mass-transfer
coefficient is obtained [60];
c air D H 2O
12 13
D v g
k xm = 2 + 0.6 drop g (4.25)
Ddrop g g DH O
2
It is further assumed that all droplets have a specified diameter D drop , and the number of
droplets in each control volume is found by dividing the total volume of the liquid phase in
each control volume by the volume of one droplet;
sat.Vcv
ND = (4.26)
1
D drop
3
In the case when the calculated relative humidity in a control volume exceeds 100%,
condensation occurs and the evaporation term is switched off.
The case of condensation is more complex, because it can occur on every solid surface
area, but the rate of condensation can be different when it takes place on a wetted surface. In
addition, the overall surface area in each control volume available for condensation shrinks
with an increasing amount of liquid water present. Berning and Djilali [60] assumed that the
rate of condensation depends only on the level of oversaturation of the gas phase multiplied
by a condensation constant. Thus, the mass flux of water undergoing condensation in each
control volume can be represented by;
x w 0 x w
m cond = C (4.27)
1 x w0
4.3.4. Membrane
The balance between the electro-osmotic drag of water from anode to cathode and back
diffusion from cathode to anode yields the net water flux through the membrane [44];
PEM Fuel Cell Modeling 361
(DW cW )
i
N W = n d M H 2O (4.28)
F
1 1
DW = 1.3 10 10 exp 2416 (4.29)
303 T
The variable cW represents the number of water molecules per sulfonic acid group
1
(i.e. mol H 2 O equivalent SO 3 ).The water content in the electrolyte phase is related to
water vapor activity via [64, 65];
xW P
a= (4.31)
Psat
The same boundary conditions are applied as in the single-phase model (section 3). Again,
symmetry boundary conditions are applied in the y and the z directions, thereby reducing the
size of the computational domain and computational costs. In the x direction, zero flux
conditions are applied at all interfaces except for the flow channels. At the inlets of the gas-
flow channels, the incoming velocity is calculated as a function of the desired current density
and stoichiometric flow ratio, as described in section 3.7. The gas streams entering the cell are
fully humidified, but no liquid water is contained in the gas stream. At the outlets, the
pressure is prescribed for the momentum equation and a zero gradient conditions are imposed
for all scalar equations.
The same computational procedure and algorithm are used as in the single-phase model
(section 3). Due to the complexity of this model with a large spatial variation in competing
transport and phase-change mechanisms, the computational time required was about three
times greater than in single-phase.
362 Maher A.R. Sadiq Al-Baghdadi
The parameters introduced to account for the multi-phase flow and phase change phenomena
are listed in Table 4.1. With the exception of the parameters listed in Table 4.1, the
physicochemical and geometric parameters are identical to those used in the single-phase
simulations (section 3).
The multi-phase model is validated by comparing model results to experimental data provided
by Wang et al. [57]. The importance of phase change to the accurate modeling of fuel cell
performance is also illustrated. Performance curves with and without phase change as well as
experimental data are shown in Figure 4.1 for the base case conditions. Comparison of the
two curves demonstrates that the effects of liquid water accumulation become apparent even
at relatively low values of current density. Furthermore, when liquid water effects are not
included in the model, the cell voltage dose not exhibit an increasingly steep drop as the cell
approaches its limiting current density. This drop off in performance is clearly demonstrated
by experimental data, but cannot be accurately modeled without the incorporation of phase
change. By including the effects of phase change, the current model is able to more closely
simulate performance, especially in the region where mass transport effects begin to
dominate.
4.8. Results
Results with and without phase change for the cell operates at nominal current density of 1.4
A/cm2 are discussed in this section. The selection of relatively high current density is due to
illustrate the phase change effects, where it becomes clearly apparent between single and
multi-phase model in the mass transport limited region.
Prior to proceeding with a detailed analysis of the results, it is useful to discuss some of
the phase-change mechanisms. The central parameter for determining the direction of phase
change is the relative humidity of the gas phase;
PH 2O
= (4.32)
Psat
i.e., the ratio of partial pressure of the water vapor in the gas-phase to the saturation pressure,
which is a function of temperature only. According to Daltons law the partial pressure of a
species is equal to its molar fraction multiplied with the total pressure of the gas phase, i.e.;
P
= xH O (4.33)
2
Psat
When the relative humidity is below 100% in the presence of liquid water, this give rise
to evaporation. Condensation, on the other hand, occurs when the relative humidity exceeds
100% in the presence of condensation surfaces, which are in abundance inside the gas-
diffusion layer. The gas diffusion layer of a PEM fuel cell is particularly interesting for phase-
change considerations, because all three parameters on the right side of Equation. 4.33 vary,
resulting in competing directions of phase change as follows:
1. The molar water fraction x H 2O increases inside the GDL, simply as a result of
reactant consumption. Provided the relative humidity of the incoming air is at 100%,
this process alone would lead to condensation.
2. The thermodynamic pressure P of the gas-phase changes inside the GDL. This is a
very interesting effect and, depending on the incoming gas condition, it can lead to
either evaporation or condensation. In the first place, there is a pressure drop inside
the GDL due again to reactants consumption. This pressure drop depends strongly on
the permeability of the gas diffusion layer, i.e., for the same amount of consumed
reactants, the pressure drop will be higher for a lower permeability. The bulk velocity
of the gas phase is directed into the GDL, and is governed by Darcys law. Thus, the
pressure drop inside the GDL depends strongly on the permeability. The partial
364 Maher A.R. Sadiq Al-Baghdadi
pressure of the water vapor decreases with the gas-phase pressure. Hence, this effect
alone would lead to undersaturation, causing evaporation.
3. The saturation pressure Psat increases with an increase in temperature caused by the
heat production term due to the electrochemical reaction. The order of magnitude of
the temperature increase depends primarily on the thermal conductivity of the gas
diffusion layer. It was found with the single-phase model that the temperature can
rise by a few degrees Kelvin and this effect alone would lead to evaporation of liquid
water.
The net phase change is a result of the balance between these competing, coupled, and
spatially varying mechanisms. It should be noted that the first two effects are also of
importance inside the gas flow channels; the oxygen depletion from inlet to outlet results in
oversaturation and condensation at the walls and channel/GDL interface, whereas the overall
pressure drop along the channel would alone cause evaporation. For the straight channel
section considered here, the total pressure drop is very small and hence the oxygen depletion
effect dominates, causing condensation. The dominant mechanisms highlighted in this
discussion are relevant to cases where the incoming air is at a high humidification level, as is
the case in practical fuel cell operation.
Figure 4.2. Gas phase velocity vectors (upper) and liquid water velocity vectors (lower) inside the
cathode GDL at a nominal current density of 1.4 A/cm2.
Figure 4.3. Gas phase velocity vectors (upper) and liquid water velocity vectors (lower) inside the
anode GDL at a nominal current density of 1.4 A/cm2.
the channels into the GDL. While the capillary pressure gradient drives the liquid water out of
the gas diffusion layers into the flow channels. Therefore, the liquid water flux is directed
from the GDL into the channel, i.e., in the opposite direction of the gas-phase velocity, where
it can leave the cell. The velocity of the liquid phase, however, is lower than for the gas
phase, which is due to the higher viscosity, and the highest liquid water velocity occurs at the
corners of the channel/GDL interface. The liquid water oozes out of the GDL, mainly at the
corners of the GDL/channel interface.
spatial variation of the saturation demonstrates again the three-dimensional nature of transport
processes in PEM fuel cells.
Clearly, liquid water saturation depends strongly on the specified capillary pressure, and
again, the permeability of the gas diffusion layer becomes the central parameter.
Figure 4.4. Liquid water saturation inside the cathode and anode GDLs at a nominal current density of
1.4 A/cm2.
368 Maher A.R. Sadiq Al-Baghdadi
Figure 4.5. Water content profiles through the MEA at a nominal current density of 1.4 A/cm2.
condensation of liquid water reduces the molar water vapor fraction in return. As a result, the
molar hydrogen fraction is increasing throughout the entire domain.
Figure 4.6. Oxygen molar fraction distribution in the cathode side predicted in single-phase model
(upper) and multi-phase model (lower) at a nominal current density of 1.4 A/cm2. Bar chart shows the
average oxygen molar fraction at the cathode CL.
370 Maher A.R. Sadiq Al-Baghdadi
The virtually constant hydrogen molar fraction is a particularly interesting feature when
considering that transport of the reactants towards the catalyst is predominantly via diffusion.
Figure 4.6. Hydrogen molar fraction distribution in the anode side predicted in single-phase model
(upper) and multi-phase model (lower) at a nominal current density of 1.4 A/cm2. Bar chart shows the
average hydrogen molar fraction at the anode CL.
PEM Fuel Cell Modeling 371
Figure 4.7. Water vapor molar fraction distribution in the cell predicted in single-phase model (upper)
and multi-phase model (lower) at a nominal current density of 1.4 A/cm2.
It can be seen from the figure that, the local current density distribution at the catalyst
layer predicted in multi-phase model has a much higher fraction of the total current, generated
under the channel area. This is due to the effects of liquid water inside the GDL, which
decreases its permeability to reactant gas flow (oxygen) and lead to the onset of pore plugging
by liquid water. This can lead to local hot-spots inside the membrane electrode assembly, and
leads to a further drying out of the membrane, thus increasing the electric resistance, which in
turn leads to more heat generation and can lead to a failure of the membrane. Thus, it is
important to keep the current density relatively even throughout the cell. Therefore, the multi-
phase model is capable of identifying important parameters for the wetting behavior of the
gas diffusion layers and can be used to identify conditions that might lead to the onset of pore
plugging, which has a detrimental effect of the fuel cell performance, especially in the mass
transport limited region.
PEM Fuel Cell Modeling 373
Figure 4.8. Diffusion overpotential distribution at the cathode catalyst layer predicted in single-phase
model (upper) and multi-phase model (lower) at a nominal current density of 1.4 A/cm2.
374 Maher A.R. Sadiq Al-Baghdadi
Figure 4.9. Diffusion overpotential distribution at the anode catalyst layer predicted in single-phase
model (upper) and multi-phase model (lower) at a nominal current density of 1.4 A/cm2.
PEM Fuel Cell Modeling 375
Figure 4.10. Dimensionless local current density distribution ( i c I ) at the cathode catalyst layer
predicted in single-phase model (upper) and multi-phase model (lower) at a nominal current density of
1.4 A/cm2.
376 Maher A.R. Sadiq Al-Baghdadi
4.9. Summary
This section presented a three-dimensional, two-phase model of the cathode and anode of a
PEM Fuel Cell. The mathematical model accounts for the liquid water flux inside the gas
diffusion layers by viscous and capillary forces and hence is capable of predicting the amount
of liquid water inside the gas diffusion layers. The physics of phase change are included in
this model by prescribing the local evaporation term as a function of the amount of liquid
water present and the level of undersaturation, whereas the condensation has been simplified
to be a function of the level of oversaturation only.
Three different physical mechanisms that lead to phase change inside the gas diffusion
layers were identified. A rise in temperature because of the electrochemical reaction leads to
evaporation, mainly at the cathode side. If the gases entering the cell are fully humidified, the
depletion of the reactants leads to an increase in the partial pressure of the water vapor, and
hence to condensation along the channel and inside the gas diffusion layers. Finally, a
decrease in the gas phase pressure inside the gas diffusion layers leads to a decrease in the
water vapor pressure, and hence causes evaporation.
The results show that the multi-phase model is capable of identifying important
parameters for the wetting behavior of the gas diffusion layers and can be used to identify
conditions that might lead to the onset of pore plugging, which has a detrimental effect of the
fuel cell performance, especially in the mass transport limited region.
5. Conclusion
This chapter is a practical summary of how to create CFD models, and how to interpret
results. A review of recent literature on PEM fuel cell modeling was presented. A full three-
dimensional computational fluid dynamics model of a PEM fuel cell with straight flow
channels has been developed. This model provides valuable information about the transport
phenomena inside the fuel cell such as reactant gas concentration distribution, liquid water
saturation distribution, temperature distribution, potential distribution in the membrane and
gas diffusion layers, activation overpotential distribution, diffusion overpotential distribution,
and local current density distribution. In addition, the hygro and thermal stresses in
membrane, which developed during the cell operation, were modeled and investigated.
The main feature of this model is the implementation of the new algorithm that allows for
the calculation of the electrochemical kinetics without simplifications. This calculation
involves the coupling of the potential field with the reactant species concentration field,
which results in an accurate prediction of local current density distribution. Results are
physically consistent and in good agreement with available experimental data.
The model is shown to be able to: (1) understand the many interacting, complex
electrochemical, phase change mechanism, and transport phenomena that cannot be studied
experimentally; (2) identify limiting steps and components; and (3) provide a computer-aided
tool for design and optimize future fuel cell with much higher power density and lower cost.
Finally, this chapter of "PEM Fuel Cell Modeling" looks at how engineers can model PEM
fuel cells to get optimal results for any application.
PEM Fuel Cell Modeling 377
References
[1] Barbir, F.; Gomez, T. Int. J. Hydrogen Energy. 1997, 22(10), 1027-1037.
[2] von Spakovsky, M.R.; Olsommer, B. Energy Conversion Manage. 2002, 43(9-12),
1249-1257.
[3] Wright, S.E. Renewable Energy J. 2004, 29(2), 179-195.
[4] Pettersson, L.J.; Westerholm, R. Int. J. Hydrogen Energy. 2001, 26(3), 243-264.
[5] Ernst, W.; Nerschook, J. Fuel Cell Review Magazine. 2004, 1(1), 25-28.
[6] Carrette, L.; Friedrich, K.A.; Stimming, U. Fuel Cells J. 2001, 1(1), 5-39.
[7] Okada, O.; Yokoyama, K. Fuel Cells J. 2001, 1(1), 72-77.
[8] Spencer, P.; Barrett, S. Int. Fuel Cell Magazine (fcFOCUS) Fuel Cell Supplement.
2003, 6-9.
[9] Francesco, D.M.; Arato, E. J. Power Sources. 2002, 108(1-2), 41-52.
[10] Laurencelle, F.; Chahine, R.; Hamelin, J.; Agbossou, K.; Fournier, M.; Bose, T.K.;
Laperriere, A. Fuel Cells J. 2001, 1(1), 66-71.
[11] Bak, P.E. Magazine Hydrogen Cars Business (H2CARSBIZ). 2004, 2(1), pp.12.
[12] Ahluwalia, R.K.; Wang, X.; Rousseau, A.; Kumar, R. J. Power Sources. 2004, 130(1-2),
192-201.
[13] Pei, P.; Ouyang, M.; Lu, Q.; Huang, H.; Li, X. Int. J. Hydrogen Energy. 2004, 29(10),
1001-1007.
[14] Schreier, P.G. COMSOL NEWS A Technical Computing Magazine. 2005, 1(1), 25-28.
[15] Litster, S.; McLean, G. J. Power Sources. 2004, 130(1-2), 61-76.
[16] Rajallakshmi, N.; Jayanth, T.T.; Thangamuthu, R.; Sasikumar, G.; Sridhar, P.;
Dhathathreyan, K.S. Int. J. Hydrogen Energy. 2004, 29(10), 1009-1014.
[17] Eikerling, M.; Kharkats, M.Y.; Kornyshev, A.A.; Volfkovich, Y.M. J. Electrochem.
Soc. 1998, 145(8), 26842699.
[18] Eikerling, M.; Kornyshev, A.A. J. Electroanalytical Chemistry. 2001, 502(1-2), 1-14.
[19] Susai, T.; Kaneko, M.; Nakato, K.; Isono, T.; Hamada, A.; Miyake, Y. Int. J. Hydrogen
Energy. 2001, 26(6), 631-637.
[20] Kolde, J.A.; Bahar, B.; Wilson, M.S.; Zawodzinski, T.A.; Gottesfeld. S. Proc.
Electrochemical Soc. 1995, 23, 193201.
[21] Eikerling, M.; Kornyshev, A.A.; Kuznetsov, A.M.; Ulstrup, J.; Walbran, S. J. Phys.
Chem. B. 2001, 105, 3646-4662.
[22] Mehta, M.; Cooper, J.S. J. Power Sources. 2003, 114(1), 32-53.
[23] Barbir, F.; Braun, J.; Neutzler, J. J. New materials Electrochemical Systems. 1999, 2(3),
197-200.
[24] Lampinen, M.J.; Fomino, M. J. Electrochem. Soc. 1993, 140(12), 35373546.
[25] Hyunchul, J.; Chao-Yang, W. J. Electrochem. Soc. 2004, 151(11), 19541960.
[26] Hirschenhofer, J.H.; Stauffer, D.B.; Engleman, R.R.; Klett; M.G. Fuel Cell Handbook;
DOE/NETL-2002/1179 (DE-AM26-99FT40575); U.S. Department of Fossil Energy,
Morgantown Energy Technology Center, Morgantown WV, 2002.
[27] Dhathathreyan, K.S.; Sridhar, P.; Sasikumar, G.; Ghosh, K.K.; Velayutham, G.;
Rajalakshmi, N.; Subramaniam, C.K.; Raja, M.; Ramya, K. Int. J. Hydrogen Energy.
1999, 24(11), 1107-1115.
[28] Cheddie, D.; Munroe, N. J. Power Sources. 2005, 147(1-2), 72-84.
378 Maher A.R. Sadiq Al-Baghdadi
[29] Standaert, F.; Hemmes, K.; Woudstra, N. J. Power Sources. 1996, 63(2), 212234.
[30] Standaert, F.; Hemmes, K.; Woudstra, N. J. Power Sources. 1998, 70(2), 181189.
[31] Springer, T.E.; Zawodzinski, T.A.; Gottesfeld, S. J. Electrochem. Soc. 1991, 138(8),
23342342.
[32] Amphlett, J.C.; Baumert, R.M.; Mann, R.F.; Peppley, B.A.; Roberge, P.R. J.
Electrochem. Soc. 1995, 142(1), 18.
[33] Pisani, L.; Murgia, G.; Valent'n, M.; D'Aguanno, B. J. Power Sources. 2002, 108(1-2),
192203.
[34] Maggio, G.; Recupero, V.; Pino, L. J. Power Sources. 2001, 101(2), 275286.
[35] Maxoulis, C.N.; Tsinoglou, D.N.; Koltsakis, G.C. Energy Conversion Manage. 2004,
45(4), 559573.
[36] Lee, W.-Y.; Park, G.-G.; Yang, T.-H.; Yoon, Y.-G.; Kim, Ch.-S. Int. J. Hydrogen
Energy. 2004, 29(9), 961-966.
[37] Al-Baghdadi, M.AR.S. Renewable Energy. 2005, 30(10), 1587-1599.
[38] Al-Baghdadi, M.AR.S.; Al-Janabi, H.AK.S. Turkish J. Eng. Env. Sci. 2005, 29(4),
235-240.
[39] Jeng, K.T.; Kuo, C.P.; Lee, S.F. J. Power Sources. 2004, 128(2); 145151.
[40] Jaouen, F.; Lindbergh, G.; Sundholm, G. J. Electrochem. Soc. 2002, 149(4),A437-A447.
[41] Grujicic, M.; Chittajallu, K.M. Chemical Engineering Science. 2004, 59(24),58835895.
[42] Grujicic, M.; Zhao, C.L.; Chittajallu, K.M.; Ochterbeck, J.M. Materials Science
Engineering B. 2004, 108(3), 241252.
[43] Francesco, D.M.; Arato, E.; Costa, P. J. Power Sources. 2004, 132(1-2), 127-134.
[44] Sui, P.; Djilali N. J. Fuel Cell Sci. Tech. ASME. 2005, 2(1), 149-155.
[45] Bernadi, D.M.; Verbrugge, M.W. AIChE J. 1991, 37(8), 1151-1163.
[46] Bernadi, D.M.; Verbrugge, M.W. J. Electrochem. Soc. 1992, 139(9), 2477-2491.
[47] Baschuk, J.J.; Li, X. J. Power Sources. 2000, 86(1-2), 181196.
[48] Fuller, T.F.; Newman, J. J. Electrochem. Soc. 1993, 140(5), 12181225.
[49] Um, S.; Wang, C.Y.; Chen, K.S. J. Electrochem. Soc. 2000, 147(12), 44854493.
[50] Ge, S.; Yi, B. J. Power Sources. 2003, 124(1), 111.
[51] Siegel, N.P.; Ellis, M.W.; Nelson, D.J.; von Spakovsky, M.R. J. Power Sources. 2004,
128(2), 173184.
[52] Siegel, N.P.; Ellis, M.W.; Nelson, D.J.; von Spakovsky, M.R. J. Power Sources. 2003,
115(1), 81-89.
[53] Zhou, T.; Liu, H. Int. J. Transport Phenomena. 2001, 3(3), 177198.
[54] Dutta, S.; Shimpalee, S.; Van Zee, J.W. Int. J. Heat Mass Transfer. 2001, 44(11),
20292042.
[55] Berning, T.; Lu, D.M.; Djilali, N. J. Power Sources. 2002, 106(1-2), 284-294.
[56] Yan, W.-M.; Chen, F.; Wu, H.-Y.; Soong, C.-Y.; Chu, H.-S. J. Power Sources. 2004,
129(2), 127137.
[57] Wang, L.; Husar, A.; Zhou, T.; Liu, H. Int. J. Hydrogen Energy. 2003, 28(11),
1263 1272.
[58] Berning, T.; Djilali, N. J. Power Sources. 2003, 124(2), 440-452.
[59] Kumar, A.; Reddy, R.G. J. Power Sources. 2003, 114(1), 5462.
[60] Berning, T.; Djilali, N. J. Electrochem. Soc. 2003, 150(12), A1589-A1598.
[61] Cha, S.-W.; O'Hayre, R.; Lee, S.J.; Saito, Y. J. Electrochem. Soc. 2004, 151(11),
A1856-A1864.
PEM Fuel Cell Modeling 379
[62] Nguyen, P.T.; Berning, T.; Djilali, N. J. Power Sources. 2004, 130(1-2), 149-157.
[63] Um, S; Wang, C.Y. J. Power Sources. 2004, 125(1), 4051.
[64] Hu, M.; Gu, A.; Wang, M.; Zhu, X.; Yu, L. Energy Conversion Manage. 2004, 45(11-
12), 18611882.
[65] Hu, M.; Gu, A.; Wang, M.; Zhu, X.; Yu, L. Energy Conversion Manage. 2004, 45(11-
12), 18831916.
[66] Hyunchul, J.; Meng, H.; Wang, C.Y. Int. J. Heat Mass Transfer. 2005, 48(7), 1303
1315.
[67] Sivertsen, B.R.; Djilali, N. J. Power Sources. 2005, 141(1), 65-78.
[68] Webber, A.; Newman, J. AIChE J. 2004, 50(12), 32153226.
[69] Tang, Y.; Santare, M.H.; Karlsson, A.M.; Cleghorn, S.; Johnson, W.B. J. Fuel Cell Sci.
Tech. ASME. 2006, 3(5), 119-124.
[70] Fuller, E.N.; Schettler, P.D.; Giddings, J.C. Ind. Eng. Chem. 1966, 58(5), 18-27.
[71] Coker, A.K. FORTRAN programs for chemical process design, analysis, and
simulation; ISBN: 0-88415-280-4; Gulf Publishing Company: Houston, Texas, 1995.
[72] De La Rue, R.E.; Tobias, C.W. J. Electrochem. Soc. 1959, 106(9), 827836.
[73] Gurau, V.; Liu, H.; Kakac, S. AIChE J. 1998, 44(11), 24102422.
[74] Parthasarathy, A.; Srinivasan, S.; Appleby, J.A.; Martin, C.R. J. Electrochem. Soc.
1992, 139(10), 28562862.
[75] Product Information, 2005, DuPont Nafion PFSA Membranes N-112, NE-1135, N-
115, N-117, NE-1110 Perfluorosulfonic Acid Polymer. NAE101.
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 381-410 2007 Nova Science Publishers, Inc.
Chapter 8
Zhigang Qi*
Plug Power Inc.
968 Albany Shaker Road, Latham, NY 12110, USA
Abstract
In order to achieve high performance and low catalyst loading, a proton-exchange
membrane (PEM) fuel cell typically employs noble metal catalysts that are dispersed on a
support such as carbon. The chemical and physical properties of carbon largely affect the
dispersion of the catalyst, the strength of the interaction between the carbon and the catalyst
particles, the making of catalyst ink formulations, and the utilization of the catalyst. An
interesting and useful aspect of carbon is that its surface can be chemically modified to render
it with certain desired properties. For example, proton conducting groups can be covalently
bonded onto the surface of carbon black such as Vulcan XC-72 to make it possess some ionic
conductance, which in turn significantly increases the catalyst utilization and the fuel cell
performance. Accompanying all the benefits, a carbon-type support also raises some potential
problems. One serious concern is the corrosion of an amorphous carbon support during the
operation of a fuel cell, which subsequently results in the loss of the catalyst-electrolyte-
reactant three-phase sites. This factor alone may prevent a PEM fuel cell from achieving a
target of 40,000 hours of operation for stationary applications. This article reviews various
aspects of surface functionalization of carbon supports for PEM fuel cells.
1. Introduction
Due to concerns of global warming and the accelerating depletion of world oil reserves,
renewable energy sources and technologies are attracting close attention worldwide [1,2].
Fuel cells have been on a fast developmental track in the past two decades because they can
*
E-mail address: zhigang_qi@plugpower.com, Tel: 1-518-738-0229 (work), Fax: 1-518-782-7914
382 Zhigang Qi
potentially offer many advantages over the conventional power generation technologies such
as the international combustion engines, generators, and turbines. The advantages comprise of
higher efficiency, lower pollution, and quieter operation.
Fuel cells are typically classified according to the types of the electrolytes they use. There
are alkaline fuel cells (operating temperature ~60-200 oC), phosphoric acid fuel cells (~ 120-
210 oC), molten carbonate fuel cells (~ 650 oC), solid oxide fuel cells (~600-1000 oC), and
proton-exchange membrane (PEM) fuel cells (RT-90 oC). A direct methanol fuel cell
(DMFC) can be considered as a special type of the PEM fuel cell that uses methanol (in either
vapor or liquid form) rather than hydrogen as the fuel. This review article focuses exclusively
on PEM fuel cells.
Thanks to its quick start up and its relatively low operating temperatures, a PEM fuel cell has
the widest application range of all types of fuel cells. The core of a PEM fuel cell is the so-
called membrane electrode assembly (MEA) that is composed of an anode, a cathode, and a
PEM. A fuel such as hydrogen is oxidized at the anode to form protons and electrons (Eq. 1).
An oxidant such as oxygen (typically from air) is reduced at the cathode by combining with
protons and electrons to form water (Eq. 2). The PEM allows protons to transport from the
anode to the cathode, physically separates the anode from the cathode, and prevents hydrogen
from mixing with the oxygen.
It is mainly the PEM that distinguishes a PEM fuel cell from all other types of fuel cells.
As its name implies, a PEM has the capability of transporting protons. It is typically made of
a solid ionomer with acidic groups such as sulfonic acid (SO3H) at the end of the polymer
side chains. Polystyrene sulfonic acid is one such ionomer, and it was used as the PEM in the
early days of the PEM fuel cell development around the 1960s. However, since the PEM fuel
cell environment is warm, corrosive, and oxidative (at cathode), an ionomer with higher
chemical and electrochemical stability is required. State-of-the-art PEMs are made of
perfluorinated polysulfonic acids, and include DuPonts Nafion.
Both the hydrogen oxidation reaction (HOR) and the oxygen reduction reaction (ORR)
are kinetically sluggish at the PEM fuel cell operating temperatures, and precious metals such
as Pt and its alloys are typically used to catalyze the reactions. In the early days, the
electrodes were made of Pt black powder and polytetrafluoroethylene (PTFE or Teflon) with
the latter functioning as a binder and a water-repelling agent. The Pt loading was 4 mg cm-2 or
higher in order to achieve acceptable performance due to the large Pt particle size (e.g.,
smaller surface area) and the low Pt utilization. For such electrodes, only the catalyst particles
located at the electrode-membrane interface have the possibility of participating in the
electrochemical reactions. The catalyst particles that are not in direct contact with the
membrane are wasted due to the lack of proton conductance throughout the catalyst layer.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 383
The situation was improved when Pt was dispersed onto a support such as carbon black that
conducts electrons. The carbon support enables the Pt particles to be highly dispersed and to
achieve a size of a few nanometers. Due to the reduction of particle size from 20 nm or larger
for Pt black to 2 nm or less for Pt supported on carbon (commonly expressed as Pt/C), the
surface area of the particles is significantly increased, which in turn lowers the required Pt
loading in the electrode. Due to the physicochemical interactions between the Pt particles and
the support, the electronic structure of the Pt atoms may be modified to result in slightly
higher catalytic activities. In addition, the interactions anchor the Pt particles on the support,
thereby retarding their aggregation process, leading to slower loss of catalyst surface area and
thus more stable electrode performance.
Various types of carbon black have been used as the support for PEM fuel cells due to
their low cost, high electronic conductivity, high corrosion resistance, and good chemical and
electrochemical stabilities. Liu et al. reviewed carbon-type catalyst supports such as carbon
black, nanostructured carbon, and mesoporous carbon for direct methanol fuel cells [3]. The
discussion is applicable to PEM fuel cells as well.
Carbon black is typically produced by thermal decomposition of organic hydrocarbon
materials. The resulting carbon black depends on the decomposition process, the hydrocarbon
material, and the post-formation treatment. Thermally treating an amorphous carbon black at
about 2000 oC in an inert gas will graphitize the surface layers. Graphitized carbon has better
corrosion resistance but lower surface area than the amorphous carbon. Depositing catalyst
particles on a graphitized carbon typically results in larger particle size than on an amorphous
carbon. It is believed that catalyst particles can only deposit on the surface defects, not on the
basal planes of the graphene sheet of carbon. Since an amorphous carbon has more surface
defects, it can provide more anchor sites for the deposition of catalyst particles, leading to a
smaller particle size. A monograph by Kinoshita has thoroughly reviewed various aspects of
carbon [4].
A new form of carbon is the single-walled and multi-walled nanotubes (CNTs) [5]. The
CNTs have a high length-to-width aspect ratio, high corrosion resistance, and good electronic
conductivity. Another form of carbon is nanofibers that are typically produced by
carbonization of polymer fibers.
Pt deposition can be carried out by a variety of methods such as chemical vapor
deposition, electrochemical reduction, and chemical processes. Chemical deposition is
probably the most popular method, and can be easily carried out by reducing a Pt precursor
such as hexachloroplatinic acid (H2PtCl6) with a reducing agent such as formaldehyde or
formic acid in the presence of a carbon black. The reducing agent sometimes can be generated
in situ as reported by Shim et al. who generated active hydrogen, the reducing agent, by a
plasma process [6].
The resulting Pt particle size is affected by factors such as the concentrations of the
reactants, type of the reducing agent, type of the catalyst precursor, type of the solvent [7],
type and characteristics of the carbon (e.g., particle size, porosity, surface area, degree of
graphitization, surface functional groups, presence of other particles [8,9], and reaction
conditions (e.g., temperature, pressure, and agitation). The surface area of carbon black can
range from less than 50 m2 g-1 to over 1000 m2 g-1. Carbon black with higher surface area
typically results in smaller Pt particles. The surface of carbon can be modified by a variety of
384 Zhigang Qi
Figure 1. High resolution TEM bright-field images of (a), (b) Pt/MWNT; and (c), (d) Pt/C [14].
Reproduced by permission of The Electrochemical Society, Inc.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 385
Some sulfur-containing groups may adversely affect the catalytic activity of the
supported Pt through surface poisoning. Swider and Rolison showed that vulcanized carbon
(VC) containing heterocyclic (e.g., thiophene-like) sulfur with a surface concentration of
0.33% atomic (or 0.88% wt.) caused poisoning to the Pt particles by the organosulfur
moieties when heated in the absence of water. The surface sulfur also slowed down the
oxidation of CO to CO2, leading to irremediable buildup of CO on the surface of the Pt
particles [13].
Alternatively, Pt colloidal particles can be produced in the absence of carbon supports.
When these Pt colloidal particles are mixed with a carbon black support, they preferentially
adsorb onto the surface of the carbon black particles.
When CNTs are used as catalyst supports (Fig. 1), their surfaces need to be treated in
order to be able to anchor the catalyst particles. The treatment is typically a chemical
oxidation process using an oxidizing acid such as nitric acid (HNO3). The oxygen-rich groups
formed on the surface of CNTs act as the anchor sites for the deposition of catalyst particles.
A greater amount of oxygen-rich groups can result in smaller catalyst particle size and higher
catalyst loading. Although a number of studies have been reported on the performance of
electrodes made of catalyzed CNTs [14,15], it is not conclusive whether CNTs can offer a
better performance than carbon black. In addition, the high cost of CNTs also limits their
usability in fuel cells.
The performance of a fuel cell is determined by the total surface area of the catalyst particles
that participate in the reactions. Ideally, all the surface of the Pt particles is used and the Pt
achieves 100% dispersion (e.g., Pt exists as individual atoms). In reality, the ideal situation
does not exist and only a small fraction of the Pt atoms can participate in the fuel cell
reaction. The reasons are that the Pt can not achieve 100% dispersion and the fuel cell
reactions require the so-called three-phase boundaries. It can be seen from Eqs. 1 and 2 that
both the anode and the cathode reactions involve protons, electrons, and reactants. So, only
the Pt surface that is accessible to protons, electrons, and the reactant is active, and such
regions are often called catalyst-electrolyte-reactant three phase boundaries as illustrated in
Fig. 2. All the other Pt surface area is basically wasted. For an electrode composed of Pt (or
386 Zhigang Qi
Pt/C) and PTFE, only the Pt at the electrode-membrane interface has the potential to provide
three-phase regions. The other Pt particles that are not in direct contact with the membrane
are simply wasted.
Catalyst
e- 3-phase
H+ region
Electrolyte
Figure 2. Illustration of catalyst-electrolyte-reactant three-phase boundary.
Pt covered by ionomer:
Less active or inactive
Pt at 3-phase
regions:
Active
Carbon
Bare Pt:
Inactive
2.2. Pt Alloys
Some Pt alloys have higher activity than Pt for the ORR. These binary or ternary alloys
typically consist of the first row transition metals such as Ti, V, Cr, Mn, Fe, Co, Ni, and Cu.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 387
The alloys can be made by either depositing the base metal onto Pt/C or depositing Pt and the
base metal(s) onto a carbon support simultaneously. The resulting materials need to be
sintered at about 1000 oC in order to form a true alloy. This process often causes coalescence
of the metal particles.
Pt alloys may also be used in the anode when the fuel contains CO. CO can strongly
absorb onto the surface of the Pt catalysts at PEM fuel cell operating temperatures. The
adsorption blocks the reaction sites and significantly slows the oxidation of hydrogen. One
way to alleviate CO poisoning is to use Pt alloys such as Pt-Ru and Pt-Mo for CO
concentrations less than about 100 ppm. In the presence of higher CO concentrations,
bleeding a few percentage of air into the anode fuel is helpful.
3. New Concepts
As shown in Fig. 3, the three-phase boundaries are scarce in a mixture of a catalyst with an
ionomer. If the catalyst is not in contact with the ionomer, it is not active due to the lack of
proton conductance. On the other hand, if the catalyst is covered fully by the ionomer, its
activity is also reduced because of the blockage that prevents the reactant from reaching the
underlying catalyst. Revolutionary approaches are needed in order to achieve higher usage of
the catalyst surface.
Qi et al. proposed that if the support can conduct both electrons and protons, the catalyst-
electrolyte-reactant three-phase regions are simplified to catalyst-reactant two-phase regions
(Fig. 4). Since most of the catalyst surface can meet this requirement, the overall catalyst
utilization should be significantly increased [22].
Figure 4. Illustration of three-phase boundary vs. two-phase boundary [22]. Reprinted from J.
Electroanal. Chem. 459, Qi, Z.; Lefebvre, M. C. & Pickup, P. G., Electron and proton transport in gas
diffusion electrodes containing electronically conductive proton-exchange polymers, 9-14, Copyright
(1998), with permission from Elsevier.
388 Zhigang Qi
* HC CH n * *
n*
N
polyacetylene H
Polypyrrole
*
N n* * NH n*
CH3
Polyaniline
Poly(3-methylpyrrole)
O O
*
S n*
*
S n *
Polythiophene
Poly(3,4-ethylenedioxythiophene)
Pt nanoparticles with ca. 4 nm diameter were deposited onto PPy/PSS particles by the
reduction of Pt(NH3)4Cl2 with formaldehyde under reflux in the presence of PPy/PSS
particles. It was observed through transmission electron microscopy (TEM) that little Pt
deposition occurred in the first 1.5 hours of reflux, while a large number of Pt nanoparticles
distributed quite homogeneously on the PPy/PSS particles formed after 2 hours of reflux (Fig.
6) [24]. Longer reflux time did not lead to increases in the average Pt particle size.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 389
PtOx nanoparticles with about 2 nm diameter were also deposited onto PPy/PSS particles
by oxidizing Na6Pt(SO3)4 with H2O2 in the presence of PPy/PSS. The electronic conductivity
of PPy/PSS was reduced by 5 orders of magnitude after the PtOx deposition, presumably due
to the overoxidation of PPy by H2O2. The overoxidation was likely to be accelerated and
worsened by the catalytically active PtOx particles.
A similar problem was encountered when conductive polyaniline/PSS was used as the
support. Fortunately, conducting poly(3,4-ethylenedioxythiophene) (PEDOT)/PSS composite
was found to be much more stable. Before Pt deposition, a PEDOT/PSS composite possessed
an electronic conductivity of 9.9 S cm-1, while the conductivity of Vulcan XC-72 was 3.0 S
cm-1 measured under the same conditions. In other words, the conductive polymer composite
was more than three times as conductive as the commercial carbon black. Pt nanoparticles
with ca. 4 nm diameter were deposited onto PEDOT/PSS composite particles through the
390 Zhigang Qi
reduction of H2PtCl6 by formaldehyde under reflux for 1 hour. The electronic conductivity of
the resulting Pt/PEDOT/PSS material was 4.0 S cm-1, a reduction of only about half of its
original conductivity. When the catalyst was evaluated in a half cell, the ORR performance
was similar to a commercial Pt/Vulcan XC-72 (Fig. 7). However, it was found that the
electronic conductivity of the Pt/PEDOT/PSS decreased to 5 mS cm-1 after 11 months of
storage in air, which indicated that inadequate long term stability of the conducting polymer
support was a significant challenge [25].
Figure 7. Polarization curves of () commercial 20% Pt/Vulcan XC-72 (0.31 mg Pt cm-2); () 37% Pt
on PEDOT/PSS (0.89 mg Pt cm-2); () 19% Pt on PEDOT/PSS (0.29 mg Pt cm-2); and () 20% Pt on
emulsion polymerized PEDOT/PSS (0.4 mg Pt cm-2) [25]. Reproduced by permission of The Royal
Society of Chemistry.
loss of the electronic conductivity is accelerated in a fuel cell environment due to the
overoxidation of the conductive polymer at the cathode, especially at potentials near the open
circuit voltage (~ 1.0 V). Fourth, the conductive polymer is reduced at the anode due to the
low voltage (~ 0.0 V) and the presence of hydrogen and thus becomes insulating. Although
the reduction is likely to be a reversible process, the composite is unable to be used at the fuel
cell anode.
4.1. Oxidation
It is well known that carbon can be oxidized by a variety of methods to achieve an oxygen-
rich surface. The oxidation can be performed using an oxidizing agent such as HNO3,
KMnO4, KClO3, and O3 or using oxygen-rich plasma treatment. The groups created are
carboxyl (-COOH), carbonyl (-C=O), phenol (-OH), quinone (=O), and ether (-O-). The
presence of these oxygen-rich groups makes the carbon surface more hydrophilic.
Pittman et al. studied carbon fiber oxidation by concentrated (70%) HNO3 at 115 oC, and
found that the generated surface acidic groups were primarily carboxylic acid (52%) and
phenolic hydroxyl (48%) [27]. They showed that the number of acidic groups increased
linearly with time although the density of the surface acidic groups reached a plateau after an
initial jump. They proposed that the oxidation process opened the structures of carbon and
thus more surface area became available for oxidation to account for the continuously
increase of surface acidic groups (Fig. 8).
Jia et al. employed an oxidized carbon support in order to increase the electrode
performance [28]. The treatment was carried out at reflux in concentrated HNO3 for 35-45
minutes. Three Pt/C samples were evaluated: Pt deposited on untreated Vulcan XC-72
(untreated), Pt deposited on HNO3-treated Vulcan XC-72 (pre-treated), and HNO3-treated
Pt/Vulcan XC-72 (post-treated). It was found that smaller Pt particles were obtained when Pt
was deposited onto the acid-treated carbon than onto the untreated carbon, and both the pre-
treated and post-treated Pt/Vulcan catalysts performed better than the untreated counterpart.
Impedance spectroscopy showed that the electrodes made using both pre-treated and post-
treated Pt/Vulcan had higher proton conductance than those made using untreated Pt/Vulcan.
The post-treated catalyst showed higher proton conductivity than the pre-treated catalyst, and
the former performed better than the latter. The improved performance was attributed to both
enhanced proton conductivity and increased wettability of the carbon support after oxidation
by HNO3 due to the formation of carboxylic acid and other oxygen-rich groups. Although it is
known that carboxyl groups can be stripped off quickly at temperatures higher than 300 oC,
decarboxylation is not expected to be significant at the temperatures that are suitable for PEM
fuel cell operation (e.g., < 100 oC).
392 Zhigang Qi
Figure 8. Illustration of severe carbon corrosion process [27]. Reprinted from Carbon 35, Pittman , Jr.,
C. U.; He, G.-R.; Wu, B. & Gardner, S. G., Chemical modification of carbon fiber surfaces by nitric
acid oxidation followed by reaction with tetraethylenepentamine, 317-331, Copyright (1977), with
permission from Elsevier.
Breakthrough was achieved when carbon particles were chemically functionalized by acidic
groups. Several different approaches have been discovered.
4.2.1. Silane
Easton et al. successfully anchored proton conducting sulfonic acid groups to the surface of
carbon or Pt-catalyzed carbon (i.e., Pt/C) via a silane linkage [29]. The linkage was achieved
by mixing Vulcan XC-72 or Pt/Vulcan-XC-72 with 2(4-chlorosulfonylphenyl)ethyl
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 393
Figure 9. Polarization curves of (, ) untreated; and (, ) sulfonated silane treated catalysts [29].
Reproduced by permission of The Electrochemical Society, Inc.
lower than that of the latter. In addition, the silanization process was not easy to control, and
toxic dichloromethane was used.
A better approach was employed by Xu et al. using a diazonium salt process [30,31]. An
alkylphosphonic or alkylsulfonic acid reacts with isobutyl nitrite [(CH3)2CHCH2ONO] in
aqueous solution to form a diazonium salt as illustrated by Eqs. 3 and 5. The diazonium salt
then reacts with carbon by replacing its surface hydrogen atoms to have the alkylphosphonic
or alkylsulfonic groups attached to the surface of carbon as shown by Eqs. 4 and 6. The alkyl
group with various length and structure can be selected, which makes this method very
versatile.
1.2 0.7
0.6
1.0
0.5
0.4
0.6
0.3
0.4
0.2
0.2
0.1
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
The optimal Nafion content for electrodes made using ethylphosphonic acid (-
CH2CH2PO3H2) functionalized 20% Pt/Vulcan XC-72 was 20% wt., while the optimal
Nafion content for electrodes made using untreated Pt/Vulcan XC-72 was 30% wt. This
50% reduction in Nafion usage translates to a lower electrode cost. Meanwhile, the former
performed better than the latter in the entire current density region (Fig. 10). For example, the
peak power of the former was 20% greater than that of the latter. Methylphosphonic acid and
propylphosphonic acid functionalized Pt/C also reduced the optimal Nafion content to 20%
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 395
wt., but the enhancement to fuel cell performance was not as good as that produced using
ethylphosphonic acid [30].
The optimal Nafion content for electrodes made using ethylsulfonic acid (-
CH2CH2SO3H) functionalized 20% Pt/Vulcan XC-72 was found to be 17.5% wt. Its
performance was significantly greater than the electrode made using untreated catalyst with
30% wt. Nafion content. For example, the peak power of the former was 60% greater than
that of the latter (Fig. 11). In addition, the fuel cell showed no performance loss in 7 days of
testing. Thermal gravimetric analysis seemed to indicate that the ethylsulfonic acid group
started to become thermally unstable at temperatures higher than 100 oC. Based on the mass
loss at about 200 oC, it was estimated that about 0.23 mmol of ethylsulfonic acid groups were
attached onto 1 gram of 20% Pt/Vulcan XC 72 for the best performing electrodes [31].
1.1 0.8
1.0
0.7
0.9
0.6
0.8
Power Density (W cm )
-2
Cell Voltage (V)
0.7 0.5
0.6
0.4
0.5
, 0.3
0.4
0.3
0.2
V-I, Untreated, 30.0% Nafion
0.2
V-I, Sulfonated, 17.5% Nafion
0.1
0.1 W-I, Untreated, 30.0% Nafion
W-I, Sulfonated, 17.5% Nafion
0.0 0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Figure 11. Performance of (, ) untreated; and (, ) ethylsulfonic acid treated Pt/Vulcan XC-72 [31].
Reproduced by permission of The Electrochemical Society, Inc.
The diazonium salt method is very versatile and can chemically link any groups onto the
surface of carbon by starting with a chemical with the following structure: H2N-R-X, where R
is an alkyl or aryl organic group and X is an end functional group such as SO3H or PO3H2.
The chain length and structure of the alkyl or aryl spacer can be changed according to needs.
For example, an aminoalkyl- or aminoaryl-sulfonic acid may have a structure of
H2N(CH2)nSO3H with n ranging from 1 and up, or a structure of H2N(CH2)nmSO3H with m
and n ranging from 0 (m and n do not equal 0 at the same time) and up ( represents aryl
groups such as -C6H4-). A longer alkyl or aryl chain gives the end group a higher mobility,
which in turn may be able to facilitate the proton transport among the SO3H end groups from
one location to another. This may improve the proton conductance within the catalyst layer.
The difference in the structure of the alkyl or aryl spacer can also be used to tailor the
properties of the end functional groups. However, if the alkyl or aryl spacer is too long and
too bulky, it may reduce the electronic conductance within the catalyst layer or reduce the
396 Zhigang Qi
free surface of the Pt particles due to the coverage of the surface of either carbon or Pt by the
long and bulky functional groups.
In order to enhance the resistance of an electrode to water flooding, Xu et al. chemically
linked hydrophobic 2,3,4,5,6-pentafluorophenyl (-C6F5) onto the surface of 20% Pt/Vulcan
XC-72 by reacting 2,3,4,5,6-pentfluoroaniline (C6F5NH2) with isobutyl nitrite in the presence
of tetrafluoroboric acid (HBF4) [32]. Eqs. 7 and 8 show the reaction steps. Thermal
gravimetric analysis indicated that 0.11 mmol of -C6F5 groups were linked onto 1 gram of the
catalyst. The groups were thermally stable at least up to 110 oC. The optimal Nafion content
was about 37% wt. for electrodes made using this modified catalyst, compared to 30% for
electrodes made using untreated catalyst. Fuel cell testing indicated a 14% increase in the
peak power for the modified catalyst over the untreated catalyst (Fig. 12). More interestingly,
although the electrode made using the modified catalyst contained more Nafion, it showed
less mass transport resistance loss than the electrode made using untreated catalyst. This
enabled Xu et al. to propose that both macro-flooding and micro-flooding occur within a fuel
cell electrode. Macro-flooding occurs within the macro-pores among carbon particles, while
micro-flooding occurs on or near the carbon particle surface. Although a higher Nafion
content can lead to higher macro-flooding for the electrode made using the hydrophobized
catalyst, the micro-flooding is significantly reduced due to the presence of the hydrophobic -
C6F5 groups. Therefore, the net result was a reduction of overall flooding.
1.1 0.8
Treated
1.0 Untreated 0.7
0.9
0.6
Power Density (W cm -2)
0.8
Cell Voltage (V)
0.5
0.7
0.4
0.6
0.3
0.5
0.2
0.4
0.3 0.1
0.2 0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Figure 12. Performance of () untreated; and () C6F5 treated 20% Pt/Vulcan XC-72 [32].
Reproduced by permission of The Electrochemical Society, Inc.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 397
Experiments indicated that using 5-10% wt. (NH4)2SO4 during the thermal decomposition
step produced a catalyst (20% Pt/Vulcan XC-72) with the best performance in the presence of
17.5% of Nafion within the electrode. The peak power was increased by 20% over the
electrode made using untreated catalyst that contained 30% Nafion (Fig. 13). However,
when 20% wt. (NH4)2SO4 was used during the thermal decomposition step, the electrode
made using the resulting catalyst performed worse than that made using the untreated catalyst,
presumably because too many SO3H groups were linked to the carbon surface, resulting in
electrode flooding.
1.2 0.7
0.6
1.0
0.4
0.6
0.3
0.4
0.2
Figure 13. Polarization curves of 20% Pt/Vulcan XC-72 treated with () 0; () 5; () 10; and () 20%
(NH4)2SO4 [33]. Reproduced by permission of The Electrochemical Society, Inc.
Another advantage of this method is that the sulfonation step can be incorporated into the
electrode fabrication process. The inorganic salt can be mixed with Nafion and Pt/C in the
398 Zhigang Qi
ink or paste formulation step. After the resulting ink or paste is applied onto a substrate such
as carbon paper or carbon cloth, the electrode undergoes a drying step. The thermal
decomposition of the inorganic salt can simultaneously occur during this electrode drying step
and get the carbon sulfonated. However, since Nafion starts to decompose at temperatures
above 180 oC, it is preferred that the decomposition temperature of the inorganic salt is lower
than this temperature.
Fig. 14 shows the impact of drying temperature on the electrode performance. Two
electrodes were made containing no (NH4)2SO4 and were dried at 135 and 235 oC,
respectively, and one electrode was made containing (NH4)2SO4 and was dried at 235 oC. It
was clear that drying at 235 oC adversely affected the performance of the electrode, which
was most likely due to the decomposition of Nafion. However, for the electrode containing
(NH4)2SO4 the decrease was much less. It was inferred that the formation of the SO3H
groups on the surface of carbon from the decomposition of (NH4)2SO4 compensated for the
damage of Nafion in the drying step.
1.2
Heated at 135 C, no ammonium sulfate
Heated at 235 C, no ammonium sulfate
1.0 Heated at 235 C, 20% ammonium sulfate
0.8
Cell Voltage (V)
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
-2
Current Density (A cm )
Figure 14. Effect of drying temperatures on fuel cell performance in the (, ) absence; and ()
presence of 20% (NH4)2SO4.
An additional advantage of this method is that the sulfonation step can be performed to
an electrode that contains either bare carbon particles or catalyzed carbon particles. A solution
of an inorganic salt can be applied to the electrode by a variety of methods such as spraying
and brushing. A thermal treatment of the resulting electrode will decompose the inorganic salt
to have the carbon particle sulfonated. This approach is expected to be extremely valuable for
PTFE-bonded electrodes that contain no Nafion because such electrodes are only active at
the electrode-membrane interface due to the lack of proton conductance within the electrodes.
The linkage of SO3H groups onto the surface of carbon from the decomposition of
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 399
(NH4)2SO4 will allow protons to transport throughout those electrodes, thereby significantly
improving the fuel cell performance and catalyst utilization.
All the functional groups discussed in this section are summarized in Fig. 15.
O
Carbon O Si CH2 CH2 S OH
O
Sulfonated silane
O
Carbon CH2 CH2 P OH
OH
Ethylphosphonic acid
O
Carbon CH2 CH2 S OH
O
Ethylsulfonic acid
F F
O
Carbon F S OH
Carbon
O
F F
Sulfonic acid
2,3,4,5,6-pentafluorophenyl
4.2.4. Plasma
Plasma is a useful technique to modify the surface of carbon and other materials. By varying
the power level, duration, and the excited gas, the surface can be modified with different
groups to variable extent. Shioyama et al. used radio frequency plasma to link hydrophobic
CF3 groups onto the surface of Vulcan XC-72 by exciting hexafluoroethane (C2F6) at 13.56
MHz [34]. The measured contact angles of the pristine and the plasma-treated carbon were 79
and 156 o, respectively, indicating a significant increase in the hydrophobicity of the carbon
after the plasma treatment. An electrode made using the treated catalyst containing no PTFE
performed similarly to that made with the pristine catalyst containing PTFE. They also
observed an enhanced oxygen reduction activity of the Pt supported on the plasma-treated
carbon, and attributed this to the increase in the 5d electron vacancy of the Pt atoms induced
by the high electro-negativity of the F atoms. They believed that the increased d-vacancy of
the Pt led to a stronger Pt-oxygen interaction that thereby weakened the O-O bond of oxygen.
400 Zhigang Qi
Carbon surface, bare or catalyzed, can also be modified by grafting a polymer through either
physical, chemical, or physicochemical interactions. A large number of polymers with
different structures and properties could be chosen for this purpose. A challenge is how to
graft the polymer onto the carbon surface effectively and stably. The grafted polymer could
reduce the mobility of Pt particles and thus lower their agglomeration rate by confining and
separating them from each other if the polymer chains (or boundless of chains) could be
arranged to selectively weave through among the Pt particles, as depicted in Fig. 16. The
amount of polymer should be carefully controlled to not cause electronic discontinuity among
the Pt/C particles.
Polymer
Pt
Carbon
Figure 17. Polarization curves of (, ) untreated; and (, ) PATBS treated Pt/Ketjen black [35]. J.
Power Sources 138, Mizuhata, H.; Nakao, S. & Yamaguchi, T., Morphological control of PEMFC
electrode by graft polymerization of polymer electrolyte onto platinum-supported carbon black, 25-30,
Copyright (2004), with permission from Elsevier.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 401
So far, there are few reports in this area. Mizuhata et al. grafted acrylamide tertiary butyl
sulfonic acid (PATBS) onto the surface of catalyzed carbon via reaction with hydroxyl groups
that had been created onto the surface of carbon in a prior step [35]. The resulting carbon
showed increase in the sulfur content and decrease in the primary pore volume, signaling the
success of the grafting process. Fuel cell tests showed that the Pt/C grafted with 1% wt.
PATBS performed slightly better than the untreated Pt/C, while the Pt/C grafted with 4% and
10% wt. PATBS performed much lower than the untreated Pt/C (Fig. 17). They suggested
that the lower performance with higher PATBS content was likely to be due to mass transport
problems such as the retardation of the reactant gas access to the active site or flooding of the
electrode.
He et al. illustrated grafting a sulfonated conducting polymer (i.e., polyaniline) onto the
surface of carbon with the polymer having both proton and electron conductance (Fig. 18)
[36]. However, neither a grafting procedure nor any experimental data was reported.
Figure 18. Illustration of carbon black modified by sulfonated polyaniline [36]. Reproduced by
permission of The Electrochemical Society, Inc.
5. Durability
For stationary applications a PEM fuel cell is required to operate for 40,000 hours or longer.
The major MEA failure modes that lead to a much shorter lifetime are the membrane breach
and the electrode decay. The former will result in reactant crossover that causes a sudden and
catastrophic failure. Ionomer chemical structure and end-group modification, reinforcement
of the membrane, incorporation of additives into the membrane, and use of a protective sub-
gasket or full-gasket are effective methods to slow down the membrane breach process. The
electrode decay causes gradual loss of the fuel cell performance and rarely causes
catastrophic failures. The performance decay lowers the fuel cell efficiency and when the
efficiency becomes lower than a predetermined value, the fuel cell reaches its end of life. In
402 Zhigang Qi
addition to efficiency, certain components and modules within a fuel cell system can only
function properly when the fuel cell voltage is above a certain value. The electrode decay is
caused by the loss of the effective catalyst surface area, which is the result of the decrease in
the three-phase boundaries, the catalyst particle size growth, and the catalyst dissolution and
re-deposition. Carbon corrosion plays a role in this decay process.
As indicated earlier, one useful function of carbon support is to anchor the Pt particles and the
interaction between carbon and Pt retards the agglomeration of the latter. During the carbon
corrosion process, which is likely to be catalyzed by the presence of Pt [37], it is expected
that the carbon surface near the Pt particles will be corroded faster and more severely. The
corrosion weakens or ceases the interaction between carbon and Pt, and thus the Pt particles
become more freely to move. Therefore, carbon corrosion can accelerate the Pt particle
growth.
Cathode e-
Membrane H+ H+
Anode e-
O 2+4H++4e- 2H2O H2 2H++2e-
~1.0 V ~0.0 V
as high as two times of the OCV, which is nearly 2.0 V (Fig. 19). At such a high voltage, it
has been found that carbon can be corroded extremely fast, and the fuel cell decays quickly
[38,39].
Cai et al. suggested that carbon oxidation proceeded in two pathways: an oxygen pathway
that leads to mass loss due to the formation of gaseous products, and a water pathway that
leads to mass gains [40]. They concluded that the surface loss of carbon from corrosion
accounted for the electrochemical active surface loss of the electrode because the size of the
Pt particles supported on the carbon did not change much under their study.
Pyun and Lee studied carbon supports through electrochemical oxidation in 85% H3PO4
solution at 145 oC and found that COOH was the major surface group formed on the carbon
surface at potentials higher than ca. 0.7 V [41]. The activity of the catalyst towards the
oxygen reduction reaction was found to become lower with the formation of the carboxylic
acid groups.
Severe carbon corrosion produces carbon dioxide and results in the loss of the carbon
material as shown by Eq. 11. For a fuel cell cathode containing 0.6 mg cm-2 of carbon, a
simple calculation according to the Faradays law shows how many hours the carbon can last
before it is completely corroded. The results are shown in Table 1. It is striking to see that the
carbon corrosion current density needs to be less than 0.15 A cm-2 in order for the carbon to
last for 40,000 hours. If we assume that the electrode will not function properly when 20% of
carbon is corroded, then a corrosion current density should be lower than 0.03 A cm-2.
Corrosion current density (A/cm2) Hours to last (hr)* Days to last (day)
0.050 107,222 4,468
0.075 71,481 2,978
0.10 53,611 2,234
0.15 35,741 1,489
0.25 21,444 894
0.50 10,722 447
5 1,072 45
25 214 9
*Assumptions: carbon loading = 0.6 mg/cm2; carbon is oxidized to CO2
It has been well established that a graphitized carbon corrodes much slower than an
amorphous carbon such as Vulcan XC-72, and the Pt particles supported on graphitic carbon
are more stable. An amorphous carbon can be converted to a graphitized carbon by a thermal
treatment of the former at ca. 2000 oC in an inert gas such as nitrogen. Since a graphitized
carbon is composed of mainly graphene sheets, its corrosion rate is low.
404 Zhigang Qi
Figure 20. Surface oxidization with time for () 20% Pt/carbon black; and () 20% Pt/CNT [42].
Reproduced by permission of The Electrochemical Society, Inc.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 405
Orignal CB Oxidized CB
Figure 21. Illustration of (A) multiwalled CNT and (B) carbon black before and after oxidation [42].
Reproduced by permission of The Electrochemical Society, Inc.
reduce the formation of hydrogen peroxide but they can accelerate its decomposition before it
has a chance to form radicals and to attack the ionomer, the purpose of protecting the ionomer
is also achieved.
Early electrodes were made of Pt black that did not contain a carbon support. Those
electrodes had good stability but unacceptable high Pt loadings. The introduction of carbon
support significantly reduced the Pt loading, but it also raises durability issues due to the
carbon corrosion. How to make high performing electrodes that do not contain carbon but
also possess low Pt loading is a challenge. Debe et al. reported a nanostructured thin film
(NSTF) electrode structure (~ 0.3 m) with oriented crystalline organic whiskers of high
aspect ratios as the support [45,46]. Those NSTF electrodes showed 75 times lower fluoride
release rate than Pt/C electrodes at 120 oC. They were highly resistant to corrosion at 1.5 V,
and only suffered a maximum of 33% surface area loss after 9000 accelerated cycles [45].
Gull et al. have demonstrated superior performance and stability of carbonless thin layer
electrodes made by a dual ion beam assisted deposition (IBAD) technique that combines
physical vapor deposition (PVD) with ion beam bombardment [47]. They found that
bilayered coatings on GDL with either a Co or a Cr inner layer (~ 50 nm thin) and a Pt outer
layer (~ 50 nm thin, and 0.08 mg Pt cm-2) showed a more than 50% higher Pt mass activity at
900 mV than a Pt single layer.
Another approach to carbonless electrode is to use a non-carbon catalyst-support that
meets the requirements of a support. Since metallic supports will corrode faster than carbon,
they can hardly be used, unless they can be completely coated by a thin layer of Pt skin to
avoid corrosion. Many ceramic materials can meet the stability requirement, but they lack of
electronic conductivity. Again, this will not be a problem if these particles can be coated by
an electronically conducting Pt thin layer. Chhina used semi-conducting indium tin oxide
(ITO) particles to support Pt and achieved an average Pt crystallite size of 13 nm. Pt
supported on ITO showed extremely high thermal stability and only lost 1% wt. of materials
versus 57% wt. for Pt supported on Hispec 4000 carbon upon heating to 1000 oC.
6. Conclusions
The incorporation of a carbon catalyst-support into the PEM fuel cell electrodes brings in
both advantages and disadvantages. The advantages include the increased catalyst dispersion
and the reduced catalyst loading. The disadvantages comprise of the promotion of peroxide
formation that accelerates the ionomer decay rate, and the problems associated with the
carbon corrosion.
Functionalizing the surface of carbon by suitable groups can add new advantages such as
reducing the catalyst particle size and enhancing the thermal stability. It has been
demonstrated that the fuel cell performance can also be significantly increased by attaching
either acidic or hydrophobic groups onto the carbon support. The presence of acidic groups is
believed to render the surface of carbon with good proton conductance that increases the
effective catalyst surface area. The optimal ionomer content within an electrode is also
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 407
significantly reduced, which can translate into a lower cost electrode since perfluorinated
ionomer such as Nafion is very expensive. The effectiveness of the hydrophobic surface
groups is believed to be the result of reduced micro-flooding in the electrode.
It is suggested that the surface functionalization of carbon can also be used for improving
other aspects of PEM fuel cells. These aspects comprise of enhanced long term durability and
stability, higher CO-tolerance, and reduced generation of peroxide.
Acknowledgements
Support from Dr. Roger Saillant (CEO) and his senior management team at Plug Power is
grateful acknowledged.
References
[1] Vielstich, W.; Lamm, A. & Gasteiger H. A. (Eds.) (2003). Handbook of fuel cells:
fundamentals technology and applications (Vols. 1-4). Chichester, England: John Wiley
& Sons.
[2] Larminie, J. & Dicks, A. (2003). Fuel cell systems explained (2nd Ed.). Chichester: John
Wiley & Sons.
[3] Liu, H.; Song, C.; Zhang, L.; Zhang, J.; Wang, H.; & Wilkinson, D. P. (2006). A review
of anode catalysts in the direct methanol fuel cell. J. Power Sources, 155, 95-110.
[4] Kinoshita, K. (1988). Carbon: Electrochemical and physicochemical properties. New
York: John Wiley & Sons.
[5] Gooding, J. J. (2005). Nanostructured electrodes with carbon nanotubes: A review on
electrochemistry and applications for sensing. Electrochimica Acta, 50, 3049-3060.
[6] Shim, J.; Joung, K. Y.; Ahn, J. H.; & Lee, W. M. (2007). Carbon-supported platinum
nanoparticles synthesized by plasma-chemical reduction method for fuel cell
applications. J. Electrochem. Soc., 154, B165-B169.
[7] Prabhuram, J.; Zhao, T. S.; Wong, C. W. & Guo, J. W. (2004). Synthesis and
physical/electrochemical characterization of Pt/C nanocatalyst for polymer electrolyte
fuel cells. J. Power Sources, 134, 1-6.
[8] Uchida, M.; Fukuoda, Y.; Sugawara, Y.; Eda, N. & Ohta, A. (1996). Effects of
microstructure of carbon support in the catalyst layer on the performance of polymer-
electrolyte fuel cells. J. Electrochem. Soc., 143, 2245-2252.
[9] Chen, J.-M.; Sarma, L. S.; Chen, C.-H.; Cheng, M.-Y.; Shih, S.-C.; Wang, G.-R.; Liu,
D.-G.; Lee, J.-F.; Tang, M.-T. & Hwang, B.-J. (2006). Multi-scale dispersion in fuel cell
anode catalysts: Role of TiO2 towards achieving nanostructured materials. J. Power
Sources, 159, 29-33.
[10] Dilsiz, N. (2000). Plasma surface modification of carbon fibers: a review. J. Adhesion
Sci. Technol., 14, 975-987.
[11] Roy, S. C.; Harding, A. W.; Russell, A. E.; & Thomas, K. M. (1997).
Spectroelectrochemical study of the role played by carbon functionality in fuel cell
electrodes. J. Electrochem. Soc., 144, 2323-2328.
408 Zhigang Qi
[12] Kim, S. & Park, S.-J. (2006). Effects of chemical treatment of carbon supports on
electrochemical behaviors for platinum catalysts of fuel cells. J. Power Sources, 159,
42-45.
[13] Swider, K. E. & Rolison, D. R. (2000). Reduced poisoning of platinum fuel-cell
electrocatalysts supported on desulfurized carbon. Electrochem. Solid-State Lett., 3, 4-6.
[14] Li, X.; Ge, S.; Hui, C. L.; & Hsing, I.-M. (2004). Well-dispersed multiwalled carbon
nanotubes supported platinum nanocatalysts for oxygen reduction. Electrochem. Solid-
State Lett., 7, A286-A289.
[15] Carmo, M.; Paganin, V. A.; Rosolen, J. M.; & Gonzalez, E. R. (2005). Alternative
supports for the preparation of catalysts for low-temperature fuel cells: the use of carbon
nanotubes. J. Power Sources, 142, 169-176.
[16] Ticianelli, E. A.; Derouin, C. R.; Redondo, A. & Srinivasan, S. (1988). Method to
advance technology of proton exchange membrane fuel cell. J. Electrochem. Soc., 135,
2209-2214.
[17] Poltarzewski, Z.; Staiti, P.; Alderucci, V.; Wieczorek, W. & Giordano, N. (1992).
Nafion distribution in gas diffusion electrodes for solid-polymer-electrolyte-fuel-cell
application. J. Electrochem. Soc., 139, 761-765.
[18] Wilson, M. S. & Gottesfeld, S. (1992). High performance catalyzed membranes of ultra-
low Pt loadings for polymer electrolyte fuel cells. J. Electrochem. Soc. 139, L28-L30.
[19] Park, C. H.; Scibioh, M. A.; Kim, H.J.; Oh, I.-H.; Hong, S.-A. & Ha, H. Y. (2006).
Modification of carbon support to enhance performance of direct methanol fuel cell. J.
Power Sources, 162, 1023-1028.
[20] Janssen, G. J. M. & Sitters, E. F. (2006). Performance of thin-film cathodes for proton-
exchange-membrane fuel cells based on high-surface-area carbon supports. J. Power
Sources, ??, ??-??. (in press)
[21] Rajalakshmi, N.; Ryu, H.; Shaijumon, M. M. & Ramaprabhu, S. (2005). Performance of
polymer electrolyte membrane fuel cells with carbon nanotubes as oxygen reduction
catalyst support material. J. Power Sources, 140, 250-257.
[22] Qi, Z.; Lefebvre, M. C. & Pickup, P. G. (1998). Electron and proton transport in gas
diffusion electrodes containing electronically conductive proton-exchange polymers. J.
Electroanal. Chem., 459, 9-14.
[23] Qi, Z. & Pickup, P. G. (1997). Size control of polypyrrole particles. Chem. Mater., 9,
2934-2939.
[24] Qi, Z. & Pickup, P. G. (1998). Novel supported catalysts: platinum and platinum oxide
nanoparticles dispersed on polypyrrole/polystyrenesulfonate particles. Chem. Commun.,
15-16.
[25] Qi, Z. & Pickup, P. G. (1998). High performance conducting polymer supported oxygen
reduction catalysts. Chem. Commun., 2299-2300.
[26] Lefebvre, M.; Qi, Z. & Pickup, P. G. (1999). Electronically conductive proton-exchange
polymers as catalyst supports for proton-exchange membrane fuel cells: electrocatalysis
of oxygen reduction, hydrogen oxidation, and methanol oxidation. J. Electrochem. Soc.,
146, 2054-2058.
[27] Pittman, Jr., C. U.; He, G.-R.; Wu, B. & Gardner, S. G. (1977). Chemical modification
of carbon fiber surfaces by nitric acid oxidation followed by reaction with
tetraethylenepentamine. Carbon, 35, 317-331.
Surface Functionalization of Carbon Catalyst-Support for PEM Fuel Cells 409
[28] Jia, N.; Martin, R. B.; Qi, Z.; Lefebvre, M. C. & Pickup, P. G. (2001). Modification of
carbon supported catalysts to improve performance in gas diffusion electrodes.
Electrochimica Acta, 46, 2863-2869.
[29] Easton, E. B.; Qi, Z.; Kaufman, A. & Pickup, P. G. (2001). Chemical modification of
proton exchange membrane fuel cell catalysts with a sulfonated silane. Electrochem.
Solid-State Lett., 4, A59-A61.
[30] Xu, Z., Qi, Z. & Kaufman, A. (2003). High performance carbon-supported catalysts for
fuel cells via phosphonation. J. Chem. Soc., Chem. Commun., 878-879.
[31] Xu, Z.; Qi, Z. & Kaufman, A. (2003). Advanced fuel cell catalysts: sulfonation of
carbon-supported catalysts using 2-aminoethanesulfonic acid. Electrochem. Solid-State
Lett., 6, A171-A173.
[32] Xu, Z.; Qi, Z.; & Kaufman, A. (2005). Hydrophobization of carbon-supported catalysts
with 2,3,4,5,6-pentafluorophenyl moieties for fuel cells. Electrochem. Solid-State Lett.,
8, A492-A494.
[33] Xu, Z.; Qi, Z.; & Kaufman, A. (2005). Superior catalysts for proton exchange membrane
fuel cells sulfonation of carbon-supported catalysts using sulfate salts. Electrochem.
Solid-State Lett., 8, A313-A315.
[34] Shioyama, H.; Honjo, K.; Kiuchi, M.; Yamada, Y.; Ueda, A.; Kuriyama, N.; &
Kobayashi, T. (2006). C2F6 plasma treatment of a carbon support for a PEM fuel cell
electrocatalyst. J. Power Sources, 161, 836-838.
[35] Mizuhata, H.; Nakao, S. & Yamaguchi, T. (2004). Morphological control of PEMFC
electrode by graft polymerization of polymer electrolyte onto platinum-supported carbon
black. J. Power Sources, 138, 25-30.
[36] He, C.; Desai, S.; Brown, G.; Bollepalli, S. (2005). PEM fuel cell catalysts: cost,
performance, and durability. Interface, 14, 41-44.
[37] Stevens, D. A.; Hicks, M. T.; Haugen, G. M. & Dahn, J. R. (2005). Ex situ and in situ
stability studies of PEMFC catalysts: Effect of carbon type and humidification on
degradation of the carbon. J. Electrochem. Soc., 152, A2309-A2315.
[38] Reiser, C. R.; Bregoli, L.; Patterson, T. W.; Yi, J. S.; Yang, J. D.; Perry, M. L.; & Jarvi,
T. D. (2005). A reverse-current decay mechanism for fuel cells. Electrochem. Solid-State
Lett., 8, A273-A276.
[39] Tang, H.; Qi, Z.; Ramani, M. & Elter, J. (2006). PEM fuel cell cathode carbon corrosion
due to the formation of air/fuel boundary at the anode. J. Power Sources, 158,
1306-1312.
[40] Cai, M.; Ruthkosky, M. S.; Merzougui, B.; Swathirajan, S.; Balogh, M. P. & Oh, S. H.
(2006). Investigation of thermal and electrochemical degradation of fuel cell catalysts. J.
Power Sources, 160, 977-986.
[41] Pyun, S.-I. & Lee, S.-B. (1999). Effect of surface groups on the electrocatalytic behavior
of Pt-Fe-Co alloy-dispersed carbon electrodes in the phosphoric acid fuel cell. J. Power
Sources, 77, 170-177.
[42] Li, L. & Xing, Y. (2006). Electrochemical durability of carbon nanotubes in
noncatalyzed and catalyzed oxidations. J. Electrochem. Soc., 153, A1823-A1828.
[43] Shao, Y.; Yin, G.; Gao, Y. & Shi, P. (2006). Durability study of Pt/C and Pt/CNTs
catalysts under simulated PEM fuel cell conditions. J. Electrochem. Soc., 153, A1093-
A1097.
410 Zhigang Qi
[44] Wang, X.; Li, W.; Chen, Z.; Waje, M. & Yan, Y. (2006). Durability investigation of
carbon nanotube as catalyst support for proton exchange membrane fuel cell. J. Power
Sources, 158, 154-159.
[45] Debe, M.; Schmoechel, A. K.; Vernstrom, G. D. & Atanasoski, R., J. (2006). High
voltage stability of nanostructured thin film, J. Power Sources, 161, 1002-1011.
[46] Debe, M. K. (2003). Novel catalysts, catalysts support and catalysts coated membrane
methods. In W. Vielstich, A. Lamm, & H. A. Gasteiger (Eds.), Handbook of fuel cells:
fundamentals technology and applications (Vol. 3, pp. 576-589). Chichester, England:
John Wiley & Sons.
[47] Gull, A. F., Saha, M. S., Allen R. J. & Muukerjee, S. (2006). Toward improving the
performance of PEM fuel cell by using mix metal electrodes prepared by Dual IBAD, J.
Electrochem. Soc., 153, A366-A371.
[48] Chhina, H.; Campbell, S. & Kesler, O. (2006). An oxidationresistant indium tin oxide
catalyst support for proton exchange membrane fuel cells. J. Power Sources, 161,
893-900.
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 411-443 2007 Nova Science Publishers, Inc.
Chapter 9
Abstract
The ideal support material for fuel-cells catalysts should have the following
characteristics: high electrical conductivity, adequate water-handling capability at the
electrode, and also good corrosion resistance under oxidizing conditions. Whereas carbon
blacks are the common support materials for electrocatalysts, new forms of carbon materials
such as graphite nanofibers (GNFs), carbon nanotubes (CNTs), ordered mesoporous carbons
(OMC) had been investigated as catalyst supports.
In the present article, the size and the loading efficiency of metal particles were
investigated by changing the preparation method of carbon-supported platinum catalysts.
First, the effect of acid/base treatment on carbon blacks supports on the preparation and
electroactivity of platinum catalysts. Secondly, binary carbon-supported platinum (Pt)
nanoparticles were prepared using two types of carbon materials such as carbon blacks (CBs)
and graphite nanofibers (GNFs) to check the influence of carbon supports on the
electroactivity of catalyst electrodes. Lastly, plasma treatment or oxyfluorination treatment
effects of carbon supports on the nano-structure as well as the electroactivity of the carbon-
supported platinum catalysts for DMFCs were studied.
Keywords: carbon supports, catalyst electrodes, fuel cells, nanoclusters, surface treatments,
electroactivity.
*
E-mail address: sunnykim@krict.re.kr, Tel: +82-42-860-7232, Fax: +82-42-861-4151 (First Author)
**
E-mail address: sjpark@inha.ac.kr, Tel: +82-42-860-7234, Fax: +82-42-861-4151 (Corresponding Author)
412 Seok Kim and Soo-Jin Park
1. Introduction
Direct methanol fuel cells (DMFCs) are attracting much more attention for their potential as
clean and mobile power sources for the near future [1-8]. Generally, platinum (Pt)- or
platinum-alloy-based nanocluster-impregnated carbon supports are the best electrocatalysts
for anodic and cathodic fuel cell reactions. These materials are very expensive, and thus there
is a need to minimize catalyst loading without sacrificing electro-catalytic activity. Because
the catalytic reaction is performed by fuel gas or fuel solution, one way to maximize catalyst
utilization is to enhance the external Pt surface area per unit mass of Pt. The most efficient
way to achieve this goal is to reduce the size of the Pt clusters.
The main routes to the synthesis of nanocluster electrocatalysts are the impregnation
method and the colloid method. The impregnation method is characterized by the step of
depositing Pt or another metal precursor as followed by a reduction step. This second step can
be chemical reduction of the electrocatalyst slurry in solution using reducing agents (liquid-
phase reduction) [9-11] or gas-phase reduction of the impregnated carbon using a relatively
high-temperature (250-600oC) flowing H2 gas stream [12,13]. Alternatively, the colloid
method [14,15] has the advantage of producing small and homogeneously distributed carbon-
supported metal nanoclusters, but the applicable methodologies are very complex. Thus, the
search for alternative routes to producing carbon-supported metal nanoclusters is a critical
theme in these research areas. Recently, alternative routes based on the polyol method [16]
and the alcohol-reduction process [17] have been developed with good results.
On the other hand, carbon is a critical material in low-temperature fuel cells, such as
polymer-electrolyte membrane fuel cells. No other material has the essential combination of
electronic conductivity, corrosion resistance, surface properties, and the low cost required for
the commercialization of fuel cells. In recent years, direct methanol fuel cells (DMFCs) have
been intensely studied because of their numerous advantages, which include high energy
density, the ease of handling a liquid, low operating temperatures, and their possible
applications to micro fuel cells. The performance and long-term stability of low-temperature
fuel cells, such as DMFCs, which event operate at room temperature, are known to be
strongly dependent on carbon supports as well as catalytically active species. Accordingly, for
the best DMFC performance, it is essential to develop good carbon materials for the catalyst
supports [18-20].
At present, electrocatalysts generally are supported on high-surface-area carbon blacks
(CBs) with a high-mesoporous distribution and graphite characteristics, and Vulcan XC-72
carbon blacks (Cabot International) are the most commonly used carbon support because of
their good compromise between electronic conductivity and Brunauer-Emmett-Tellers
(BETs) surface area. However, the effect of the surface characteristics of the various carbon
materials had not been fully studied to our best knowledge.
The main routes for the synthesis of Pt-Ru/carbon black electrocatalysts can be grouped into
two types: impregnation and colloidal procedures. The first is characterized by a deposition
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 413
black can easily accommodate a high amount of metal phase, with a high degree of
distribution but, at the same time, the significant amount of micropores on the carbon support
will not allow a homogeneous distribution of the electrocatalytic phase through the support,
which could lead to mass transport limitations of the reactant as well as limited access to the
inner eletrocatalytic sites. Vulcan XC appears to be the best compromise with the presence of
a small amount of micropores and a reasonably high loading of the metal phase. Up to
accommodate a high loading of the metal phase. Up to now, this carbon black is the most
used carbon support for the preparation of DMFC catalysts.
Many researchers have reported that new kinds of carbon materials and the structural and
surface modification of carbon supports had been studied to improve the electroactivities of
catalyst electrodes.
Li et al. [27] reported that MWNT-supported Pt catalysts show enhanced ORR activities
and superior cell performance in comparison to XC72-supported Pt catalysts. The high
electrocatalytic activity may be attributed to the unique structural and higher electrical
properties and the small amount of organic impurities of MWNTs. Gonzalez [28] had
compared the effect of three kinds of carbon materials. Noble metal catalysts in the form of
nanoparticles supported on high surface area carbon exhibit characteristics that depend
strongly on the nature of the support. Results for the DMFC showed power densities
exceeding 100 mW cm-2 at 90 C and 0.3 Mpa and the activity of the anodes followed the
sequence: PtRu/MWNT > PtRu/Vulcan > PtRu/SWNT.
A (Pt-Ru)/herringhone GCNF(graphite carbon nanofibers) nanocomposites had been
prepared using a single-source molecular precursor as a source of metal [29]. When this
nanocomposites is used as an anode catalyst in working DMFCs, fuel cell performance is
enhanced by ~50% relative to that recorded for an unsupported Pt-Ru colloid anode catalyst.
C.A. Bessel [30] reported that 5 wt.% Pt on platelet and ribbon type GNF showed improved
oxidation activities of 400% when compared to Vulcan (XC-72) in methanol oxidation
studies. They also showed the reduced CO poisoning effect. This improvement in
performance is believed to be the fact that the metal particles adopt specific crystallographic
orientations when dispersed on the highly tailored graphite nanofibers structures.
J. Yi [31] had studied that a carbon support with a high-surface area and high
mesoporosity served the best in terms of a high dispersion of Pt nanoparticles. In a unit cell
test of the PEMFC, a Pt catalyst with a high mesoporosity and fine dispersion of metal
showed an enhanced performance. The findings indicate that the surface area combined with
the mesoporosity had a positive influence on the metal dispersion and the distribution of
ionomer, leading to the enhanced cell performance.
MWNT and high surface area mesoporous carbon xerogel were prepared and used as
supports for mono-metallic Pt and bi-metallic PtRu catalysts by J.L. Figueiredo [32]. A
remarkable increase in the activity was observed when PtRu catalysts were supported on the
oxidized xerogel. From the XPS results, it had been shown that the oxidized support helps to
maintain the metals in the metallic state, as required for the electro-oxidation of methanol.
Recently, W.C. Choi [33] had demonstrated that pitch-based porous carbons having
sulfonic acid groups on their surfaces were synthesized by carbonization of mesophase pitch
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 415
Ideal support material should have the following characteristics: high electrical conductivity,
adequate water-handling capability at the electrode, and also good corrosion resistance under
oxidizing conditions. Zhou et al. [39] have prepared uniform Pt nanoclusters of an average
diameter of 2.9 nm on carbon by the polyol method. Those catalysts have exhibited higher
electro-catalytic activities in the oxygen reduction reaction than those of commercial E-TEK
catalysts. Lee et al. [40-42] have reported that Pt or Pt alloys produced by microwave
irradiation and assisted by stabilizing agents showed high electro-catalytic activity in the
methanol oxidation reaction. Besides, electrocatalysts can be prepared by electrochemical
deposition methods [43,44].
Recently, W. Li et al. [45] have reported a modified polyol synthesis method for the
preparation of multi-wall carbon nanotubes (MWNTs)-supported Pt catalysts, finding that the
particle size can be controlled by adjusting the ratio of water to ethylene glycol (EG) in the
preparation process. It has also been reported that the particle size and the loading level of Pt-
based catalysts are key factors determining the electrochemical activity and cell performance
of DMFCs [6,11,43]. Whereas carbon blacks are the common support materials for
electrocatalysts [46,47], new forms of carbon materials such as carbon or graphite nanofibers
(GNFs) [48,49] and carbon nanotubes (CNTs) [45] have been investigated as catalyst
supports.
In the present study, the size and the loading efficiency of metal particles were
investigated by changing the preparation method of carbon-supported platinum catalysts.
Furthermore, acid/base treatment effects of carbon blacks on the nano-structure as well as the
electroactivity of the carbon-supported platinum catalysts for DMFCs were studied.
416 Seok Kim and Soo-Jin Park
and C1s spectra were analyzed using a vendor-supplied curve-fitting program (VGX900) for
peak deconvolution.
Electrochemical experiments were carried out using a potentiostat/galvanostat from
AUTOLAB/PGSTAT30 (Eco Chemie, Netherlands). A conventional three-compartment
electrochemical cell was used with an Ag/AgCl (saturated KCl) electrode as the reference, a
platinum plate (1 cm2 area) as the counter electrode, and a glassy carbon (GC) electrode
covered with a layer of catalysts as the working electrode. To prepare the electrodes for the
above measurements, dispersed Pt electrocatalysts were applied to the GC disk electrode (3
mm diameter) in the form of ink. The ink was made by mixing 5mg of the catalyst and 0.2ml
of Nafion (5 wt.%) solution. The mixture was agitated in an ultrasonic bath for 1 hr to make a
suspension. Two microliters of this suspension was placed on the GC electrode and dried at
room temperature. This homogeneous film coating attached firmly onto the GC electrode
surface.
C. Preparation Methods
1000
Pt(200)
Pt(220)
Intensity / a.u.
(a)
500
(b)
(c)
(d)
0 20 40 60 80
2 Theta
Figure 1. Powder X-ray diffraction patterns of Pt/CBs, prepared in (a) 3.0M, (b) 1.0M, (c) 0.5M, and
(d) 0.1M NaOH aqueous solutions.
418 Seok Kim and Soo-Jin Park
In the case of the 0.1 M NaOH solution, it is clear, given the absence of any prominent
peak, that the Pt-reduction reaction did not fully occur. It is interesting to note that the
broadness of a peak is dependent on the NaOH concentration. In the case of the 1.0 M
solution, the Pt (220)-peak broadness of the samples is rather large, meaning that the average
size of the Pt nanoclusters is small. Conversely, in the case of the 0.5 M solution, the peak
broadness of the sample is rather small, meaning that the average size of the Pt nanoclusters is
large. The average Pt nanocluster size was calculated with a Scherrer equation [52-54] and is
demonstrated in Table 1. The detailed Pt (220) peaks were curve-fitted by mixed Gausiann-
Lorentzian methods according to Radmilovic et al [53].
K
LC = (1)
B cos
where Lc is the mean size of the Pt particles, K is the Scherrer constant (=0.89), =0.154 nm,
B is the half-height width of the (220) diffraction line, and is the radian-unit Bragg angle.
A 1.0 M NaOH concentration is the best condition for obtaining the small Pt nanocluster
size of 3.07 nm. In the case of the 0.5 M NaOH concentration, the large Pt nanocluster size of
11.4 nm was obtained. In the case of the 0.1 M NaOH concentration, the Pt size could not be
calculated because the Pt (220) peak was not clearly observed. Consequently, it was
concluded that the average Pt nanocluster size was dependent on the concentration of base
material in the Pt-precursor-and-carbon-dispersed solution.
A Pt 100% loading efficiency means that the deposited Pt weight percent is 10% against
the carbon materials. These values are summarized in Table 1. The 1.0 M NaOH
concentration shows the highest loading efficiency, 90%. This result supports the conclusion
that the 1.0 M concentration is the best condition for obtaining a high Pt loading efficiency. In
the case of the 0.1 M concentration, the Pt/CBs show a low Pt loading efficiency, 31%. This
can be explained by the fact that the Pt clusters were not effectively deposited on the carbon
surface, which is hardly detected by the XRD patterns in Figure 1. Furthermore, slight peaks
for oxygen, sodium and sulfur elements also appeared. The sodium and sulfur probably
originated from NaOH and the impurities of virgin carbon blacks (VCB), respectively. The Pt
loading efficiency was also measured independently using the ICP-AES method, the results
of which, in comparison with those of the EDS method, are shown in Table 1. The values for
the two methods are slightly different, owing to the small inhomogeneous distribution of the
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 419
Pt nanoclusters. However, the values of Pt loading efficiency for the samples show an
identical tendency.
We prepared virgin carbon blacks (VCB), neutral-treated carbon blacks (NCB), base-treated
carbon blacks (BCB), and acid-treated carbon blacks (ACB) to examine the effect of chemical
treatment. That was expected to change the surface characteristics or functional group of the
carbon surface. Table 2 shows the results for the surface and adsorption properties of the CBs
before and after the chemical treatments. It can clearly be seen that the chemical treatments
influence the pH, acid-base surface values, and specific surface areas (SBET) of the CBs. This
seems to be a consequence of the changes of functional group as well as microstructure
changes of the treated CBs. As shown in Table 2, the pH, base surface value, and specific
surface area of the ACB sample show prominent changes compared with VCB. This result
indicates that a strong acid-base reaction is occurred on the original base-like carbon surface,
and that the acidic solution catalyzed the micro-pore degradation or micro-pore blocking of
the VCB. By contrast, BCB shows an increased base value and a slightly enhanced specific
surface area.
Table 3. Structural parameter results for carbon blacks obtained by XRD analysis
Scherrer equation was used to calculate the Lc from the width of the (002) reflection [55].
These chemical-treated CB parameters are summarized in Table 3. The interlayer spacing for
VCB (0.3648 nm) was increased to 0.3665 nm for ACB. At the same time, the Lc increased
to 5.86 nm for ACB, as the half-height width of the (002) diffraction line, B, decreased. As
mentioned above in Table 1, it is thought that acid treatment leads to an increased degree of
420 Seok Kim and Soo-Jin Park
(a)
Transmittance
(b)
(c)
(d)
Figure 2. FT-IR spectra of (a) VCB, (b) NCB, (c) BCB, and (d) ACB.
To analyze the functional groups on the carbon surface, IR spectroscopy was conducted
on the treated CBs. The IR spectra are shown in Figure 2. VCB shows characteristic peaks at
3430 cm-1, 2910 cm-1, 1630 cm-1 and 1240 cm-1, which can be assigned to hydroxyl group (O-
H), aliphatic group (C-H), carboxyl/ester groups (C=O), and ether group (C-O-C), and
respectively. It has previously been reported that CBs produced at the high temperature of
800C showed oxidized groups such as pyrone or chromene and base functional groups [56].
ACB shows an enhanced peak intensity at 2910 cm-1, 1630 cm-1, and 1240 cm-1, because the
oxidized groups were developed by acid treatment on a rather base carbon surface. BCB
shows a slightly increased peak intensity due to the oxidized reaction of the carbon surface
with oxygen in a base solution. In contrast, NCB shows a decreased peak intensity for all
characteristic peaks. Consequently, the functional groups on the carbon surface were
effectively changed by the acid/base chemical treatments.
Figure 3 shows the XRD patterns of the chemical-treated CBs-supported Pt catalysts after
Pt deposition. Pt deposition was performed in 1.0 M NaOH solution, the best condition for
small size and high loading efficiency. Pt/VCB, Pt/NCB and Pt/BCB show a sharp and
intense peak at 40, 46, and 67, which can be assigned as Pt (111), Pt (200), and, Pt (220),
respectively. In comparison, Pt/ACB shows no peak for Pt deposition, which reflects the fact
that the Pt was not effectively deposited to the carbon surface. This, in turn, probably resulted
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 421
from the Pt reduction reaction being incomplete due to the high-acidic characteristics of the
CB surfaces.
4000
Pt(111)
Pt(200)
Pt(220)
Intensity / a.u.
3000
2000
1000
0
0 10 20 30 40 50 60 70 80
2 Theta
Figure 3. (A) Powder X-ray diffraction patterns of (a) Pt/VCB, (b) Pt/NCB, (c) Pt/BCB, and (d)
Pt/ACB.
It has been reported that the base condition is preferable in the preparation of Pt-
nanoclusters-deposited carbon supports because the Pt precursor (H2PtCl6) has to be reduced
by a reducing agent [43,44], which reduction process is best effected under base conditions
(e.g., pH>10) in precursor-containing solutions. Furthermore, the Pt average nanoparticle size
is smaller when the preparation condition has characteristics of a more base nature. By
contrast, ACB shows the smallest surface BET area of all samples due to the high degree of
aggregation. In this view, Pt deposition of carbon support is not easily available in the case of
decreased ACB surface areas.
It is readily apparent that in the case of Pt/ACB, Pt (220) peak does not exist. Pt/VCB
shows a rather broad peak at 67.6. In the case of Pt/BCB, the broadness of the peak is rather
increased, reflecting the size decrease of the Pt nanoparticles. From this peak, we can
422 Seok Kim and Soo-Jin Park
calculate the average crystalline size of the Pt nanoparticles with a Scherrer equation. The
average Pt nanoparticle sizes are summarized in Table 4. In the case of Pt/VCB, the average
Pt size is 3.01 nm, which the typical average range is 3~5 nm. In the case of Pt/BCB, the
average size decreases to the smallest size of 2.65 nm.
The calculated platinum loading efficiencies are summarized in Table 4. Pt/BCB shows
the highest value, 97 %, whereas Pt/ACB shows the lowest value, 38 %, indicating clearly
that Pt deposition to the carbon surface is enhanced in the case of base-treated CBs.
Consequently, base treatment affects a size decrease for Pt nanoclusters, and produces the
positive effect of enhancing the loading efficiency of Pt deposition to the carbon surface. It is
thought that the high loading efficiency of Pt is related to a probable Pt-reduction reaction
with the carbon surfaces manifesting base characteristics. The small Pt nanocluster size for
Pt/BCB might have originated with well-distributed base functional groups (e.g., the hydroxyl
group), with which Pt reduction can occur. Furthermore, small oxygen and sulfur peaks are
also shown. The sulfur probably originated from the impurities of virgin carbon blacks
(VCB). By chemical treatment of CBs, we could obtain a smaller size and a higher loading
efficiency of Pt nanoclusters-deposited to carbon supports.
(a) (b)
1500
1500 II
II
I
I
Intensity (cps)
Intensity (cps)
1000
1000
IV IV
III 500 III
500
VI VI
V
V
0 0
64 66 68 70 72 74 76 78 80 82 84 64 66 68 70 72 74 76 78 80 82 84
Binding Energy (eV) Binding Energy (eV)
(c) (d)
I II II
1000
1000
Intensity (cps)
I
Intensity (cps)
IV
IV 500 V
500 III VI
VI III
V
0 0
64 66 68 70 72 74 76 78 80 82 84 64 66 68 70 72 74 76 78 80 82 84
Figure 4. XPS Pt4f spectra of (a) VCB, (b) NCB, (c) BCB, and (d) ACB.
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 423
(a) (b)
I
80000
I
80000
Intensity (cps)
Intensity (cps)
40000
40000
II
II
III III
0 0
276 278 280 282 284 286 288 290 292 294 276 278 280 282 284 286 288 290 292 294
Binding Energy (eV) Binding Energy (eV)
(c) (d)
80000
I I
80000
Intensity (cps)
II II
III III
0 0
276 278 280 282 284 286 288 290 292 294 276 278 280 282 284 286 288 290 292 294
Binding Energy (eV) Binding Energy (eV)
Figure 5. XPS C1s spectra of (a) VCB, (b) NCB, (c) BCB, and (d) ACB.
In establishing the chemical state of the catalysts, XPS was used to determine the surface
oxidation state of the metallic clusters and supports. As most of the atoms in nanocluster
particle are surface atoms, it is expected that the measured oxidation state well reflects the
bulk oxidation state. Pt4f and C1s spectra for the chemical-treated CBs-supported Pt catalysts
are shown in Figures 4 and 5. The binding energies (BE) and relative peak areas are
summarized in Table 5.
The Pt4f signal consisted of three pairs of doublets. The most intense doublet, assigned to
peaks I and II, reflected metallic Pt. The second set of doublets, assigned to peaks III and IV
and which was observed at a BE of approximately1.1 eV higher that that for Pt (0), could be
attributed to the Pt (II) chemical state, as in PtO or Pt(OH)2. The third set of doublets,
assigned to peaks V and VI, was of the weakest intensity and occurred at even higher BEs,
which indicate that those peaks were likely caused by a small amount of Pt (IV) species on
the surface. The shifts of Pt4f signal by about 0.4 eV were observed with respect to previously
investigated unsupported Pt catalysts of the same bulk composition and lattice parameter [57].
Such an effect may be tentatively attributed to metal-support interaction.
424 Seok Kim and Soo-Jin Park
Table 5. Binding energies and relative peak areas obtained from (a) Pt4f and (b) C1s
spectra of Pt/VCB, Pt/NCB, Pt/BCB, and Pt/ACB
(a)
(b)
and 72.40%, respectively. These facts indicate that a portion of graphitic carbon was
decreased by some kind of reaction with oxygen or another reaction after chemical treatment.
(a) (b)
20nm
(c) (d)
Figure 6. TEM images of (a) Pt/VCB, (b) Pt/NCB, (c) Pt/BCB, and (d) Pt/ACB (scale bar for (a), (b),
and (c) is 5 nm, but scale bar for (b) is 20 nm)
Figure 6 shows the TEM images of (a) Pt/VCB, (b) Pt/NCB, (c) Pt/BCB, and (d) Pt/ACB.
From these images, it is observed that Pt nano-sized particles successfully deposited on the
chemical-treated carbon blacks. This is consistent with the above XRD, EDS, and XPS
results. In the case of Pt/VCB, Pt/NCB, and Pt/BCB, Pt nanoparticles of 2~6 nm were
observed. By contrast, the TEM images of Pt/ACB shows a rather scarce appearance of Pt
particles ranging from 5nm to 20nm, meaning a low loading level by ineffective Pt reduction
onto acid-treated carbon blacks.
Figure 7 shows the cyclic voltammetry (CV) results for the catalyst electrodes in
electrolytes of 1.0 M methanol + 0.5 M sulfuric acid. In all of the curves, at the forward scan,
the current began to rise and showed an anodic peak I at a potential of about 540mV. The
curves for Pt/BCB and Pt/NCB began to rise at about 300 mV, which could be considered the
onset potential, 100 mV lower than that of Pt/VCB. This implies that Pt/BCB and Pt/NCB
show easier and faster electrochemical reactions due to the reduction of the overpotential in
426 Seok Kim and Soo-Jin Park
methanol oxidation. At the backward scan, another anodic peak II appeared at a potential of
about 290 mV. This peak was attributed to the removal of incompletely oxidized
carbonaceous species (e.g. CO poisoning) formed in the forward scan [59,60]. For peak I, the
current density followed this order: Pt/BCB > Pt/NCB > Pt/VCB > Pt/ACB. For peak II, the
current density followed this order: Pt/NCB > Pt/BCB > Pt/VCB ~ Pt/ACB. Consequently, it
is thought that the electroactivity was improved in the cases of Pt/BCB and Pt/NCB.
Furthermore, Pt/BCB shows a suppressing effect for the CO poisoning, resulting from the
rather small anodic peak II compared to the anodic peak I.
3
c
a
Current Density (mA/mg)
2 b
I
II
1 b
c
d
0
Figure 7. CV graphs of (a) Pt/VCB, (b) Pt/NCB, (c) Pt/BCB, and (d) Pt/ACB, measured in 1.0M
CH3OH + 0.5M H2SO4.
Unique properties of graphite nanofibers (GNFs), which are one of the most popular carbon
materials have generated an intense interest in an application of these new carbon materials
toward a number of applications including energy storages, polymer reinforcements, and
catalyst supports. Recently, a study is undertaken to explore the physicochemical effects of
GNFs-supported metallic particles on the electrocatalytic oxidation of methanol when
compared with a traditional supports medium, carbon blacks [28,29]. Carbon is not only used
to conduct electrons and serves as a catalyst support but also helps to the stabilization of
three-phase boundary and morphology of an electrode for fuel cells [61].
The usage of various graphite and carbon black materials as a single catalyst support for
electrodes has been reported. However, with a single carbon support it may not easy to
control an electrode structure to achieve a combination of high conductivity, high porosity,
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 427
good morphology, and suitable hydrophobicity. Sakaguchi et al. [62] and Watanabe et al. [63]
approached an idea of adopting binary carbon supports for the use in liquid electrolyte fuel
cells. They have demonstrated an improved utilization of Pt catalyst using this approach.
The main routes for the synthesis of metallic nanoparticle electrocatalysts can be grouped
into an impregnation method and a colloidal method. An impregnation method is
characterized by a deposition step of Pt or other metal precursors followed by a reduction
step. This can be the chemical reduction of the metallic catalysts slurry in solution by using
reducing agents (liquid-phase reduction) [61,35,62] or gas-phase reduction of the metallic
particles impregnated carbon using a flowing H2 gas stream at a rather high temperature about
250-600oC [63,64]. In comparison to this, the colloidal methods [14,65] have the advantages
to produce small and homogeneously distributed carbon-supported metallic nanoparticles,
however, the methodologies are very complex.
It is reported that the particle sizes and the loading levels of Pt-based catalysts are key
factors that determine their electrochemical activity and cell performance for DMFCs [68-73].
However, the effect of the preparation method and the surface characteristics of various
carbon materials have not fully studied to our best knowledge.
In this paper, we had employed binary carbon supports to fabricate thin film electrodes in
DMFCs. The roles of binary carbon supports and an optimal mixing ratio will be evaluated
and characterized through cyclic voltammetry measurements. It will be shown that with the
usage of two carbon supports, electrochemical activities and loading contents of catalysts can
be enhanced. This improvement is further exemplified by the enhanced electrode kinetics of
methanol oxidation for a binary carbon support-electrode in comparison to a single support-
electrode.
The objective of this study is to investigate the structural effect of binary carbon supports
consisting of GNFs and CBs on the electrochemical properties of the carbon-supported
metallic nanoparticles. The sizes and loading levels of metallic nanoparticles have been
measured by changing the mixing ratio of two types of carbon materials. By changing the
mixing ratio of GNFs/CBs, the specific surface area and the morphological structure of the
mixed carbon materials are controlled.
B. Preparation
The carbon blacks supplied by Korea Carbon Black Co. were used in our experiments. These
have an average particle size of 24 nm, DBP adsorption of 153 (cc-100g-1) and specific
surface area of 112 (m2 g-1). The GNFs were supplied by Showa Denko Co. (Japan). These
carbon fiber materials have a diameter of 100~150 nm and a length of 5~50 m, resulting a
large aspect ratio. They show a rather small specific surface area of 30 (m2 g-1).
Prior to a preparation of Pt-based catalyst, the residual chemicals of carbon materials
used were removed by a Soxhlet extraction by boiling with acetone at 80 C for 3h. Two
types of carbon materials were washed several times with distilled water and dried in a
vacuum oven at 90 C for 12h. Binary carbon supports were prepared by mixing CBs and
GNFs with changing the mixing ratio. The samples were assigned as GNF0, GNF10, GNF30,
GNF50, and GNF100 by changing the weight content of GNFs like 0, 10, 30, 50, and 100%.
In proportional to this change, the weight content of CBs changed like 100, 90, 70, 50, and
0%, respectively.
428 Seok Kim and Soo-Jin Park
To assess the effect of the usage of a binary carbon support on the preparation and
electrochemical behaviors of carbon-supported Pt nanoparticles, we had prepared the mixed
carbon supports consisting of CBs and GNFs with changing mixing ratio. After Pt
incorporation into carbon materials, the average crystalline sizes of Pt nanoparticles were
obtained by XRD measurements. Figure 8 shows the powder X-ray diffraction patterns of Pt
catalysts deposited on (a) GNF0, (b) GNF10, (c) GNF30, (d) GNF50, and (d) GNF100. With
an increase of GNFs contents, a sharp peak at 2 = 26 and a small peak at 2 = 54 increase
gradually. This gradual changes of the peak intensity can be clearly explained by the fact that
Pt deposited GNF100 shows a strong peak and Pt deposited on GNF0 shows no peak at this
position. This means that GNFs in this study have a crystalline graphitic structure. All
samples show the typical Pt crystalline peaks of Pt(111), Pt(200), Pt(220) and Pt(311).
This indicates that Pt particles prepared on various carbon materials shows a similar
loading level and similar crystalline structure. From this result, it is found that Pt
nanoparticles have been successfully deposited on a mixed binary carbon support regardless
of the mixing ratio of CBs and GNFs.
The average size of Pt nanoparticles was calculated by using a Scherrer equation and was
demonstrated in Table 6. The average size of Pt increased from 3.58nm to 6.41nm gradually
by changing the GNF content from 0% to 100%. It is interesting to note that a sharpness of
peak is enhanced with an increase of GNFs contents. The larger the sharpness of peaks is, the
larger the average crystalline size of Pt particles is. The increase of Pt average size could be
related to a decrease of specific surface area by increasing the portion of GNFs which have
the smaller specific surface area rather than CBs. This is due to the fact that CBs have a
specific surface area of 112 m2 g-1 and the GNFs have a specific surface area of 30 m2 g-1.
Due to the decrease of specific surface area, it is expected that the deposited Pt particles are
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 429
more aggregated and become larger. From this result, it is concluded that the particle size can
be changed by controlling the specific surface area by using the mixed carbon supports with a
different specific surface area.
6000 Pt(111)
Pt(200)
Pt(220)
Pt(311)
Intensity / a.u.
4000
(e)
2000
(d)
(c)
(b)
0
0 20 40 60 80 (a)
2 Theta
Figure 8. Powder X-ray diffraction patterns of Pt catalysts deposited on (a) GNF0, (b) GNF10, (c)
GNF30, (d) GNF50, and (e) GNF100.
Platinum loading levels can be calculated by considering the relative peak intensity of Pt
against the peak intensity of carbon with an atomic ratio. This value is summarized in Table
7. In the case of GNF0, Pt loading level shows a value of 88%. With an increase of GNFs
contents upto 30%, Pt loading level is enhanced to a value of 98%. However, a further
increase of GNFs contents over 30% leads to a decrease of loading level. In the case of
GNF100, Pt loading level shows the smallest value of 73%. This can be explained by the fact
that Pt nanoparticles is not fully impregnated on carbon surface. Consequently, Pt loading
level shows the highest value of 98%, when the GNFs content is 30%. It is expected that the
loading level of Pt deposition be related to a reaction condition of Pt reduction and surface
characteristics of carbon supports. In other words, the loading level is probably dependent on
the changes of surface functional groups related with Pt reduction or changes of specific
430 Seok Kim and Soo-Jin Park
surface area by changing the mixing ratio of two carbon materials. Furthermore, slight peaks
of oxygen and sulfur atoms are also shown. Sulfur is probably originated from the side
product of carbon blacks. From these results, sulfur content is reduced by increasing the
content of GNFs. It is thought that the side products such as sulfur can deteriorate a function
of catalysts.
Pt Pt Sulfur
Carbon
Loading levela Loading levelb Contenta
Samples
(%) (%) (wt.%)
GNF0 88 86 1.12
GNF10 93 89 0.99
GNF30 98 95 0.75
GNF50 85 82 0.53
GNF100 73 72 0.02
a: EDS measurements
b: ICP-AES measurements
D. Electroactivity of Catalysts
Figure 9 shows the electroactivity of Pt catalyst supported on the mixed carbon supports by
cyclic voltammograms. The voltammetric features are consistent with the previous reports
[77,78]. Anodic peaks for a methanol oxidation were shown at 640 ~ 740 mV for a forward
scan and another anodic peaks were shown at 390 ~ 510 mV for a backward scan for each
sample. The latter anodic peaks are known to be related to the removal of incompletely
oxidized carbonaceous species formed in the forward scan [77]. This peak current can be
reduced by alloying Pt catalyst with another metal catalysts for anode catalysts of DMFC
[78]. The peak current and potential were described in Table 8. The current density of anodic
peaks increased from 0.489 mA cm-2 to 0.989 mA cm-2 by increasing GNF contents up to
30%. However, a further increase of GNF content over 30% has brought a decrease of current
density and a large positive shift of anodic peak potential from 667mV to 734mV. This means
that the electroactivity has been decayed when the GNFs content is over 30%. From this
result, it can be concluded that the electroactivity are the best when the mixing ratio of GNFs
and CBs is 30:70. However, this GNF30 sample shows a rather large current density for
backward scan. GNF10 shows a rather small current density for backward scan.
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 431
Consequently, the loading level of Pt incorporation and electroactivity for MeOH oxidation
could be changed by controlling the mixing ratio of binary carbon supports. The enhancement
of the loading level or the electroactivity is originated from the change of carbon specific
surface area and morphological changes of carbon supports by controlling the mixing ratio of
GNFs and CBs.
1.4
1.2
1.0
0.8
I (mA/cm )
2
0.6 (a)
0.4
0.2 (e)
0.0
-0.2
200 400 600 800 1000 1200
V (V vs. SCE)
Figure 9. Cyclic voltammograms of Pt catalysts deposited on (a) GNF0, (b) GNF10, (c) GNF30, (d)
GNF50, and (e) GNF100, which is measured in 0.5M H2SO4 + 1.0M CH3OH. (Sweep rate: 20mV sec-1,
each graph is obtained at first sweep of multiple sweeps).
Figure 10 shows SEM micrographs of Pt-impregnated carbon supports which have a different
mixing ratio of CBs and GNFs. The diameter of GNFs is in the range of 100 ~ 150 nm. These
GNFs have a large aspect ratio of > 500 and a high electrical conductivity of 0.1 S cm-1 due to
the well-ordered graphitic structure. Pt nanoparticles are not clearly visible in the SEM
432 Seok Kim and Soo-Jin Park
micrographs because their average sizes are nanometer scale. By increasing GNFs contents in
the order of (a) 10, (b) 30, (c) 50, and (d) 100 %, it is clearly shown that the content of the
powder-like carbon black particles becomes smaller. Accordingly, it is clearly known that the
specific surface area of carbon supports becomes smaller. This can be one of the reasons why
the average size of Pt particles becomes larger. The decrease of specific surface area of
carbon supports can lead to the aggregation of Pt nanoparticles, resulting a decrease of
effective electrochemical reaction site. However, the volume of vacancy between the carbon
particles is larger and larger, which can give a free channel of liquid or gas mass transfer.
This can function as a beneficial effect on the enhancement of electroactivity. It is thought
that the above two adverse effect on the electrochemical activity can explain the existence of
an optimum ratio of two different carbon materials.
(a) (b)
(c) (d)
Figure 10. SEM micrographs of Pt catalysts deposited on (a) GNF10, (b) GNF30, (c) GNF50, and (d)
GNF100 (Each scale bar means 2m length).
Figure 11 shows impedance plots of electrocatalysts in 0.5M H2SO4 + 1.0M MeOH. These
plots were obtained by measuring AC impedance with changing a frequency from 1 MHz to
0.1 Hz. The plots show a semi-circle like part, though not a perfect shape at a higher
frequency range and a linear part at a lower frequency range. When the imaginary part of
impedance value is almost zero, the real part of impedance value indicates the charge-transfer
resistance for catalyst electrode. These resistance values obtained by means of equivalent
circuit analysis were demonstrated as a function of GNFs contents in Figure 12. With an
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 433
increase of GNFs content from 0% to 30%, the resistance change from 19 Ohm cm2 to 11
Ohm cm2. With the increase of GNFs content from 30% to 100%, the resistances increase
from 11 Ohm cm2 to 65 Ohm cm2. These resistance behaviors as a function of GNFs contents
show a similar tendency with the above electroactivity results. From this result, the resistance
decrease in the case of GNF30 can be one of the reasons for the enhancement of the
electroactivity of Pt catalysts. This change of charge-transfer resistance can be explained
primarily by two reasons. The one reason is the difference of an electrical conducting nature
of carbon materials themselves. The other one is the change of an electrical contact point
between carbon materials, which is originated from the morphology change of Nafion
polymer electrolyte-coated catalyst electrodes.
300
0.53 Hz
200
-Im (Ohm cm )
2
8.5 Hz
100 (a)
(b)
(c)
(d)
(e)
0
0 100 200 300
2
Re (Ohm cm )
Figure 11. Impedance plots Pt catalysts deposited on (a) GNF0, (b) GNF10, (c) GNF30, (d) GNF50,
and (e) GNF100, which is measured in 0.5M H2SO4 + 1.0M CH3OH.
70
60
Resistance (Ohm cm )
2
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
Content of GNFs (%)
Figure 12. Resistance changes of Pt catalysts deposited on (a) GNF0, (b) GNF10, (c) GNF30, (d)
GNF50, and (e) GNF100 obtained from the impedance plots of Figure 11.
434 Seok Kim and Soo-Jin Park
Plasma treatment is one of the most popular treatments of carbon materials. This treatment
takes place only on the carbon black surfaces without changing its bulk properties. Besides, it
was possible to process under oxidative, reductive, or inactive, atmospheres [79]. Therefore,
in this work, the effects of the plasma treatment for CBs on the modification of surface
functional groups and the deposition of Pt catalyst were investigated.
Plasma treatment for the carbon blacks was carried out using a radio frequency for N2 gas
(Tegal Plasmod). The radio frequency (13.56 MHz) generated by N2-plasma was operated at
300 W. The input treatment time for N2-plasma treatment varied between 0 and 50 sec,
namely, P0, P10, P20, P30, and P50 under a pressure of about 0.75 torr and a flow rate of
about 100 ml/min.
B. Structural Properties
Fig. 13 shows FT-IR spectra of the carbon blacks as a function of N2-plasma treatment time.
As shown in Fig. 1, functional groups of the N2-plasma treated CBs were observed at peak of
650 cm-1, 1050 cm-1, 1490 cm-1, 1550 cm-1, 1630 cm-1, and 3430 cm-1, such as, -NH, C-N,
=NH, NH3+, C=N, and NH2. The band intensity at 1050 cm-1, 1630 cm-1, and 3430 cm-1 of
the P0 was rather weak than that of the N2-plasma treated CBs. In case of the N2-plasma
treated CBs samples, the intensity of 1050 cm-1, 1630 cm-1, and 3430 cm-1 was increased with
N2-plasma treatment time, due to the changes of functional groups. XPS is used to determine
the elemental composition on CBs surfaces. Fig. 14 shows the XPS survey scan spectra of the
Transmittance (%)
P0
P10
P20
P30
P50
N2-plasma treated CBs. The C1s and O1s peaks of N2-plasma treated CBs are found at the
binding energy of about 284.6 and 532.8 eV, respectively [80]. Also, the peak of Pt4f is
observed 74 eV. The intensity of Pt4f peak on the N2-plasma treated CBs is increased. This
indicates that the intensity of Pt4f is increased due to the reaction between the radical on N2-
plasma treated CBs and Pt4f. However, after N2-plasma treatment, the peak of N1s is hardly
observed because the content of the nitrogen is very low. Therefore, to estimate the content of
nitrogen, elemental analysis (EA) is used.
O1s
N1s C1s
Pt/P0
Intensity (a. u.)
Pt/P10
Pt4f
Pt/P20
Pt/P30
Pt/P50
Figure 14. XPS survey scan spectra of the N2-plasma treated Pt/carbon blacks catalysts.
Elemental analysis
Sample
C1s N1s
P0 98.1 1.2
Table 9 shows the EA results of the N2-plasma treated CBs surfaces. The content of
nitrogen are increased by the N2-plasma treatment. This is clearly attributed to the increase of
nitrogen-containing functional groups on the CBs surfaces. The result can also be explained
that the N2-plasma treatment produces various nitrogen-containing functional groups, i.e., -
NH, C-N, =NH, NH3+, C=N, and NH2 groups on the CBs surfaces. Therefore, it can be
seen that the N2-plasma treatment of CBs surfaces lead to the increase of nitrogen-containing
functional groups, resulting in improving the deposited capacity of Pt catalyst.
C. Electrochemical Properties
Fig. 15 shows linear sweep voltammograms (LCV) of N2-plasma treated Pt/CBs catalysts in
0.5 M H2SO4 containing 1.0 M CH3OH. Pt/P30 show the best electroactivity among the
samples due to its highest current density of anodic peak. This is evidence that the N2-plasma
Pt/P30 has high activity towards methanol oxidation.
-4
-7.0x10
Pt/P0
-4 Pt/P10
-6.0x10
Pt/P20
-4 Pt/P30
-5.0x10
Pt/P50
-4
-4.0x10
I (A)
-4
-3.0x10
-4
-2.0x10
-4
-1.0x10
0.0
-4
1.0x10
400 600 800 1000
mV (Ag/AgCl)
Figure 15. Cyclic voltammograms of the N2-plasma treated Pt/carbon blacks catalysts.
Anodic peaks of methanol oxidation are observed at about 750 ~ 850 mV. The current
density of anodic peak is greatly enhanced and the peak potential is shifted negatively when
the plasma treatment time is 30 sec. It means the higher electroactivity and better redox
reversibility. However, further treatment has brought the decrease of current density and
positive shift of anodic peak potential. This means that the electrocatalytic activity and redox
reversibility has been decayed. From this result, it can be concluded that the electrocatalytic
activity are the best when the treatment time is 30 sec.
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 437
B. Oxyfluorination
MWNTs were oxyfluorinated under several conditions. The oxyfluorination reaction was
performed with F2, O2, and N2 gases in a batch reactor made of nickel with an outer electric
furnace. After evacuation, fluorine and oxygen mixtures were introduced to the reactor at
room temperature, and then the reactor was heated to the treatment temperature. After the
reaction, the samples were cooled to room temperature, and then the reactive gases were
purged from the reactor with nitrogen. The pressure was 0.2 MPa and the reaction time was
15 min at the treatment temperature.
C. Structural Properties
Fig. 16 shows the FT-IR results for oxyfluorinated MWNTs. The carboxyl/ester group (C=O)
at 1632 cm-1 and hydroxyl group (OH) at 3450 cm-1 are observed for the oxyfluorinated
MWNTs. Particularly, the intensity of the OH group peak of MWNTs exhibits the highest
value at an oxyfluorination temperature of 100 . In case of the oxyfluorinated MWNT
samples, the intensity at 1632 cm-1, and 3450 cm-1 increased with oxyfluorination
temperature, due to the changes of functional groups.
This result may lead to improve electrochemical properties of Pt/MWNTs due to the
increase of specific polarity and the H-bonding of the MWNTs. However, after a specific
reaction time, oxyfluorination hardly influence, change of surface functional groups of
MWNTs, due to the disappearance of free radical.
438 Seok Kim and Soo-Jin Park
300-MWNTs
200-MWNTs
Transmittance (%)
100-MWNTs
R.T-MWNTs
neat-MWNTs
Pt(111)
Pt(200) Pt(220)
Pt/300-MWNTs
Intensity (a. u)
Pt/200-MWNTs
Pt/100-MWNTs
Pt/R.T MWNTs
Pt/neat MWNTs
20 40 60 80
The powder XRD patterns for oxyfluorinated Pt/MWNTs are shown in Fig. 17. Fig. 17
shows that Pt deposited on both neat-MWNTs and oxyfluorinated MWNTs forms a face
centered cubic (fcc) structure and has major peaks at about 2 = 39.7 (111), 46.2 (200), and
67.4 (220). The broader diffraction peaks for the catalysts also lead to a smaller alloy particle
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 439
size calculated by the Scherrer equation. The calculation results estimated the average size of
Pt catalysts. The sharpening of the XRD peaks indicates an increase of the average crystallite
size with increasing the oxyfluorination temperature. In the case of Pt/100-MWNTs, it is the
optimum condition for the small size of Pt nanoclusters of 3.58 nm. Pt/300-MWNTs shows
the large size of Pt nanoclusters of 3.7 nm. The particle size of Pt for Pt/200-MWNTs cannot
be calculated because the Pt (220) peak is too broad. It is therefore concluded that the average
size of Pt nanoclusters is strongly dependent on the functional groups in the precursor and
carbon dispersed solution.
D. Electrochemical Properties
-0.002
-0.001
0.000
mV (Ag/AgCl)
The current density of anodic peak is greatly enhanced and the peak potential is shifted
negatively at the treatment temperature of 100 , thus leading to higher electroactivity and
better redox reversibility. However, further treatment decreased the current density and
positive shift of anodic peak potential, resulting in loss of electrocatalytic activity and redox
reversibility. From this result, it is concluded that the high electrocatalytic activity is achieved
when the treatment temperature is 100 .
440 Seok Kim and Soo-Jin Park
7. Conclusion
The size and the loading efficiency of deposited metal particles as fuel-cells catalysts were
investigated by changing the preparation method of various carbon materials supported
catalysts. Furthermore, chemical and physical treatment effects of carbon supports on the
nano-structure as well as the electroactivity of the carbon-supported platinum catalysts for
DMFCs were studied.
First, Carbon-supported metal nanoclusters as fuel-cell catalyst electrodes were prepared
by a chemical method of H2PtCl6 reduction on acid/base-treated carbon blacks (CBs). The
size and the loading efficiency of the metal clusters were dependent on the preparation
method and the surface characteristics of the CBs. In the process of catalyst preparation, the
NaOH concentration was a critical factor in obtaining Pt nanoclusters of small size and high
loading efficiency. When the NaOH concentration was 1.0 M, carbon-supported Pt
nanoclusters of 3.1 nm average size and 90% loading efficiency were obtained. The Pt
clusters deposited on base-treated CBs showed the smallest particle size, 2.65 nm, and the
highest loading efficiency, 97%. By contrast, the Pt clusters were ineffectively deposited on
acid-treated CBs. Accordingly, the Pt catalysts deposited on the base-treated CBs showed
enhanced electroactivity compared with those deposited on virgin CBs.
Secondly, binary carbon-supported platinum (Pt) nanoparticles were prepared by a
chemical reduction method of Pt precursor on two types of carbon materials such as carbon
blacks (CBs) and graphite nanofibers (GNFs). Average sizes and loading levels of Pt metal
particles were dependent on a mixing ratio of two carbon materials. The highest
electroactivity for methanol oxidation was obtained by preparing the binary carbon supports
consisting of GNFs and CBs with a weight ratio of 30:70. Furthermore, with an increase of
GNFs content from 0% to 30%, a charge-transfer resistance changed from 19 Ohm cm2 to 11
Ohm cm2. The change of electroactivity or the resistance for catalyst electrodes was attributed
to changes of specific surface area and morphological changes of carbon supported catalyst
electrodes by controlling the mixing ratio of GNFs and CBs.
Thirdly, in order to improve the dispersion of platinum catalysts deposited on carbon
materials, the effects of surface plasma treatment of carbon blacks (CBs) were investigated.
The surface characteristics of the CBs were determined by fourier transformed-infrared (FT-
IR), X-ray photoelectron spectroscopy (XPS), and Boehms titration method. The
electrochemical properties of the plasma-treated CBs-supported Pt (Pt/CBs) catalysts were
analyzed by linear sweep voltammetry (LSV) experiments. From the results of FT-IR and
acid-base values, N2-plasma treatment of the CBs at 300 W intensity led to a formation of a
free radical on the CBs. The peak intensity increased with increase of the treatment time, due
to the formation of new basic functional groups (such as C-N, C=N, NH3+, NH, and =NH)
by the free radical on the CBs. Accordingly, the basic values were enhanced by the basic
functional groups. However, after a specific reaction time, N2-plasma treatment could hardly
influence on change of the surface functional groups of CBs, due to the disappearance of free
radical. Consequently, it was found that optimal treatment time was 30 second for the best
electro activity of Pt/CBs catalysts and the N2-plasma treated Pt/CBs possessed the better
electrochemical properties than the pristine Pt/CBs.
Lastly, multi-walled carbon nanotubes (MWNTs) were oxyfluorinated at several different
temperatures. The results indicated that graphitic carbon was the major carbon functional
A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts 441
component on the MWNT surfaces and other functional groups, such as CO, C=O, HO
C=O, CFx, were also present after oxyfluorination. The changes of surface properties of
oxyfluorinated MWNTs were investigated using FT-IR, EDS and XRD. From the surface
analysis, it was found that surface fluorine and oxygen contents increased with increasing
oxyfluorination temperature and showed a maximum value at 100 C. Consequently, it was
found that optimal treatment temperature was 100 C for the best electro activity of
Pt/MWNT-100 catalyst and the oxyfluorinated Pt/MWNTs possessed the better
electrochemical properties than the pristine Pt/MWNTs.
References
[1] B.D. McNicol, D.A.J. Rand, and K.R. Williams, J. Power Sources, 83, (1999) 15.
[2] S. Wasmus and A. Kuver, J. Electroanal. Chem., 45, (1999) 19.
[3] A.S. Arico, S. Srinivasan and V. Antonucci, Fuel Cells, 2, (2001) 1.
[4] A.S. Arico, P. Creti, E. Modica, G. Monforte, V. Baglio, and V. Antonucci Electrochim.
Acta, 45, (2000) 4319.
[5] X.M. Ren, P. Zelenay, A. Thomas, J. Davey, and S. Hottesfeld, J. Power Sources, 86,
(2000) 111.
[6] Hamnett, Catal. Today, 38, (1997) 445.
[7] A.D. Anderson, G.A. Deluga, D.L. Boxall, M.J. Vergne, D.M. Hercules, E.A. Kenik,
C.M. Lukehart, J. Nanosci. Nanotech. 4, (2004) 809.
[8] E.S. Steigerwalt, G.A. Deluga, C.M. Lukehart, J. Nanosci. Nanotech. 3, (2003) 247.
[9] Y.C. Liu, X.P. Qiu, Y.Q. Huang, and W.T. Zhu, J. Power Sources, 111, (2002) 160.
[10] Pozio, M.D. Francesco, A. Cemmi, F. Cardellini, and L. Giorgi, J. Power Sources, 105,
(2002) 13.
[11] S.J. Park, H.J. Jung, and C.W. Nah, Polymer (Korea), 27, (2003) 46.
[12] O.V. Cherstiouk, P.A. Simonov, and E.R. Savinava, Electrochim. Acta, 48, (2003) 3851.
[13] A.S. Arico, V. Baglio, E. Modica, A.D. Blasi, and V. Antonucci, Electrochem.
Commun., 6, (2004) 164.
[14] H. Bonnemann, R. Brinkmann, W. Brijoux, E. Dinjus, T. Joussen, and B. Korall,
Angew. Chem., Int. Ed. Engl., 30, (1991) 1312.
[15] M. Gotz and H. Wendt, Electrochim. Acta, 43, (1998) 3637.
[16] L.K. Kurihara, G.M. Chow and P.E. Schoen, Nanostruc. Mater., 5, (1995) 607.
[17] N. Toshima and T. Yonezawa, New J. Chem., 22, (1998) 1179.
[18] K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, John Wiley,
New York (1988)
[19] Carbon Materials for Advanced Technologies (Ed: T.D. Burchell), Pergamon, New
York, 1999
[20] Sciences of Carbon Materials (Eds.: H. Marsh, F. Rodriguez-Reinoso), Publicaciones de
la Universidad de Alicante, Alicante, Spain, 1997
[21] V. Jalan, Extended Abstracts, Electrochemical Society, Los Angeles, CA, Fall Meeting,
Abstract 192 (October 1979)
[22] H. Bonnemann, R. Brinkmann, W. Brijoux, E. Dinjus, T. Joussen, B. Korall, Angew.
Chem., 103 (1991) 1344.
[23] H.G. Petrow, R.J. Allen, US Patent, 3,992,331 (1976)
442 Seok Kim and Soo-Jin Park
[55] M.J. Wang and S. Wolff, Carbon Black Science and Technology edited J.B. Donnet,
R.C. Bansal and M.J. Wang, Marcel Dekker, New York (1993) p. 229.
[56] S.J. Park and K.D. Kim, J. Colloid Interface Sci., 212, (1999) 186.
[57] A.S. Arico, G. Monforte, E. Modica, P.L. Antonucci, and V. Antonucci, Electrochem.
Commun., 2, (2000) 466.
[58] R. Woods, J. Electroanal. Chem., 9, (1976) 1.
[59] Z. Liu, X.Y. Ling, X. Su, J.Y. Lee, J. Phys. Chem. B, 108, (2004) 8234.
[60] L. Gao, H. Huang, C. Korzeniewski, Electrochim. Acta, 49, (2004) 1281.
[61] S.J. Park, H.J. Jung, C.W.Nah, Polymer (Korea), 27, (2003) 46.
[62] M. Sakaguchi, K. Uematsu, A. Sakata, Y. Sato, M. Sato, Electrochim. Acta, 34,
(1989) 625.
[63] M. Watanabe, K. Makita, H. Usami, S. Motoo, J. Electroanal. Chem., 197 (1986) 195
[64] Pozio, M.D. Francesco, A. Cemmi, F. Cardellini, L. Giorgi, J. Power Source, 105,
(2002) 13
[65] O.V. Cherstiouk, P.A. Simonov, E.R. Savinava, Electrochim. Acta, 48, (2003) 3851
[66] A.S. Arico, V. Baglio, E. Modica, A.D. Blasi, V. Antonucci, Electrochem. Commun., 6,
(2004) 164.
[67] M. Gotz, H. Wendt, Electrochim. Acta, 43, (1998) 3637
[68] Hamnett, Catal. Today, 38, (1997) 445
[69] H. Tang, J.H. Chen, Z.P. Huang, D.Z. Wang, Z.F. Ren, L.H. Nie, Y.F. Kuang, S.Z. Yao,
Carbon, 42, (2004) 191.
[70] Z. Liu, X. Lin, J.Y. Lee, W. Zhang, M. Han, L.M. Gan, Langmuir, 18, (2002) 4054.
[71] Z. Liu, J.Y. Lee, W. Chen, M. Han, L.M. Gan, Langmuir, 20, (2004) 181.
[72] Z. Jusys, J. Kaiser, R.J. Behm , Langmuir, 19, (2003) 6759.
[73] T.C. Deivaraj, J.Y. Lee, J. Power Source, 142, (2005) 43
[74] T.J. Schmidt, M. Noeske, H.A. Gasteiger, R.J. Behm, J. Electrochem. Soc., 145,
(1998) 925
[75] M. Watanabe, H. Igarashi, K. Yosioka, Electrochim. Acta, 40, (1995) 329
[76] T.J. Schmidt, H.A. Gasteiger, G.D. Stab, P.M. Urban, D.M. Kolb, R.J. Behm, J.
Electrochem. Soc., 145, (1998) 2354.
[77] Z. Liu, X.Y. Ling, X. Su, J.Y. Lee, J. Phys. Chem. B, 108, (2004) 8234.
[78] L. Gao, H. Huang, C. Korzeniewski, Electrochim. Acta, 49, (2004) 1281.
[79] S.J. Park, J.S. Kim, J. Colloid Interface Sci., 224 (2001) 336.
[80] S.D. Gardner, C.S.K. Singamsetty, G.L. Booth, G.R. He, C.U. Pittman Jr., Carbon, 33
(1995) 587.
[81] S.J. Park and Y.S. Jang, J. Colloid Interface Sci., 237 (2001) 91
[82] S. Berchmans and R. Vijayavalli, Langmuir, 11 (1995), 286
[83] R. Blyth, H. Buqa, F.P. Netzer, M.G. Ramsey, J.O. Besenhard and P. Golob, Appl. Surf.
Sci., 167 (2000) 99
[84] K.C. Walter, H.H. Kung and C.J. Maggiore, Appl. Phys. Lett., 71 (1997) 1320
[85] T. Muhl, J. Kretz, I. Monch, C.M. Schneider, H. Bruckl and G. Reiss, Appl. Phys. Lett.,
76 (2000) 786
In: Fuel Cell Research Trends ISBN: 1-60021-669-2
Editor: L.O. Vasquez, pp. 445-471 2007 Nova Science Publishers, Inc.
Chapter 10
Abstract
Alkaline solid polymer electrolyte membranes (ASPEM) have been extensively studied
for the varied electrochemical device applications due to the thinner thickness, lower
electrolyte permeability, higher ionic conductivity and ease in processibility. We prepared a
series of ASPEM based on polyvinyl alcohol (PVA) polymers such as PVA/KOH,
PVA/PEO/KOH, PVA/KOH/glass-fiber-mat, PVA/PECH, PVA/TEAC, PVA/PAA, and
PVA/PAA/PP/PE composites. These new material systems were introduced with unique
improvements for fuel cell applications. They have not only the potential to lower the
processing cost but also can provide high ionic conductivity. The PVA/PAA polymer blend
system exhibited high ionic conductivity of 0.301 S cm-1 and the anionic transport number
could reach 0.99, both at room temperature. For Zn/air and Al/air battery application, the
power density was as high as 90-110 mW cm-2. The PVA-based composite SPE has great
potential for use in alkaline battery systems. The progressive advancements in the science and
technology of solid polymer electrolyte membranes are presented in this chapter.
We will demonstrate the preparation techniques for ASPEM and the characterization
results. The relationship between structure and properties will be discussed and compared.
The double-layer carbon air cathodes were also prepared for solid-state alkaline metal fuel cell
fabrication. The alkaline solid state electrochemical systems, such as Ni-MH, Zn-air fuel cells,
*
E-mail address: wu@mail.cgu.edu.tw
446 G.M. Wu, S.J. Lin and C.C. Yang
Al-air fuel cells, Zn-MnO2 and Al-MnO2 cells, were assembled with anodes, cathodes and
alkaline solid polymer electrolyte membranes. The electrochemical cells showed excellent cell
power density and high electrode utilization. Therefore, these PVA-based solid polymer
electrolyte membranes have great advantages in the applications for all-solid-state alkaline
fuel cells. Some other potential applications include small electrochemical devices, such as
supercapacitors and 3C electronic products.
Introduction
In 1973, Wright et al. [1-3] discovered a class of solid polymer electrolytes (SPE) composed
of polyethylene oxide (PEO) and alkali metal ions that exhibited conductivity at room
temperature. Later, Armand et al. [4] proposed the use of SPE in lithium battery. Since then,
many researchers have developed different polymer-based systems with improved electrolyte
characteristics in lithium polymer batteries for 3C electronic products, and even for electric
vehicles [5]. For the use of SPE in lithium battery, the polymer electrolytes usually contain
lithium salts such as LiClO4, LiBF4, LiPF6, LiCF3SO3 [6,7] and a non-aqueous characteristic
is necessary. During the charge and discharge processes, the charge-carrier ion of Li-polymer
electrolyte is transported by lithium. However, the alkaline SPE system is totally different
from the lithium polymer electrolytes. It is an aqueous system and contains water. The
dissolved hydroxide (OH-) ion is transported from cathode to anode in SPE during the
discharge process and the electric current is thus generated. Fig. 1 shows the schematic
illustration of alkaline H2/O2 fuel cell and Table 1 summarizes the characteristics for different
fuel cell systems. The alkaline fuel cell and Zn/air battery can be operated at ambient
temperature with stable electrochemical properties and lower cost.
e- e-
V
e-
Electrolyte
Operating
Mobile
Fuel cell type temperature Characteristics
ion
(oC)
Low cost, low operating temperature and
Alkaline fuel cell (AFC) R.T.-150 OH-
relatively low power applications ~50 kW.
Low cost, low operating temperature,
Zn/air, Al/air
R.T.-100 OH- mechanically rechargeable and low power
Metal/air semi-fuel cell
applications ~10 kW.
Proton exchange membrane Relatively low operating temperature, higher
80-100 H+
fuel cell (PEMFC) cost.
Solid oxide fuel cell High operating temperature and wide power
500-1000 O2-
(SOFC) applications ~100 MW.
Molten carbonate fuel cell High operating temperature and wide power
700-1000 CO32-
(MCFC) applications ~100 MW.
The traditional aqueous alkaline batteries have some disadvantages that inhibit its
advanced applications: (1) leakage and evaporation of liquid electrolyte, (2) corrosion of
electrode, (3) instability in chemical and electrochemical characteristics, and (4) large size
and heavy weight that reduce the energy and power density of battery. On the other hand, the
alkaline SPEs have low cost, thin thickness, high flexibility, good film-forming characteristic,
reasonable mechanical strength, no electrolyte leakage problem, and the high ionic
conductivity at room temperature has been developed. For the past decade, compared to the
lithium battery SPE, there are only few literatures to discuss these novel alkaline SPE
materials. Fauvarque et al. [8-13] developed water-containing alkaline solid polymer
electrolytes based on PEO and KOH by a solvent casting method. The ionic conductivity was
up to 10-3 S cm-1. They demonstrated good cyclability of 60 cycles without short circuit and
the faradic efficiency yields could be up to 87% in Ni/Cd and Ni/Zn cells. Lewandowski et al.
[14] prepared polyvinyl alcohol (PVA)-KOH alkaline SPE and discussed the membrane
properties by cycle voltammetry. We have also developed a series of novel polyvinyl alcohol-
based SPEs for use in alkaline metallic-air fuel cell. The PVA/polyacrylic acid (PAA)-KOH
with interpenetrating network (IPN) [15] showed high ionic conductivity and very
encouraging electrochemical performance. However, further developments are still required
to improve the ionic conductivity, alkaline sustainability, mechanical properties, and water
holding stability. This chapter describes the recent advances in alkaline SPEs primarily based
on PVA polymers. The developments in PEO-based alkaline SPE and gel-type alkaline SPE
are also briefly reviewed. The applications in electrochemical cells, especially metal-air fuel
cells, are discussed.
Although Hassan et al. [16] continued to study PEO/KOH SPE, the ionic conductivity
reached only about 3.110-5 S cm-1. The effect of KOH concentration on the ionic
conductivity of PEO polymer electrolyte could be attributed to the decreased crystallinity of
448 G.M. Wu, S.J. Lin and C.C. Yang
PEO polymer with the increasing KOH concentration. This also increased the concentration
of conductive OH- ions in the polymer matrix. The optimum KOH concentration was
controlled at 40-60 wt.% in PEO matrix. The amorphous phase in polymer matrix is therefore
favorable for ion transportation and movement.
For the first time, a totally solid-state electric double layer capacitor (EDLC) was
fabricated using PEO-KOH-H2O as the SPE and the polymer electrolyte could replace large
amount of liquid KOH electrolyte [17,18]. The ideal rectangular shape of cyclic voltammetry
result for this solid-state EDLC was obtained, and the real value of specific capacitance was
90 F g-1. It was only slightly lower than that of liquid electrolyte supercapacitor, and it might
be related to the electrode material and structure.
Alkaline PEO-based SPE may be limited by the high crystallinity and poor thermal
stability. Vassal et al. [11] intended to use amorphous copolymer, poly(epichlorohydrin-co-
ethylene oxide) P(ECH-co-EO), and presented good electrochemical properties without
PEOs disadvantages. The ionic conductivity value at room temperature was around 10-3 S
cm-1. The all solid-state alkaline Ni/MH and Zn/air electrochemical cells also exhibited stable
behavior.
Inasmuch as the use of epichlorohydrin concept, Agel et al. [13] developed a new and
cheap type of anion exchange membranes (AEM) by preparing the polyepichlorohydrin
(PECH) graft quaternary amines (DABCO, TEA) for use in alkaline cells. Its a quasi-gas
impervious polymer membrane. The ionic conductivity was much improved to 10-2 S cm-1
due to the low crystallinity and the anion exchange between Cl- and OH- ions on the polymer
side chains. For the first time, the alkaline SPE employed in alkaline fuel cell, the test results
exhibited good performance and could tolerate at high temperature up to 120 oC.
The same concept was applied on the anion exchange membranes. Yu et al. [19] used the
AEM MORGANE-ADP membranes (Solvay, S.A.), a cross-linked fluorinated polymer with
the exchange group of quaternary ammonium, in direct methanol fuel cell (DMFC) and they
investigated the properties for preparing membrane electrode assembly. It was described that
the DMFC with anionic exchange membrane in alkaline media showed better performance
than in acid media. This was attributed to that methanol oxidation catalyst has been less
structure sensitive and it could approach to use the lower-cost non-precious metal catalysts.
Although the power density of the DMAFC with AEM was close to 10 mW cm-2, when
compared to Nafion membrane, it still showed higher resistance and needs to solve the main
hindrance in improving fuel cell performance.
To improve the disadvantages of common liquid electrolytes, gelled polymer electrolytes
(GPE) have ionic conductivity close to the aqueous electrolyte and have well contact with the
electrodes. But the poor mechanical property, high evaporation and easy leakage stalled the
wide use in battery systems. Feuillade [20] studied PAN gel polymer electrolyte with physical
crosslinking. Tsuchida et al. [21] prepared PVDF gel polymer electrolyte and found higher
ionic conductivity at 10-3 S cm-1 at ambient temperature. More recently, Girish [22] reported
PAN gelled polymer electrolyte with propylene (PC), ethylene carbonate (EC) and zinc
trifluoromethane sulfonate (ZnTfs). The ionic conductivity was about 210-3 S cm-1 and the
gelled polymer electrolyte exhibited good charge/discharge cyclibility for Zn|GPE|-MnO2
cell. The successful preparation of gelled polymer electrolyte required a cross-linking
procedure, physical or chemical [23]. However, physical crosslinking gel may encounter
serious problems at high temperature environment and under long-term storage, like
dissolution and phase separation. The gelled polymer electrolyte can be prepared from low
Developments of Alkaline Solid Polymer Electrolyte Membranes 449
molecular weight poly(vinyl alcohol), PVA and poly(acrylic acid), PAA. There exists an
interesting preparation method by the freezing/thawing techniques on PVA/PAA polymer
blends [24,25]. They can absorb water more than 5 times of the polymer weight, but until
now there have been no literature data for alkaline SPE preparation. For the chemical
crosslinking preparation, we have previously introduced a free radical polymerization method
for PAA. It requires a crosslinking agent with bi-functional or tri-functional groups to react
with PAA and to form a stable structure of gelled polymer. Iwakura et al. [26] used
commercial crosslinked PAA (Aldrich, #43532-5) to absorb KOH aqueous solution and
applied it on Ni-MH battery.
For the experimental results, the ionic conductivity value of gel polymer electrolytes has
been as high as that of the KOH aqueous solution (about 0.6 S cm-1). This is the typical
characteristic for gelled polymer electrolyte because the crosslinked structure can absorb and
hold a large amount of KOH aqueous solution. The higher ionic conductivity is good for the
battery application during high rate discharge.
Recently, some research has been focused on the preparation of solid-state supercapacitor
[27, 28]. For example, the sulfonated polypropylene (PP) separator was impregnated with
crosslinked PAA/KOH. The hydrogel polymer electrolytes have been used on the electric
double layer capacitor. The solid-state capacitor remained high discharge capacitance of 80 F
g-1 after 20,000 charge/discharge cycle. For low temperature tests, Zhang et al. [29] used PVA
SPE as gelled polymer electrolyte (=0.097 S cm-1) in the NiO/activated hybrid
supercapacitor. The system showed higher specific capacitance of 65 F g-1 at 40 oC and they
even obtained 30 F g-1 at low temperature of -20 oC.
80,000. Fig. 2 displays typical AC impedance spectra for the alkaline PVA/KOH SPE at
different temperatures ranging from -20 to 90 oC. The ionic conductivity is highly dependent
on the temperature and is followed by the Arrhenius equation: = o exp(-Ea/RT).
8
-20oC o
70 C
-10oC
7
0oC o
o 50 C
10 C 60oC
6 o
20 C o
90 C
o
30 C
o
5 40 C
o
50 C 20oC
-Zim''/ohm
o
60 C o
30 C
4 70oC
o
80oC 40 C
o
3 90oC 80 C
2
o
-20 C
O
1 -10 C
o
0C
10oC
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Zre / ohm
Z ''/ohm
re
The best way to verify the applicability of SPE is to assemble it in A real practical battery
system. The alkaline zinc manganese dioxide cell (Zn/MnO2) has many advantages of high
power density, longer shelf life, and better dimensional stability. It can be used under
continuous or intermittent demand, and has been advanced to the portable battery market. The
disadvantages that may limit its use in some devices are the slopping discharge curve and
lower energy density. The Zn/air battery uses oxygen directly from the atmosphere to produce
electrochemical energy, and is thus classified as semi-fuel cell. Much research has been
focused on this system due to its high energy density and flat discharge voltage. Aluminum
electrode provides another potential battery material because of its high theoretical Ah
capacity, high voltage and high specific energy. In this chapter, we assembled alkaline
aluminum manganese dioxide cell (Al/MnO2) and alkaline zinc or aluminum air cell (Zn/air,
Al/air) to investigate the applicability of these cells. Fig. 3 shows the solid state Zn/air cell
structure with alkaline SPE membrane.
The preparation procedure is briefly described for carbon-based air cathode and Zn gelled
anode. The carbon slurry for the gas diffusion layer was prepared with a mixture of 70 wt.%
Shawinigan acetylene black (AB50) and 30 wt.% PTFE (Teflon-30 suspension) as a binder.
The slurry was coated on a Ni-foam current-collector. The gas diffusion layer was sintered for
Developments of Alkaline Solid Polymer Electrolyte Membranes 451
about 30 min at 300 oC. The active layer of the air electrode was then sprayed with a mixture
of 15 wt.% PTFE binder, 1025 wt.% of KMnO4+MnO2 (2:1), 60 wt.% Vulcan XC-72R, and
appropriate amount of isopropyl alcohol onto the gas-diffusion layer. The air electrode with
both the gas diffusion layer and the active layer was finally sintered for about 30 min at 360
o
C for binding. The thickness of the air cathode electrode was about 0.4-0.6 mm. On the other
hand, Zn anode gel was prepared according to the following formulation: 65 wt.% Zn
powders, 0.5-1 wt.% Carbopol 940 gelling agent, 32 wt.% KOH, and some metal additives
used as hydrogen inhibitor.
Air cathode:
PVA-based alkaline solid
O2+ 2H2O + 4e ?4OH polymer electrolyte
EO = 0.40 V membrane
Figure 3. The scheme of a solid state Zn/air cell with alkaline PVA SPE membrane.
The prepared PVA/KOH/H2O SPE was employed for both Ni/MH and Zn/air batteries. Fig. 4
shows typical charge and discharge curves of all solid-state Ni/MH battery. The results
exhibited the advantage of flat plateau discharge curve and the battery had average 82%
current efficiency after ten cycles [33]. In addition, the PVA/KOH SPE was successfully
assembled into Zn/air battery with a high zinc utilization of 83%.
For the Ni/MH and Zn/air batteries with alkaline SPE under high rate or long-term
discharge, the water content might lose quickly and still cause deformation of SPE
membrane. Therefore, the composite PVA/KOH alkaline SPE with glass-fiber-mat (GF) was
developed to prevent the film from deformation and could enhance the mechanical properties
[34]. It was further evidenced that the reinforcing GF improved membrane dimensional
stability without significantly affecting its ionic conductivity value. We compared the
discharge curves of solid state Zn/air cells with PVA/KOH/GF SPE and commercial PP/PE
and cellulose separators at C/10. The results showed that at the high discharge rate, the Zn/air
cell with PVA/KOH/GF SPE could reach higher Zn utilization of 86% and higher plateau
voltage than those of the other commercial separators.
452 G.M. Wu, S.J. Lin and C.C. Yang
1500
1400
1300
1000
900
800
700
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
time (hrs)
1600
1500
1400
1300
1200
1100
E (mV)
1000
900
800
700
600
500
400
300
0 5 10 15 20 25 30 35 40 45 50 55
time (hrs)
Figure 4. The typical charge and discharge curves of solid-state Ni/MH battery with alkaline PVA-
based SPE (a) single cycle charge and discharge curve and (b) ten cycles of charge and discharge
curves.
The PEO-PVA-H2O alkaline SPE was further developed and applied in the Zn/air and
Ni/MH batteries [35]. It was found that the new alkaline SPE with the composition of PEO:
PVA = 2:8 had the higher ionic conductivity of 0.0608 S cm-1. For the Ni/MH battery, after
40 cycles of charge/discharge tests, the Ni(OH)2 active material still had an average capacity
density of 250 mAh g-1. However, the PVA-PEO blend could form a porous and brittle
structure membrane. This is not favorable for Zn/air battery application, because the zinc
dendrite would penetrate through the air cathode and cause short-circuit during the discharge.
Developments of Alkaline Solid Polymer Electrolyte Membranes 453
The composite model of PEO-PVA-H2O SPE with GF could be employed to improve the
stability [36]. The zinc utilization for Zn/air cell was achieved at 84% when discharged at 150
mA.
In this section, PVA was blended with polyepichlorohydrin (PECH) in DMSO solution to
prepare the PVA/PECH blend polymer membrane. The blend membrane was immersed in 6
M KOH aqueous solution to form the alkaline PVA/PECH SPE. It was improved in chemical,
mechanical, and electrochemical properties [37]. The optimal blend ratio of PVA:PECH was
found to be 1:0.2. This polymer blend formed a uniform and homogeneous film. High PECH
content, such as PVA:PECH (1:1), resulted in phase separation morphology. The solid-state
Zn/air batteries with PVA/PECH blend polymer electrolytes have been assembled and the test
results are listed in Table 2.
Table 2. The results of solid-state Zn-air cells with PVA-PECH solid polymer
electrolytes at C/10 discharge rate.
Zn-air cell
Item PVA-PECH PVA-PECH PVA-PECH
(1:0) SPE (1:0.2) SPE (1:0.5) SPE
Theo. capacity (mAh) 1,476 1,476 1,476
Discharge current (mA) 150 150 150
Discharge time (hr) 8.9 8.6 7.7
Real capacity (mAh) 1338 1290 1160
Zn utilization (%) 89 86 77
If the PECH content was further increased to the blend ratio of 1:0.5 to 1:1, phase
separation was observed and the film exhibited a sticky surface. These issues posed problems
for PVA/PECH SPE for alkaline battery system, although it had lower activation energy of 1-
3 kJ mol-1 and excellent mechanical strength of 58.9 MPa. As a result, it was found that the
blend ratio of 1:1 sample exhibited the lowest KOH content and thus low ionic conductivity.
The anionic exchange concept has been employed on PVA polymer electrolyte to improve the
ionic conductivity of alkaline PVA-based SPE. The solvent casting method started by
blending PVA and quaternary amine, such as tetraethyl ammonium chloride (TEAC) in
dimethyl sulfoxide (DMSO) to form PVA/TEAC blend polymer membrane [38]. After
immersed in 32 wt.% KOH solution, the alkaline PVA/TEAC SPE membranes were obtained.
When more TEAC was added, the crystalline intensity of PVA from XRD peak was greatly
reduced. It was suggested that the addition of TEAC into the PVA polymer matrix increased
the domain of amorphous region and then the hydrophilicity. The XRD spectra are evidenced
in Fig. 5.
454 G.M. Wu, S.J. Lin and C.C. Yang
500
400 (c)
300
(d)
200
100
10 20 30 40 50 60 70 80
2 (degree)
Blend ratio
T (oC) PVA:PECH PVA:PECH PVA:PECH
=1:0.2 =1:0.5 =1:1
30 0.0074 0.0042 0.0012
40 0.0081 0.0049 0.0018
50 0.0083 0.0053 0.0025
60 0.0091 0.0058 0.0029
70 0.0096 0.0066 0.0032
Blend ratios
T (oC) PVA/TEAC PVA/TEAC PVA/TEAC
= 1 0.2 =1 1 =1 2
30 0.0106 0.0219 0.0459
40 0.0138 0.0223 0.0462
50 0.0164 0.0233 0.0464
60 0.0201 0.0249 0.0480
70 0.0224 0.0259 0.0497
These SPE films also exhibited uniform and homogeneous surface. No phase separation
was observed for all the blend PVA/TEAC polymer films, and all SPE films appeared
translucent. As a result, the conductivity value of the alkaline PVA/TEAC SPE of PVA:
TEAC = 1:2 ( = 0.0459 S cm-1) was close to that of alkaline PVA polymer electrolyte at
Developments of Alkaline Solid Polymer Electrolyte Membranes 455
ambient temperature. The ionic conductivity of blend PVA/PECH and PVA/TEAC with
different blend ratios at different temperatures are displayed in Tables 3 and 4, respectively.
In order to obtain higher ionic conductivity and suitable mechanical strength for alkaline PVA
SPE, we need to further enhance its water absorbility and hydrophilicity. The preparation idea
was to blend two highly hydrophilic polymers, polyvinyl alcohol (PVA) and polyacrylic acid
(PAA) to form a cross-linked (interpenetrating) SPE membranes. The high hydrophilicity of
PAA is contributed by the strong hydrogen bonding between the pendant carboxylic acid (-
COOH) groups in the side chains [39,40]. PAA has low film-formability and is not popular
for polymer membrane applications. However, its high water absorbing ability and strong
water holding characteristic make it widely used as a super-absorbing material in the diaper,
medicine and agriculture industry [41]. Many efforts have been proposed to improve the
applicability of PAA for membranes: including grafting, crosslinking and blending. Habert
[42] reported grafting PAA onto polysulfone membrane for dialysis and desalination via
reverse osmosis. Ito et al. [43] grafted PAA onto the polycarbonate membranes to reduce the
pore size. For the preparation of crosslinked PAA, the chain mobility and relaxation mode
were studied [44]. Only few studies focused on the behavior and structure of PVA/PAA blend
polymers [45-48]. We synthesized cross-linked PAA and blended it with PVA to form the
IPN structure of PVA/PAA membrane.
For the preparation procedure, the acrylic acid monomer with cross-linker was firstly
blended with PVA polymer. A free radical polymerization was carried out. The solid-state
PVA/PAA polymer membranes were obtained by a solution casting method. The appropriate
weight ratios of acrylic acid monomer (PVA:PAA varied from 10:3 to 10:7.5 PAA) and 1
wt.% (vs. AA monomer) cross-linking agent, such as triallylamine, N,N-dimethyl acrylamide,
glutaraldehyde or N,N-methylene bisacryalmide (MBAC) [44,49], were directly dissolved
and mixed in distilled water with agitation for 12 h at 60 oC. The AA monomer solution was
added with 75 mol.% KOH (vs. AA monomer). The resulting mixture was then blended with
PVA polymer aqueous solution until completely homogeneous. The appropriate weight
percent (10 wt.% by AA monomer) of initiator, such as ammonium persulfate (APS),
potassium persulfate (KPS), 2,2-azobisisobutyronitrile (AIBN) or benzoyl peroxide (BPO),
was slowly added. The mixture solution for the free radical polymerization was under
continuous stirring condition at 90 oC for 2 h. The resulting homogeneous polymer solution
was then poured onto a PTFE or glass plate. The excess water solvent was evaporated slowly
at 80 oC in a vacuum oven. The resulting PVA/PAA blend membranes showed homogeneous
and semi-transparent structure.
The absorption of KOH electrolyte by the SPE membrane is very crucial, because the
higher KOH electrolyte content is beneficial for higher ionic conductivity. It would be
necessary to have higher concentration of hydroxyl ions (OH-) in polymer matrix. After PVA
was blended with PAA, the typical (101) crystalline peak of PVA at 2=19o was significantly
reduced [50,51]. The XRD spectra for the PVA/PAA membranes were shown in Fig. 6. This
is very helpful to enhance the ionic transport property since there is more amorphous domain
available for ion transport. It also improves the absorbility of KOH solution in PVA/PAA
membrane due to the higher hydrophilicity property of PAA.
456 G.M. Wu, S.J. Lin and C.C. Yang
1000
(a) Pure PVA film
(a)
(b) PVA : PAA = 10 : 3
(c) PVA : PAA = 10 : 5
(d) PVA : PAA = 10 : 7.5
800
600
(b)
400
(c)
200 (d)
0
10 15 20 25 30 35 40 45 50 55 60
2 / degree
The PVA/PAA blend membranes exhibited novel absorption behavior that is sensitive to
the aqueous media and the composition of blend polymer. Table 5 lists the absorption
percentage data of PVA/PAA polymer membranes under various solutions. It has been
evidenced that PVA:PAA=10:7.5 sample had very high absorption in 7 M KOH solution
(413%). It increased nearly ten times from that of PVA:PAA=10:0 sample (49%). This is
attributed to the IPN structure and hydrophilic groups in the PVA/PAA polymer system.
Table 5. The absorption (%) results for the PVA/PAA polymer membranes at room
temperature.
Blend ratios
Medium PVA:PAA PVA:PAA PVA:PAA PVA:PAA
(10:0) (10:3) (10:5) (10:7.5)
6 M CH3OH 37 103 127 206
2 M H2SO4 --- 161 164 224
7 M KOH 49 207 255 413
D.I. water 41 68 98 195
The higher absorption of KOH directly influenced the ionic conductivity of the
PVA/PAA SPE. Table 6 summarized the ionic conductivity data. Very good ambient
temperature ionic conductivity was achieved. The results also demonstrated that the IPN
structure had good electrolyte holding capability even at higher temperature. Water channels
may be provided for hydroxyl ion (OH-) to move from the cathode side to the anode side
during discharge [39, 49, 52].
Developments of Alkaline Solid Polymer Electrolyte Membranes 457
Blend ratios
T (oC) PVA : PAA PVA : PAA PVA : PAA PVA : PAA
(10:0) (10:3) (10:5) (10:7.5)
25 0.072 0.142 0.256 0.301
40 0.096 0.151 0.289 0.341
50 0.108 0.163 0.300 0.381
60 0.121 0.178 0.321 0.422
70 0.132 0.190 0.352 0.456
80 0.142 0.201 0.379 0.488
90 - 0.211 0.410 0.510
Cyclic voltammetry was used to evaluate the electrochemical stability of the alkaline
PVA/PAA SPE. The sweep potential was firstly scanned in the positive-going direction and
then reversed. Fig. 7 shows the cyclic voltammograms for the Zn| PVA/PAA| Zn symmetric
cell with the PVA/PAA SPE of 10:5 for cyclic sweeping in the 0.50 to + 0.50 V range. The
cathodic and anodic peaks are symmetrical and well-defined at the position between Eca=-
0.091V and Ean=0.091V. The reduction current density and the oxidation current density
increase apparently with increasing cycles. Fig. 8 shows the results with scanning rate from 1
to 50 mV s-1. The ions mass transport along the water channels in the polymer matrix is
unobstructed, facile and quite reversible [53]. On the other hand, the study on Al/air cell was
not as impressive. The magnitude of the peak reduction and peak oxidation current densities
decreased slightly with increasing sweeping cycle. A deactivation layer (oxide) might be
formed on Al electrode while the sweeping cycle was increased.
80
cycle 12
60
cycle 9
cycle 6
40 cycle 3
cycle 2
20
i (mA cm )
cycle 1
-2
-20
-40
-60
-80
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
E (V)
150
scan rate : 1 mV/s
scan rate : 5 mV/s
scan rate : 10 mV/s
100 50 mV/s
scan rate : 20 mV/s
scan rate : 50 mV/s
20 mV/s
50 10 mV/s
5 mV/s
i (mA cm-2)
1 mV/s
0
-50
-100
-150
-0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30
E (V)
Figure 8. The cyclic voltammograms for Zn| PVA/PAA| Zn symmetric cell with increasing scanning
rate.
1.6
1.4
1.2
1.0
-Zim / ohm
0.8
0.6
0.4
0.2
(a) PVA : PAA = 10 : 3
0.0 (b) PVA : PAA = 10 : 5
(c) PVA : PAA = 10 : 7.5
-0.2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Zre / ohm
a)
1.2
1.0
0.8
-Zim / ohm
0.6
0.4
0.2
b)
Figure 9. AC impedance spectra of Zn-air cells (a) before and (b) after C/10 discharge.
Developments of Alkaline Solid Polymer Electrolyte Membranes 459
It was found that the bulk resistance (Rb) of Zn/air cells with the different blend ratios of
PVA/PAA polymer electrolyte membrane was in the range of 0.21-0.26 ohm before the C/10
discharge, as shown in Fig. 9. However, the bulk resistance increased slightly to 0.26-0.28
ohm after the discharge.
Fig. 10 shows the electrochemical performance of solid-state Zn/air cell as a function of
discharge current. The results also includes the corresponding power density curves. The peak
power density of Zn/air cell (over 50 mW cm-2) is much higher than that of gelled polymer
Zn/air cell (only 6-7 mW cm-2) [54,55].
60
2.4
1.2
20
1.0
(c) PVA : PAA = 10 : 7.5
0.8
10
0.6 (b) PVA : PAA = 10 : 5
0.4
(a) PVA : PAA = 10 : 3 0
0.2
0.0
0 10 20 30 40 50 60 70 80 90 100
i (m A cm -2 )
Figure 10. The cell potential of Zn/air cells vs. discharge current density using the various PVA/PAA
SPE.
The ionic transport numbers for alkaline PVA-based SPE are very important because the high
anionic transport number of SPE can limit the carbonation problems [56]. The anionic
transport number (t-) was measured by the dynamic Hittorfs method [57]. A test cell, as
shown in Fig. 11, with two Pt electrodes was made for electrolysis and the electrolysis current
was imposed by a power supply. All PVA-based polymer electrolyte membranes were located
and fixed at two separated compartments with the same 1 M KOH solution. The reaction
occurred at the Pt-cathode, producing O2, H2 and OH- while consuming water and OH-. The
balance of OH- ions in each compartment led to OH- transport number after a fixed amount of
charge was passed through the polymer membranes.
After the electrolysis, the concentration of OH- ion in both compartments was determined
by titration using 1 M HCl stock solution. The cationic and anionic transport numbers could
be calculated by the following two equations:
460 G.M. Wu, S.J. Lin and C.C. Yang
t+ = nF / It (C0V0 C+V+)
t = nF / It (CV C0V0)
where n is the charge of OH ion, F is Faradays number, I is the applied current during
electrolysis, t is the electrolysis time, C0, C+, C are the OH ion concentration at the initial
and at the final of the electrolysis in the anodic and cathodic compartments, V0 is the initial
volume in each compartment, and V+ and V are the final volume in the anodic and cathodic
compartments, respectively.
Pt-cathode Pt-anode
H2 O2
2H2O+2e-?H2+2OH- 4OH-?O2+4e-+2H2O
SPE film
Figure 11. The experimental apparatus for dynamic transport number measurements.
Alkali solution
Type of membrane
LiOH NaOH KOH
PVA 0.82 0.85 0.95
PVA/Glass-fiber 0.82 0.84 0.95
PVA/PECH (1:1) 0.73 0.80 0.86
PVA/TEAC (1:2) 0.93 0.95 0.99
PVA/PAA (10:5) 0.92 0.96 0.99
PVA/PAA/PP/PE (10:5) 0.88 0.95 0.98
PP/PE (after sulfonation) 0.80 0.84 0.89
PP/PE separator 0.73 0.77 0.79
Developments of Alkaline Solid Polymer Electrolyte Membranes 461
1.00
LiOH(aq)
KOH(aq)
0.95 NaOH(aq)
0.90
0.85
-
t
0.80
0.75
0.70
0 1 2 3 4 5 6
-1
Concentration (mol L )
Figure 12. The alkali metal salt effect on ionic transport number for alkaline PVA solid polymer
electrolyte.
Experimental results indicate that the ionic transport property for the PVA-based SPE is
highly dependent on both the alkali salt and the solution concentration. All the values of
anionic transport numbers for different PVA-based SPEs with KOH solution are much higher
than those with alkaline NaOH and LiOH solutions. This trend is consistent with literature
results [58] that ion movement in polymer is related to polymer segmental motion, and the
order of ionic conductivity has been K+>Na+>Li+.
Recently, PVA-based polymers have been successfully prepared to form alkaline SPE
[59-62], and the experimental results confirmed the unique properties, such as high ionic
conductivity and high anionic transport number. These membranes consist of a polymer
matrix swollen with water and KOH electrolyte solution, and this may directly influence the
SPE on cell performance.
Non-woven separators play important role in the alkaline battery system [63]. It slightly
absorbs alkaline electrolyte solution and prevents the contact between anode and cathode,
which causes short-circuit of the battery. But the large pore size in the non-woven separator
cannot completely prevent Zn dendrite growth from penetrating through the other side, and
this may cause the battery system in danger. The higher KOH electrolyte content in these
separators may easily cause leakage of electrolyte or dry-out of electrolyte. This would limit
the shelf-life of battery. Although multi-layers of non-woven cloth separators may slightly
solve some problems, the thicker multi-layers could still result in less electrode active
materials which decreases electrochemical performance.
In the following section, we present a simple method to prepare high ionic conducting
membrane separators with lower cost non-woven polypropylene/polyethylene (PP/PE)
462 G.M. Wu, S.J. Lin and C.C. Yang
membrane through sulfonation [64-68]. The non-woven PP/PE separator membrane was in
sheet form and was composed of numerous filaments of 10-20 m in diameter. The filament
has a core-and-shell structure, with polypropylene as the core and polyethylene as the shell.
The sulfonation treatment was carried out using a highly concentrated sulfuric acid. FTIR
spectra are shown in Fig. 13 for the PP/PE membrane samples before and after sulfonation
treatment. Several additional transmittance bands appeared for the sulfonated membranes.
The peaks at around wavenumber of 11501200 cm-1 were from SO3- asymmetric stretching
modes [69]. The characteristic absorption bands at 550585 cm-1 can be assigned to the
deformation of SO bonds [70]. These band regions are in accordance with those for
alkanesulfonic acid, CSO3H. These peaks indicated that the sulfonic acid groups had been
successfully incorporated to the surface of polymer membrane. No significant change was
observed for the peaks at higher wavenumbers, such as 28003000 cm1, which were mainly
from CC and CH bonds. In addition, the sulfonation reaction might happen on both the
surface of polyethylene on the shell and the exposed polypropylene, originally in the core
[71,72].
5.0
(a) Sulfonation time 0 h
4.5 (b) Sulfonation time 3 h
(c) Sulfonation time 9 h
4.0 (d) Sulfonation time 72 h
3.5
Transmittance (arbitrary unit)
3.0
(d)
2.5
2.0 (c)
1.5
(b)
1.0
0.5
(a)
0.0 550~585 cm-1 1150~1200 cm
-1
-0.5
400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600
Wavenumber (cm-1)
The SEM micrograph of the highly sulfonated PP/PE membrane is shown in Fig. 14. The
non-woven membrane sheet had numerous 1020 m filaments that were bound together.
The uniform porosity provided adequate channels for the transport of ions. After the
sulfonation, the diameter of the filaments was slightly reduced. The surface polyethylene
layer was somewhat etched away. The longer the sulfonation time, the deeper the sulfonation
reaction could take place. The degree of sulfonation was thus enhanced by more sulfonic acid
groups grafted on the surface.
Developments of Alkaline Solid Polymer Electrolyte Membranes 463
300 m
Fig. 15 displays the KOH electrolyte absorption curves for the non-woven PP/PE
membranes after immersion in 32 wt.% KOH solution. These sulfonated non-woven
membranes can reach the equilibrium absorption within 1 h. The sulfonation treatment helped
increasing the percent of absorption from 100% to 185%. The AC impedance spectra for the
non-woven PP/PE membrane at 25oC with different sulfonation times are shown in Fig. 16.
220
200
(d)
180
(c)
160
Absorption ratio (%)
140
120
(b)
100
(a)
80
60
(a) sulfonation time 0 h
40 (b) sulfonation time 3 h
(c) sulfonation time 72 h
20 (d) sulfonation time 128 h
0
0 5 10 15 20 25 30 35 40
Immersion time (h)
Figure 15. The absorption curves of non-woven membranes using 32 wt.% KOH solution.
464 G.M. Wu, S.J. Lin and C.C. Yang
20
18
16
14
12
-Z''im / ohm
10
The values of bulk resistance for the PP/PE membrane separators were in the range of 1
3 ohm. The ionic conductivity was highly dependent on the sulfonation time. It was
evidenced that the ionic conductivity for the sulfonated PP/PE separator sample increased
with sulfonation time. As a result, the ionic conductivity of the highly sulfonated PP/PE
separator was as high as 0.0175 S cm-1. It was approximately two times that of the non-
sulfonated sample ( = 0.0088 S cm-1). The sulfonated non-woven PP/PE membranes
exhibited high ionic conductivity in the alkaline systems.
Although the alkaline PVA/PAA SPE has high ionic conductivity, it also exhibits lower
mechanical strength and large deformation after absorbing KOH electrolyte. On the other
hand, the sulfonated PP/PE separator membrane shows high hydrophilicity with moderate
ionic conductivity. Therefore, we tried to prepare a new class of composite polymer
electrolyte membranes based on sulfonated non-woven PP/PE sheets and PVA/PAA
polymeric blends. This composite polymer membrane would offer high mechanical strength
and good ionic conductivity. The composite membrane is composed of PP/PE separator
sheets as reinforcements that are coated with homogeneous polymer PVA/PAA blends. The
composite PVA/PAA polymer membranes were also immersed in 32 wt.% KOH solution to
become freestanding composite solid polymer electrolyte membranes [73].
The SEM micrographs for the PVA/PAA (10:5) composite polymer membranes with
highly sulfonated PP/PE separator are shown in Fig. 17. It has been found no phase
separation. The improvement in hydrophilicity of PP/PE separator produced no visible voids
on the surface and the cross-section. The binding between PP/PE and PVA/PAA polymer has
been very well.
Developments of Alkaline Solid Polymer Electrolyte Membranes 465
(a) (b)
Figure 17. SEM micrographs for the PVA/PAA composite polymer membrane reinforced with
sulfonated PP/PE separator (a) top surface and (b) cross-section.
150
cycle 1
cycle 10
cycle 20 cycle 100
100 cycle 30 increasing sweeping
cycle 40 cycle
cycle 50
cycle 60
50 cycle 70 cycle 10
cycle 80
cycle 90
i (mA cm )
cycle 1
-2
cycle 100
0
-50
-100
-150
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
E (V)
(a)
200
(a) s-PP/PE / PVA:PAA=10:3
(c)
150 (b) s-PP/PE / PVA:PAA=10:5
(c) s-PP/PE / PVA:PAA=10:7.5
(b)
100
(a)
50
i (mA cm-2)
-50
-100
-150
-200
-250
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5
E (V)
(b)
Figure 18. The cyclic voltammograms for (a) the Zn| PVA/PAA/PP/PE| Zn cell with PVA/PAA (10:5),
and (b) various composition of PVA/PAA at the 100th sweep cycle.
466 G.M. Wu, S.J. Lin and C.C. Yang
In the cyclic voltammetric study, the sweep potential was firstly scanned in the positive-
going direction and then reversed. Fig. 18 shows the cyclic voltammograms for the
PVA/PAA (10:5) composite solid polymer electrolyte membranes for cyclic sweeping in the
0.50 to + 0.50 V range on Zn| PVA/PAA/PP/PE| Zn symmetric cell at a scan rate of 10 mV
s-1. It can be clearly evidenced that even after 100 cycles, the cathodic and anodic peaks are
still symmetrical between Eca= -0.12V and Ean= 0.12V. This suggests high electrochemical
stability at zinc electrode. The entire sweeping curves are stable and smooth, and the
reduction current density and the oxidation current density apparently increase with
increasing sweeping cycle. In addition, the CV results at the 100th sweeping cycle suggest
higher reduction and oxidation current densities for the higher PAA content sample. The
symmetrical cathodic and anodic peaks would be obtained.
The solid-state Zn/air cell using only sulfonated PP/PE separator showed discharge
capacity of 728 mAh and lower Zn anode utilization. This indicates that during the discharge
process, the expansion of zinc anode may produce Zn dendrite that would easily penetrate
through the porous PP/PE separator and cause short circuit. However, the improved solid-
state Zn/air cell with the novel composite solid polymer electrolyte membrane shows the very
high discharge capacity of 1507 mAh. The zinc utilization is alos as high as 96%. This
indicates that the composite PVA/PAA/PP/PE SPE has uniform morphology that can
effectively impede Zn dendritic penetration and thus extend cell life. However, it is necessary
to keep in mind that the Zn/air or Al/air cell is a semi-open system. The cells electrolyte or
water can easily evaporate into the ambient during long-term operation. Nevertheless, it will
directly influence the composite PVA/PAA polymer membrane based very high PAA content
(PVA:PAA=10:7.5). The serious curling problem for this composite sample was probably
caused by the unbalanced tensile stress inside the membrane, while KOH electrolyte
evaporates and loses during high-rate or long-term discharge operation.
2.0 100
1.8
(b) s-PP/PE / PVA:PAA=10:5
80
1.6
Power density (mW cm )
-2
60
1.4
1.2 40
1.0
20
(b) s-PP/PE / PVA:PAA=10:5
0.8
(c) s-PP/PE / PVA:PAA=10:7.5
0
0.6 (a) s-PP/PE / PVA:PAA=10:3
0.4 -20
0 20 40 60 80 100 120 140 160
i (mA cm-2)
Figure 19. The electrochemical performance of Zn/air cells using composite PVA/PAA/PP/PE SPE.
Developments of Alkaline Solid Polymer Electrolyte Membranes 467
Figs. 19 and 20 show the electrochemical performance of the solid-state Zn/air cells and
Al/air cells, respectively. The results in cell power density curves are also displayed. It was
found that the solid-state Zn/air cell with the composite solid polymer electrolyte membrane
with composition ratio of PVA/PAA=10:5 had the highest power density of 90 mW cm-2. It is
about two times higher than that of the Zn/air cell assembled from PVA/PAA solid polymer
electrolyte (only about 40-50 mW cm-2). The value is also much higher than that of gelled
Zn/air cell (only 6-7 mW cm-2). In addition, the results of Al/air cell exhibited tremendous
improvement from 1-2 mW cm-2 to 70 mW cm-2. It can be attributed to the good binding
between sulfonated PP/PE and PVA/PAA polymer blends. Thus, the excellent dimensional
stability of composite solid polymer electrolyte membrane can sustain higher discharge
current density in battery applications. In addition, Table 8 summarizes the characteristic
properties of different alkaline PVA-based solid polymer electrolytes and commercial non-
woven PP/PE separator.
3.00 80
2.50 60
2.25
1.75
(c) s-PP/PE / PVA:PAA=10:7.5
E (V)
1.25
1.00 0
(b) s-PP/PE / PVA:PAA=10:5
(a) s-PP/PE / PVA:PAA=10:3
0.75
0.50 -20
(c) s-PP/PE / PVA:PAA=10:7.5
0.25
0.00 -40
0 10 20 30 40 50 60 70 80 90 100 110 120
-2
i (mA cm )
Figure 20. The electrochemical performance of Al/air cells using composite PVA/PAA/PP/PE SPE.
The applications of composite PVA/PAA/PP/PE SPE were also studied using Zn/MnO2
and Al/MnO2 cells. The electrochemical cell performance results as functions of discharge
current density are shown in Figs. 21 and 22. Unlike the semi-open system in metal-air cells,
the Zn/MnO2 or Al/MnO2 cell is a closed system without the water evaporation problem, if
the cell is firmly sealed. It has been found that the power density of the Zn/MnO2 and
Al/MnO2 cells is increased when the PVA/PAA blend ratio is increased. This was due to the
high ionic conductivity of the alkaline composite PVA/PAA/PP/PE solid polymer electrolyte
membranes. For Al/MnO2 cell, it was found that the cell had the highest power density of 110
mW cm-2 at 120 mA cm-2. This particular cell was assembled with PVA/PAA (10:7.5)
468 G.M. Wu, S.J. Lin and C.C. Yang
membrane and its cell performance of Al/MnO2 cell is better than that of the Zn/MnO2.
Aluminum material has very high theoretical capacity, high cell voltage, and high specific
energy density. These solid-state alkaline cells with alkaline composite PVA/PAA/PP/PE
solid polymer electrolyte membranes indeed exhibit excellent electrochemical performance.
2.6 100
2.4
2.2 80
1.2
20
1.0 (c) s-PP/PE/PVA:PAA=10:7.5
0.8
(a) s-PP/PE/PVA:PAA=10:3
0
0.6
(b) s-PP/PE/PVA:PAA=10:5
0.4 -20
0.2
0.0 -40
0 10 20 30 40 50 60 70 80 90 100 110 120
-2
i (mA cm )
Figure 21. The electrochemical performance of solid-state Zn/MnO2 cells using composite
PVA/PAA/PP/PE SPE.
3.0 120
2.8 110
(c) s-PP/PE / PVA:PAA=10:7.5
2.6
100
2.4
2.2 90
(mW cm2-2))
Power density (mW/cm
2.0 80
1.8 70
(b) s-PP/PE / PVA:PAA=10 : 5
1.6
E (V)
60
1.4
1.2 50
(a) s-PP/PE / PVA:PAA=10 : 3
1.0 40
0.8 30
(c) s-PP/PE / PVA:PAA=10 : 7.5
0.6
20
0.4 (b) s-PP/PE / PVA:PAA=10 : 5
0.0 0
0 20 40 60 80 100 120 140 160 180 200 220
i (mA cm-2)
Figure 22. The electrochemical performance of solid-state Al/MnO2 cells using composite
PVA/PAA/PP/PE SPE.
Developments of Alkaline Solid Polymer Electrolyte Membranes 469
Property
Utilization Max.
Mechanical
SPE type t- Zn/air Zn/air cell power
strength
(S cm-1) (1M KOH) at C/10 density
(MPa)
(%) (mW cm-2)
PVA
0.047 0.95 3.8 82 30
(PVA:KOH=30:40)
PVA/Glass-fiber 0.047 0.92 9.5 84 30
3. Conclusion
The preparation and the characteristic properties of different alkaline solid polymer
electrolytes have been presented and discussed. The PEO-based SPE was firstly proposed and
used for the alkaline SPE. Although the ionic conductivity can be improved by increasing
KOH content, he brittle characteristics and porous structure may limit its practicability in
electrochemical applications. The alkaline PVA-based polymer SPE has been successfully
introduced and developed since 2000. When carefully prepared, the ionic conductivity can be
as high as 0.301 S cm-1 for the PVA/PAA solid polymer electrolyte. The alkaline composite
solid polymer electrolyte membrane (CSPEM) based on sulfonated PP/PE separator and
PVA/PAA may solve the problems in deformation, shrinkage or expansion. It still shows high
ionic conductivity and good mechanical strength. For Zn/air battery application, the power
density was as high as 90 mW cm-2, and thus the PVA-based composite SPE has high
potential for use in alkaline battery system. The gelled polymer electrolyte (GPE) has the
highest ionic conductivity of 0.60 S cm-1 at room temperature, which is close to KOH
aqueous solution. For the cell with GPE application, it can remain high cell performance,
even at long-term charge/discharge cycle on Ni-MH battery and supercapacitor. The major
issues for GPE lie in low mechanical strength, low thermal stability and high aqueous
evaporation and leakage rates. The electrochemical cell may require firm sealing or packing
for applications.
470 G.M. Wu, S.J. Lin and C.C. Yang
References
[1] Fenton, D. E.; Parker, J. M.; Wright, P. V. Polymer 1973, 14, 589.
[2] Wright, P. V. Brit. Polym. J. 1975, 7, 319.
[3] Wright, P. V. Electrochim. Acta 1998, 43, 1137.
[4] Armand, M. B.; Chabagno, J. M.; Duclot, M. J. Fast ion transport in solids; Duclot, M.
J.; Vashishta, P.; Mundy, J. M.; Shenoy G. K.; Ed, Electrodes and Electrolytes,
1979, 131.
[5] Song, J.Y.; Wang, Y. Y.; Wan, C. C. J. Power Sources 1999, 77, 183.
[6] Zahurak, S. M.; Kaplan, M. L.; Rietman, E. A.; Murray, D. W.; Cava, R. J.
Macromolecules 1988, 21, 654.
[7] Vallee, A.; Besner, S.; Prudhomme, J. Electrochim Acta 1992, 37, 1579.
[8] Fauvarque, J. F.; Guinot, S.; Bouzir, N.; Salmon, E.; Penneau, J. F. Electrochim. Acta
1995, 40, 2449.
[9] Guinot, S.; Salmon, E.; Penneau, J. F.; Fauvarque, J. F. Electrochim Acta 1998,43, 1163.
[10] Guinot, S.; Salmon, E.; Penneau, J. F.; Fauvarque, J. F. J. Appl. polym. Sci. 1997,
65, 601.
[11] Vassal, N.; Salmon, E.; Fauvarque, J. Electrochim Acta 2000, 45, 1527.
[12] Vassal, N.; Salmon, E.; Fauvarque, J. F. J. Electrochemical Society 1999, 146, 20.
[13] Agel, E.; Bouet, J.; Fauvarque, J. F.; Yassir, H. Ann. Chim. Sci. Mat. 2001, 26, 59.
[14] Lewandowski, A.; Skorupska, K.; Malinska, J. Solid State Ionics 2000, 133, 265.
[15] Wu, G. M.; Lin, S. J.; Yang, C. C. J. Membr. Sci. 2006, 275, 127.
[16] Hassan, M. F.; Arof, A. K. Phys. Stat. Sol. (a) 2005, 202, 2494.
[17] Gamby, J.; Tabema, P. L.; Simon, P.; Fauvarque, J. F.; Chesneau, M. J. Power Sources
2001, 101, 109.
[18] Lewandowski, A.; Zajder, M.; Frackowiak, E.; Beguin, F. Electrochim. Acta 2001, 46,
2777.
[19] Yu, E. H.; Scott, K. J. Power Sources 2004, 137, 248.
[20] Feuillade, G..; Perche, P. J. Appl. Electrochem. 1975, 5, 63.
[21] Tsuchida, E.; Ohno, H.; Tsunemi, K. Electrochim. Acta 1983, 28, 591.
[22] Kumar, G. G.; Sampath, S. Solid State Ionics 2003, 160, 289.
[23] Murata, K.; Izuchi, S.; Yoshihisa, Y. Electrochim. Acta 2000, 45, 1501.
[24] Hickey, A. S.; Peppas, N. A. Polymer 1997, 38, 5931.
[25] Ricciardi, R.; Auriemma, F.; Gaillet, C.; Rosa, G. D.; Laupretre, F. Macromolecules
2004, 37, 9510.
[26] Iwakura, C.; Nohara, S.; Furukawa, N.; Inoue, H. Solid State Ionics 2002, 148, 487.
[27] Kalpana, D.; Renganathan, N. G.; Pitchumani, S. J. Power Sources 2006, 157, 621.
[28] Wada, H.; Nohara, S.; Furukawa, N.; Inoue, H.; Sugoh, N.; Iwasaki, H.; Morita, M.;
Iwakura, C. Electrochim. Acta 2004, 49, 4871.
[29] Yuan, C.; Zhang, X.; Wu, Q.; Gao, B. Solid State Ionics 2006, 177, 1237.
[30] Hodge, R. M.; Edward, G. H.; Simon, G. P. Polymer 1996, 37, 1371.
[31] Martinelli, A.; Matic, A.; Jacobsson, P.; Borjesson, L.; Navarra, M. A.; Fernicola, A.;
Panero, S.; Scrosati, B. Solid State Ionics 2006, in press.
[32] Bunn, C. W. Nature 1948, 161, 929.
[33] Yang, C. C.; Lin, S. J. J. Appl. Electrochem. 2003, 33, 777.
Developments of Alkaline Solid Polymer Electrolyte Membranes 471
alcohols, 212
A algorithm, x, 162, 169, 170, 180, 186, 201, 205, 274,
300, 301, 302, 303, 304, 318, 320, 361, 376
access, 249, 285, 401, 414
alkaline media, 254, 256, 263, 264, 269, 448
accounting, 165, 297, 301, 358
alkane, 159
accuracy, 12, 13, 66, 303, 320
alloys, 220, 248, 255, 263, 264, 266, 267, 269, 382,
acetic acid, 147
386, 387, 390, 415
acetone, 416, 427
alternative(s), viii, 10, 121, 135, 136, 137, 148, 157,
achievement, 5
223, 224, 233, 236, 241, 247, 249, 279, 297, 348,
acid, viii, ix, x, 10, 11, 135, 136, 137, 138, 139, 140,
412, 437
142, 143, 144, 145, 146, 147, 148, 149, 150, 151,
alternative hypothesis, 121
152, 153, 154, 156, 159, 163, 173, 220, 221, 223,
aluminum, 220, 286, 450
224, 225, 226, 231, 233, 235, 236, 237, 247, 248,
ambient air, 237, 326, 368
255, 256, 264, 266, 269, 285, 361, 382, 383, 385,
amines, 448
388, 391, 392, 394, 395, 396, 399, 401, 403, 404,
ammonia, 175
408, 409, 411, 414, 415, 416, 419, 420, 425, 428,
ammonium, 159, 397, 398, 448, 453, 455
440, 447, 448, 449, 455, 462
anion, 448
acidity, 137, 151, 156, 384
anisotropy, 221
acrylic acid, 449, 455
ANOVA, 76, 79
activation, x, 14, 73, 80, 93, 95, 96, 105, 107, 109,
aqueous solutions, 264, 417, 418
110, 111, 113, 114, 115, 203, 264, 265, 274, 289,
aromatics, 157
290, 291, 292, 295, 299, 300, 302, 303, 304, 311,
Arrhenius equation, 450
314, 317, 318, 337, 348, 360, 376, 385, 453
asbestos, 248
activation energy, 114, 314, 337, 453
aspect ratio, 221, 232, 383, 406, 427, 431
active site, 256, 401, 405, 415
assessment, 322, 348
adaptation, 79
assignment, 175, 207, 268
additives, 401, 451
assumptions, 94, 128, 169, 292, 305, 355
adhesion, 227, 252
atmospheric pressure, 154, 164, 165
adjustment, 198, 321
atoms, 147, 219, 256, 259, 269, 383, 385, 394, 397,
ADP, 448
399, 423, 430, 437
adsorption, 4, 96, 208, 254, 255, 256, 263, 264, 265,
attention, 10, 11, 73, 77, 80, 137, 291, 294, 381, 412
268, 269, 387, 413, 416, 419, 427
automobiles, 10, 214
aerospace, 175
autonomy, 212, 213
agent, 251, 382, 383, 390, 391, 421, 449, 451, 455
availability, 66, 72, 75, 192, 314
aggregation, 383, 420, 421, 432
aggregation process, 383
agriculture, 455 B
AIBN, 455
alcohol, xi, 233, 251, 412, 445, 447, 449, 451, 455 bandwidth, 176
474 Index
complement, 179 133, 165, 204, 291, 294, 297, 298, 314, 325, 326,
complexity, 113, 164, 216, 223, 247, 249, 269, 294, 338, 363, 368, 403
361, 413 contaminants, 222
complications, 249, 269 contamination, 175
components, vii, 3, 4, 9, 10, 11, 18, 19, 20, 21, 29, continuity, 297, 298, 305, 309, 355, 359
30, 31, 34, 44, 53, 54, 65, 98, 149, 150, 165, 179, control, 5, 18, 37, 72, 76, 78, 79, 80, 88, 113, 133,
180, 183, 192, 194, 200, 211, 213, 219, 222, 230, 139, 153, 154, 162, 163, 166, 167, 169, 170, 171,
279, 283, 288, 305, 313, 358, 376, 402 172, 173, 175, 176, 179, 180, 183, 184, 185, 192,
composites, xi, 158, 388, 445 198, 200, 201, 202, 204, 205, 206, 207, 208, 241,
composition, 72, 94, 95, 102, 103, 137, 162, 164, 250, 262, 265, 266, 301, 313, 316, 359, 360, 394,
171, 173, 174, 201, 236, 249, 255, 269, 306, 307, 400, 408, 409, 426
318, 413, 423, 434, 449, 452, 456, 465, 467 convergence, 300, 317, 318, 320
compounds, 157, 174, 175, 220, 413, 415 conversion, vii, 10, 75, 165, 174, 203, 278, 279, 302
computation, 12, 179, 207 cooling, 12, 83, 88, 89, 90, 216, 286, 300, 301
computational fluid dynamics, 300, 376 copolymers, 157, 159
computational grid, 301 copper, 12, 232
computers, 11, 211 correlation, 91, 92, 96, 109, 110, 113, 308, 309, 357,
computing, 162, 179, 301, 304, 317 358
concentration, vii, 4, 9, 12, 14, 18, 21, 22, 23, 25, 26, correlations, 91, 292
27, 31, 32, 35, 36, 37, 48, 51, 53, 54, 56, 57, 67, corrosion, x, 175, 230, 232, 233, 254, 286, 381, 383,
95, 111, 139, 144, 145, 147, 150, 152, 164, 203, 392, 402, 403, 404, 405, 406, 409, 411, 412, 415,
221, 258, 259, 260, 262, 265, 268, 274, 275, 277, 447
284, 285, 289, 290, 291, 292, 294, 295, 297, 298, costs, 136, 361, 413
300, 301, 302, 303, 314, 319, 320, 322, 325, 326, coupling, 93, 167, 168, 376
338, 348, 359, 360, 368, 376, 385, 417, 418, 420, coverage, 256, 257, 261, 263, 265, 266, 396
440, 447, 448, 455, 459, 460, 461 cristallinity, 139, 144, 153
conception, 213, 214, 216, 245 crystal structure, 449
condensation, 141, 175, 220, 275, 300, 329, 355, crystalline, 406, 413, 417, 419, 420, 422, 428, 453,
357, 360, 363, 364, 366, 368, 369, 376 455
conditioning, 75, 170, 176, 227 crystallinity, 447, 448, 449
conduction, 145, 284, 286, 299, 314 current limit, viii, 9, 15, 17, 54
conductivity, viii, x, xi, 58, 135, 136, 137, 142, 144, CVD, 217, 218, 219, 229
145, 147, 148, 149, 150, 151, 152, 153, 154, 156, cycles, 162, 163, 200, 207, 208, 209, 287, 294, 406,
159, 180, 232, 233, 254, 275, 277, 285, 286, 287, 447, 451, 452, 457, 466
291, 292, 297, 299, 301, 305, 308, 309, 310, 312, cycling, 5
319, 320, 354, 364, 367, 383, 388, 389, 390, 391,
393, 406, 411, 412, 413, 415, 426, 431, 437, 445,
446, 447, 448, 449, 450, 451, 452, 453, 454, 455, D
456, 457, 461, 464, 467, 469
damping, 176, 177
conductor, 145, 226, 305
danger, 461
confidence, 108, 109
data collection, 79
configuration, 148, 163, 266, 268
DBP, 416, 427
Congress, 132
decay, 4, 5, 267, 401, 402, 405, 406, 409
connectivity, 225
decomposition, 143, 144, 147, 383, 397, 398, 406
conservation, 165, 355, 356
decomposition temperature, 398
constant rate, 186, 198
deconvolution, 417
constraints, 310
defects, 4, 383, 404
construction, 291
definition, 176, 214
consumers, 222
deformation, 222, 347, 348, 451, 462, 464, 469
consumption, vii, viii, 9, 10, 20, 30, 53, 67, 71, 72,
degradation, 75, 113, 137, 138, 142, 144, 157, 287,
73, 76, 77, 78, 79, 80, 81, 82, 83, 84, 86, 87, 88,
409, 419
89, 90, 91, 92, 93, 94, 102, 103, 104, 110, 113,
dehydration, 145, 149, 294, 299
delivery, 165, 222, 224, 236
476 Index
demand, 170, 176, 186, 194, 195, 197, 198, 199, 104, 107, 108, 109, 110, 111, 113, 114, 151, 161,
200, 201, 202, 203, 302, 449, 450 162, 163, 171, 179, 180, 186, 188, 192, 193, 194,
Denmark, 133, 206 195, 196, 198, 200, 201, 202, 205, 206, 224, 231,
density values, viii, 71, 73 250, 274, 295, 298, 300, 302, 303, 304, 310, 311,
Department of Energy, 206 318, 325, 326, 327, 329, 331, 333, 334, 335, 337,
dependent variable, viii, 71, 72, 73, 76, 77, 78, 79, 338, 339, 341, 342, 344, 345, 347, 348, 350, 355,
80, 81, 113, 115, 116 360, 367, 368, 369, 370, 371, 372, 373, 374, 375,
depolarization, 11 376, 408, 412, 414, 418, 430, 449
deposition, 153, 214, 216, 217, 218, 219, 220, 222, divergence, 305
227, 240, 253, 254, 285, 348, 383, 384, 385, 388, DMF, 139, 146, 147, 148, 150
389, 390, 393, 402, 406, 412, 413, 415, 420, 421, doping, viii, 135, 136, 149, 216
422, 427, 429, 434, 437 dry ice, 251
deposition rate, 218 drying, 174, 249, 321, 331, 354, 367, 372, 398
deposits, 175 durability, 3, 5, 247, 347, 405, 406, 407, 409
designers, 13, 291, 295, 297, 303 duration, 399
desorption, 4, 96 dynamical properties, 169
deviation, 81, 83, 85, 86, 169, 171, 172, 179, 202, dynamical systems, 194
204, 313
devolatilization, 174
dialysis, 455 E
dielectrics, 218
education, 274
differential equations, 165, 193, 298
election, 201
diffraction, 416, 417, 418, 419, 421, 428, 429, 438
electric conductivity, 312
diffusion, ix, 12, 13, 14, 93, 95, 96, 97, 98, 99, 100,
electric current, 446
101, 102, 107, 111, 113, 115, 216, 232, 236, 247,
electric field, 314
249, 250, 251, 258, 259, 260, 262, 263, 265, 266,
electrical conductivity, x, 254, 286, 411, 415, 431
267, 268, 269, 275, 277, 281, 282, 283, 284, 285,
electrical power, 193, 194, 202, 203
286, 287, 290, 294, 295, 297, 298, 299, 300, 301,
electrical resistance, 240
303, 304, 305, 306, 309, 310, 312, 314, 315, 316,
electricity, vii, 9, 10, 18, 148, 162, 163, 168, 173,
318, 320, 326, 329, 335, 337, 338, 342, 347, 348,
192, 208, 219, 278
354, 355, 357, 358, 359, 360, 363, 364, 366, 367,
electrocatalysis, 249, 408
368, 370, 371, 372, 376, 387, 408, 409, 450, 451
electrocatalyst, 227, 249, 250, 254, 255, 409, 412,
diffusion region, 297, 301
413
diffusivities, 107, 310
electrochemical impedance, 267
diffusivity, 96, 100, 114, 312, 315, 326, 338, 359,
electrochemical reaction, vii, 9, 13, 78, 96, 156, 250,
361, 368
254, 283, 286, 288, 289, 291, 295, 301, 304, 314,
dilute gas, 98
323, 325, 331, 337, 355, 357, 364, 376, 382, 425,
dimensionality, 318
432
dimethylformamide, 139
electrochemistry, 154, 156, 303, 407, 437
dimethylsulfoxide, 139
electrode/electrolyte interfaces, 13, 15
direct measure, 103
electrodeposition, 217, 253
disaster, ix, 273, 280
electrodes, vii, ix, x, 4, 5, 10, 12, 17, 21, 53, 96, 100,
disclosure, 81
106, 156, 163, 173, 213, 221, 222, 224, 225, 226,
discontinuity, 183, 400
227, 228, 230, 232, 237, 247, 248, 249, 250, 251,
discretization, 320
252, 253, 254, 263, 264, 265, 266, 267, 268, 269,
dispersion, x, 4, 152, 233, 254, 381, 385, 388, 404,
285, 291, 297, 301, 302, 304, 305, 310, 314, 335,
406, 407, 414, 440
337, 338, 355, 382, 385, 386, 387, 389, 391, 393,
displacement, 347, 348, 351, 353
394, 395, 396, 398, 399, 404, 406, 407, 408, 409,
dissociation, 136, 145, 259
410, 411, 414, 415, 417, 425, 426, 427, 428, 433,
dissolved oxygen, 295
440, 448, 459
distilled water, 251, 416, 427, 455
electrolysis, 237, 459, 460
distribution, viii, ix, x, 4, 19, 20, 29, 71, 72, 73, 74,
electrolyte, vii, viii, ix, xi, 4, 5, 10, 11, 13, 14, 15,
77, 79, 80, 81, 82, 84, 87, 88, 89, 90, 92, 102, 103,
41, 58, 78, 106, 114, 135, 136, 142, 144, 149, 152,
Index 477
159, 163, 173, 180, 233, 236, 247, 248, 249, 250, ethylene, 415, 416, 428, 448
255, 264, 265, 266, 268, 269, 280, 285, 291, 301, ethylene glycol, 415, 416, 428
314, 361, 367, 385, 386, 387, 400, 407, 408, 409, Europe, 214
412, 427, 433, 445, 446, 447, 448, 449, 451, 453, evacuation, 437
454, 455, 456, 459, 460, 461, 463, 464, 466, 467, evaporation, 219, 220, 275, 277, 300, 303, 357, 359,
468, 469 360, 362, 363, 364, 376, 447, 448, 467, 469
electron, 10, 139, 154, 155, 220, 239, 286, 289, 290, evolution, 225, 258, 262, 263, 294, 296
291, 314, 337, 388, 399, 401 experimental condition, 103, 249, 258, 268
electron microscopy, 388 exposure, 217
electronic structure, 383 expulsion, 389
electrons, 4, 106, 154, 155, 156, 163, 174, 203, 225, extraction, 416, 427
276, 281, 283, 284, 285, 289, 337, 382, 383, 385, extrapolation, 262, 269
387, 388, 426
electroplating, 219, 220
embossing, 222 F
emergence, 212, 214
fabrication, ix, xi, 4, 158, 211, 212, 213, 214, 216,
emission, 163
225, 226, 227, 228, 229, 231, 236, 237, 240, 250,
encouragement, 274
252, 269, 273, 303, 397, 445
endothermic, 84, 88, 90, 282
failure, 302, 331, 372, 401, 405
energy, vii, viii, ix, 9, 10, 11, 12, 13, 16, 17, 18, 20,
feature selection, 201
21, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35,
feedback, 169
36, 37, 39, 40, 41, 42, 43, 44, 46, 47, 48, 50, 51,
fibers, 4, 383, 407
53, 55, 56, 57, 59, 60, 62, 63, 64, 65, 66, 67, 75,
filament, 462
76, 97, 114, 135, 154, 156, 161, 162, 165, 173,
fillers, 149
180, 195, 196, 212, 213, 218, 221, 223, 231, 247,
film, 151, 216, 217, 219, 220, 224, 225, 226, 229,
273, 275, 278, 279, 281, 288, 289, 298, 302, 306,
232, 241, 249, 251, 252, 265, 269, 406, 408, 410,
313, 314, 331, 337, 355, 356, 359, 381, 412, 426,
417, 427, 428, 437, 447, 449, 451, 453, 454, 455,
435, 447, 450, 453, 468
456, 460
energy consumption, vii, 9, 20, 53
film thickness, 449
energy density, vii, 9, 11, 16, 17, 18, 21, 24, 25, 26,
films, 214, 216, 217, 218, 219, 226, 229, 454
27, 28, 29, 30, 31, 32, 33, 34, 35, 39, 41, 42, 43,
filters, 76, 214
44, 46, 47, 50, 51, 53, 56, 57, 59, 60, 62, 63, 64,
filtration, 252, 393
65, 67, 223, 450, 468
financial performance, viii, 161
energy efficiency, vii, viii, 9, 10, 11, 13, 17, 21, 24,
flexibility, 137, 148, 162, 447
25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 39,
floating, 266
40, 41, 42, 43, 44, 46, 47, 50, 51, 53, 55, 56, 57,
flocculation, 252
59, 60, 62, 63, 65, 67
flood, 354, 367
England, 205, 407, 410
flooding, 236, 286, 295, 299, 300, 301, 321, 392,
entropy, 276, 288, 311, 331
396, 397, 401, 407
environment, ix, 72, 135, 222, 254, 264, 265, 269,
flow field, x, 230, 232, 233, 273, 274, 286, 287, 298,
270, 273, 287, 291, 302, 303, 331, 382, 391, 393,
299, 300, 301, 303, 305, 317, 318, 322
448
fluid, ix, 83, 88, 89, 104, 273, 298, 300, 301, 302,
environmental impact, ix, 163, 273, 302
303, 316, 376
epitaxial growth, 219
fluidized bed, 174
equilibrium, 83, 87, 89, 159, 193, 195, 197, 203,
fluorine, 437, 441
257, 259, 260, 265, 289, 290, 291, 310, 313, 314,
focusing, 137
355, 356, 359, 416, 463
foils, 230, 231, 233, 241
equipment, 16, 18, 34, 175, 302
forgetting, 171, 202
ester, 420, 437
formaldehyde, 383, 388, 390, 393, 428
estimating, 126, 206
fossil, 10, 76, 205, 281
etching, 214, 216, 217, 220, 221, 222, 227, 230, 233,
fossil fuels, 10, 281
237, 239, 240
fouling, 449
ethanol, 223, 224, 227, 235, 237, 251
France, 133, 211, 214, 377, 378, 441, 443
478 Index
free energy, 154, 275, 288, 289 429, 434, 436, 437, 439, 440, 441, 449, 455, 456,
freedom, 120, 121, 122, 126, 128, 129, 130 462
freezing, 449 growth, 211, 219, 402, 404, 461
FTIR, 440, 462 growth rate, 219
FT-IR, 420, 434, 437, 438 guidance, 5
FT-IR, 440 guidelines, 348
FT-IR, 441 gyroscope, 214
fuel efficiency, 148
functionalization, x, 150, 381, 405, 407
H
G Hawaii, 133
H-bonding, 147, 437
gallium, 219 HE, 139
gas diffusion, 12, 13, 249, 250, 251, 266, 267, 268, heat, vii, ix, x, 4, 9, 10, 11, 18, 27, 41, 75, 76, 80, 88,
277, 282, 283, 285, 286, 287, 294, 295, 297, 298, 154, 155, 162, 163, 168, 173, 174, 203, 213, 216,
299, 300, 301, 303, 304, 305, 309, 310, 314, 316, 250, 254, 273, 275, 276, 281, 282, 286, 299, 300,
318, 335, 342, 348, 354, 355, 357, 359, 363, 364, 301, 302, 303, 304, 309, 310, 311, 316, 331, 332,
366, 367, 368, 371, 372, 376, 408, 409, 450, 451 335, 347, 359, 360, 364, 372
gas phase, 174, 255, 276, 277, 295, 310, 329, 355, heat capacity, 275, 309
356, 357, 358, 359, 360, 363, 366, 376 heat release, 301
gases, 95, 98, 134, 165, 174, 200, 203, 204, 212, heat removal, 286
218, 219, 221, 237, 250, 276, 284, 285, 286, 289, heat transfer, 88, 216, 276, 299, 300, 303, 304, 310,
297, 298, 299, 300, 301, 304, 305, 355, 376, 392, 359
416, 437 heating, 115, 175, 203, 204, 220, 275, 301, 406
gasification, 162, 173, 174, 175, 201, 208 height, 318, 418, 419
gasoline, 136 high power density, 230, 232, 331
gel, 12, 150, 154, 251, 447, 448, 449, 451 hip, 216
generation, viii, ix, 19, 20, 75, 153, 161, 162, 173, homogeneity, 81, 90, 92
188, 192, 196, 201, 205, 206, 207, 227, 248, 276, Hong Kong, 225
278, 297, 301, 302, 303, 311, 331, 335, 347, 372, hot pressing, 252
382, 405, 407 humidity, 5, 11, 16, 18, 144, 145, 153, 277, 294, 297,
Georgia, 205, 206 298, 299, 301, 302, 303, 313, 319, 355, 359, 360,
germanium, 218 363
Germany, 75, 206, 207, 243 hybrid, 11, 150, 151, 153, 159, 161, 162, 163, 168,
glass, xi, 142, 214, 217, 222, 225, 227, 228, 232, 170, 176, 201, 206, 207, 208, 252, 449
241, 445, 451, 455 hydrazine, 390
glass transition, 142 hydrides, 212, 223
glass transition temperature, 142 hydrocarbons, 11, 255, 278, 279
global markets, 11 hydrofluoric acid, 220
glycol, 415, 416, 428 hydrogen, vii, ix, 4, 10, 96, 103, 114, 115, 132, 135,
goals, 5, 413 139, 142, 144, 147, 148, 154, 156, 157, 163, 174,
gold, 220, 221, 225, 226, 227, 230, 231, 232, 241, 178, 182, 202, 203, 223, 236, 247, 249, 250, 255,
265 256, 258, 259, 260, 261, 262, 263, 264, 265, 266,
graph, 87, 103, 431 267, 268, 274, 275, 276, 281, 283, 285, 292, 295,
graphene sheet, 383, 403 297, 299, 306, 307, 311, 314, 315, 318, 322, 326,
graphite, x, 12, 20, 233, 286, 287, 305, 310, 313, 338, 360, 366, 368, 369, 370, 382, 383, 387, 391,
411, 412, 414, 415, 426, 440 394, 397, 402, 405, 406, 408, 413, 449, 451, 455
gravimetric analysis, 395, 396 hydrogen atoms, 147, 256, 394, 397
groups, x, 136, 137, 138, 139, 142, 143, 144, 145, hydrogen bonds, 144, 147
147, 148, 149, 150, 151, 152, 153, 156, 157, 217, hydrogen gas, 281
236, 237, 285, 381, 382, 383, 384, 385, 391, 392, hydrogen peroxide, 236, 405, 406
393, 394, 395, 396, 397, 398, 399, 401, 403, 404, hydrolysis, 153, 393
405, 406, 407, 409, 414, 415, 416, 420, 422, 424, hydrophilicity, 137, 142, 148, 151, 453, 455, 464
Index 479
lithography, 216, 217, 221, 229, 232 microfabrication, ix, 211, 213, 214, 216, 217, 221,
location, 93, 165, 176, 200, 288, 395 224, 225, 231, 236, 237, 240
low temperatures, ix, 273 micrometer, 236
microscope, 3, 239
microstructure, 148, 154, 407, 419, 437
M microstructures, 96, 137, 138, 148, 156
migration, 144
management, 163, 213, 225, 227, 241, 250, 269, 292,
military, vii
295, 297, 331, 348, 354, 355, 407
miniaturization, ix, 211, 213, 216, 236, 240
manganese, 450
mixing, 251, 382, 386, 392, 417, 427, 428, 430, 431,
manifolds, 228
440, 449
manufacturing, 6, 214, 216, 225, 238, 252, 253, 287
mobility, 147, 151, 395, 400, 455
market, 211, 212, 229, 450
modeling, viii, x, 13, 71, 72, 76, 134, 162, 164, 165,
markets, 11
168, 173, 193, 201, 206, 207, 273, 291, 292, 294,
masking, 217, 220, 226, 237
295, 296, 298, 299, 318, 362, 376
mass loss, 403
models, ix, x, 66, 71, 72, 76, 77, 78, 79, 81, 96, 101,
material surface, 437
107, 116, 122, 132, 133, 162, 165, 169, 179, 180,
matrix, 118, 119, 139, 144, 151, 152, 153, 163, 171,
185, 186, 192, 195, 201, 206, 207, 208, 209, 269,
202, 208, 248, 285, 310, 448, 453, 455, 457, 461
273, 274, 291, 292, 294, 295, 297, 298, 300, 301,
Maxwell equations, 295, 358
302, 303, 321, 371, 376
measurement, 16, 72, 103, 170, 197, 249, 389, 416
modules, 180, 402
measures, 127
modulus, 277, 319, 320
mechanical energy, 195
moisture, 174, 222, 232, 313, 393
mechanical properties, 137, 139, 144, 150, 151, 156,
moisture content, 313
447, 451
mold, 222
mechanical stress, 5, 302
mole, 203, 277, 292, 306, 319, 329
mechanical structure, 214
molecular mass, 97, 202
media, 174, 236, 241, 254, 256, 263, 264, 268, 269,
molecular weight, 16, 97, 114, 137, 308, 449
309, 310, 448, 456
molecules, 97, 98, 114, 148, 150, 218, 221, 236, 237,
melt, 173
361, 393
melting, 219, 449
molybdenum, 254, 255
membranes, viii, xi, 10, 11, 135, 136, 137, 142, 144,
momentum, 165, 305, 310, 355, 356, 357, 361
145, 147, 148, 149, 150, 151, 152, 153, 154, 155,
monograph, 383
157, 158, 159, 160, 214, 222, 223, 226, 236, 237,
monomer, 138, 140, 141, 144, 146, 388, 455
238, 239, 241, 347, 408, 445, 446, 448, 453, 455,
monomers, 136, 138
456, 459, 461, 462, 463, 464, 466, 467, 468
morphology, 3, 392, 413, 415, 426, 427, 433, 453,
memory, 216
466
metal oxide, 212, 287
motion, 193, 314, 461
metal oxides, 212
motivation, 10, 249
metal salts, 219
movement, 448, 461
metals, 213, 220, 223, 254, 255, 286, 382, 386, 414
multidimensional, ix, 273
methane, 83, 89, 163, 174, 178
multiplication, 119
methanol, vii, 9, 11, 12, 13, 14, 15, 16, 18, 19, 20,
21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,
34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 46, 47, N
48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60,
61, 62, 63, 64, 65, 66, 67, 135, 136, 139, 148, 149, Na+, 461
152, 153, 154, 157, 158, 159, 223, 225, 227, 228, nanocomposites, 414
233, 235, 236, 280, 297, 382, 383, 407, 408, 412, nanofibers, x, 383, 411, 414, 415, 426, 440
413, 414, 415, 425, 426, 427, 430, 436, 439, 440, nanomaterials, 415
448 nanometer, 214, 432
methyl groups, 147 nanometer scale, 214, 432
microelectronics, ix, 211, 213, 214, 229, 230, 437 nanometers, 383
Index 481
nanoparticles, x, 388, 389, 407, 408, 411, 414, 415, oxygen, vii, viii, 9, 10, 12, 13, 14, 31, 37, 41, 42, 48,
421, 422, 425, 427, 428, 429, 430, 431, 432, 440 51, 58, 66, 67, 93, 114, 115, 147, 148, 154, 156,
nanostructured materials, 407 163, 168, 174, 202, 203, 213, 233, 236, 258, 275,
nanotube, 410 276, 281, 283, 285, 291, 292, 295, 299, 300, 301,
nanotubes, x, 218, 383, 404, 407, 408, 409, 411, 415, 306, 307, 311, 314, 315, 318, 321, 322, 323, 325,
440 326, 338, 348, 354, 360, 364, 366, 367, 368, 369,
natural gas, 10, 72, 74, 75, 76, 132, 163 371, 372, 382, 385, 391, 399, 403, 408, 415, 418,
natural resources, 10, 278 420, 422, 424, 425, 430, 437, 441, 450
negativity, 399 oxygen consumption, 326
neglect, 295 oxygen consumption rate, 326
Netherlands, 75, 417, 428 ozone, 237
network, 150, 185, 192, 193, 202, 206, 222, 294, 447
neural network, 294
nickel, 10, 163, 220, 254, 437 P
nitric acid, 385, 408
PAA, xi, 445, 447, 449, 455, 456, 457, 458, 459,
nitrogen, 48, 159, 175, 307, 318, 371, 384, 393, 403,
460, 464, 465, 466, 467, 468, 469
435, 436, 437
packaging, 216, 230
nitrogen oxides, 175
palladium, 226, 227, 254
NMR, 139, 140
PAN, 448
noble metals, 223, 254, 255
parameter, 73, 77, 80, 95, 96, 106, 107, 116, 137,
nodes, ix, 247, 249
144, 145, 148, 164, 169, 175, 177, 178, 195, 201,
noise, 171, 176, 279
203, 277, 294, 319, 321, 362, 363, 367, 413, 419,
normal distribution, 171
423
nucleation, 219, 384
parameter estimation, 80, 96, 107
null hypothesis, 121
parents, 274
partial differential equations, 165
O particles, x, 4, 5, 173, 228, 250, 252, 254, 269, 285,
288, 381, 382, 383, 384, 385, 386, 387, 388, 389,
observations, 116, 117, 118, 119, 120, 122, 123, 125, 390, 391, 392, 396, 397, 398, 400, 402, 403, 404,
127, 130 406, 408, 411, 414, 415, 418, 425, 426, 427, 428,
oil, 76, 381 432, 440
operator, 76, 78, 81, 186 passivation, 221
optimization, vii, ix, 9, 76, 132, 133, 175, 273, 294, passive, 226, 227, 231, 235, 237
303, 413 patents, 250
ordinary differential equations, 165 pathways, 145, 286, 403
organic compounds, 413 patterning, 221, 222, 230, 232
organic solvent, 144, 153, 388 pedigree, viii, 71, 73, 81, 112, 113, 114
organic solvents, 144, 153, 388 PEEK, viii, 135, 136, 137, 138, 139, 140, 141, 142,
orientation, 219 143, 144, 145, 146, 147, 148, 149, 150, 151, 152,
osmosis, 455 153, 154, 155, 156, 157, 158, 159
output, viii, ix, 9, 12, 20, 21, 29, 30, 31, 33, 34, 41, periodicity, 316
44, 46, 47, 51, 54, 55, 56, 57, 64, 67, 80, 154, 155, permeability, xi, 136, 146, 148, 149, 152, 154, 158,
162, 166, 167, 169, 170, 171, 172, 180, 182, 183, 275, 299, 300, 319, 320, 357, 363, 367, 368, 371,
184, 185, 187, 193, 194, 195, 197, 198, 200, 202, 372, 445
203, 204, 227, 273, 280, 294, 314, 315 permeation, 148, 157, 159
oxidation, ix, 10, 11, 13, 21, 27, 37, 40, 42, 77, 150, permit, 144, 219, 268
163, 174, 213, 216, 217, 219, 247, 249, 255, 257, peroxide, 236, 405, 406, 407, 455
258, 259, 262, 265, 266, 267, 283, 382, 384, 385, pH, 181, 182, 236, 251, 257, 260, 261, 416, 419,
387, 391, 392, 402, 403, 404, 405, 408, 410, 414, 421, 428, 449
415, 423, 426, 427, 430, 431, 436, 437, 439, 440, phase boundaries, 385, 386, 387, 402
448, 457, 466 phenol, 144, 391
oxides, 153, 156, 157, 175, 212 photoelectron spectroscopy, 416, 440
oxyfluorination, x, 411, 437, 438, 439, 441
482 Index
photolithography, 216, 217, 218, 222, 227, 232, 233, polymeric membranes, 136
236, 237, 240 polymerization, 152, 157, 388, 400, 409, 449, 455
photoluminescence, 221 polymerization time, 152
physical mechanisms, 355, 376 polymers, viii, xi, 135, 136, 138, 139, 142, 144, 145,
physical properties, x, 217, 250, 269, 381 146, 148, 149, 153, 156, 157, 158, 213, 217, 231,
physicochemical properties, 271, 407 233, 236, 241, 387, 388, 390, 400, 408, 445, 447,
physico-chemical properties, 413 455, 461
physics, 295, 300, 302, 303, 357, 376 polypropylene, 449, 461, 462
pitch, 414 polystyrene, 10
planning, 78, 192 polyvinyl alcohol, xi, 445, 447, 455
plants, viii, ix, 161, 162, 163, 169, 206, 209, 248, poor, 17, 20, 21, 31, 40, 44, 122, 144, 149, 150, 195,
278 197, 223, 225, 226, 233, 286, 448
plasma, x, 218, 220, 221, 383, 384, 391, 399, 407, porosity, 4, 115, 152, 250, 251, 286, 292, 294, 299,
409, 411, 434, 435, 436, 437, 440 319, 383, 413, 426, 462
platinum, x, xi, 10, 41, 135, 163, 213, 220, 221, 225, porous materials, 286
248, 249, 253, 254, 255, 263, 264, 266, 267, 269, porous media, 241, 309, 310
285, 295, 400, 407, 408, 409, 411, 412, 415, 417, porphyrins, 10
418, 421, 422, 429, 430, 440 potassium, 173, 220, 248, 455
PMMA, 231, 232, 233, 235 potassium persulfate, 455
Poincar, 211 power, vii, viii, ix, xi, 9, 10, 11, 12, 13, 19, 20, 21,
poison, 405 22, 26, 27, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38,
polar groups, 151 41, 42, 44, 46, 47, 48, 51, 53, 54, 56, 57, 58, 60,
polarity, 437 61, 63, 65, 66, 67, 74, 75, 76, 113, 155, 157, 161,
polarization, viii, 5, 71, 73, 80, 88, 93, 96, 102, 104, 162, 163, 165, 166, 167, 168, 169, 170, 171, 172,
105, 114, 115, 133, 134, 154, 164, 203, 255, 260, 173, 176, 179, 180, 181, 183, 184, 185, 186, 188,
282, 290, 291, 295, 320, 321, 322, 348, 415 192, 193, 194, 195, 196, 197, 198, 200, 201, 202,
pollution, 76, 279, 382 203, 204, 205, 206, 207, 208, 209, 211, 216, 221,
polycarbonate, 455 222, 224, 225, 227, 228, 229, 230, 231, 232, 233,
polycondensation, 146 240, 241, 248, 273, 276, 278, 279, 280, 281, 282,
polydimethylsiloxane, 231 285, 294, 300, 302, 315, 320, 331, 348, 354, 376,
polyether, 158 382, 394, 395, 396, 397, 399, 412, 414, 415, 445,
polyetheretherketone, viii, 135, 136, 159 446, 447, 448, 450, 459, 467, 469
polyetheretherketone (PEEK), viii, 135, 136 power generation, 20, 162, 192, 206, 207, 227, 248,
polyimide, 233 302, 382
polymer, viii, ix, xi, 10, 14, 15, 58, 135, 136, 137, prediction, x, 132, 186, 202, 205, 274, 302, 303, 304,
138, 139, 140, 142, 143, 144, 145, 148, 149, 150, 318, 376, 385
151, 152, 153, 154, 157, 158, 159, 212, 214, 220, prediction models, 302
222, 225, 230, 232, 233, 241, 247, 248, 264, 266, president, 274
268, 282, 283, 284, 285, 298, 312, 348, 355, 361, pressure, vii, viii, 4, 9, 20, 31, 48, 58, 67, 93, 94, 96,
367, 382, 383, 388, 389, 390, 391, 400, 401, 407, 97, 98, 99, 114, 115, 122, 154, 164, 165, 168, 169,
408, 409, 412, 426, 428, 433, 445, 446, 447, 448, 175, 181, 182, 203, 204, 208, 214, 221, 227, 250,
449, 451, 453, 454, 455, 456, 457, 459, 461, 462, 268, 276, 289, 291, 294, 295, 297, 299, 300, 303,
464, 465, 466, 467, 468, 469 305, 306, 311, 313, 315, 316, 317, 319, 355, 356,
polymer blends, 467 358, 360, 361, 363, 364, 366, 367, 376, 383, 416,
polymer chains, 400 434, 437
polymer electrolytes, 446, 448, 449, 453, 467 priming, 227
polymer film, 151, 454 probability, 127, 156
polymer films, 454 production, 27, 76, 137, 174, 214, 285, 291, 294,
polymer matrix, 151, 153, 448, 453, 455, 457 301, 311, 331, 338, 354, 360, 364, 413
polymer solubility, 139 production costs, 413
polymer structure, 137 program, ix, 165, 273, 274, 417
polymeric blends, 464 proportionality, 266
polymeric chains, 145 propylene, 448
Index 483
protons, 13, 14, 137, 139, 145, 148, 155, 156, 260, reduction, 10, 13, 37, 41, 42, 80, 84, 87, 122, 144,
281, 283, 284, 285, 314, 338, 355, 382, 385, 387, 151, 174, 195, 201, 213, 241, 258, 301, 354, 357,
388, 389, 391, 399 382, 383, 384, 388, 389, 390, 391, 393, 394, 396,
prototype, 72, 73, 230 399, 403, 405, 407, 408, 412, 413, 415, 416, 418,
PTFE, 232, 233, 235, 250, 251, 252, 253, 265, 267, 421, 422, 425, 427, 428, 429, 440, 457, 466
382, 386, 389, 398, 399, 450, 451, 455 reflection, 419
pulse, 12, 30, 183, 186, 196 reforms, 163
pumps, vii, 9, 18, 227 regional, 193
pure water, 358 regression, 71, 72, 76, 77, 78, 79, 81, 109, 111, 112,
PVA, xi, 445, 446, 447, 449, 450, 451, 452, 453, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122,
454, 455, 456, 457, 458, 459, 460, 461, 464, 465, 126, 127, 129, 132, 133
466, 467, 468, 469 regression analysis, 114
pyrolysis, 144, 174 regulation, 89, 192, 193, 206
regulators, 190, 193
reinforcement, 5, 401
Q relationship, xi, 11, 40, 44, 76, 116, 120, 156, 183,
249, 292, 445
quality improvement, 206
relationships, 18, 19, 182, 183, 263, 265, 292, 294
quartz, 214
relaxation, 455
quasi-equilibrium, 259
relevance, 11, 226, 236
quaternary ammonium, 448
reliability, vii, viii, 12, 13, 161, 302
quinone, 391
renewable energy, 381
reparation, 440
R replacement, 76, 241, 281
replication, 221, 222
radial distribution, 192 reserves, 381
radiation, 217, 227, 416 residuals, 119, 120, 127
radical polymerization, 449, 455 resistance, x, 4, 5, 15, 16, 17, 20, 22, 34, 37, 41, 53,
radio, 220, 221, 399, 434 65, 67, 81, 96, 106, 107, 142, 150, 153, 155, 225,
random errors, 118 233, 240, 254, 255, 267, 268, 286, 287, 292, 299,
range, 4, 20, 21, 38, 51, 60, 63, 65, 72, 79, 104, 105, 301, 312, 337, 338, 372, 383, 390, 392, 396, 404,
111, 134, 137, 145, 151, 167, 168, 170, 173, 180, 405, 411, 412, 415, 432, 433, 440, 448, 449, 459,
181, 187, 194, 222, 241, 248, 254, 259, 278, 285, 464
294, 301, 382, 383, 422, 428, 431, 432, 449, 457, resolution, 221, 320, 384
459, 464, 466 resources, ix, 10, 161, 273, 278, 304
reactant, 4, 5, 10, 13, 76, 95, 96, 197, 203, 250, 280, response time, 180, 198, 203
285, 286, 289, 295, 297, 298, 299, 300, 304, 313, retardation, 401
314, 337, 363, 368, 371, 372, 376, 385, 386, 387, retention, 145, 151
401, 414 RIE, 221, 225, 226, 227, 229, 230, 237
reactants, vii, ix, 4, 13, 15, 94, 115, 154, 156, 165, rings, 139, 228, 406
180, 182, 222, 224, 249, 250, 254, 273, 283, 285, risk, 213
288, 291, 295, 303, 306, 311, 314, 322, 348, 363, robustness, 172, 176, 178
370, 376, 383, 385 rolling, 251, 252, 285
reaction mechanism, 263, 289 Romania, 71
reaction order, 96 room temperature, xi, 145, 147, 151, 225, 232, 235,
reaction rate, 250, 262, 289, 291, 301 267, 412, 417, 437, 445, 447, 448, 469
reaction time, 144, 437, 440 roughness, 437
reaction zone, 249 Royal Society, 389, 394
reactivity, 249, 257, 267, 268, 437 rural areas, 192
reality, 5, 155, 385 ruthenium, 27, 40, 41, 254, 255
recovery, 168, 205, 208
recycling, 18
484 Index