1600216692 fuel cell

FUEL CELL RESEARCH TRENDS

FUEL CELL RESEARCH TRENDS

L.O. VASQUEZ EDITOR

Nova Science Publishers, Inc. New York

Copyright © 2007 by Nova Science Publishers, Inc. 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

NOTICE TO THE READER The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of information contained in this book. The Publisher shall not be liable for any special, consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or reliance upon, this material. Any parts of this book based on government reports are so indicated and copyright is claimed for those parts to the extent applicable to compilations of such works. 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. LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA Fuel cell research trends / L. O. Vasquez (editor). p. cm. Includes index. ISBN-13: 978-1-60692-750-2 1. Fuel cells. I. Vasquez, L. O. TK2931.F7845 2007 621.31'2429--dc22 2007011125

Published by Nova Science Publishers, Inc. New York

CONTENTS

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 Fuel Cell (DMFC) from Single Cell, Fuel Cell Stack to DMFC System

9

Rongzhong Jiang

Chapter 2 Experimental Activity on a Large SOFC Generator 71 M. Santarelli, P. Leone, M. Calì 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 Fuel Cells

211

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 Fuel Cells: A Review

381

Zhigang Qi

L.O. Vasquez vi

Chapter 9 A Study on the Role of Carbon Support Materials for Fuel-Cell Catalysts

411

Seok Kim and Soo-Jin Park

Chapter 10 Developments of Alkaline Solid Polymer Electrolyte Membranes Based on Polyvinyl Alcohol and Their Applications in Electrochemical Cells

445

G.M. Wu, S.J. Lin and C.C. Yang

Index 473

PREFACE A fuel cell is an electrochemical energy conversion device. It produces electricity from

external supplies of fuel (on the anode side) and oxidant (on the cathode side). These react in the presence of an electrolyte. Generally, the reactants flow in and reaction products flow out while the electrolyte remains in the cell. Fuel cells can operate virtually continuously as long as the necessary flows are maintained.

Fuel cells differ from batteries in that they consume reactants, which must be replenished, while batteries store electrical energy chemically in a closed system. Additionally, while the electrodes within a battery react and change as a battery is charged or discharged, a fuel cell's electrodes are catalytic and relatively stable.

Fuel cells are very useful as power sources in remote locations, such as spacecraft, remote weather stations, large parks, rural locations, and in certain military applications. A fuel cell system running on hydrogen can be compact, lightweight and has no major moving parts. Because fuel cells have no moving parts, and do not involve combustion, in ideal conditions they can achieve up to 99.9999% reliability. This equates to less than one minute of down time in a six year period.

This book presents new leading-edge research in the field. 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. Chapter 1 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

L.O. Vasquez viii

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 λox), 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 <100°C 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

L.O. Vasquez x

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

on the nano-structure as well as the electroactivity of the carbon-supported platinum catalysts for DMFCs were studied.

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. The authors 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.

In Chapter 10, the authors 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, 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.

EXPERT COMMENTARY

In: Fuel Cell Research Trends ISBN: 1-60021-669-2Editor: L.O. Vasquez, pp. 3-6 © 2007 Nova Science Publishers, Inc.

QUALIFICATION OF FUEL CELL MEMBRANEELECTRODE ASSEMBLIES

Zhigang QiPlug 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 cellsystems.

Ideally, each MEA should be qualified according to the following categories before it isassembled into a fuel cell stack:

1. Thickness and uniformity of thickness of catalyst layersThe catalyst layer thickness typically ranges between 5 and 20 μm for proton exchange

membrane (PEM) fuel cells. A variation of ±2 μm that can hardly be observed under anoptical microscope could mean a thickness difference of 40 to 10%. A 40% difference willcause a significant variation in the fuel cell performance and durability. It is absolutelyunacceptable if there are membrane areas in the active region that are not covered by thecatalyst at all.

2. Surface morphology of catalyst layersHow smooth is the catalyst surface? Are there any cracks? If so, how big and how severe

are the cracks? Cracks can result from the segregation of catalyst layer components, or fromthe catalyst layers being dried too fast, or from the catalyst layers being made too thick.Smooth surfaces with no or only a few small cracks are expected to offer more stable anddurable performance.

Zhigang Qi4

3. Porosity and pore size distribution of catalyst layersThe pores are for the transport of fuel cell reactants and product(s). Optimal porosity and

pore size distribution can facilitate the mass transport process to minimize the fuel cellperformance loss due to concentration overpotential. If some pores are more hydrophobicthan others, what is the relative distribution? Is the distribution of pore sizes andhydrophobicity within the allowable range?

4. Distribution of catalyst particles and other components within a catalyst layerA catalyst layer typically contains at least a catalyst and one other component such as an

ionomer for proton conductance. It is preferred that the two components mix togetherhomogeneously so that the number of catalyst-electrolyte-reactant three-phase regions will bemaximized. Their dispersion is mainly controlled by the catalyst ink formulation processwhere adequate sonication and agitation are normally applied.

5. PinholesPinholes 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 wasteof fuels, and faster decrease in MEA life. Combustion may occur between hydrogen and airaround the pinholes and the heat generated could enlarge the pinholes quickly, which in turnleads to higher reactant crossover rates and more severe combustion. Pinholes may arise fromthe original defects in the membrane, or be created during the MEA fabrication process, orgradually form due to localized decay of the membrane materials. It is critical to havepinhole-free virgin MEAs for longer fuel cell lifetime.

6. Electrode shortingIf 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 ofan MEA is not high enough, it is an indication that the anode and the cathode may be incontact with each other at some points. The fuel cell OCV will be lower due to shorting, asymptom similar to that caused by pinholes. One effective way to distinguish shorting frompinholes is to monitor the OCV change with the pressure difference between the fuel and theair. If the OCV drops significantly with an increase in the pressure difference, it is more likelyto 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 catalystclumps being pushed through the membrane during the MEA fabrication process. So, it couldbe very helpful to smooth the surfaces of electrodes before they are bonded onto a membrane.

7. Electrochemical active surface area of electrodesThe electrochemical active surface area (ECSA) reflects the total catalyst surface that has

the potential to participate in the fuel cell reaction. It is typically measured by the hydrogenadsorption/desorption peak area or the CO oxidative stripping peak area. A larger ECSAnormally gives better fuel cell performance. The ratio of ECSA to the mass of the catalyst isan indication of how effectively the precious metal catalyst is used. The ratio of ECSA to thetotal geometrical surface area of the catalyst estimated by the particle size is an indication ofhow effectively the catalyst surface is used. The latter ratio can be used to gauge how well(high ratio) or bad (low ratio) a catalyst layer is made.

Qualification of Fuel Cell Membrane Electrode Assemblies 5

8. Variation of ESCA among electrodesIt is preferred that little variation of the ECSA exists among electrodes so that a more

uniform performance is achieved among all the cells in a stack. A fuel cell stack with uniformindividual cell performance is easier to control and longer life is expected. On the other hand,if one cell has abnormally low performance, it increases the difficulty of controlling the fullstack, and could drag down the neighboring cells, resulting in shorter stack life.

9. Dimensional change accompanying relative humidityThe dimension of a membrane used in a PEM fuel cell changes with the relative humidity

(RH). Repeated swell-shrink cycling of the MEA poses significant mechanical stress on themembrane, which can cause it to wear down or tear apart quickly. A membrane with littledimensional change offers better mechanical stability, and this is typically achieved bymembrane reinforcement.

10. Hydrophobicity of catalyst layersHydrophobicity affects the catalyst-electrolyte-reactant three-phase regions and the mass

transport resistance of the catalyst layers. If the hydrophobicity is too low, the pores withinthe catalyst layer will be flooded easily, leading to high mass transport resistance. In contrast,if the hydrophobicity is too high, the catalyst particles may not be easily wetted by water oran electrolyte, which results in lower proton conductance. The hydrophobicity should bevaried based on the fuel cell operating conditions.

11. PerformanceOne of the goals of these above characterizations is the achievement of high and uniform

fuel cell performance. These characterizations provide the necessary knowledge about whatparameters should and can be tailored in order to achieve higher performance. Once a goodand uniform performance is consistently achieved, some or all of these characterizations canbe eliminated. The best way to gauge the performance of an MEA is to collect voltage –current polarization curves.

12. DurabilityAnother goal of these characterizations is the achievement of durable fuel cell

performance. Durability can be gauged by the fuel cell performance decay rate and thelikelihood of premature failures. If a fuel cell suffers frequent premature failures, which aremostly 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 predicthow 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 themass transport resistance of the catalyst layer. Testing an MEA under a properly chosenaccelerated condition can offer some guidance on its decay rate under regular fuel cellconditions.

In reality, nearly none of the above characterizations are performed routinely onindividual MEAs partly due to time constraint. In addition, most of the characterizations willaffect the MEA either physically or chemically, which makes it less usable afterwards.Typical practice is to perform most of the above characterizations to establish an acceptable

Zhigang Qi6

MEA manufacturing process and assume that future MEAs from that process are good unlesstesting in a fuel cell indicates that they are not. At that time, the entire fuel cell stack has to beopened to replace the bad MEAs. This is a time-consuming process and frequently damagessome of the previously good MEAs.

Before a simple universal method that can quickly and non-destructively evaluate thequality of an MEA is established, the best practical approach in the moment is to optimize theMEA manufacturing process so that MEAs with consistent and controlled properties are madeevery 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

POWER AND ENERGY EFFICIENCY ANALYSIS OF DIRECT METHANOL FUEL CELL (DMFC)

FROM SINGLE CELL, FUEL CELL STACK TO DMFC SYSTEM

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.

Rongzhong Jiang 10

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 system’s 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

Rongzhong Jiang 12

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 and Concepts

The principle of the method in this study is to use the experimental DMFC single cell discharge data and the corresponding methanol crossover results at running time to calculate the power and energy efficiencies of DMFC single cells, stacks and systems at various methanol concentrations, operating temperatures and oxygen partial pressures. The simulated data of DMFC systems and stacks are compared with the experimental data of the DMFC single cells in order to determine the deviations of the computation and the accuracy of the


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.

1. DMFC Single Cell

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

At the anode −+ ++→+ eHCOOHOHCH 66223 Eo = 0.02V (1)

At the cathode OHeHO 22 36623

→++ −+ Eo = 1.23V (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

Rongzhong Jiang 14

[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.

Fuel In Air In

Fuel Out Air Out

- V +

1 5 6 7234

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.

The cell potential of a DMFC can be expressed by [65-66]

crossmassRact EEEEEEi −−−−= 0 (3)

00r0 i log b E E += (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


cell; and i0 (A) and b0 (mV/dec) are the exchange current and the Tafel slope at i0, respectively.

CH3OH + H2O

CO2 + 6H+ + 6e-

Methanol

H2O

Air

3/2 O2 + 6H+ + 6e-

3H2O

MeOH

H +

CO2

H2O

O2

H2O

e-

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,

1000)( cellcell

cellopencelliRibLogEE ⋅

−−= (5)

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 (Ω) is Ohmic resistance of the fuel cell.

Rongzhong Jiang 16

According to the literature [63-64], for the measurement of methanol crossover in a running DMFC single cell, the equivalent current (ie, mA/cm2) of methanol crossover rate can be approximately expressed as a linear equation,

celleoe iBii ⋅−= (6) 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 (ζcell ), energy

efficiency (ηcell), and energy density (εcell, Wh/Kg) of a DMFC single cell from the experimental voltage-current curves.

eii

i

cell

cellcell +

=ζ (7)

)(21.1 eiiEi

cell

cellcellcell +⋅

⋅=η (8)

)(6.0)

36006)(1000(

)( ee iiMFEiF

MiiEi

cellMeOH

cellcell

MeOHcell

cellcellcell +

⋅=

+⋅

=ε (9)

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, ηcell, in eq. (8) is the product of voltage efficiency, Ecell/1.21, and Faradic efficiency, ζcell.

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 cell’s 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


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

stackstackcellstack iRnEE −= (10)

Fuel In

Fuel Out

Air In

Air Out

Single Cell

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 (Ω) 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.

Rongzhong Jiang 18

stackestack

stackstack ii

i

−+=ζ (11)

)(21.1 stackestack

stackstackstack iin

Ei

−+⋅

=η (12)

)36006)(1000(

)(F

MiinEi

MeOHstackestack

satckstackstack

−+⋅

=ε (13)

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.


Methanol Sensor

Fuel CellStack

Air Compressor

Fuel Pump 1

Fuel Reservoir

Fuel Mixer

Air Humidifier

HeatExchanger

Electric Fans

Electronic

Controller

Fuel Pump 2

Water Reservoir

Fuel Flow

Controller

Air Flow Controller

Rechargeable Battery

Thermometers

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.

Fuel Cell Stack Load

Internal Power Users Including Pumps and Electronic Controllers

RechargeableBattery

+

-

+ -

DCDC converter

i i 1i2

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.

Rongzhong Jiang 20

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 0.05Ω·cm-2·cell-

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

AnRstack ρ= (14)


Here, ρ (Ω·cm-2·cell-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

ernalstackstack

ernalstacksystem ii

EPii int

int −=−= (15)

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

)(21.1 int stackeernalsystem

systemsystemsystem iiin

Ei

−++⋅

=η (16)

)36006)(1000(

)( int

FMiiin

Ei

MeOHstackeernalsystem

systemsystemsystem

−++

⋅=ε (17)

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.

Results and Discussion

1. Effect of Methanol Concentration by Using Air as Oxidant

1.1. Single Cell Results 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-

Rongzhong Jiang 22

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

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 50 100 150 200 250 300 350 400

Current (mA/cm2)

Volta

ge (V

)

0.5M MeOH

1.0M MeOH

2.0M MeOH

3.0M MeOH

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.

Table 1. Electrochemical parameters for simulation of voltage-current curves of a DMFC single cell at 60 oC using air as oxidant and methanol as fuel with different

concentrations.

Methanol Concentration (M) Parameters 0.5 1.0 2.0 3.0

Eopen (V) 0.66 0.63 0.58 0.56 b (V/Dec) 0.097 0.080 0.076 0.115 Rcell (Ω) 0.62 0.63 0.62 0.62 Pmax-cell (mW/cm2) * 60 73 61 30

* Pmax-cell is the peak power of DMFC single cell.


0

10

20

30

40

50

60

70

80

0 50 100 150 200 250 300 350 400

Current (mA/cm2)

Pow

er (m

W)

0.5M MeOH

1.0M MeOH

2.0M MeOH

3.0M MeOH

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.

0

100

200

300

400

500

0 50 100 150 200 250 300 350 400

Cell Current (mA/cm2)

Equi

vale

nt C

urre

nt o

f Met

hano

l Cr

osso

ver (

mA/

cm2 )

3.0M MeOH

2.0M MeOH

1.0M MeOH

0.5M MeOH

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.

Rongzhong Jiang 24

Table 2. Electrochemical parameters for simulation of methanol crossover of a DMFC single cell at 60 oC, and the peak energy efficiency and energy density using air as

oxidant and methanol as fuel with different concentrations.


ieo (mA/cm2) 60 105 255 402 B 0.36 0.44 0.60 0.70 ηcell 0.283 0.243 0.142 0.063 εcell (Wh/Kg) 1726 1481 863 374

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 50 100 150 200


Cell

Volta

ge (V

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Fara

dic

Eff a

nd E

nerg

y Ef

f

Cell Voltage

Faradic Eff

Egergy Eff

A

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 50 100 150 200 250 300


Cell

Volta

ge (V

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Fara

dic

Eff a

nd E

nerg

y Ef

f

Cell Voltage

Faradic Eff

Egergy Eff

B

Figure 9. Continued on next page.


0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 50 100 150 200 250 300 350


Cell

Volta

ge (V

)

0.0

0.2

0.4

0.6

0.8

1.0

Fara

dic

Eff a

nd E

nerg

y Ef

f

Cell Voltage

Faradic Eff

Egergy Eff

C

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0 100 200 300 400

Cell current (mA/cm2)

Cel

l Vol

tage

(V)

0.0

0.2

0.4

0.6

0.8

1.0

Fara

dic

Eff a

nd E

nerg

y E

ffCell Voltage

Faradic Eff

Egergy Eff

D

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

Rongzhong Jiang 26

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.

0

200

400

600

800

1000

1200

1400

1600

1800

0 50 100 150 200 250 300


Ener

gy D

ensi

ty (W

h/K

g)

Energy Density(0.5M MeOH))

Energy Density(1.0M MeOH)



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.

1.2. DMFC Stack Results 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


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.

0

4

8

12

16

20

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Stack Current (A)

Vola

tge

(V)

Stack Voltage (0.5M MeOH) Stack Voltage (1.0M MeOH)

Stack Voltage (2.0MMeOH) Stack Voltage (3.0M MeOH)

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.

Rongzhong Jiang 28

0

10

20

30

40

50

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Stack Current (A)

Stac

k Po

wer

(W)

Stack Power (0.5M MeOH)




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

400

800

1200

1600

0.0 1.0 2.0 3.0 4.0 5.0 6.0Stack Current (A)

Ener

gy D

ensi

ty (W

h/K

g)

0.00

0.04

0.08

0.12

0.16

0.20

0.24

0.28

Ener

gy E

ffici

ency

Stack Energy (0.5M MeOH) Stack Energy (1.0M MeOH)Stack Energy (2.0M MeOH) Stack Energy (3.0M MeOH)

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.


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.


Pmax-stack * (W) 36 40 33 15

ηstack 0.278 0.237 0.133 0.046 εstack (Wh/Kg) 1690 1438 806 281 Eopen (V) 18.4 17.6 16.2 15.7

* Here, Pmax-stack is peak stack power.

1.3. DMFC System Results

The discharge performance of a DMFC system is relatively more complex than that of a single cell or a stack. Especially, the short term performance is significantly affected by the internal electronics and the rechargeable battery. Therefore, in this simulation, we assume that the fuel cell stack will not charge the internal battery, because our goal is to obtain the energy efficiency and energy density of the DMFC system with long-term operation. Actually, for the long term performance the effect of internal battery in the DMFC system on the power and energy outputs can be neglected because the charged and discharged electric energies by the battery can be balanced in a relatively long time. The output current of the DMFC system can be calculated with equation (15). According to the power distribution and the electrical connections among the components shown in Figure 5, the system output voltage would be the same as that of stack voltage if the system auxiliary components would not take powers (i.e., if zero current were bypassed to the auxiliary components). Figure 14A shows plots of output power versus output current between the DMFC stack and system. The simulated DMFC contains a 28-cell stack with 22 cm2 electrode area operated at 60 oC with 1.0M

0

10

20

30

40

50

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Output Current (A)

Powe

r (W

)

Power, DMFC System

Power, DMFC StackA


Rongzhong Jiang 30

0.2

0.3

0.4

0.5

0.6

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Output Current (A)

Inte

rnal

Cur

rent

Con

sum

ptio

n (A

)

B

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.


Pmax-system * (W) 32 36 29 11

ηsystem 0.243 0.211 0.116 0.035 εsystem (Wh/Kg) 1438 1285 707 204 Eopen (V) 15.3 14.8 13.5 11.5

* Here, Pmax-system is peak system power.


0

2

4

6

8

10

12

14

16

18

0.0 1.0 2.0 3.0 4.0 5.0Output Current (A)

Volta

ge (V

)

Sys V-out (0.5M MeOH) Sys V-out (1.0M MeOH)

Sys V-out (2.0M MeOH) Sys V-out (3.0M MeOH)

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

Rongzhong Jiang 32

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.

0

5

10

15

20

25

30

35

40

0.0 1.0 2.0 3.0 4.0 5.0

Output Current (A)

Pow

er (W

)

Sys Power (0.5M MEOH)




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.

0

400

800

1200

1600

0.0 1.0 2.0 3.0 4.0 5.0

Output Current (A)

Ener

gy D

ensi

ty (W

h/Kg

)

0 .00

0.04

0.08

0.12

0.16

0.20

0.24

Ener

gy E

ffici

ency

Sys Energy (0.5M MeOH) Sys Energy (1.0M MeOH)

Sys Energy (2.0M MeOH) Sys Energy (3.0M MeOH)

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.


1.4. Comparison of Energy Densities under Various Operating Conditions 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.

0

200

400

600

800

1000

1200

1400

1600

0.0 1.0 2.0 3.0 4.0 5.0

Output Current (A)

Ener

gy D

ensi

ty (W

h/Kg

) A

0

200

400

600

800

1000

1200

1400

0 10 20 30 40

Output Power (W)

Ener

gy D

ensi

ty (W

h/Kg

) B


Rongzhong Jiang 34

400

600

800

1000

1200

1400

9.0 9.5 10.0 10.5 11.0 11.5 12.0 12.5 13.0

Output Voltage (V)

Ener

gy D

ensi

ty (W

h/Kg

) C

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 oC 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 mW·cm-2⋅cell-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 mW·cm-2·cell-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


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 cell’s peak energy efficiency (0.284) is considered as 100%, the DMFC system’s 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.

0.25

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0 50 100 150 200 250 300

Normalized Output Current (mA/cm2)

Nor

mal

ized

Vol

tage

(V/C

ell)

DMFC Stack

DMFC System

DMFC Single Cell

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.

Peak Power Output (mW/cm2⋅cell) Methanol (M) 0.5 1.0 2.0 3.0 Single Cell 60 73 61 30 DMFC Stack 58 65 54 24 DMFC System 52 58 47 17

Rongzhong Jiang 36

0

10

20

30

40

50

60

70

80

0 50 100 150 200 250 300

Normalized Output Current (mA/cm2)

Norm

aliz

ed P

ower

Out

put (

mW

/cm

2 .Cel

l)

Sys Output Power vs. Cell Current

Sys Output Power vs. Sys Output Current

Stack Power vs. Stack Current

Cell Power vs. Cell Current

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 oC. 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.

Peak Energy Efficiency Methanol (M) 0.5 1.0 2.0 3.0 Single Cell 0.284 0.243 0.142 0.058 DMFC Stack 0.278 0.237 0.133 0.046 DMFC System 0.243 0.211 0.116 0.035

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.

Peak Energy Density (Wh/Kg) Methanol (M) 0.5 1.0 2.0 3.0 Single Cell 1726 1481 864 349 DMFC Stack 1690 1438 806 281 DMFC System 1438 1285 707 204

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


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.

2. Effect of Operating Temperature by Using Air as Oxidant

2.1. Single Cell Results 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

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 100 200 300 400 500 600

Current (mA/cm2)

Volta

ge (V

)

Voltage at 80 oC

Voltage at 60 oC

Voltage at 40 oC

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.

Rongzhong Jiang 38

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 mW·cm-2·cell-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.

0

20

40

60

80

100

120

0 100 200 300 400 500 600

Current (mA/cm2)

Pow

er (m

W)

Power at 80 oC

Power at 60 oC

Power at 40 oC

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.

Table 8. Electrochemical parameters for simulation of voltage-current curves of a DMFC single cell using 1.0M methanol as fuel and air as oxidant at various operating

temperatures.

Temperature (oC) Parameters 40 60 80

Eopen (V) 0.63 0.63 0.66 b (V/Dec) 0.095 0.080 0.075 Rcell (Ω) 0.85 0.63 0.53 Pmax-cell (mW/cm2) * 43 73 101


0

50

100

150

200

0 50 100 150 200 250 300 350 400


Equi

vale

nt C

urre

nt o

f Met

hano

l C

ross

over

(mA

/cm

2 )

80 oC MeOH Crossover



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.

Table 9. Electrochemical parameters for simulation of methanol crossover of a DMFC single cell using 1.0M methanol as fuel and air as oxidant at various operating

temperatures; and the peak energy efficiency and energy density obtained.


ieo (mA/cm2) 74 105 181 B 0.47 0.44 0.40 ηcell 0.218 0.243 0.217 εcell (Wh/Kg) 1327 1481 1322

Rongzhong Jiang 40

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 20 40 60 80 100 120 140 160


Cel

l vol

tage

(V)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Fara

dic

Eff

abd

Ene

rgy

Eff

Voltage at 40 oC

Faradic Eff at 40 oC

Egergy Eff at 40 oC

A

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 100 200 300 400


Cel

l vol

tage

(V)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Fara

dic

Eff a

nd E

nerg

y Ef

f

Voltage at 80 oC

Faradic Eff at 80 oC

Egergy Eff at 80 oC

B

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


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.

0

200

400

600

800

1000

1200

1400

1600

0 50 100 150 200 250 300

Cell Current (m A/cm2)

Ener

gy D

ensi

ty (W

h/K

g)

Energy Density (40 oC)



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.

2.2. DMFC Stack Results 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

Rongzhong Jiang 42

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.

0

4

8

12

16

20

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

Stack Current (A)

Vola

tge

(V)

Stack Voltage (80 oC)



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.


Pmax-stack (W) 23 40 57

ηstack 0.214 0.237 0.209 εstack (Wh/Kg) 1299 1438 1272 Eopen (V) 17.6 17.6 18.5


0

10

20

30

40

50

60

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

Stack Current (A)

Stac

k Po

wer

(W)

Stack Power (80 oC)

Stack Power (60 oC)

Stack Power (40 oC)

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

400

800

1200

1600

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Stack Current (A)

Ener

gy D

ensi

ty (W

h/Kg

)

0.00

0.04

0.08

0.12

0.16

0.20

0.24

0.28

Ener

gy E

ffici

ency

Stack Energy (40 oC)



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.

Rongzhong Jiang 44

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.

2.3. DMFC System Results According to the relationship of power distributions in a DMFC system as described in Figure 5, the voltage-current curves at different operating temperatures can be calculated with eqs. (10), (14) and (15). Figure 29 shows voltage-current curves of a DMFC system that contains a 28-cell stack with 22 cm2 electrode area operated at different temperatures. The auxiliary components consume an average of 4W to run the fuel pump, air pump and electronic controllers, which cause a parasitic current loss. Even under idling condition, the internal system has to consume a small amount of power. Consequently, the open circuit voltage of the DMFC system is much lower than that of the corresponding DMFC stack; and the voltage -current curves at different operating temperatures are well separated from the beginning of providing power, to the time of supplying a high current output to the external load. From 80 to 40 oC, the open circuit voltage is decreased from 15.8 to 13.4V. With increasing discharge current, the voltage difference of the voltage-current curves at different operating temperatures becomes more significant due to the increased electrode kinetic rate at the higher operating temperature. Figure 30 shows power-current curves of the DMFC system under the same operating conditions as that in Figure 29. With increasing discharge current, the output power increases. The peak powers are obtained at 8.0V stack voltage, where the stack current is turned off to avoid poor mass transfer at too low cell voltages (< 0.3V). Table 11 lists open circuit voltages and peak powers for the 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. For every 20 degree of temperature increase, there is a 17W increase in peak power for the DMFC system.

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.


Pmax-system * (W) 19 36 53

ηsystem 0.176 0.211 0.193 εsystem (Wh/Kg) 1070 1285 1174 Eopen (V) 13.4 14.8 15.8


0

4

8

12

16

20

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0Output Current (A)

Volta

ge (V

)

Sys V-out (80 oC)

Sys V-out (60 oC)

Sys V-out (40 oC)

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.

0

10

20

30

40

50

60

0.0 2.0 4.0 6.0 8.0

Output Current (A)

Pow

er (W

)

Sys Power (80 oC)

Sys Power (60 oC)

Sys Power (40 oC)

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.

Rongzhong Jiang 46

0

200

400

600

800

1000

1200

1400

0.0 1.0 2.0 3.0 4.0 5.0 6.0Output Current (A)

Ener

gy D

ensi

ty (W

h/K

g)

0.00

0.04

0.08

0.12

0.16

0.20

Ener

gy E

ffici

ency

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.


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


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.

Peak Power Output (mW/cm2⋅cell) Temperature (oC) 40 60 80 Single Cell 43 73 101 DMFC Stack 37 65 93 DMFC System 31 58 86

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.

Peak Energy Efficiency Temperature (oC) 40 60 80 Single Cell 0.218 0.243 0.217 DMFC Stack 0.213 0.236 0.209 DMFC System 0.176 0.211 0.193

Rongzhong Jiang 48

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.

Peak Energy Density (Wh/Kg) Temperature (oC) 40 60 80 Single Cell 1328 1481 1322 DMFC Stack 1299 1438 1272 DMFC System 1070 1285 1174

3. Effect of Oxygen Partial Pressure by Using Different Methanol Concentrations

3.1. Single Cell Results 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 DMFC’s 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 DMFC’s 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.


0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 100 200 300 400 500

Current (mA/cm2)

Vol

tage

(V)

O2 and 0.5M MeOH

O2 and 1.0M MeOH

O2 and 2.0M MeOH

O2 and 3.0M MeOH)

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.

0

20

40

60

80

100

120

0 100 200 300 400 500

Current (mA/cm2)

Pow

er (m

W)

O2 and 0.5M MeOH

O2 and 1.0M MeOH

O2 and 2.0M MeOH

O2 and 3.0M MeOH

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.

Rongzhong Jiang 50

Table 15. Electrochemical parameters for simulation of voltage-current curves of a DMFC single cell at 60 oC using O2 as oxidant and methanol as fuel with different

concentrations.


Eopen (V) 0.625 0.650 0.620 0.600 b (V/Dec) 0.039 0.058 0.058 0.072 Rcell (Ω) 0.86 0.66 0.54 0.58 Pmax-cell (mW/cm2) * 81 95 102 75

Table 16. Electrochemical parameters for simulation of methanol crossover of DMFC single cell at 60 oC, and the peak energy efficiency and energy density using O2 as



ieo (mA/cm2) 60 105 255 402 B 0.36 0.44 0.60 0.70 ηcell 0.328 0.294 0.209 0.127 εcell (Wh/Kg) 1996 1786 1272 771

0

400

800

1200

1600

2000

0 50 100 150 200 250 300 350 400


Ener

gy D

ensi

ty (W

h/K

g)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Ener

gy E

ffici

ency

O2 and 0.5M MeOHO2 and 1.0M MeOHO2 and 2.0M MeOHO2 and 3.0M MeOH

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


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.

3.2. DMFC Stack Results 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



Pmax-stack * (W) 48 55 57 41

ηstack 0.322 0.285 0.198 0.117 εstack (Wh/Kg) 1959 1736 1207 728 Eopen (V) 17.5 18.2 17.4 16.8

Rongzhong Jiang 52

0

4

8

12

16

20

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

Stack Current (A)

Vol

atge

(V)

O2 and 0.5M MeOH

O2 and 1.0M MeOH

O2 and 2.0M MeOH

O2 and 3.0M MeOH

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.

0

10

20

30

40

50

60

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

Stack Current (A)

Stac

k Po

wer

(W)

O2 and 0.5M MeOH

O2 and 1.0M MeOH

O2 and 2.0M MeOH

O2 and 3.0M MeOH

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.


0

400

800

1200

1600

2000

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0Stack Current (A)

Ener

gy D

ensi

ty (W

h/K

g)

0.00

0.04

0.08

0.12

0.16

0.20

0.24

0.28

0.32

Ener

gy E

ffici

ency

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)

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.

3.3. DMFC System Results 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

Rongzhong Jiang 54

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.

0

4

8

12

16

20

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

Output Current (A)

Volta

ge (V

)

Sys V-out (O2 and 0.5M MeOH)Sys V-out (O2 and 1.0M MeOH)Sys V-out (O2 and 2.0M MeOH)Sys V-out (O2 and 3.0M MeOH)

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.


0

10

20

30

40

50

60

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

Output Current (A)

Pow

er (W

)

Sys Power (O2 and 3.0M MEOH)




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.

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0Output Current (A)

Ener

gy D

ensi

ty (W

h/K

g)

0.00

0.04

0.08

0.12

0.16

0.20

0.24

0.28

0.32

Ener

gy E

ffici

ency

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)

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.

Rongzhong Jiang 56

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.


Pmax-system * (W) 44 51 53 37

ηsystem 0.289 0.262 0.184 0.109 εsystem (Wh/Kg) 1756 1591 1120 660 Eopen (V) 16.0 16.0 15.3 14.1

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.


When O2 is used as oxidant in a DMFC system, significant performance differences from those of using air as oxidant result. Table 19 lists the peak powers for the DMFC single cell, stack and system operated at 60 oC using O2 as oxidant and methanol as fuel with different concentrations. With increasing methanol concentration the peak power increases. The highest peak power is obtained at 2.0M methanol using O2 as oxidant. This phenomenon is different from that of using air as oxidant, in which the peak power is obtained at 1.0M methanol. From DMFC single cell to stack, there are 3.7%, 6.3%, 8.8% and 10.7% decrease in power for methanol concentrations at 0.5, 1.0, 2.0 and 3.0M, respectively. The higher power loss at higher methanol concentration is attributed to methanol crossover. However, from DMFC stack to system the power loss is almost the same (6 to 7 mW·cm-2·cell-1) under varying methanol concentration. Table 20 summarizes the peak energy efficiency for the DMFC single cell, stack and system under the same operating conditions as that shown in Table 19. At 0.5M methanol the highest energy efficiency 0.322 is obtained. Increasing methanol concentration will decrease the energy efficiency. At 3.0M methanol, the peak energy efficiency is only 0.117. From DMFC stack to system the peak energy efficiency is decreased by 7% to 10% depending on methanol concentration. For example, at 0.5M methanol, the peak energy efficiency is 0.322 for the DMFC stack, but only 0.289 for the DMFC system, decreasing about 10%. Table 21 summarizes peak energy density for DMFC single cell, stack and system under the same conditions as those described above. At 0.5M


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.

Peak Power Output (mW/cm2⋅cell) Methanol (M) 0.5 1.0 2.0 3.0 Single Cell 81 95 102 75 DMFC Stack 78 89 93 67 DMFC System 71 83 86 60

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.

Peak Energy Efficiency Methanol (M) 0.5 1.0 2.0 3.0 Single Cell 0.328 0.293 0.209 0.127 DMFC Stack 0.322 0.285 0.198 0.117 DMFC System 0.289 0.262 0.184 0.106

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.

Peak Energy Density (Wh/Kg) Methanol (M) 0.5 1.0 2.0 3.0 Single Cell 1996 1787 1271 771 DMFC Stack 1959 1736 1207 712 DMFC System 1756 1591 1120 642

4. Effect of Oxygen Partial Pressure by Varying Operating Temperature

4.1. Single Cell Results 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

Rongzhong Jiang 58

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 DMFC’s 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

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 100 200 300 400 500 600 700 800

Current (mA/cm2)

Vol

tage

(V)

O2 and 80 oC

O2 and 60 oC

O2 and 40 oC

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.

Table 22. Electrochemical parameters for simulation of voltage-current curves of DMFC single cell using 1.0M methanol as fuel and O2 as oxidant at various operating

temperatures.


Eopen (V) 0.61 0.65 0.68 b (V/Dec) 0.068 0.058 0.058 Rcell (Ω) 0.85 0.66 0.46 Pmax-cell (mW/cm2) 59 95 147


0

20

40

60

80

100

120

140

160

0 100 200 300 400 500 600 700 800

Current (mA/cm2)

Pow

er (W

)

O2 and 80 oC

O2 and 60 oC

O2 and 40 oC

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.

0

400

800

1200

1600

2000

0 50 100 150 200 250 300Cell Current (mA/cm2)

Ener

gy D

ensi

ty (W

h/Kg

)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

Ener

gy E

ffici

ency

Energy Density (O2 and 40 oC)



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.

Rongzhong Jiang 60

Table 23. Electrochemical parameters for simulation of methanol crossover of DMFC single cell using 1.0M methanol as fuel and O2 as oxidant at various operating

temperatures; and the peak energy efficiency and energy density obtained.


ieo (mA/cm2) 74 105 181 B 0.47 0.44 0.40 ηcell 0.271 0.294 0.275 εcell (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.

4.2. DMFC Stack Results

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.


Pmax-stack * (W) 34 55 83

ηstack 0.266 0.285 0.259 εstack (Wh/Kg) 1614 1736 1574 Eopen (V) 17.1 18.2 18.5


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.

0

4

8

12

16

20

0.0 2.0 4.0 6.0 8.0 10.0 12.0

Stack Current (A)

Vola

tge

(V)

Stack Voltage (O2 and 80 oC)



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.

0

20

40

60

80

100

0.0 2.0 4.0 6.0 8.0 10.0 12.0

Stack Current (A)

Stac

k P

ower

(W)

Stack Power (O2 and 80 oC)



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.

Rongzhong Jiang 62

0

200

400

600

800

1000

1200

1400

1600

1800

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Stack Current (A)

Ener

gy D

ensi

ty (W

h/K

g)

0.00

0.04

0.08

0.12

0.16

0.20

0.24

0.28

Ener

gy E

ffici

ency

Stack Energy (O2 and 40 oC)



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.

4.3. DMFC System Results 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


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.

0

4

8

12

16

20

0.0 2.0 4.0 6.0 8.0 10.0 12.0

Output Current (A)

Volta

ge (V

)

Sys V-out (O2 and 80 oC)Sys V-out (O2 and 60 oC)Sys V-out (O2 and 40 oC)

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.


Pmax-system * (W) 30 51 79

ηsystem 0.231 0.262 0.242 εsystem (Wh/Kg) 1403 1591 1476 Eopen (V) 14.8 16.0 17.0

Rongzhong Jiang 64

0

20

40

60

80

100

0.0 2.0 4.0 6.0 8.0 10.0 12.0

Output Current (A)

Pow

er (W

)

Sys Power (O2 and 40 oC)



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

400

800

1200

1600

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Output Current (A)

Ener

gy D

ensi

ty (W

h/K

g)

0.00

0.04

0.08

0.12

0.16

0.20

0.24

0.28

Ener

gy E

ffici

ency

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.


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.


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.

Rongzhong Jiang 66

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.

Peak Power Output (mW/cm2⋅cell) Temperature (oC) 40 60 80 Single Cell 59 95 147 DMFC Stack 55 89 135 DMFC System 49 83 128

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.

Peak Energy Efficiency Temperature (oC) 40 60 80 Single Cell 0.272 0.293 0.275 DMFC Stack 0.265 0.286 0.258 DMFC System 0.230 0.262 0.242

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.

Peak Energy Density (Wh/Kg) Temperature (oC) 40 60 80 Single Cell 1651 1787 1627 DMFC Stack 1614 1736 1573 DMFC System 1403 1591 1475

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 cell’s 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.


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 mW⋅cm-2⋅cell-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 mW·cm-2·cell-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 Wh·Kg-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 mW⋅cm-2⋅cell-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 mW·cm-2·cell-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 Wh·Kg-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.

Rongzhong Jiang 68

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.


[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.

Rongzhong Jiang 70

[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.


Chapter 2

EXPERIMENTAL ACTIVITY ON A LARGE SOFCGENERATOR

M. Santarelli*, P. Leone, M. Calì and G. Orsello1

Dipartimento di Energetica. Politecnico di Torino. Corso Duca degli Abruzzi 24,10129 Torino (Italy)

1TurboCare S.p.A., Corso Romania 661, Torino (Italy)

Abstract

The Multiscale Analysis Group of the Politecnico di Torino (Italy) is involved in theexperimental analysis and modeling of the CHP-100 SOFC Field Unit built by Siemens PowerGeneration Stationary Fuel Cells (SPG-SFC). The experimental analysis of a large SOFCgenerator in operation is a complex task, due to the large number of variables which affect itsoperation, the limited number of measurements points in the generator volume, the necessityto avoid malfunctions in the real operation. As a consequence, the experimental analysis of theCHP-100 SOFC Field Unit has been developed with methods of Design of Experiments, andwith a statistical analysis of the collected data. The experimental sessions have been designedin order to investigate the effect of two important operation factors (the overall fuelconsumption FC and the air stoichs λox), in order to characterize the operation of the singlesectors of the SOFC generator, and to obtain the sensitivity maps of the main investigateddependent variables. Furthermore, the main result is the estimation of the local values of fuelutilization of the various sectors of the generator, through the combination of the experimentalvoltage sensitivity analysis to overall FC and an analytical model of polarization, to outlinethe distribution of fuel inside the generator. Finally, the sectors of the generator, of differentpedigree and position, are compared in terms of the polarization effects, showing how thelocal fuel utilization and temperature affect the estimated local anode exchange currentdensity values.

Keywords: large SOFC generator, experimental activity, design of experiment, regressionmodels, sensitivity analysis, polarization effects.

* E-mail address: [email protected]

M. Santarelli, P. Leone, M. Calì et al.72

1. Introduction

The Multiscale Analysis Group of the Politecnico di Torino (Italy) is involved in theexperimental analysis and modeling of the CHP-100 SOFC Field Unit built by SiemensPower Generation Stationary Fuel Cells (SPG-SFC). The generator has been installed in aSOFC laboratory at TurboCare S.p.A. (a subsidiary of SPG-SFC) since June 19, 2005, and todate has shown the record availability of 99.5%. The SOFC CHP-100 is the first to utilize thecommercial prototype air electrode supported cells and in-stack reformers. The generator isfed with natural gas from the grid. The stack is composed of 1152 single tubes, arranged in 48cell bundles (4 cell bundles are connected in series to form a bundle row, and 12 bundle rowsare 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, theexperimental environment is completely different compared to a laboratory, because it isdifficult 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, inorder 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 designof the experiment coupled with a consequent statistical analysis of the collected data, in orderto be able to outline the significance of every observed effect.

Moreover, the large plant is characterized by a not uniform distribution of its physicaland chemical variables (e.g. temperatures, chemical composition of mass flows, etc.), and atthe same time the measurement points are discrete: therefore, in many parts of the plant thedata are not available. As a consequence, a combination of some experimental techniques andof analytical models could be used to deduce the distribution, inside the generator volume, ofsome important variables.

Finally, the generator is in real operating conditions, and therefore its experimentalanalysis has to avoid malfunctions and dangerous operations; usually, one possibility is theperturbation of some independent variables in a safety experimental range, and the analysis ofthe 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 beendeveloped with methods of Design of Experiments, and with a statistical analysis of thecollected data.

A design of experiments procedure has been used to study in a rigorous manner thecollected experimental data and also to relate, through analytical expressions, the investigateddependent variables with the control factors (independent variables). The design ofexperiments approach allows one to study the main effect of the factors and also theirinteraction effects. The aim is the description of the analytical relations which express thedependent variables as a function of the examined control factors through first (simplefactorial) or second-order (spherical CCD) regression models.

The experimental sessions have been designed in order to investigate the effect of twoimportant operation factors: the overall fuel consumption (FC) and the air stoichs (λox). Themain expectation has been the characterization of the operation of the single sectors of theSOFC 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 interms of sensitivity maps of the main investigated dependent variables (voltages).

The sensitivity analysis of the local voltage to the overall fuel consumption modificationscan 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 byderiving the polarization curve respect to FU, it is possible to link the distribution of voltagesensitivities to the overall FC to the distribution of the local FU. Therefore, the main purposeof the previous sensitivity analysis is to estimate the local values of fuel utilization of thestack sectors, through the combination of the experimental voltage sensitivity analysis to FCand the analytical model. This is useful to outline its drawbacks on the distribution of fuelinside the generator: the results are used to debate on the issue of fuel distribution systems inthe generator.

Finally, the sectors of the generator, of different pedigree and position, are compared interms of the main polarization effects. The polarization behavior has been investigated atfixed setpoint temperature and overall fuel consumption. On this basis, a parameterestimation, using a polarization model of a single cell, has been performed and used for theanalysis of the local activation effects. It has been shown how the local fuel utilization andtemperature affect the estimated local anode exchange current density values.

2. Description of the Plant

The CHP-100 kWe SOFC Field Unit (Siemens Power Generation-Stationary Fuel Cells) is thefirst to utilize the commercial prototype air electrode supported cells (22 mm diameter, 150

Figure 1. Picture of the SOFC CHP-100 test site in TurboCare (Torino, Italy).


cm active length, 834 cm2 active area) and in-stack reformers. The generator is fed withnatural gas from the grid distribution. In Figure 1 the picture representing the CHP 100 testsite in TurboCare (Torino, Italy) is shown.

The upper level of the system hierarchy after the single cell is the cell bundle, whichconsists of a 24-cell array arranged as 8 cells in electrical series by 3 cells in electricalparallel. Two bundles in series form a sector; two sectors are connected in series to form arow, 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 cellstack arrangement is shown in Figure 2 [1], outlining the position of the fuel ejectors and thepower leads. In this Chapter, the results will be discussed dividing ideally the primarygenerator in four main zones (North; South; Power Leads side and Ejectors side), and also inthe central zone (in the middle between the Power Leads and the Ejectors sides).

NORTH SOUTH

Figure 2. Schematic of the stack arrangement.

Figure 3. Simplified flow schematic of the SOFC CHP-100 BoP.


The system is completed by the balance of plant (BoP), with five major skids: GeneratorModule, 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) theelectrochemical generator; (b) the Power Conditioning System (PCS): an inverter, whichoperates 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 theSOFC 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 someoffices of the TurboCare factory.

The simplified flow schematic of the SOFC CHP-100 BoP is shown in Figure 3, while inFigure 4 the simplified schematic of the primary generator structure is reported.

Figure 4. Simplified schematic of the primary generator structure.

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 runhours of around 32,000 (including run hours in the Netherlands and Germany); a averagestack temperature of 954°C; 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% ofthe workshop requirement); a heat generation of 60 kWt (hot water @85°C). Othercharacteristics 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, onannual basis); 4 stops (one operation error, three inverter failures); 24 successful operationtransition to power dissipator due to utility (AEMD, Azienda Energetica Metropolitana


Torino Distribuzione) or TurboCare grid failures; one complete thermal cycle (August 2006);a very dependable automatic operation (no operator in control room); a remote controlcapability via modem; a easy maintenance (replacement of air filters and of reactantdesulphurization).

In one year of operation the SOFC CHP-100 fuelled by natural gas avoids the productionof 272 tons of CO2, the environmental pollution of 1035 kg of NOx, the import of 121 tons ofoil 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 energyproduced by a fuel cell generator (even if using a fossil fuel): the energy production estimatein 2006 is 900 MWh (18 certificates).

Politecnico of Torino has developed several activities, like the design of the SOFCthermal plant [4], the modeling of the generator and of the BoP [5-8], the safety design andanalysis [9]. In [10] a model of the operation of the SOFC CHP-100 has been developedusing a 0D approach and validated through a first session of experimental tests. In [11,12] theeffect of the setpoint generator temperature and fuel consumption factors on severaldependent variables (i.e. DC and AC electric power, recovered heat, electric and globalefficiency, efficiency of the pre-reforming process) is analyzed in form of screening tests andthe process responses are treated in form of response surface plots. In [13] the regressionmodels have been used in constrained optimization procedures to maximize differentobjective functions (AC electric power, recovered heat, etc.).

3. Methodology and Description of the Experimental Session

3.1. Regression Models and Sensitivity Analysis of the Voltage DistributionInside the Plant

3.1.1. Design of Experiment Method

To develop the study, the factorial analysis and the response surface method (RSM) havebeen applied, and first and second-order regression models linking the dependent variables tothe 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 ofinterest to model and explore this relationship. Suppose that there is a single dependentvariable or response Y that depends on k independent or regressor variables kxxx ,...,, 21 . Therelationship between these variables is characterized by a mathematical model, called aregression model, that fits a set of sample data. In some instances, the experimenter knows theexact form of the true functional relationship between Y and kxxx ,...,, 21 . However, in mostcases, the true functional relationship is unknown and the experimenter can choose anappropriate function in order to approximate it. Low-order polynomial models are widelyused as approximating functions. Thus, a regression model can be written in the followingform:


εββββ +++++= kk xxxY ...22110 (1)

It is called a multiple linear regression model with k regressor variables. The parametersβj (j=0,1…,k) are called the regression coefficients. The parameter βj represents the expectedchange in the response Y per unit change in xj when all the remaining independent variables xi

(i≠j) are held constant. Thus, the regression coefficient of a regressor variable represents thesensitivity parameter of the dependent variable to changes in the regressor variable(independent variable). The method of least squares is typically used to estimate theregression coefficients in a multiple linear regression model.

If an experimental session is designed by applying the methodology of the Design ofExperiments, then the investigated dependent variables can be described by regressionmodels. Different forms of regression models can be obtained from an experimental sessionaccording to the chosen design of experiments, for example first (equation (1)) or secondorder regression models (equation (2)):

εββββ ++++= ∑∑∑∑=<=

k

jjjj

jijiij

k

jjj xxxxY

1

2

10 (2)

Because of the reduced experimental domain (due to safe operation of the real plant), theanalysis described in this Chapter will deal with results obtained by a first-order design.

The collected data are first analyzed through the Yate’s method (analysis of the factorsignificance on the dependent variable); after, first-order regression models are obtained.Particular attention has been addressed in the description of the operation in terms ofsensitivity maps.

To better understand the methods of Design of Experiment used in the experimentalsessions, in the Appendix the statistical concepts and the methods used in the analysis arebriefly resumed. The analysis is described in the following Paragraphs.

3.1.2. Experimental Session

The experimental session has been designed in order to evaluate the main and interactioneffects of two independent variables (factors), the overall fuel consumption (FC) and the airstoichs (λox), on the distribution of voltages in different sectors of the generator.

Fuel Utilization and Fuel ConsumptionThe overall fuel utilization factor is defined as the ratio of the fuel mass flow which operatesthe electrochemical oxidation on the anode surface and the total fuel mass flow entering in thegenerator.

totf

echf

GG

FU,

,= (3)


The control variable overall fuel consumption (FC) factor is defined in the same way, butat the numerator it takes into account also the fuel mass flow which is chemically oxidized onthe anode side by the air leakage from cathode to anode, plus the fuel leakage through theelectrolyte layer (in the order of 20÷30 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 fuelutilization value is evaluated by the control system of the generator. An indicative differencebetween overall fuel utilization and overall fuel consumption is as follows: imposing anoverall fuel consumption value of 83.75%, the control overview shows an overall fuelutilization value of 79%.

Air stoichsThe air stoichs factor is defined as the ratio of the total air mass flow which enters in thegenerator and the air mass flow actually used in the electrochemical reaction on the cathodesurface.

echair

totairox G

G

,

,=λ (4)

In other words, 4.5 stoichs means an introduction inside the generator of the 450% of airmass flow compared to the actually used value; an increase of 0.1 stoichs means an increaseof 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 speedof the air blowers, named AIRO (speed of air primary blower, expressed as percentage of itsmaximum value).

ProcedureIn the experimental session, the average current density has been kept constant at the nominalvalue of 0.2 A/cm2 (500 A generator current).

The generator setpoint temperature (TGEN) is the highest temperature measured by fivethermocouples placed in the central zone of the generator. This variable can be controlled bythe operator, but in the experimental sessions it has not been considered as a independentvariable (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 (λox).

A design of experiments procedure has been used in the planning of the experimentalsession; this has been done with the aims to study in a rigorous manner the collectedexperimental data and also to relate, through analytical expressions, the investigateddependent variables with the control factors (independent variables). The design ofexperiments 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 applyingsimple 22 factorial analysis. The aim is the description of the analytical relations whichexpress the dependent variables as function of the two control factors through first-order(simple 22) regression models.


Each single combination of the factors has been replicated twice to evaluate thesignificance of the factor modification on every dependent variable. Furthermore, the factorialanalysis 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 factorshave been found and analyzed with an ANOVA.

The ANOVA both of the experimental data and of the first-order regression models havebeen developed using the factorial of Table 1, where the treatment combinations and therange of variation of each factor are reported:

• 8.46.4 ÷=oxλ• %25.8475.81 ÷=FC

The range of variation of each factor has been imposed in agreement with TurboCare andSPG-SFC, in order to avoid malfunctions of the generator.

Table 1. Treatment combinations and levels of the two factors FC and oxλ

Treatment Factor level

λox=4.6(1)

FC=81.75%

λox=4.8(a)

FC=81.75%

λox=4.6(b)

FC=84.25%

λox=4.8(ab)

FC=84.25%

Due to the slow thermal modification of the generator, the air stoichs were imposed everyday at 6 p.m., to reach a stabilized temperature of the generator during the night. During theday, the tests with constant oxλ and variable FC were performed, thanks to the fastadaptation of the generator to the overall fuel consumption factor modifications.

Imposed the FC value, the data collection started after 60 min (to stabilize the signaldata) for a length of 30 min, with a rate of 1 min. The single value of a variable is an averageof 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 Yate’s method (analysis of the factorsignificance on the dependent variable).

Second, first-order regression models have been obtained, with the consequent responsesurfaces and contour plots.

λox


Third, particular attention has been addressed in the description of the operation in termsof sensitivity maps of the investigated dependent variables (voltages) to the air stoichs (λox),and the overall fuel consumption (FC).

3.2. Analysis of the Polarization Effects of the Various Sectors at Variable LocalFuel Utilization and Temperature

In the experimental session we have obtained also the polarization curves that describe theelectrochemical behaviour of the various sectors of the SOFC generator. The curves havebeen obtained varying the values of the overall fuel consumption in two different sequentialranges of current:

• I=435÷475 A with overall fuel consumption FC=84.25% and setpoint temperatureTGEN=967°C

• I=475÷495 A with overall fuel consumption FC=83.75% and setpoint temperatureTGEN=967°C

The tests have been performed maintaining constant the generator setpoint temperature atthe value of 967°C: this value of temperature is automatically maintained from the systemacting on the air flow. A current variation causes a variation of the inlet air temperature: thisis controlled by a by-pass valve of the low temperature heat recuperator. Besides, a currentvariation causes a variation in the thermodynamics condition of the stack: to maintainconstant 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 timesteps 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 fuelconsumption was imposed, and after the load was increased with the same regular ramp (5 Aevery 30 min), up to the value of 495 A.

The process data are averaged out on the whole time interval where the current isconstant. 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 toacquire the data every second. The length of acquisition for the test has been 30 minutes: onthis 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 setpointtemperature, and the overall fuel consumption in two different values. All the other variableswere 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 theoverall fuel consumption (Paragraph 3.1), and this has outlined how the fuel distributionaffects the polarization behavior.

Then, a deeper analysis has been made by performing a parameter estimation, allowingan analysis of the local anode activation effects with variable local fuel utilization andtemperature: the anode exchange current density of every sector has been the estimatedparameter, and its strict relation to the local fuel utilization of every sector (Paragraph 3.1)has been outlined.


Finally, the multi-pedigree of the stack arrangement has been characterized by comparingthe different estimated anode exchange current densities of the various sectors of differentpedigree and position inside the generator.

4. Voltage Sensitivity Analysis

In this Paragraph the results of the sensitivity analysis of the voltage of the various sectors ofthe generator (local voltage) to the overall fuel consumption and air stoichs modifications areshown. As said in the Paragraph 3.1, the coefficients of the regression models represent thesensitivity coefficients of the analyzed dependent variable (the local voltage) to the factors. Inthis case, the coefficients are expressed in mV/%FC (sensitivity to overall fuel consumption,considering the sensitivity to the 1% modification of the overall FC variable) andmV/0.1stoichs (sensitivity to air stoichs, considering the sensitivity to the 10% modificationof the λox variable), being the FC and λox the factors (inputs of the operator).

2 4 6 8 10 12 14 16 18 20 22 240.61

0.62

0.63

0.64

0.65

0.66

0.67

0.68Voltage distribution in the main stack sectors

Cell voltage/V

Sectors

Figure 5. Distribution in the stack sectors of the voltage of the single cell (I=435 A, setpointtemperature TGEN=967°C, FC=84.25%, λox=4.6)

In order to better understand the discussion about voltage sensitivity tests, some resultsare presented about the voltage distribution in the stack [17]. In Figure 5 the distribution inthe stack sectors of the voltage of the single cell is shown for the following operatingcondition (I=435 A, setpoint temperature TGEN=967°C, FC=84.25%, λox=4.6). The completevoltage distribution data is not shown due to a non disclosure agreement with Siemens PowerGeneration.

The average cell voltage is around 14666 ± mV, and the estimated average area

specific resistance is around 208.004.1 cm⋅Ω± , comparable with the data reported insome literature papers [3,18]. The performance homogeneity (obtained by the estimatedstandard deviation) of the single cell voltage is then 2%, which is a very good index of theoperation; during different experimental sessions, values of performance homogeneity of thesingle cell voltage have been found up to 5%. It is also evident that the edge sectors show alower cell voltage.


4.1. Voltage Sensitivity to Overall Fuel Consumption

In the next graphs the distribution of the sensitivity of the voltage of the various sectors to theoverall 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 andEjectors side; in Figure 8 the value of the voltage sensitivity of the sectors is divided in Northand South side.

side

Bundle row

Voltage sensitivity to fuel consum ption (absolute value)

Voltage sensitivity [mV/%FC]

Power leads EjectorSouth

North

38

40

42

44

46

48

50

52

54

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 Ejectorsside.


Board Center-56

-54

-52

-50

-48

-46

-44

-42

-40

-38

-36

Sectors

Voltage sensitivity [mV/%FC]

Voltage sensitivity to fuel consumption

North sideSouth side

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 thesectors to the overall fuel consumption are reported.

Table 1. Mean and standard deviation for the voltage sensitivity of the sectors to overallfuel consumption divided in total, Power Leads and Ejectors side, North and South side

Voltage sensitivity of the sectors to overall fuel consumptionSide SideTotal Power Leads Ejectors North South

mean [mV/%FC] -45.02 -44.71 -45.33 -42.68 -47.37standard deviation [mV/%FC] 4.69 4.35 5.17 3.26 4.83

The mean sensitivity value for the Power Leads side is similar to the one of the Ejectorsside, but in this side the values are more dispersed.

The most evident difference is between the sectors placed at the center and the twoboundaries of the generator: the sectors at the boundaries have a higher voltage sensitivity tothe overall fuel consumption. The different behaviour could be related to the fluid-dynamiceffects caused by the position of the two ejectors: the segments more far from the ejectorsseem to be more sensitive to variation of the overall FC factor, and this is probably linked tolower local values of fuel utilization.

Another explanation could be linked to the fact that the hydrocarbon fractions (essentiallymethane) are not totally converted in the reforming boards. Therefore, part of the reformingreactions occurs at the beginning of the tubes, that is at the bottom plane. This causes a localcooling effect. A variation of the overall FC factor determines a modification of the chemicalequilibrium inside the reformer boards, and as a consequence a modification of the coolingeffect at the bottom plane: when increasing the overall fuel consumption, there is a decrease


of the residual reforming reaction at the bottom plane of the tube; because this reaction isendothermic, this causes a increase of the local temperature. This local overheating causes acurrent density gradient along the tube, with an increase of overpotentials and, as a finalconsequence, this fact can reduce the voltage of the segment. This is emphasized at theboundary rows, where probably the reforming reactions are reduced (the reformer boardsplaced at the external boundaries of the generator receive the water vapour rich-fuel gasthrough orifices with lower diameter, while the reformers placed between the internal bundlerows receive the fuel gas through orifices with higher diameter): thus the voltage reductioneffect 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 ofthe two ejectors: probably in the South side a minor amount of fuel gas arrives (especially atthe boundary), and the operation is characterized by a local FC factor higher than the nominalone. 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 fuelconsumption produces a decrease of around 4.5% in the voltage of the sectors.

4.2. Voltage Sensitivity to Air Stoichs

In the next graphs the distribution of the sensitivity of the voltage of the various sectors to theair stoichs factor is shown: in Figure 9 the voltage sensitivity map; in Figure 10 the value ofthe voltage sensitivity of the sectors divided in Power Leads and Ejectors side; in Figure 11the value of the voltage sensitivity of the sectors divided in North and South side.

side

Bundle row

Voltage sensitivity to stoichs (absolute value)Voltage sensitivity [mV/0.1 stoichs]


North

8

10

12

14

16

18

20

22

24

26

28

Figure 9. Distribution map of the sectors voltage sensitivity to air stoichs.


1 2 3 4 5 6 7 8 9 10 11 12-35

-30

-25

-20

-15

-10

-5

Rows

Voltage sensitivity [mV/0.1 stoichs]

Voltage sensitivity to stoichs

Power lead sideEjector side

not significant

Figure 10. Sectors voltage sensitivity to air stoichs divided in Power Leads and Ejectors side.

Board Center-35

-30

-25

-20

-15

-10

-5

Sectors

Voltage sensitivity [mV/0.1 stoichs]

Voltage sensitivity to stoichs

South sideNorth sidenot significant

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 thesectors to the air stoichs are reported.


Table 3. Mean and standard deviation for the voltage sensitivity of the sectors to airstoichs divided in total, Power Leads and Ejectors side, North and South side

Voltage sensitivity of the sectors to air stoichsSide SideTotal Power lead Ejector North South

mean [mV/stoichs] -206.77 -234.38 -179.17 -234.38 -179.17standard deviation [mV/stoichs] 72.42 41.33 87.15 69.11 67.28

The Power Leads side is more sensitive and the values are less dispersed than at theEjectors side. The North side is more sensitive and the values are more dispersed than in theSouth side.

From the graphs, the center rows result to be the more sensitive to a variation of thefactor; this can be linked to the fact that the air is introduced from the piping in the center of aplenum over the stack, and next it is sent to the sectors: in this way, the central sectors receivea greater flow whereas the edge sectors receive less air. This effect is less pronounced in thePower Leads side.

2 4 6 8 10 12 14 16 18 20 22 24-60

-50

-40

-30

-20

-10

0

Sectors

FC sensitivity mV/%FC

sensitivity mV/0.1Stoichs

Segm ent distribution of voltage sensitivity

FCFC regStoichs/10Stoichs reg

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, thesesectors are characterized by a high sensitivity to the overall fuel consumption, which tends topredominate on the other experimental factors such as the air stoichs.

The average sector sensitivity is around –200 mV/stoichs. The increase of the airstoichiometry λox leads to a significant decrease of voltage in all the sectors (except for theones placed at the generator edges: but we have evaluated that the effect is statistically notsignificant). Note that the central zone of the generator shows a maximum of sensitivity of


voltage to air stoichiometry: it is interesting to notice the parabolic shape of the sensitivitydistribution along the bundle rows and the complementary behaviour between the voltagesensitivity to air stoichs and overall fuel consumption.

As the SOFC CHP-100 is an air-cooled system, the decrease of voltage, due to anincrease of air stoichiometry, is mainly due to the variation of the equilibrium temperature: anincrease of λox causes a reduction of the equilibrium temperature with consequent increase ofovervoltages (mainly ohmic contribution). This decrease of the temperature is almost uniformin all the bundle rows of the generator, then also at the edges. Thus, it would be expected thatalso the edge sectors will present the same behaviour, that is a reduction of voltage withincreasing of λox. The negligible effect of this operation on these sectors is explained with theobservation that the local fuel consumption is the dominant factor of the operation in the edgesectors.

4.3. Voltage Sensitivity: Correlation between Factors

In Figure 12 the graph of the voltage sensitivity of the various sectors to overall fuelconsumption and air stoichs is shown. The values of the voltage sensitivity to air stoichs havebeen divided by 10 to better show and compare the results.

Fuel consum ption

Side

Bundle row

Voltage sensitivity m ap (absolute value)

FC sensitiv

ity m

V/%

FC sensitiv

ity m

V/0.1Stoichs


North Stoichs/10

SidePower leads EjectorSouth

North

5

10

15

20

25

30

35

40

45

50

not significant

Figure 13. Distribution map of the sectors voltage sensitivity to overall fuel consumption and airstoichs.

In Figure 13 the voltage sensitivity map is reported.The main effect of the voltage sensitivity to overall fuel consumption and air stoichs is

negative.The effect of the air stoichs is coincident to a modification of the setpoint temperature of

the generator; therefore, where the voltage sensitivity to air stoichs is high, it means that thevoltage sensitivity to a modification of the setpoint is high. Of course, this is due to the fact


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 temperatureand consequently of the local voltage, while an increase of air stoichs means a decrease of thelocal temperature and consequently of the local voltage. Concerning the distribution insidethe generator, it seems that the cooling effect is more effective in the center of the generatorcompared to the boundaries, and this could be linked to the distribution of the cooling air, andto the fact that the center is the zone with the highest average temperatures due to the lowerheat transfer to the outside.

The effect of the overall fuel consumption is not properly a thermal effect, but especiallya fluid-dynamic effect which influences the Nernst contribution to the polarization. It is veryevident at the boundaries of the generator, where we have observed (see [19]) that the fueldistribution is not optimised and the reforming reactions in the reformer boards have a lowerkinetic.

It could be interesting to analyze the fact that the boundary rows (1 and 12) have thereformer unit on the external side working with less fuel gas flow (the reformer boards placedat the external boundaries of the generator receive the water vapour rich-fuel gas throughorifices with a lower diameter, while the reformers placed between the internal bundle rowsreceive the fuel gas through orifices with a higher diameter); this could mean that, inparticular, the external sides of the rows 1 and 12 are not optimally cooled by the endothermicreforming reaction. This probably could affect the values of temperature and voltageassociated to the sectors at the boundary zones.

2 4 6 8 10 12-55

-50

-45

-40

-35

-30

-25

-20

-15

Rows

FC sensitiv

ity m

V/%

FC

sensitiv

ity m

V/0.1Stoichs

Voltage sensitivity of the segm ents placed at the Power Leads side

2 4 6 8 10 12-60

-50

-40

-30

-20

-10

0

Voltage sensitivity of the segm ents placed at the Ejectors side

Rows


notsignificant

Figure 14. Sectors voltage sensitivity to overall fuel consumption and air stoichs divided in PowerLeads and Ejector side.

Where the voltage sensitivity to overall fuel consumption is higher the voltage sensitivityto air stoichs is not significant: in these sectors (at the boundaries of the generator) the effectof 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 biginfluence on fuel distribution.

It is possible to notice the symmetry between the distribution of the voltage sensitivity tooverall fuel consumption and air stoichs: where the sensitivity to overall fuel consumption ishigh 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 voltagechanging these two regulation factors: during a regulation of one factor, the other one couldbe 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.

2 4 6 8 10 12-50

-45

-40

-35

-30

-25

-20

-15

-10

Rows

Voltage sensitivity of the segm ents placed at the North side


2 4 6 8 10 12-60

-50

-40

-30

-20

-10

0

Voltage sensitivity of the segm ents placed at the South side

Rows

FC sensitiv

ity m

V/%

FC

sensitiv

ity m

V/0.1Stoichs

Figure 15. Sectors voltage sensitivity to overall fuel consumption and air stoichs divided in North andSouth side.

The voltage sensitivity distribution to the factors for the North and South side is nearlythe same: every part of the generator seems to have the same trend and it seems that thereisn’t 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 thiscould be linked to fluid-dynamic considerations, especially concerning the fuel distributioncoming from the ejectors. In particular, the Ejectors side is more sensitive to overall FCvariation. 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 thereforming reactions occurs at the beginning of the tubes, that is at the bottom plane. Thiscauses a local cooling effect. A variation of the overall FC factor determines a modificationof the chemical equilibrium inside the reformer boards, and as a consequence a modificationof the cooling effect at the bottom plane: when increasing the overall fuel consumption, there


is a decrease of the residual reforming reaction at the bottom plane of the tube; because thisreaction is endothermic, this causes a increase of the local temperature. This local overheatingcauses a current density gradient along the tube, with an increase of overpotentials and, as afinal consequence, this fact can reduce the voltage of the segment. This is emphasized at theboundary rows, where the reforming reactions are reduced (the reformer boards placed at theexternal boundaries of the generator receive the water vapour rich-fuel gas through orificeswith lower diameter, while the reformers placed between the internal bundle rows receive thefuel gas through orifices with higher diameter), and therefore the cooling effect is moreevident on this part of the generator.

Finally, another difference from Power Leads and Ejectors side is the sensitivity to airstoichs: at the Ejectors side at the boundaries they show a not significant effect; the samesectors are the ones with the highest sensitivity to overall FC, which could represent apredominant effect. Moreover, the center rows result to be the most sensitive to a variation ofthe air stoichs; this can be linked to the fact that the air is introduced from the piping in thecenter of a plenum over the stack, and next it is sent to the sectors: in this way, the centralsectors receive a greater air flow whereas the edge sectors receive less air. This effect is lesspronounced 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 thevoltage sensitivity of the single cell to both overall fuel consumption and air stoichs, outlinedfor cells placed at the Ejectors and Power Leads sides.

The estimation of the single cell sensitivity to overall fuel consumption is around -2.8mVcell/%FC, comparable with values of some literature papers [16]; the homogeneity of theoverall FC sensitivity among the various cells is around 10%. The single cell sensitivity to air


stoichs has been estimated in around -1.3 mVcell/0.1stoichs. Also from this figure it is clearthat 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 correlationswith the generator setpoint temperature (Figure 18: correlation between the sensitivity to airstoichs and generator setpoint temperature; Figure 19: correlation between the sensitivity tooverall fuel consumption and generator setpoint temperature).

-56 -54 -52 -50 -48 -46 -44 -42 -40 -38 -36-35

-30

-25

-20

-15

-10

-5

1

2

3

4

5

6

78

9

10

1112

13

14

15 16

17

18

19

20

21

2223

24

Voltage sensitivity

Voltage sensitivity to fuel consum ption [m V/% FC]

Volta

ge sensitiv

ity to air stoichs [m

V/0.1 stoichs]

not significant valuesof sensitivity to

stoichs

Figure 17. Correlation between voltage sensitivity to air stoichs and to overall fuel consumption.

-3 -2 -1 0 1 2 3 4 5 6-35

-30

-25

-20

-15

-10

-5

1

2

3

4

5

6

78

9

10

11 12

13

14

1516

17

18

19

20

21

2223

24

Voltage sensitivity

Voltage sensitivity to setpoint tem perature [m V/°C]

Volta

ge sensitiv

ity to air stoichs [m

V/0.1 stoichs]

not significant valuesof sensitivity to

stoichs

not significantvalues of

sensitivity tosetpoint

temperature

Figure 18. Correlation between voltage sensitivity to air stoichs and to generator setpoint temperature.


-56 -54 -52 -50 -48 -46 -44 -42 -40 -38 -36-3

-2

-1

0

1

2

3

4

5

6

1

2

3

4

5

6

7

8

9

10

11

1213

14

15 16

17

1819

20

21

2223

24

Voltage sensitivity

Voltage sensitivity to fuel consum ption [m V/% FC]

Volta

ge sensitiv

ity to setpoint temperature [mV/°C]

not significantvalues of

sensitivity tosetpoint

temperature

Figure 19. Correlation between voltage sensitivity to overall fuel consumption and generator setpointtemperature.

As expected, it seems to occur a linear correlation between the voltage sensitivity to airstoichs and to generator setpoint temperature (Figure 18). It is a direct correlation (sametrend), because both cause a thermal effect on the generator.

Concerning the correlation between the other two factors, the correlation seems veryweak, and it appears as an inverse correlation.

Concerning the sectors 2, 3, 6, 22 and 23, where the voltage sensitivity to air stoichs isnot significant, they appears also to be the only segments with negative sensitivity to thegenerator setpoint temperature. In particular, they have the highest value of sensitivity tooverall fuel consumption. As already commented, in these sectors (at the boundaries of thegenerator) the effect of the overall fuel consumption is very high and the other factors do nothave a significant influence on the voltage. In particular, these sectors are close to theejectors, which have a big influence on the fuel distribution.

5. Deduction of the Local Fuel Utilization Distribution

The following results were obtained from the experimental analysis: stack terminal voltagesensitivity -1.06 V/%FC; sector sensitivity -44 mVsector/%FC; single cell sensitivity -2.8mVcell/%FC. It has been shown in Figure 5 that the homogeneity of the voltage sensitivity tooverall FC among the various cells is around 10%. The different behavior of the voltagesensitivity is mainly addressed to the local fuel utilization operation, thus to the design of thefuel distribution system.

In this Paragraph, the relation between the sensitivity of the sector voltage to the overallfuel consumption and the local fuel utilization of the sectors is described. The local fuel


utilization of the sectors is estimated coupling the experimental voltage sensitivity to overallfuel 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 singlecell voltage sensitivity to FU has been developed. Each term of the polarization model isdiscussed and then some hypothesis are assumed for the particular case study of a tubularcathode supported cell (used in the SOGC CHP-100 plant).

5.1. Analytical Expression of the Sensitivity of the Cell Voltage to the FuelUtilization

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 thefuel utilization does not cause any effect on the ohmic overpotential inside the cell layers, andtherefore the ohmic overpotential is not affected by the fuel utilization term.

The cell voltage of a tubular SOFC is expressed by the equation:

caconcohmcaactNernstc VV /,/, ηηη −−−= (5)

where cV is the cell voltage, NernstV is the Nernst potential, caact |,η is the activation

overpotential, ohmη is the ohmic overpotential, caconc |,η is the diffusion overpotential.

First, if the performance of a cell is mainly limited by the ohmic contribution and theother terms of polarization play a marginal role, then the sensitivity of the cell voltage to FUis 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 followingequation:

( ) ( )( ) ⎟

⎟

⎠

⎞

⎜⎜

⎝

⎛ ⋅⋅

⋅⋅

+=OH

OHNernst y

yyFTREV

2

22

2/1

0 ln2

(6)

Under the assumption that the oxygen utilization is negligible, that is the oxidant is fed tothe cell far in excess of the stoichiometrically required amount, the Nernst potential is afunction of the operating temperature and of the axial location along the cell length (x), atfixed pressure:

( ) ( )( ) ⎟

⎟

⎠

⎞

⎜⎜

⎝

⎛ ⋅⋅

⋅⋅

+=)(

)(ln

2)(),(

2

22

2/1

0 xyyxy

FTRTETxV

OH

OHNernst (7)

The average Nernst potential is the x-averaged Nernst potential, that is:


∫ ⋅=L

NernstNernst dxxVL

V0

)(1(8)

According to [16,21], it is shown that the sensitivity of voltage to FU is dependent on theaverage current density, due to the variation with current of the consumption of reactantsalong the cell axis. In particular, the sensitivity decreases with the increase of current. In orderto model this behaviour, some assumptions are now made. At high current densities it can beassumed that the Nernst potential is linear with the cell axial position:

( ) ( )( ) BxA

xyyxy

FTRTETxV

OH

OHNernst +⋅=⎟

⎟⎠

⎞⎜⎜⎝

⎛ ⋅⋅

⋅⋅

+=)(

)(ln

2)(),(

2

22

2/1

0 (9)

This refers to an operation with a variation of the reactants partial pressure along the celllength according to the exponential equation:

xy

LyL

iii

i

eyxy⋅⎟⎟⎠

⎞⎜⎜⎝

⎛⋅

⋅= )0()(

log1

)0()( (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:

0

0

22

22

)0(

)0(

OHyy

yy

OH

HH

=

=(11)

in the fuel mixture (anode inlet composition).At x=L (open end), that is at the cell outlet, it can be written:

FUyyLy

FUyLy

HOHOH

HH

⋅+=

−⋅=00

0

222

22

)(

)1()((12)

The x-averaged Nernst potential at high current densities is thus written in the form:

( ) ( )( ) ⎟

⎟

⎠

⎞

⎜⎜

⎝

⎛

⋅+

⋅−⋅⋅

⋅⋅

+=FUyyyFUy

FTRTETV

HOH

OHHCNernst 00

2/10

0,22

22)1(

ln4

)()( (13)

At low current densities, a significant length of the cell operates close to the exit Nernstpotential, that is most of the fuel is consumed very close to the fuel inlet (closed end of the


cell). As a consequence, at low current densities the average Nernst potential is approximatedto the exit Nernst potential and maintained constant along the cell tube:

( ) ( )( ) ⎟

⎟

⎠

⎞

⎜⎜

⎝

⎛

⋅+

⋅−⋅⋅

⋅⋅

+=FUyyyFUy

FTRTETV

HOH

OHLCNernst 00

2/10

0,22

22)1(

ln2

)()( (14)

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 thesensitivity of the Nernst potential to the same parameter. With this assumption it is possible towrite:

dFUdV

dFUdV Nernstc = (15)

With this assumption, the first derivative of the cell voltage with respect to the fuelutilization (that is, the voltage sensitivity to the fuel utilization) at low current densities istwice the same term at high current densities: in fact, at high current density the voltagesensitivity is given by the equation (16), while at low current density it is given by theequation (17).

⎥⎥⎦

⎤

⎢⎢⎣

⎡

+⋅−−=

∂∂

)()1(1

4 0022FUyyFUF

RTFU

V

HOHHC

Nernst (16)

⎥⎥⎦

⎤

⎢⎢⎣

⎡

+⋅−−=

∂∂

)()1(1

2 0022FUyyFUF

RTFU

V

HOHLC

Nernst (17)

5.1.1. Contribution of the Activation Overpotential

To refine the model, some considerations about the effect of the fuel utilization on theactivation overpotentials are now discussed.

The anode and cathode activation overpotentials are described by the Butler-Volmerequation [22]:

⎥⎦

⎤⎢⎣

⎡−=

⋅⋅ actcacta RTF

RTF

eeIIηϑηϑ

0 (18)

The exchange current I0 is expressed as a function of the composition of the reactantgases. 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


simulation of the characteristic curves at high fuel utilization. Due to contradictory datadiscussed 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 thestoichiometry of the anode electrochemical reaction (the resulting polarization resistance isinversely proportional to the partial pressure of the reactant, and directly proportional to thepartial pressure of the reaction product):

⎟⎟⎠

⎞⎜⎜⎝

⎛−

−

⋅⎟⎟⎠

⎞⎜⎜⎝

⎛⋅⎟⎟⎠

⎞⎜⎜⎝

⎛= RT

E

ref

OH

ref

Hanodeanode

anodeact

epp

pp

I,

2221

,0 γ (19)

Mogensen et al. [24] investigated Ni/YSZ cermet electrodes at 1273 K and found theelectrode impedance to be formed by a low-frequency contribution, with resistance decreasedby increasing both the partial pressure of hydrogen and water, and a high-frequencycontribution with resistance and capacitance almost independent of the partial pressures ofhydrogen and water. The proposed model does not respect the stoichiometry of theelectrochemical reaction:

⎟⎟⎠

⎞⎜⎜⎝

⎛−

⋅⎟⎟⎠

⎞⎜⎜⎝

⎛⋅⎟⎟⎠

⎞⎜⎜⎝

⎛= RT

E

ref

OH

ref

Hanodeanode

anodeact

epp

pp

I,

22,0 γ (20)

Anyway, from both these models it is possible to express the anode activationoverpotential as a function of the local partial pressures of fuel and products, expressing thusthe dependence of the activation overpotential on the fuel utilization factor.

Nevertheless, the activation contribution to the cell voltage sensitivity to FU is neglectedin this part of the analysis because of its small effect of the voltage drop [16,21]. In the nextParagraph 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 fuelutilization will be outlined.

5.1.2. Contribution of the Diffusion Overpotential

Concerning the diffusion overpotential, the main relation with the fuel utilization is present atthe 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 binarydiffusivity is a function of the fundamental binary diffusivity OHH DD

22− and the

microstructural parameters of the anode. In electrode microstructures with very small poresize, the possible effects of Knudsen diffusion, adsorption/desorption and surface diffusionmay also be present.

The fundamental binary diffusivity is evaluated using the Chapman-Enskog model. Thismodel describes the diffusion in binary gas mixtures at low to moderate pressures and itresults from solving the Boltzmann equation. The derived working equation is:


DDAB

ABAB f

nMTkD ⋅Ω⋅⋅⋅

⋅⋅⋅= 2

2/1)/4(163

σππ (21)

where, MA, MB are the molecular weights of the chemical species A an B, n is the numberdensity of molecules in the mixture, k is the Boltzmann constant, T the absolute temperature,and the term MAB is expressed as:

[ ] 1)/1()/1(2 −+⋅= BAAB MMM (22)

Moreover, ΏD is the collision integral for diffusion (depending on temperature and on theintermolecular force law between colliding molecules), and σAB is a characteristic length.Finally there is a correction term (fD) which is in the order of unity and particularly liesbetween 1 and 1.02 for molecules with comparable molecular weights, while it is between 1and 1.1 if the molecular masses are very unequal. If fD is chosen as unity and n is expressedby the ideal gas law, then:

DABABAB MP

TDΩ⋅⋅⋅

⋅= 22/1

2/300266.0σ

(23)

The diffusion coefficient DAB is expressed in cm2/s, the temperature T is expressed in K,the pressure in bar and the characteristic length, σAB, is expressed in Å; the diffusion collisionintegral ΏD is dimensionless.

The intermolecular energy ψ between two molecules, to the distance r, can be expressedby the following equation:

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−⎟

⎠⎞

⎜⎝⎛⋅⋅=

612

4rrσσεψ (24)

where ε and σ are the characteristic Lennard-Jones energy and length, respectively. It ispossible to demonstrate that:

( )

⎟⎠⎞

⎜⎝⎛ +

=

=

2

2/1

BAAB

BAAB

σσσ

εεε(25)

and that the collision integral is only depending upon the term T*=kT/ εAB. The relation ofNeufield is:

)()()(* ***)( THTFTDBDe

Ge

Ee

CT

A⋅⋅⋅

+++=Ω (26)


with:

AB

TkTε⋅

=* (27)

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-Jonesrelation is used. In this case the collision integral can be evaluated according to the Brokawequation [27,28]:

*

2

)()()(*

19.0)( *** Te

Ge

Ee

CT

A ABTHTFTDBD

δ⋅++++=Ω

⋅⋅⋅(28)

According to the considered equations, it is found that 82.029322 =−

KOHHD (Lennard-

Jones) or 84.029322 =−

KOHHD (Brokaw) or 89.0293

22 =−K

OHHD (Fuller); moreover,

20.029322=−

KNOD (Lennard-Jones). The literature [16,21,25,26,27,28] experimental values are

91.029322 =−

KOHHD and 22.0293

22=−

KNOD .

In terms of physical measurable parameters, an analytical expression for the anodediffusion overpotential is proposed. The expression of the limiting current density at theanode side is:

a

effabH

as tTRDFUpF

i⋅⋅

⋅⋅⋅= )()(2

2 (29)

It is obtained by imposing to zero the following equation:

ceffa

abH

aH i

DFtTR

FUpp ⋅⋅⋅⋅⋅

−=)(2

)(22

(30)

where the pressure of fuel at the bulk depends on the fuel utilization.The effective diffusion coefficients have been evaluated according to the equation:

5.1293

/

/)(/ )

293( TDD K

jica

caeffca ⋅⋅= −τ

ε(31)

The diffusion overpotential due to the anode diffusion is then given by:


⎟⎟⎠

⎞⎜⎜⎝

⎛−

⋅⋅

=as

adiff i

iFTRV 1log

2(32)

The tubular cell used in the SOFC CHP-100 is a cathode-supported cell: then, the maineffect of diffusion is mainly addressed to the cathode layer. In terms of physical measurableparameters, the cathode limiting current density can be evaluated in the form:

c

bO

effcbO

cs

tTRppp

DpFi

⋅⋅⋅⎟⎟⎠

⎞⎜⎜⎝

⎛ −

⋅⋅⋅=

2

2 )(4(33)

(This equation is an approximation for anode supported cells with very small cathodethickness). In general, the limiting current density at the cathode is given by imposing to zerothe equation:

ceffc

C iDpFtTR

bO

cO eFUpppp

⋅⋅⋅⋅⋅⋅

⋅−−= )(

22

4)),(( λ (34)

where the partial pressure of the oxidant at the bulk depends on fuel utilization and the airstoichiometry according to the following equation [5]:

00

0

2

2

2

2

HO

HO

yFUy

yFUy

⋅−

⋅−=

λλ

(35)

The diffusion overvoltage due to the cathode diffusion is then given by:

⎟⎟⎠

⎞⎜⎜⎝

⎛−

⋅⋅

=cs

cdiff i

iFTRV 1log

4(36)

In this model, the drop of voltage due to a variation of fuel utilization is modelled alsoconsidering the effect on the cathode diffusion. The effect of diffusion on the cell voltagesensitivity to fuel utilization has a significant effect when the limiting current density isapproached, or for low value of the air stoichiometry. It should also be noticed that in theequations (34) and (35), as posed in [3], the variable λ has not the meaning of excess ofoxidant with respect to the stoichiometric requirement, and so it has a different meaningcompared to the air stoichiometry λox as defined and used in this Chapter.

As a result, it is possible to write, with respect to equations (32) and (36):


FUi

iiii

FTR

iFUi

i

iiF

TRFUV as

ascas

c

as

asc

as

c

adiff

∂∂

⋅⋅−⋅

⋅=

⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜

⎝

⎛∂∂

⋅

−⋅⋅

=∂

∂

)(21

12 2 (37)

FUi

iiii

FTR

iFUi

i

iiF

TRFUV cs

csccs

c

cs

csc

cs

c

cdiff

∂∂

⋅⋅−⋅

⋅=

⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜

⎝

⎛∂∂

⋅

−⋅⋅

=∂

∂

)(41

14 2 (38)

It should be noted that:

)(/

/ 1

effCa

cadiff

DFUV

∝∂

∂(39)

ca

cadiff t

FUV

/

/

∝∂

∂(40)

The effect of the mass diffusion on the sensitivity of voltage to the fuel utilization isinversely proportional to the effective binary diffusivity on the electrodes layers and directlyproportional to the electrodes thickness.

The parameters used for solving the equations of the contribution of the diffusionoverpotential to the analytical expression of the sensitivity of the cell voltage to the fuelutilization are been listed in Table 4:

Table 4. Parameters used in the model of the diffusion overpotential

Parameter Anode Cathode)(| effcaD , effective binarydiffusivity

5.129322)( )

293( TDD K

OHHeffa5.1293

)( )293

(22

TDD KNOeffc

ji DD ,fundamental binarydiffusivity (cm2/s)

84.029322

KOHHD 20.0293

22

KNOD

3.0 3.0 353 5

5.1.3. Contribution of the Air Leakages

An important role is played by eventual leakages of air at the anode side; in fact some fuelcan burn with the oxidant with the consequence that it does not react at the anode surface togenerate current. If it would occur, then the effective fuel utilization will be higher than theexpected and the voltage sensitivity to a FU perturbation will be very high, because the


situation approaches the one with high values of fuel utilization. This is the case of the tubularcathode-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 UtilizationThe results of the model are shown in Figure 20. Two models are considered: (1) thesimplified 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% airleakage).

M odel of voltage sensitivity to FU

0123456789101112131415

70 75 80 85 90 95

FU (% )

dV/dFU (mV/%FU)

com plete m odel:low current com plete m odel:high current

nernst contribution:low current nernst contribution:high current

Figure 20. Model of the cell voltage sensitivity to fuel utilization for a tubular cathode-supported SOFC

dV/dFU vs FU

0

2

4

6

8

10

12

14

70 72 74 76 78 80 82 84 86 88 90 92 94 96

FU, %

dV/d

FU, m

V/%

FU

100 mA/cm2 400 mA/cm2 model approx model low current approx

Figure 21. Experimental data of the cell voltage sensitivity to fuel utilization for a tubular cathode-supported SOFC.


The curves are drawn for an operating condition of 1000°C and for a fuel with 89% H2

and 11% H2O of composition. In Figure 21 some experimental results (continuous lines) ofsingle 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.

It is interesting to outline how experimental sensitivities of Figure 21 cannot be describedby the simple model considering just the contribution of the sensitivity of the Nernst potentialto FU (dashed lines in Figure 21). In contrast, in Figure 22 an experimental investigation on adifferent design of cell (tubular anode-supported cell) shows a good agreement between theexperimental sensitivity to FU and the simple model (only sensitivity of the Nernst potential):this indicates that, for this typology of cell, the diffusion terms and the air leakages can beneglected, and this means that its polarization performance is not limited by the massdiffusion and the air leakages.

5.2. Deduction of the Local Fuel Utilization Distributionin the SOFC CHP-100 Plant

Once the model is developed, it can be used in order to explain the different values of localsensitivity to fuel utilization and to analyse the shape of the distribution of the local fuelutilization along the various segments of the SOFC CHP-100 generator.

The values of the voltage sensitivity of each sector to the overall fuel consumption havebeen estimated experimentally through the experiments described in the Paragraph 4.


Now, the experimental values of the voltage sensitivity of the various sectors to theoverall fuel consumption can be used together with the analytical model of the sensitivity ofthe cell voltage to the fuel utilization, in order to deduce the distribution of the local values offuel utilization pertaining to every sector of the generator.

According to the model equations, the voltage sensitivity to the fuel utilization dependson the local temperature in the sectors and on the fuel inlet conditions at the anode in terms ofhydrogen and water vapour partial pressures. The effect of the local temperature is taken intoaccount in the analysis, because it has been directly measured on the generator. The effect ofthe fuel inlet conditions is taken into account by direct measurement in the generator, but withthe 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 modelin terms of fuel utilization.

0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.951

2

3

4

5

6

7

8

9x 10

-3

Fuel consumption

voltage sensitivity to fuel consumption [mV/%FC]

Theoretical curveExperimental data

Figure 23. Graph of the theoretical curve (200 mA/cm2) and the experimental measurements of localvoltage sensitivity to the overall fuel consumption

A graph with the curve of the voltage sensitivity to the fuel utilization, obtained by themodel, and expressed as a function of the overall fuel consumption, can be developed. In thisgraph, the measurements of the voltage sensitivity of every sector is inserted. In this way, thelocal values of fuel utilization are deduced from the graph. In Figure 23 the graph of thetheoretical curve (for a current density of 200 mA/cm2, that is the experimental conditions ofthe generator) and the experimental measurements of local voltage sensitivity to the overallfuel consumption is shown. From the curve, the local fuel utilization values are deduced.


Estimation of the local fuel utilization

64

66

68

70

72

74

76

78

80

82

1 2 3 4 5 6 7 8 9 101112131415161718192021222324

cell sector

Loca

l Fue

l Ut

ilizati

on %

-0.004

-0.0035

-0.003

-0.0025

-0.002

-0.0015

-0.001

-0.0005

0

FU S

ensit

ivity/

V/

%FU

Local FU Local FU voltage sensitivity Local FU voltage sensitivity Local FU

Figure 24. Estimation of the distribution of the local values of fuel utilization inside the generator.

The results of the estimation procedure are shown in Figure 24.The higher is the local voltage sensitivity to the overall fuel consumption, the higher is

the local fuel utilization.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. As already written as anhypothesis, this behavior could be linked to a fluid-dynamic problem of not uniformdistribution of fuel especially at the boundary rows; moreover, here probably the reformingreactions are reduced (the reformer boards placed at the external boundaries of the generatorreceive the water vapor rich-fuel gas through orifices with lower diameter, while thereformers placed between the internal bundle rows receive the fuel gas through orifices withhigher diameter). This can explain the higher values of local fuel utilization at the boundaryrows.

The boundaries of the generator work at high fuel utilization, and this can also explain thelower voltages observed at the edge’s sectors.

6. Polarization Analysis

In the experimental session, a polarization analysis was performed in order to investigate theelectrical behavior of the sectors of the generator.

The analysis was performed during real operating conditions of the generator, thus onlyshort-range V-I characteristics were measured. The current was changed between 435A and475A, at fixed overall fuel consumption of 84.25% and setpoint temperature of 967°C.


ASR

cm

2

Figure 25. Polarization behaviour of two sectors of the generator.

The polarization behavior has shown significant differences among the various sectors ofthe generator: different slopes in the current range imposed for the polarization. As anexample, in Figure 25 the polarization curves of two sectors are shown. The polarizationcurve 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 SpecificResistance (ASR).

6.1. Polarization Model

The polarization analysis consisted in the fitting of the experimental V-I curves with aproposed polarization model. The model has been already presented in equation (5).

6.1.1. Nernst Potential

The Nernst potential is expressed by the equation (6) and evaluated according to equations (7)and (10).

6.1.2. Activation Overpotential

The activation overpotential is modelled considering the Butler-Volmer equation presented inequation (18).

The cathode exchange current density has been calculated according to the followingequation:


⎟⎟⎠

⎞⎜⎜⎝

⎛−

⋅⎟⎟⎠

⎞⎜⎜⎝

⎛= RT

E

ref

Ocathodecathode

cathodeact

epp

I,

2

25.0

,0 γ (41)

The anode exchange current density has been considered as a fitting parameter. For theestimation of its first attempt value in the fitting analysis, the equation (19) has beenconsidered.

The parameters used in equations (18), (19) and (41) have been assumed by the currentliterature: γa=8.2 107 [mA/cm2] (first attempt value in the fitting analysis); γc=1.8 107

[mA/cm2]; Eact,anode=120000 [J/mol]; Eact,cathode=120000 [J/mol]; θa,anode=2, θc,anode=1,θa,cathode=1.4, θa,cathode=0.6.

6.1.3. Ohmic Overpotential

Despite the ionic resistance of the electrolyte plays an important role in the cell ohmicresistance, it is the contribution of the anode and cathode resistance which determines thehigh 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 thecurrent literature [30,31] and the ohmic overpotential has been evaluated according to theequations (42) and (43):

iRiRiRiR elancatohmic ⋅+⋅+⋅+⋅= intη (42)

jjj pathR ⋅= ρ (43)

Ionic resistivity of 8YSZ

0

20

40

60

80

100

120

140

160

700 750 800 850 900 950 1000

Temperature/°C

Res

istiv

ity/o

hm *

cm

Bessette et Al. [17] Siem ens Patent [19] Costam agna et Al. [9]

Figure 26. Ionic resistivity of 8YSZ as a function of temperature.


At 1000°C the ohmic area specific resistance of tubular cathode-supported cells is around0.40 Ώ cm2; for an HPD5 cell is around 0.25 Ώ cm2; for an HPD10 cell is around 0.15 Ώ cm2;for a Delta 9 cell is around 0.18 Ώ cm2 [32]. In Figure 26 the ionic resistance of 8YSZ isdrawn 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 cell’s layers are listed in thefollowing and have been assumed by the literature [3, 21,22,26,30,31,32,33,34]:

⎟⎟⎠

⎞⎜⎜⎝

⎛−

⋅= cellTcat e

1392

00298.0ρ (44)

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅= cellTan e

600

044.0ρ (45)

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅= cellTel e

10350

00294.0ρ (46)

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅= cellTe4690

int 1256.0ρ (47)

The description of the materials used in these cells is largely available in literature[32,35].

6.1.4. Diffusion Overpotential

The diffusion overpotentials are described according to equations (32) and (36). Thefundamental binary diffusivities have been calculated using the previously described Brokawmodel and the values of microscopic parameters have been assumed by literature and arelisted in Table 4.

6.2. Estimation of the Local Anode Exchange Current Density

The parameter estimation procedure involved the anode exchange current density, consideredin the expression of the anode activation overpotential (equation (18)). As already noticed, inliterature there are some expressions which describe the anode exchange current (equations(19) and (20): according to these equations, this parameter should be significantly affected bythe fuel utilization values, and the aim is to outline how the distribution of the local fuelutilization (that is, the distribution of fuel) inside the generator affects the activation of thereaction at the anode side in the various sectors.

The procedure of parameter estimation consisted in the evaluation of the values of theanode exchange current density in each sector of the generator, in order to obtain itsdistribution inside the stack. In the estimation, the local values of fuel utilization evaluated in


the previous Paragraph 5, and the measured local values of temperature, have been associatedto each sector.

1 2 3 4 5 6 7 8 9 10 11 1250

100

150

200

250

300

Sector

Anode exc

hange current density [mA/cm2]

Anode exchange current density distribution

Power leadEjector

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 isshown:

0 5 10 15 20 250

50

100

150

200

250

300

Sector

Ano

de a

xcha

nge

curr

ent d

ensi

ty [m

A/c

m2 ]

95% confidence region for estimated value of anodic exchange current density

Lower limitCalculated valueUpper limit

Figure 28. 95% confidence region for the anode exchange current density estimation.


It is possible to notice a symmetric trend centered on the row 6 for the Power Leads sideand 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, andhigher values for the center sectors.

This behaviour is the direct and logic consequence of the distribution of fuel utilizationinside the generator. There are high values of fuel utilization at the boundaries of thegenerator 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 thekinetic is reduced at the boundary rows. This behaviour is expressed by the values of theanode exchange current density.

In Figure 28 the 95% confidence region for the anode exchange current densityestimation 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 currentdensity.

In Figure 29 the correlation between the local anode exchange current density and thelocal 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 logicconsequence of the distribution of fuel utilization inside the generator. There are high valuesof fuel utilization at the boundaries of the generator, and therefore the kinetic is reduced at theboundary rows: this behaviour is expressed by the lower values of the anode exchange currentdensity.

0 5 10 15 20 250

50

100

150

200

250

300

Sector

Anode axc

hange current density [mA/cm2]

95% confidence region for estim ated value of anodic exchange current density

Lower lim itCalculated valueUpper lim it

2 4 6 8 10 1280

100

120

140

160

180

200

220

240

260

280

300

Sector

Anode exch

ange current d

ensity [m

A/cm2]

Correlation between anode exchange current density and fuel utilization power leads side

2 4 6 8 10 120.75

0.76

0.77

0.78

0.79

0.8

0.81

0.82

0.83

0.84

0.85

Fuel u

tilizatio

n

2 4 6 8 10 1280

100

120

140

160

180

200

220

240

260

280

300

Sector

Anode exch

ange current d

ensity [m

A/cm2]

Correlation between anode exchange current density and fuel utilization ejector side

2 4 6 8 10 120.75

0.76

0.77

0.78

0.79

0.8

0.81

0.82

0.83

0.84

0.85

Fuel u

tilizatio

n

Figure 29. Correlation between the local anode exchange current density and the local fuel utilization.


2 4 6 8 10 1280

100

120

140

160

180

200

220

240

260

280

300

Sector

Anode exch

ange current d

ensity [mA/cm2]

Correlation between anode exchange current density and fuel utilization power leads side

2 4 6 8 10 12

900

Temperature [°C]

2 4 6 8 10 1280

100

120

140

160

180

200

220

240

260

280

300

Sector

Anode exch

ange current d

ensity [mA/cm2]

Correlation between anode exchange current density and fuel utilization ejector side

2 4 6 8 10 12

900

Temperature [°C]

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 itis also influenced by the local temperature. A higher temperature should have a increasingeffect on the anode exchange current density. Therefore, in Figure 30 the correlation betweenthe local anode exchange current density and the local temperature (as usual, divided inPower Leads and Ejectors side) is shown. It seems that the correlation with the localtemperature is less evident than the correlation with the local fuel utilization, especially at theEjectors side (fuel inlet). Thus, it seems that the effect of the local fuel utilization is moresignificant 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 temperatureis high (at the boundary rows of the generator) the fuel utilization is high too. This behaviourcan compensate the effect of temperature (positive) and fuel consumption (negative) on theanode exchange current density. It can be observed that the temperature effect attenuates thedependence of the anode exchange current density from the fuel utilization (in fact, thedistribution of temperature is symmetric compared to the fuel utilization distribution): thenegative activation effect at the boundary rows due to high fuel utilization is reduced byhigher local values of temperature, and the positive activation effect at the center (low fuelutilization) is reduced too by lower local values of temperature.

As a conclusion, the effect of fuel utilization distribution on the anode exchange currentdensity is predominating, while the temperature distribution attenuates this behaviour.


5 10 15 200

10

20

30

40

50

60

Sector

percentual loss

[%]

Percentual loss on sector

activation

concentrationohm ic

activation concentration ohm ic0

10

20

30

40

50

60M ean percentual loss

Percentual loss

Figure 31. Values of the overpotentials for every sector at about 180 mA/cm2.

6.3. Contribution of the Overpotential Terms

Based on the proposed electrochemical model, the percentage values of the overpotentialcontributions of the considered electrochemical model (with the anode activationoverpotential calculated using the estimated values of the anode exchange current density) atabout 180 mA/cm2 are shown. In Figure 31 the percentage values of the overvoltages forevery sector at about 180 mA/cm2 are shown.

As expected, it is possible to notice that the activation overpotential and the ohmicoverpotential are the main losses factor (their sum is about the 90% of the losses), while theconcentration (diffusion) overpotential is less relevant in the low current range considered forthe experiment.

In particular, the activation overpotentials are quite high: this is a consequence of thealready highlighted not optimal fuel distribution inside the generator, with high values oflocal fuel utilization especially at the boundary rows of the stack.

In Figure 32 the percentage weight of the overpotential contributions at 180 mA/cm2 areshown. A comparison is made: (1) contribution evaluated with the value of anode exchangecurrent density find in literature; (2) contribution evaluated with the value of anode exchangecurrent density estimated by the regression procedure on the SOFC CHP-100 in two differentsectors of the generator (the highest and the lowest values of anode exchange current densityare considered).


Literature I_0_,_a=295.15 [mA/cm 2] I_0_,_a=91.75 [mA/cm 2]0

10

20

30

40

50

60

70

80

90

100

Percentual loss

ActivationOhmicConcentration

Figure 32. Percentage weight of the overvoltage contributions at 180 mA/cm2 (evaluated for the valueof anode exchange current density find in literature and the values find by the regression on the SOFCCHP-100).

90060

80

100

120

140

160

180

200

220

240

1

2

3

4

5

6

7

8

910

11

12

13

14

15

16

17

18 1920

21

22

23

24

Tem perature [°C]

Anode exch

ange current d

ensity [mA/cm2]

Pedigree distribution ofanode exchange current density

0-10000 hours10000-16000 hours

O ver 16000 hours

0-10k h 10k-16k h over 16k h60

80

100

120

140

160

180

200

220

240

Running hours

Anode exch

ange current d

ensity [mA/cm2]

M ean value of anode exchange current densitydivided by sector running hours

Figure 33. Correlation between the cell anode exchange current density and its pedigree.


It is possible to notice that the results of the regression procedure on the SOFC CHP-100determine lower values of the activation overpotential and higher values of the ohmic anddiffusion overpotentials, compared to the results obtained with literature data.

Probably, it could be due to the fact that the literature equations are proposed for planarcells, and their behaviour is towards an underestimation of the anode exchange currentdensity, with the consequence of overestimating the activation losses.

Finally, in Figure 33 a correlation between the anode exchange current density and thelocal temperature (grouped by sector running hours), and the mean values for the runninghours (the cell pedigree), are shown.

As a general comment, it can be noticed that in the new sectors the anode exchangecurrent density is lower compared to the old sectors. Therefore, it is possible to conclude thatthe effect of the position of the sector inside the generator predominates on its pedigree: infact, the boundary sectors (sectors 1 and 24) have a very low anode exchange current densitycompared to other sectors with the same pedigree disposed in different positions in thegenerator.

7. Conclusion

The Chapter dealt with the description of the experimental analysis of the CHP-100 SOFCField 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 SOFCgenerator in operation, with the complexity linked to the large number of variables whichaffect its operation, the limited number of measurements points in the generator volume, thenecessity to avoid malfunctions in the real operation.

As a consequence, the experimental analysis of the CHP-100 SOFC Field Unit has beendeveloped with methods of Design of Experiments, and with a statistical analysis of thecollected data.

The following conclusions can be outlined:

• the design of experiments procedure is a powerful tool to study in a rigorous manner thecollected experimental data and also to relate, through analytical expressions, theinvestigated dependent variables with the control factors (independent variables); thedesign of experiments approach allows one to analyse the significance of the main andinteraction effects of the factors;

• the experimental analysis of the SOFC generator is presented in terms of sensitivityanalysis of the local voltages with respect to the overall fuel consumption and the airstoichiometry control factors;

• the use of local voltage sensitivity tests to the overall fuel consumption, coupled with theanalytical model of the sensitivity of the cell voltage to the fuel utilization, is a usefulprocedure to estimate the local fuel utilization of the various sectors of a real largegenerator during its operation: the indication of the fuel distribution inside the generatorvolume has therefore been found;

• high local fuel utilization have been detected: they can lead to local overheating whichcan limit the operation (power density, efficiency and degradation) of the generator;


• further, through a regression analysis on the polarization curves, a distribution of theanode exchange current density has been estimated, and correlated to the distribution offuel utilization and temperature, and to the cell pedigree.

Nomenclature

BoP Balance of PlantCHP Cogeneration Heat and PowerDi-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 (%)

totairG , air mass flow (kg s-1)

echairG , electrochemical air mass flow utilization (kg s-1)

echfG , electrochemical fuel mass flow utilization (kg s-1)

totfG , fuel mass flow (kg s-1)

HC High CurrentHPD High Power DensityI generator current (A)

0I exchange current density (mA/cm2)

anodeI ,0 anode exchange current density (mA/cm2)

cathodeI ,0 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 CurrentMi molecular weights of the chemical species (g mol-1)n number density of molecules in the mixturep stack pressure (Pa)

aHp

2hydrogen pressure at the anode/electrolyte interface (Pa)

bHp

2hydrogen pressure at the anode bulk (Pa)

aOp

2 oxygen pressure at the cathode/electrolyte interface (Pa)


bOp

2 oxygen pressure at the cathode bulk (Pa)R universal gas constant (J mol-1 K-1)

jR cell ohmic resistivity (Ω cm2)

SOFC Solid Oxide Fuel CellT 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)

NernstV Nernst potential (V)

NernstV x-averaged Nernst potential (V)

HCNernstV , average Nernst potential, high current approximation (V)

LCNernstV , average Nernst potential, low current approximation (V)

x axial position along the tubular cell (m)xi regressor variablesYi dependent variable

iy molar fraction of reactants at the electrodes’ bulk0

2Hy hydrogen molar fraction at fuel inlet condition0

2OHy water molar fraction at fuel inlet condition

Greek

βi regression coefficientγ pre-exponential coefficient in activation polarization equation (A cm-2)ε electrode porosity

caact |,ηactivation overpotential at anode and cathode (V)

caconc |,η concentration (diffusion) overpotential anode and cathode (V)

ohmη ohmic overpotential (V)θ transfer coefficient in activation polarization equationλair stoichiometric excess of oxidantρj ohmic resistivity of materials (Ω cm)σAB characteristic length of Chapman-Enskog model (Å)ΏD collision integral for diffusionτ electrode tortuosity


Appendix A. Basics of Regression Models and Factorial Design

A1. Regression Models

A1.1. IntroductionIn many problems two or more variables are related and it is of interest to model and explorethis relationship.

Suppose that there is a single dependent variable or response Y that depends on kindependent or regressor variables, for example kxxx ,...,, 21 . The relationship betweenthese variables is characterized by a mathematical model, called a regression model, that is fitto a set of sample data. In some instances, the experimenter knows the exact form of the truefunctional relationship between Y and kxxx ,...,, 21 , say ( )kxxxY ,...,, 21φ= .

However, in most cases, the true functional relationship is unknown and the experimenterchooses an appropriate function to approximateφ . Low-order polynomial models are widelyused as approximating functions.

In general, the response variable Y may be related to k regressor variables. The model:

εββββ +++++= kk xxxY ...22110 (A1)

is called a multiple linear regression model with k regressor variables. The parameters jβ( kj ,...,1,0= ) are called the regression coefficients. This model describes a hyperplane in

the k-dimensional space of the regressor variables. The parameter jβ represents the expected

change in response Y per unit change in jx when all the remaining independent variables ix

( )ji ≠ are held constant.

A1.2. Estimation of the Parameters in Linear Regression ModelsThe method of least squares is typically used to estimate the regression coefficients in amultiple linear regression model. Suppose that kn > observations on the response variableare available, say nYYY ,...,, 21 . Along with each observed response iY , an observation on

each regressor variable is available; let ijx denote the i-th observation or level of variable jx .

Table A1. Data for multiple linear regression

Y 1x 2x ... kx

1Y 11x 12x … kx1

2Y 21x 22x … kx2

… … … …… … … …

nY 1nx 2nx … nkx


The model equation (A1) can be written in terms of the observations as:

nixY

xxxYk

jiijji

iikkiii

,...,2,1

...

10

22110

=++=

+++++=

∑=

εββ

εββββ(A2)

Assume that the error term iε in the model has ( ) 0=iE ε and ( ) 2σε =iV . The method

of least squares chooses the coefficients jβ in equation (A2) so that the sum of the squares of

the errors iε is minimized. The least squares function is:

2

1 10

1

2 ∑ ∑∑= ==

⎟⎟⎠

⎞⎜⎜⎝

⎛−−==

n

i

k

jijji

n

ii xYL ββε (A3)

The function L is to be minimized with respect to kβββ ,...,, 10 . The least squares

estimators, say kβββ ˆ,...,ˆ,ˆ10 , must satisfy

∑ ∑= =

=⎟⎟⎠

⎞⎜⎜⎝

⎛−−−=

∂∂ n

i

k

jijji xYL

k1 1

0ˆ,...,ˆ,ˆ0

0ˆˆ210

βββ

βββ

(A4a)

kjxxYL n

iij

k

jijji

jk

,...2,10ˆˆ21 1

0ˆ,...,ˆ,ˆ

10

==⋅⎟⎟⎠

⎞⎜⎜⎝

⎛−−−=

∂∂ ∑ ∑

= =

βββ

βββ

(A4b)

Simplifying the equations (A4):

∑ ∑∑∑= ===

=++++n

i

n

ii

n

iikki

n

ii Yxxxn

1 1122

1110

ˆ...ˆˆˆ ββββ

∑ ∑∑∑∑

∑ ∑∑∑∑

= ====

= ====

=++++

=++++

n

i

n

iiik

n

iikkiik

n

iiik

n

iik

n

i

n

iii

n

iikikii

n

ii

n

ii

Yxxxxxxx

Yxxxxxxx

1 11

222

111

10

1 11

11212

1

211

110

ˆ...ˆˆˆ

...............

...............

ˆ...ˆˆˆ

ββββ

ββββ

(A5)

The equations (A5) are called the least squares normal equations. Note that there are1+= kp normal equations, one for each of the unknown regression coefficients. The


solution to the normal equations will be the least squares estimators of the regression

coefficients kβββ ˆ,...,ˆ,ˆ10 .

The normal equations are solved in matrix notation. The model in terms of theobservations (A2) may be written in matrix notation as:

εβ += XY

where:

⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢

⎣

⎡

=

nY

YY

Y...

2

1

,

⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢

⎣

⎡

=

nknn

k

k

xxx

xxxxxx

X

...1...............

...1

...1

21

22221

11211

,

⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢

⎣

⎡

=

kβ

ββ

β...

1

0

,

⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢

⎣

⎡

=

kε

εε

ε...

1

0

In general Y is an ( )1×n vector of the observations, X is an ( )pn× matrix of the

levels of the independent variables, β is a ( )1×p vector of the regression coefficients and

ε is an ( )1×n vector of random errors.

The vector of least squares estimators β has to be found, which minimizes:

( ) ( )ββεεε XYXYLn

ii −⋅−=== ∑

=

''1

2

Note that L may be expressed as:

βββββββ XXYXYYXXXYYXYYL ''''2''''''' +−=+−−= (A6)

because YX ''β is a ( )11× matrix, or a scalar, and its transpose ( ) YXYX ''''' ββ = is thesame scalar.

The least squares estimators must satisfy:

0ˆ'2'2ˆ

=+−=∂∂ ββ β

XXYXL

which simplifies to:

β'' XXYX = (A7)


Equation (A7) is the matrix form of the least squares normal equations. To solve thenormal equations, multiply both sides of equation (A7) by the inverse of XX ' . Thus, theleast squares estimator of β is:

( ) YXXX ''ˆ 1−=β (A8)

It is easy to see that the matrix form of the normal equations is identical to the scalarform. Writing out equation (A7) in detail, the following form is obtained:

⎥⎥⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢⎢⎢

⎣

⎡

=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

⎥⎥⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢⎢⎢

⎣

⎡

∑

∑

∑

∑∑∑∑

∑∑∑∑

∑∑∑

=

=

=

====

====

===

n

iiik

n

iii

n

ii

kn

iik

n

iiik

n

iiik

n

iik

n

iiki

n

iii

n

ii

n

ii

n

iik

n

ii

n

ii

yx

yx

y

xxxxxx

xxxxxx

xxxn

1

11

1

1

0

1

2

12

11

1

11

121

1

21

11

112

11

ˆ

ˆˆ

.

...

...

...

...

β

ββ

If the indicated matrix multiplication is performed, the scalar form of the normalequations (A5) will result. In this form it is easy to see that XX ' is a ( )pp× symmetric

matrix and YX ' is a ( )1×p column vector. Note the special structure of the XX ' matrix:

the diagonal elements of XX ' 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 columnsof X .

Furthermore, note that the elements of YX ' are the sums of cross-products of thecolumns of X and the observations Y . The fitted regression model is:

βˆ XY = (A9)

In scalar notation, the fitted model is:

nixYk

iijji ,...,2,1ˆˆˆ

10 =+= ∑

=

ββ

The difference between the actual observation iY and the corresponding fitted value iY

is the residual, iii YYe ˆ−= . Then the ( )1×n vector of residuals is denoted by:

YYe ˆ−= (A10)


A1.3. Estimating the Variance σ2

To develop an estimator of the variance 2σ , it is necessary to consider the sum of squares ofthe residuals:

( )∑ ∑= =

==−=n

i

n

iiiiE eeeYYSS

1 1

22'ˆ

Substituting βˆ XYYYe −=−= :

βββ

ββββββ

ˆ''ˆ''ˆ2'

ˆ''ˆˆ'''ˆ'ˆ'ˆ

XXYXYY

XXXYYXYYXYXYSS E

+−=

=+−−=⎟⎠⎞⎜

⎝⎛ −⋅⎟

⎠⎞⎜

⎝⎛ −=

Because YXXX 'ˆ' =β , this last equation becomes:

YXYYSSE ''ˆ' β−= (A11)

Equation (A11) is called the error or residual sum of squares and it has pn − degrees offreedom associated with it, where n is the total number of observations and 1+= kp .

It can be shown that:

( ) ( )pnSSE E −= 2σ

so an unbiased estimator of 2σ is given by:

pnSSE

−=2σ (A12)

A1.4. Analysis of Variance of the RegressionThe test for significance of regression is a test to determine whether a linear relationshipexists between the response variable Y and a subset of the regressor variables kxxx ,...,, 21 .The test of hypothesis are:

joneleastatforHH

j

k

0:0...:

1

210

≠====

ββββ

(A13)


Rejection of 0H implies that at least one of the regressor variables kxxx ,...,, 21

contributes significantly to the model. The test procedure involves an analysis of variance,partitioning the total sum of squares TSS into a sum of squares due to the model (or toregression) and a sum of squares due to residual (or error):

ERT SSSSSS += (A14)

A computational formula of the sums of squares can be easily found. Because

( ) ( )n

YYY

nY

YSSn

i in

i

n

i iiT

2

1

1

2

12 ' ∑∑ ∑ =

=

= −=−= (A15)

the equation (A11) can be written as:

( ) ( )RT

n

i in

i iE SSSS

nY

YXn

YYYSS −=

⎥⎥

⎦

⎤

⎢⎢

⎣

⎡−−−= ∑∑ ==

2

1

2

1 ''ˆ' β

Therefore, the regression sum of squares is:

( )n

YYXSS

n

i iR

2

1''ˆ ∑ =−= β (A16)

and the error sum of squares is:

YXYYSS E ''ˆ' β−= (A17)

If the null hypothesis 0H is true, then 2σRSS is distributed as 2kχ , where the number

of degrees of freedom for 2χ is equal to the number of regressor variables in the model.

The test procedure for 0...: 210 ==== kH βββ is to compute:

( ) E

R

E

R

MSMS

knSSkSS

F =−−

=10 (A18)

and to reject 0H if 1,,0 −−> knkFF α . From the expected mean squares it could be shown that,

in general, EMS is an unbiased estimator of 2σ . Also, under the null hypothesis, RMS is

an unbiased estimator of 2σ . Therefore, under the alternative hypothesis, the expected valueof the numerator of the test statistic is greater than the expected value of the denominator, and


0H should be rejected for values of the test statistic that are too large. 0H would be rejected

and concluded that there are differences in the treatment means if 1,,0 −−> knkFF α .

Table A2. Analysis of variance for significance of regression

Source ofvariation

Sum ofsquares

Degrees offreedom

Meansquare 0F

Regression RSS k RMS ER MSMSF =0

Error or residual ESS 1−− kn EMS

Total TSS 1−n

In the analysis the coefficient of multiple determination R2 is calculated:

T

E

T

R

SSSS

SSSSR −== 12 (A19)

The coefficient R2 is a measure of the amount of reduction in the variability of Y obtainedby using the regressor variables kxxx ,...,, 21 in the model. Adding a variable to the modelwill always increase R2, regardless of whether the additional variable is statisticallysignificant or not. Thus, it is possible for models that have large values of R2 to yield poorpredictions of new observations or estimates of the mean response.

A2. Factorial Design

A2.1. Basic Definitions and PrinciplesFactorial designs are widely used in experiments involving several factors where it isnecessary to study the joint effect of the factors on a response.

The factorial approach is more efficient than the approach one-factor-at-a-time and it isindispensable when there is an interaction between the factors.

The effect of a factor is defined to be the change in response produced by a change in thelevel of the factor. This is frequently called a main effect because it refers to the primaryfactors of interest in the experiment.

In some experiments it is found that the difference in response between the levels of onefactor is not the same at all levels of the other factors. When this occurs, there is aninteraction between the factors.

Factorial designs are widely used in experiments involving all the combinations of factorsand levels. There are a levels of the factor A, b levels of the factor B and each replicatecontains the ba× treatment combinations.

The most important of these special cases is that of k factors, each at only two levels.These levels may be quantitative, such as two values of temperature, pressure or time; or they


may be qualitative, such as two machines, two operators, the “high” and “low” levels of afactor, or perhaps the presence and absence of a factor. A complete replicate of such a designrequires k22...222 =×××× observations and is called k2 factorial design.

The k2 design is particularly useful in the early stages of experimental work, when manyfactors are likely to be investigated. It provides the smallest number of runs with which kfactors can be studied in a complete factorial design. Consequently, these designs are widelyused in factor screening experiments.

A2.2. The 22 DesignThe first design in the k2 series is one with only two factors, say A and B, each run at twolevels. This design is called a 22 factorial design. The levels of the factors may be arbitrarilycalled “low” and “high”. Consider an experiment replicated three times. The data obtained areas follows:

Table A3. 22 design with 3 replicates

Factor Replicate

A BTreatment combination

I II IIITotal

- - A low, B low IBAy −− IIBA

y −− IIIBAy −− ( )1

+ - A high, B low IBAy −+ IIBA

y −+ IIIBAy −+ a

- + A low, B high IBAy +− IIBA

y +− IIIBAy +− b

+ + A high, B high IBAy ++ IIBA

y ++ IIIBAy ++ ab

By convention, the effect of a factor is denoted by a capital Latin letter. Thus “A” refersto the effect of factor A, “B” refers to the effect of factor B, and “AB” refers to the ABinteraction. In the 22 design the low and high levels of A and B are denoted by “-“ and “+”,respectively, on the A and B axes.

Fattore B

+

_

Fattore A+

_( ) a1

abb

Figure A1. Treatment combination in the 22 design.


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 thetreatment combination is denoted by the corresponding lowercase letter and that the low levelof 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 lowlevel, b represents A at the low level and B at the high level, and ab represents both factors atthe 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 thechange in response produced by a change in the level of that factor averages over the levels ofthe other factor. Also, the symbols (1), a, b, ab now represent the total of all n replicates takenat the treatment combination.

The effect of A at the low level of B is ( )[ ] na 1− and the effect of A at the high level of

B is [ ] nbab − . Averaging these two quantities yields the main effect of A:

[ ] ( )[ ] ( )[ ]1211

21

−−+=−+−= baabn

ababn

A (A20)

The average main effect of B is found from the effect of B at the low level of A,( )[ ] nb 1− , and at the high level of A, [ ] naab − , as:

[ ] ( )[ ] ( )[ ]1211

21

−−+=−+−= ababn

baabn

B (A21)

The interaction effect AB is defined as the average difference between the effect of A atthe high level of B and the effect of A at the low level of B. Thus:

[ ] ( )[ ] ( )[ ]baabn

ababn

AB −−+=−−−= 1211

21

(A22)

Alternatively, AB may be defined as the average difference between the effect of B at thehigh 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. Theeffect of A can be found as the difference in the average response of the two treatmentcombinations in which A is at high level ( +A

y ) and the two treatment combinations in which

A is at low level ( −Ay ), say:

( ) ( )[ ]121

21

2−−+=

+−

+=−= −+ baab

nnb

naabyyA

AA


The effect of B is found as the difference between the average of the two treatmentcombinations in which B is at high level ( +B

y ) and the average of the two treatment

combinations in which B is at low level ( −By ), say:

( ) ( )[ ]121

21

2−−+=

+−

+=−= −+ abab

nna

nbabyyB

BB

Finally, the interaction effect AB is the average of the right-to-left diagonal treatmentcombinations in the square in Figure A1, minus the average of the left-to-right diagonaltreatment combinations (a e b), or:

( ) ( )[ ]baabnn

ban

abAB −−+=+

−+

= 121

221

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 generallybe used to confirm this interpretation.

Consider the sums of squares for A, B and AB. We define the total effect or contrast of Aas:

( )1−−+= baabContrastA (A23)

The contrast sum of squares is equal to the contrast squared divided by the total numberof observations in the design:

( )[ ]nbaabSS A 4

1 2−−+= (A24)

( )[ ]nababSSB 4

1 2−−+= (A25)

( )[ ]n

baabSS AB 41 2−−+

= (A26)

The total sum of squares is found in the usual way, that is:

∑∑∑= = =

−=2

1

2

1 1

2...2

4i j

n

kijkT n

yySS (A27)


where TSS has 14 −n degrees of freedom. The error sum of squares, with ( )14 −n degreesof freedom, is usually computed by subtraction as:

ABBATE SSSSSSSSSS −−−= (A28)

Furthermore, based on the p-values, it is possible to conclude that the main effects arestatistically 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 Yates’s order.

The contrast coefficients used in estimating the effects are in Table A4.

Table A4. Contrast coefficients

Effects ( )1 a b abA -1 +1 -1 +1B -1 -1 +1 +1AB +1 -1 -1 +1

Note that the contrast coefficients for estimating the interaction effect are just the productof the corresponding coefficients for the two main effects. The contrast coefficient is alwayseither +1 or -1, and a table of plus and minus signs such as in Table A5 can be used todetermine the proper sign for each treatment combination.

The column headings in Table A5 are the main effects (A and B), the AB interaction, andI, which represents the total or average of the entire experiment. To find the contrast forestimating any effect, simply multiply the signs in the appropriate column of the table by thecorresponding treatment combination and add.

Table A5. Algebraic signs for calculating effects in the 22 design

Factorial effectTreatmentcombination I A B AB

( )1 + - - +a + + - -b + - + -ab + + + +

For instance, considering the column of A in Table A5, it is immediately possible tocalculate the effect expression according to the treatment combinations, because it issufficient to associate the elements of each combination column with those of the column ofthe considered factor, that is ( ) abbaA +−+−= 1 .

In a 22 factorial design it is possible to express the results of the experiment in terms of aregression model:


εβββ +++= 22110 xxY (A29)

where 1x , 2x are coded variables and the β ’s are regression coefficients.

The intercept 0β is the grand average of all observations, and the regression coefficients

1β and 2β are one-half the corresponding factor effect estimates. The regression coefficientis one-half the effect estimate because a regression coefficient measures the effect of a unitchange 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 theplots of normal probability and residuals versus fitted values. Furthermore, response surfaceplots can be realized using the regression model expression.

A2.3. The 2k DesignThe methods of analysis that we have presented thus far may be generalized to the case of a2k factorial design, that is a design with k factors each at two levels. The statistical model fora 2k design would include k main effects,

⎟⎟⎠

⎞⎜⎜⎝

⎛2k two-factor interactions,

⎟⎟⎠

⎞⎜⎜⎝

⎛3k three-factor

interactions, …, and one k -factor interaction. That is, for a 2k design the complete modelwould contain 12 −k effects.

The treatment combinations can be written in standard order by introducing the factorsone 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 TableA6.

The first step is to estimate factor effects and examine their signs and magnitudes. Thisgives the experimenter preliminary information regarding which factors and interactions maybe important, and in which directions these factors should be adjusted to improve theresponse.

Table A6. Analysis procedure for a 2k design

1. Estimate factor effects

2. Form initial model

3. Perform statistical testing

4. Refine model

5. Analyze residuals

6. Interpret results


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 beenreplicated. The analysis of variance is used to formally test for significance of main effectsand interaction. Table A7 shows the general form of an analysis of variance for a 2k factorialdesign with n replicates. Step 4, refine the model, usually consists of removing any notsignificant variables from the full model. Step 5 is the usual residual analysis to check formodel adequacy and to check assumptions. Sometimes model refinement will occur afterresidual 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, orresponse surface and contour plots.

In general, the contrast for an effect are determined by expanding the right-hand side ofequation (A30):

( )( ) ( )111 ±±±= kbaContrast KAB (A30)

The sign in each set of parentheses is negative if the factor is included in the effect andpositive if the factor is not included. For example, consider a 23 factorial design and thecontrast for AB would be:

( )( )( ) ( ) babcaccababccbaContrast AB −−−−+++=+−−= 1111

Once the contrasts for the effects have been computed, the effects can be estimated andthe sums of squares can be computed according to:

( )KABk Contrastn

KAB22

=

( )22

1KABkKAB Contrast

nSS

⋅=

where n denotes the number of replicates.

Table A7. Analysis of variance for a k2 design

Source of variability Sum of squares Degrees of freedom

k main effectsA ASS 1

B BSS 1

K KSS 1


Table A7. Continued

Source of variability Sum of squares Degrees of freedom

⎟⎟⎠

⎞⎜⎜⎝

⎛2k two-factor interactions

AB ABSS 1

AC ACSS 1

JK JKSS 1

⎟⎟⎠

⎞⎜⎜⎝

⎛3k three-factor interactions

ABC ABCSS 1

ABD 1

IJK IJKSS 1

⎟⎟⎠

⎞⎜⎜⎝

⎛kk 1= k-factor interaction

KABC KABCSS 1

Error ESS ( )12 −nk

Total TSS 12 −kn

A2.4. The Addition f Center Points to the 2k

Consider the first order regression model. If interaction terms are added to a main effects orfirst-order model, resulting in

∑ ∑∑= <

+++=k

j jijiijjj xxxY

10 εβββ (A31)

then there is a model capable of representing some curvature in the response function. Thiscurvature, of course, results from the twisting of the plane induced by the interactionterms jiij xxβ . 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:

εββββ ++++= ∑∑∑∑=<=

k

jjjj

jijiij

k

jjj xxxxY

1

2

10 (A32)


where the jjβ represent pure second-order or quadratic effects.

In running a two-level factorial experiment, usually first there is a fitting of the first-ordermodel, but being alert to the possibility that the second-order model is more appropriate.There is a method of replicating certain points in a 2k factorial that will provide protectionagainst curvature from second-order effects as well as allow an independent estimate of errorto be obtained. The method consists of adding center points to the 2k design. These consist ofnc replicates run at the points ( )kixi ,...,2,10 == . One important reason for adding thereplicate runs at the design center is that center points do not affect the usual effect estimatesin a 2k design.

Consider a 22 design with one observation at each of the factorial points ( )−−, , ( )+−, ,

( )−+, and ( )++, and cn observations at the center point ( )0,0 . Let Fy be the average of

the four runs at the four factorial points, and let Cy be the average of the cn runs at the

center point. If the difference CF yy − is small, then the center points lie on or near the plane

passing through the factorial points, and there is no quadratic curvature. If CF yy − is large,then quadratic curvature is present. A single-degree-of-freedom sum of squares for purequadratic curvature is given by:

( )CF

CFCFpuraquad nn

yynnSS+−

=2

(A33)

where Fn is the number of factorial design points. This sum of squares may be compared tothe error mean square to test for pure quadratic curvature.

When points are added to the center of the k2 design, then the test for curvature actuallytests the following hypotheses:

∑

∑

=

=

≠

=

k

jjj

k

jjj

H

H

11

10

0:

0:

β

β

Furthermore, if the factorial points in the design are unreplicated, one may use the Cncenter points to construct an estimate of error with 1−Cn degrees of freedom.

The mean square for pure error is calculated from the center points as follows:

( )

11int

2

−

−=

−=

∑c

spocentrali

C

EE n

yy

nSSMS (A34)


A2.5. Spherical Central Composite DesignAssuming a second-order model:

εββββββ ++++++= 2222

2111211222110 xxxxxxY

it is not possible to estimate the six parameters β because there are only 5 points in thedesign, 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.

cc

2x

1x

( )1 a

b ab

+xα−xα

+yα

−yα

1x

2x

cc

bc

b ab

abc

+xα

( )1 a

c ac

−xα

+yα

−yα

+zα

−zα

3x

Figure A2. Central Composite Design with k = 2 and k = 3.

In a CCD there are Cn+8 runs if 2=k or Cn+14 runs if 3=k . In general, the CCD

consists of a 2k factorial with Fn runs, k2 axial or star runs, and Cn center runs.There are two parameters in the design that must be specified: the distance α of the axial

runs from the design center and the number of center points Cn .The choice of α depends mainly on the region of interest. When the region of interest is

spherical, some central runs are included in the design in order to obtain a reasonably stablevariance of the predicted response. Generally, three to five center runs are recommended toobtain a good estimate of the experimental error.

A2.6. Face-Centered Central Composite DesignIn many situations, the region of interest is cuboidal rather than spherical. In these cases, auseful variation of the central composite design is the face-centered central composite designor the face-centered cube, in which 1=α .

This design locates the star or axial points on the centers of the faces of the cube asshown in Figure A3. This variation of the central composite design requires only three levelsof each factor, and in practice it is frequently difficult to change factor levels.


Figure A3. Face-centered central composite design for k = 3.

The face-centered central composite design is not rotatable, but it does not require asmany center points as the spherical CCD: 32÷=Cn is sufficient to provide a good varianceof prediction throughout the experimental region.

References

[1] Kabs H., Operational Experience with Siemens-Westinghouse SOFC CogenerationSystems, 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 systemfor a SOFC CHP100, Eco-efficiency 2005 – HYSYDays 1st World Congress of YoungScientists on Hydrogen, 18-20 May 2005, Torino, Italy.

[5] Leone P., Santarelli M., Calì M., Computer experimental analysis of the CHPperformance of a SOFC stack by a factorial design, J. Power Sources, 156 (2006),400-413.

[6] Calì M., Santarelli M.G., Leone P., Design of experiments for fitting regression modelson the tubular SOFC CHP100 kWe: screening test, response surface analysis andoptimization, accepted for publication on International Journal of Hydrogen Energy,2006.

[7] Leone P., Santarelli M., Computer Experimental analysis of a tubular SOFC CHP toevaluate 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] Calì M., Santarelli M., Leone P., Comparison of the behavior of the CHP-100 SOFCField Unit fed by natural gas or hydrogen through a computer experimental analysis,World Hydrogen Technology Convention, Singapore, 2005.

[9] Calì M., Fontana E., Giaretto V., Orsello G., Santarelli M., The EOS Project: a SOFCpilot plant in Italy- safety aspects, HySafe - International Conference on HydrogenSafety, Pisa (Italy), 2005.


[10] Cali M., Debenedictis F., Leone P., Santarelli M., Verda V., Benchmark characterizationof a tubular SOFC CHP generator: model description and experimental validation,ASME 1st European Fuel Cell Technology and Applications Conference, Rome (Italy),December 2005.

[11] Calì M., Leone P., Santarelli M., Orsello G., Disegna G., Operation of the tubular SOFCCHP100 kWe Field Unit in Italy. General topics and operation description by means ofregression models, World Hydrogen Energy Conference WHEC 2006, Lyon (France),June 2006.

[12] Calì M., Orsello G., Santarelli M., Leone P., Experimental activity on the tubular SOFCCHP100 kWe Field Unit in Italy: factor significance, effects and regression modelanalysis, Proceedings of ESDA2006, 8th Biennial ASME Conference on EngineeringSystems Design and Analysis, Torino (Italy), July 2006.

[13] Santarelli M., Leone P., Calì M., De Benedictis F., Experimental activity on the tubularSOFC CHP100 generator in Italy: regression models analysis and optimization, LucerneFuel 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) 183–188.

[17] Santarelli M., Leone P., Gariglio M., Calì M., Spinelli P., Orsello G., De Benedictis F.,Experimental analysis of the polarization effects at variable local temperature and fuelconsumption 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 MaterialsScience: 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] Calì M., Santarelli M., Leone P., Experimental analysis of the behavior of someperformance 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 ReportSemi Annual, Acumentrics Advanced Power & Energy Technologies, February 2006.

[21] Draper R., DiGiuseppe G., Optimal design of current take off bus bars for tubular solidoxide fuel cells, Proceedings of ASME European Fuel Cell Conference 2005, Rome(Italy), 2005.

[22] Costamagna P., Honneger K., Modeling of Solid Oxide Heat Exchanger IntegratedStacks 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 TheirElectrochemical Properties in H2-H2O Atmospheres, Journal of ElectrochemicalSociety, Vol. 141, No. 8, pp. 2129-2134, 1994.

[24] Mogensen M. et al., H2-H2O-Ni-YSZ Electrode Performance, Journal ofElectrochemical Society, Vol. 151, No. 9, pp. A1436-A1444, 2004.


[25] Kim J., Virkar A.V., Fung K.Z., Metha K., Singhal S.C., Polarization effects inintermediate temperature, Anode-Supported Solid Oxide Fuel Cells, J. ElectrochemicalSoc., Vol. 146 (1), pp. 69-78, 1999.

[26] Zhao F., Anil Virkar V., Dependence of polarization in anode-supported solid oxide fuelcells on various cell parameters, Journal of Power Sources, Vol. 141, pp. 79–95, 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 solidoxide fuel cell modeling, Journal of Power Sources, Vol. 110, pp. 186–200, 2002.

[29] Vora S.D., Advances in Solid Oxide Fuel Cell Technology (SOFC) at SiemensWestinghouse, 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 fuelcell, Journal of Electrochemical Society, Vol. 142 (11), pp. 3792-3800, 1995.

[31] Campanari S., Iora P., Definition and sensitivity analysis of a finite volume SOFCmodel 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 7thEuropean SOFC Forum, Lucerne (Switzerland), 2006.

[33] US Patent n°6,207,311 B1, Inventors: Baozhen L., Ruka R., Singhal S.C., Solid OxideFuel Cell operable over wide temperature range, Assigne: Siemens Westinghouse PowerCorporation, 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.


Chapter 3

ALTERNATIVE SULFONATED POLYMERS TO NAFION FOR PEM FUEL CELL

Angelo Basile* and Adolfo Iulianelli Institute on Membrane Technology, ITM-CNR, c/o University of Calabria, Via P. Bucci,

Cubo 17/C, 87030 Rende (CS), Italy

Abstract

Commonly, polymer electrolyte membrane fuel cells operate at temperature <100°C 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: [email protected], Phone: (+39) 0984 492013, Fax: (+39) 0984 402103

Angelo Basile and Adolfo Iulianelli 136

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:

a) the system can be simpler since it does not need a reformer, b) ambient temperature start-up is possible, c) infra structure built for gasoline can be used without significant variations.

Vice versa, some disadvantages are present to prevent DMFC from commercialization:

a) high cost of the polymeric membranes, b) high methanol crossover.

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:

a) PEEK electrophilic or nucleophilic sulfonation, b) S-PEEK and non-functional polymers and/or solids blending, c) S-PEEK, heteropolycompounds and polyetherimide doping with inorganic acids, etc.

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 (σ), 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.

The Pure S-PEEKs’ Synthesis and the Variables Affecting Their Performances

The distinct differences between the S-PEEK and the Nafion are qualitatively explained by differences in the microstructures and in the acidity of the sulfonic acid functional groups [1-3]. A perfluorosulfonic polymer such as Nafion naturally combines, in one macromolecule, the extremely high hydrophobicity of the perfluorinated backbone with the extremely high hydrophilicity of the sulfonic acid functional groups. Especially in the presence of water, this gives rise to a hydrophobic/hydrophilic regions’ separation. The sulfonic acid functional groups aggregate to form a well connected hydrophilic domain, which is responsible for the protons and water transport. Vice versa, the hydrophobic domain provides the polymer with the morphological stability and prevents the polymer dissolving in water. For the S-PEEKs, a different situation takes place with respect to the Nafion for what concerns the transport properties and the morphological stability. In fact, as a result of the smaller hydrophilic/hydrophobic difference (the backbone is less hydrophobic and the sulfonic acid functional group is less acidic and, therefore, also less polar) and of the smaller polymer backbone flexibility, the separation into a hydrophilic and a hydrophobic domain is less pronounced for the S-PEEKs than the Nafion [4]. As schematically illustrated in Figure 1 the water filled channels in sulfonated PEEK are narrower compared to those of the Nafion. They are less separated and more branched with more dead-end “pockets”. These features correspond to the larger hydrophilic/hydrophobic interface and, therefore, also to a larger average separation of neighbouring sulfonic acid functional groups [4].

However, focusing the attention on the PEEK polymer, it can be said that its mechanical properties are affected by several variables such as molecular weight, polymer composition, etc., while the membrane characteristics depend on the polymer structure, the casting solvents, the membrane uniformity, etc, but, the most important parameter affecting the S-PEEK behaviours is the polymer sulfonation procedure, which can be carried out by using different approaches evidencing each one a series of advantages and disadvantages.


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.

The S-PEEK polymers can be synthesized by means of a direct electrophilic sulfonation via concentrated sulphuric acid or via chlorosulphuric acid.

When 95-98% of H2SO4 is used, the PEEK polymer degradation and the cross-linking reactions, which occur using 100% H2SO4 or chlorosulphuric acid, can be avoided. The cross-linking presumably involves the sulphone groups formation. Therefore, using H2SO4 containing a few percent of water it is possible to decompose the aryl pyrosulfate intermediate, required for sulfone groups formation, by means of the water’s action.

Furthermore, the PEEK sulfonation can be realized introducing directly the sulfonic acid group onto the polymer back-bone [5-7] by modification or by polymerizing sulfonated monomers [8,9].

The S-PEEK direct synthesis from sulfonated monomer seems to be more advantageous than the post-sulfonation because it avoids the cross-linking formation and other side


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 80–90°C and for these reasons, they are not suitable for the DMFC applications [14].

Figure 2. Structure and atom numbering of S-PEEK, x + y = n, y/(x + y) = DS.

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,A’B,B’,C,D), represents the DS


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


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.

Table 1. S-PEEK sulfonation degree dependence by the Monomer M/Monomer K ratio in the nucleophilic aromatic substitution reaction [10]

Polymer Ratio (Monomer M/Monomer K)

DS (calculated by titration) [%]

S-PEEK 2:8 44 4:6 76 6:4 110 10:0 169


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. a) The glass transition temperature is a thermo-resistance indicator of the polymers.

PEMFC and DMFC normally work at temperature ranging between 80-120°C and, therefore, Tg higher than these values indicate if the electrolyte polymer is temperature resistant and if any thermo-degradation takes place during the fuel cell operations. As illustrated in Figure 6, the Tg increases monotonically with the DS, i.e. with the gradual introduction of –SO3H groups into the PEEK polymer. The mechanism explaining the Tg increasing with the DS is related to the increased intermolecular interaction due to the hydrogen bonding of SO3H groups (ionomer effect) and to the increased molecular bulkiness.

Figure 6. Glass transition temperature depending on DS for a S-PEEK sample. Reprinted from [18] with permission from Elsevier.

b) The enhancement of hydrophilicity due to the sulfonation of PEEK can be followed by water absorption as a function of the DS. In Figure 7 is illustrated an example of W.U. of S-PEEK membranes plotted against the number of sulfonic acid groups. As shown, the water uptake increases with DS and reaches 120% for a S-PEEK membrane with 80% DS (0.8


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.


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 520°C. 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 300°C 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 > 80°C. 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:

• Crosslinking of sulfonated polymers. • Blending of cation-exchange polymers which show a specific interaction with the

cation exchange group. • Blending with polymers capable of hydrogen bonds formation. • Compatibilization of blend polymers by hydrogen bonds formation. 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,


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-PEEK’s 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 85°C 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 100°C. 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 20–50°C, more rapidly between 50 and 100°C and then gradually again up to 145°C [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 water’s lost velocity induces the conductivity decrease at temperature of almost 50°C. 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 145°C. Therefore, not only the capacity of S-PEEK’s 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

MeOH permeability [cm2/s] Conductivity [S/cm]

Membrane type S-PEEK DS (%)

W.U.25°C (%)

IEC (meq/g) T = 25 °C T = 80 °C T = 25

°C T = 80 °C β [Ss/cm3]

Casting solvent Sulfonation PEEK procedure

S-PEEK [10] 80 37 1.312 5.0·10-8 3.0·10-7 2.8·10-3 6.7·10-3 4.3 DMF Direct synthesis by sulfonated monomer

S-PEEK [16] 82 - - - - 5.5·10-3 2.4·10-2 - DMAc Concentrated H2SO4 S-PEEK [21] - 75 2.48 - - - 4.0·10-3 - NMP Concentrated H2SO4 S-PEEK [14] 47 100 - - 1.5·10-7 5.0·10-3 1.7·10-2 5.1 DMF Concentrated H2SO4 S-PEEK [20] 93 67 2.37 1.54 · 10-6 - 9.3·10-3 5.5·10-2 3.8 DMAc Concentrated H2SO4

S-PEEK [46] - 92 - - - 8.0·10-2 1.6·10-1 - DMAc Aromatyc nucleophilic polycondensation

S-PEEK [4] 70 - 1.43 - - 7.5·10-2 1.5·10-1 - - Concentrated H2SO4 S-PEEK [22] 69 23 - - - 3.8·10-2 - - DMF Concentrated H2SO4 S-PEEK [13] 65 - - - - 1.0·10-5 7.5·10-4 f DMF Concentrated H2SO4 S-PEEK [17] 79 32 - - - 1.6·10-2 - - DMAc Concentrated H2SO4 Si-SPPSU/S-PEEK [26] 90 50a - - - - 6.4·10-2 a - DMA - SiS-PEEK [25] 90 240 - - - 2.0·10-2 - - DMSO - S-PEEK/BPO4 [32] 50-80 >80 - - - 7.5·10-2 - - DMAc - HPA/S-PEEK [18] 80 600 - - - - 9.5·10-2 b - DMAc - S-PEEK/PANI [34] 44.5 42.2c 1.21 2.08·10-7 - - 1.8·10-3 b 3.9 DMF - S-PEK/ZP-ZrO2 [35] 60 35.7 - - 1.0·10-2 - - DMF - S-PEEK-WC-3h [39,40] 40 14 0.76 - - - 1.7·10-2 a - DMA Concentrated H2SO4 S-PEEK-WC [38] 82 - 1.47 1.9·10-6 - - 2.5·10-2 d 4.1 DMF Chlorosulphuric acid Zr/S-PEEK-WC [42] 104 60 4.7·10-8 - - - - DMF - NAFION 117 - 38 0.91 - - 1.0·10-1 1.5-1.9·10-1 4.4e - -

a 120 °C b 100 °C c 80 °C d 115 °C e [42] f 60 °C


In Table 2, several literature data concerning proton conductivity for different pure S-PEEKs are reported at 25 and 80°C 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 25°C 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 × 10−5 S/cm at room temperature and a DS equal to 65% and at 60°C it is ranged between 7 × 10−4 and 8 × 10−4 S/cm, while D.J. Jones et al. [46] found a conductivity of 3.5 × 10−3 S/cm with a 65% DS at 80°C. 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.

Figure 9. Reaction scheme of DMF or DMAc transformation in presence of sulphuric acid.

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.


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].


Figure 11. The selectivity β of a S-PEEK sample and Nafion117. Reprinted from [20] with permission from Elsevier.

High proton conductivity (σ) 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 β (=log(σ/P)) [24], a comparison can be more easily carried out within different S-PEEK membranes. In Figure 11, the β factor comparison between a S-PEEK sample and Nafion117 membranes is shown. Increasing the DS, the S-PEEK’s selectivity decreases because, overcome DS = 70%, the sulfonated polymer becomes more soluble in methanol and then methanol permeability increases [20].

Although generally the Nafion’s proton conductivities are higher than those of the S-PEEKs, the Nafion methanol permeability increases with increasing temperature and the β factor decreases making the S-PEEKs more attractive than Nafion for DMFC applications.

Modified Sulfonated PEEK Membranes for PEMFC and DMFC

As already discussed, it is well known that in polyaryl etherketone-based systems such as S-PEEK, the absence of a significant hydrophilic/hydrophobic separation results in very narrow and poorly connected water channels and large separation between the sulfonic acid groups. As a consequence, dehydration causes a strong decrease in conductivity and poor morphological stability. Several approaches can be proposed to attain the correct balance between the hydrophilic and hydrophobic components, among them: the use of inorganic fillers for the development of composite systems, acid or basic doping and association of polymers.

An interesting method to improve the S-PEEK performances is to prepare physically cross-linked membranes by blending ionomeric polymers with different mechanical properties. The blending technique makes the advantage of combining the positive features of each component with a very simple procedure. Better electrochemical characteristics are


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 (80–200°C) and most of them, being based on composite materials or acid–base 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 10−3 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


(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 2×10-2 – 1×10-2 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 1.9×10−1 S/cm and its sodium form has a conductivity of about 10−2 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 250°C. 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 80°C, improving also the conductivity’s 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


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 acid–base 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


deposition, a cluster formation on polymer matrix is responsible, due to its chemical interaction with the methanol, for the lower methanol crossover with respect to the pure S-PEEK and the Nafion [34].

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 PEEK’s 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,


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 conductivity’s 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).

Electrochemical Characteristic of PEMFC and DMFC

The polarization curve is the most important characteristic of a fuel cell and it is used for diagnostic purpose, as well as for sizing and control of a fuel cell. The volt–ampere characteristic of a PEM or of a DMFC can be, in principle, explained from basic thermodynamic and electrochemistry. Imagining to analyse a proton exchange fuel cell as a fully reversible process, the maximum voltage generated (denoted as E) should correspond to the electromotive force (EMF) generated at the output. The process is fully reversible if, combining hydrogen (or any suitable fuel) and oxygen into water in the fuel cell, it does not make unrecoverable losses (for example it does not produce heat). In the reversible case, it can be assumed that the energy change in the system (Gibbs free energy of formation) is converted into electrical energy. Then

ΔG(T, P, fn) = nNeE or

E = ΔG(T, P, fn)/nNe, where ΔG 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 Avogadro’s 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 80°C, 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


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

Reference Polymer OCV [V] R [Ωcm2] Li et al. [14] S-PEEK 0,64 20.6a

Basile et al. [40] S-PEEK-WC 0,78 4,5 Ren et al. [23] S-PEEK - 1.34b

Wakizoe et al. [47] Nafion 0,96 0,168 a [11] b 60 °C

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

ΔVr = 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


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

ΔVOV = 0

2.3 ln( )( , , )

RT inF i M T Sα

where R is the gas constant, T temperature, α 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


PEEK modification procedure such as chemical polymerization with polyaniline or PEEK chemical modification as the PEEK-WC, useful in PEMFC/DMFC application in sulfonated form or like blend with zirconia oxides.

Several kind of the reported S-PEEK based composite membranes show also to reduce the hydrophilic/hydrophobic separation and the hydrodynamic solvent transport (water and probably also methanol permeation). Therefore, polymers based on pure S-PEEK and its blends are not only interesting low-cost alternative membrane materials for PEMFC applications but they may also help to reduce the problems associated with high water and methanol cross-over in DMFCs.

For what concerns the pure and blended S-PEEK characterisation in terms of voltage and power density, unfortunately, not many studies have been carried out and, then, only the few data did not allow to discuss deeply on the advantages of different type of S-PEEK over Nafion in PEMFC and DMFC applications.

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


[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) 17–34

[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 (110-160°C), 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. D’Epifanio, D. Marani, M. Trombetta, E. Traversa, S. Licoccia, SPEEK/PPSU-based organic–inorganic membranes: proton conducting electrolytes in anhydrous and wet environments, J. Membrane Sci., 279 (2006) 186–191


[26] M.L. Di Vona, D. Marani, A. D’Epifanio, E. Traversa, M. Trombetta, S. Licoccia, A covalent organic/inorganic hybrid proton exchange polymeric membrane: synthesis and characterization, Polymer, 46 (2005) 1754-1758

[27] K.D. Kreuer, Proton conductivity: materials and applications, Chem. Mater., 8 (1996) 610-641

[28] S.D. Mikhailenko, S. Kaliaguine, J.B. Moffat, Electrical Impedance studies of the ammonium salt of 12-tungsto-phosphoric acid in presence of liquid water, Sol. State Ion., 99 (1997) 281-286

[29] S.D. Mikhailenko, S.M.J. Zaidi, S. Kaliaguine, Development of zeolite based proton conductive membranes for use in direct methanol fuel cells, Report in Natural Resources Canada, Ottawa, 1997

[30] M. Drzewinkski, W.J. Macnight, Structure and properties of sulfonated polysulfone ionomers, J. Appl. Pol. Sci., 30 (1985) 4753-4770

[31] Y. Sone, P. Ekdunge, D. Simonsson, Proton conductivity of Nafion117 as measured by a four electrode AC Impedance method, J. Electrochem. Soc., 143 (1996) 1254-1259

[32] S.D. Mikhailenko, S.M.J. Zaidi, S. Kaliaguine, Sulfonated poly ether ether ketone based composite polymer electrolyte membranes, Catal. Tod., 67 (2001) 225-236

[33] K. Miyatake, K. Oyaizu, E. Tsuchida, A.S. Hay, Synthesis and properties of novel sulfonated arylene ether/fluorinated alkane copolymers, Macromol., 34 (2001) 2065-2071

[34] R.K. Nagarale, G.S. Gohil, V.K. Shahi, Sulfonated poly(ether ether ketone)/polyaniline composite proton-exchange membrane, J. Membrane Sci., 280 (2006) 389-396

[35] B. Ruffmann, H. Silva, B. Sculte, S.P. Nunes, Organic/inorganic composite membranes for application in DMFC, Sol. State Ion., 162-163 (2003) 269-275

[36] E. Drioli, H.-C Zhang, A study of polyetheretherketone and polyarylsulfone ultra-filtration membranes, Chimica Oggi, 11 (1989) 59-63

[37] F. Trotta, E. Drioli, G. Moraglio, E. Baima Poma, Sulfonation of polyetheretherketone by chlorosulfuric acid, J. Appl. Pol. Sci., 70 (1998) 477-482

[38] E. Drioli, A. Regina, M. Casciola, A. Oliveti, F. Trotta, T. Massari, Sulfonated PEEK-WC membranes for possible fuel cell applications, J. Membrane Sci., 228 (2004) 139-148

[39] L. Paturzo, A. Basile, A. Iulianelli, J. C. Jansen, I. Gatto, E. Passalacqua, High temperature proton exchange membrane fuel cell using a sulfonated membrane obtained via H2SO4 treatment of PEEK-WC, Catal. Tod., 104 (2005) 213-218

[40] A. Basile, L. Paturzo, A. Iulianelli, I. Gatto, E. Passalacqua, Sulfonated PEEK-WC membranes for proton-exchange membrane fuel cell : Effect of the increasing level of sulfonation on electrochemical performances, J. Membrane Sci., 281 (2006) 377-385

[41] L. Jia, X. Xu, H. Zhang, J. Xu, Sulfonation of polyetheretherketone and its effects on permeation behavior to nitrogen and water vapor, J. Appl. Pol. Sci., 60 (8) (1996) 1231-1237

[42] A. Regina, E. Fontananova, E. Drioli, M. Casciola, M. Sganappa, F. Trotta, Preparation and characterization of sulfonated PEEK-WC membranes for fuel cell applications A comparison between polymeric and composite membranes, J. Power Sou., 160 (2006) 139-147

[43] B. Baranowski, Non-equilibrium thermodynamics as applied to membrane transport, J. Membrane Sci, 57 (1991) 119-159


[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. Rozière, 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


Chapter 4

A POTENTIAL ALTERNATIVE IN THE ELECTRICUTILITY

Francisco Jurado*

University of Jaén, Department of Electrical Engineering, 23700 Alfonso X,nº 28, EPS Linares (Jaén), Spain

Abstract

A fully mature fuel cell industry constitutes a potential opportunity to electric utilities. Itcould meliorate the technical and financial performance of existing distribution lines byimproving service quality and reliability.

When fuel cells connect to the power system, both the owner of the energy resource andthe central power system benefit. Reliability increases for both because they can support eachother.

However, the interconnection of fuel cell plants to the grid is still hindered by restrictiveconditions and procedures for grid connection. Problems arise with regard to determination ofthe point of connection, safety and stability issues. Most important it is needed to establish astandardized technical interface for allocation of connection that take into account possiblepositive effects of distributed generation on transmission and distribution losses.

At last, the aim would be the substitution of conventional coal-fired generation withIntegrated Gasification Combined Cycle plants that integrate fuel cells and gas turbines. Thischapter studies the use of distributed resources for ancillary services and simulating theimpact of distributed resources on utility distribution networks.

Keywords: Biomass, distributed generation, electric system stability, fuel cells, hybridsystems.

1. Introduction

The interaction realized by fuel cell—microturbine hybrids derive primarily from using therejected thermal energy and combustion of residual fuel from a fuel cell in driving the gas

* E-mail address: [email protected], Telephone: +34-953-648518, Fax: +34-953-648586

Francisco Jurado162

turbine. This leveraging of thermal energy makes the high temperature Molten CarbonateFuel Cells (MCFCs) ideal candidates for hybrid systems. Use of a recuperator contributes tothermal efficiency by transferring heat from the gas turbine exhaust to the fuel and air used inthe system.

Traditional control design approaches, consider a fixed operating point in the hope thatthe resulting controller is robust enough to stabilize the system for different operatingconditions. On the other hand, adaptive control incorporates the time-varying dynamicalproperties of the model (a new value of gas composition) and considers the disturbancesacting 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 percentefficiency and near-zero emissions. Fuel flexibility enables the use of low-cost indigenousfuels, renewables, and waste materials. However, the characteristics of gas from biomassgasification may vary significantly. In many practical cases, the designer has a model of thesystem, but the system parameters are subject to uncertainty. In this case robust control can beused to design a control system with a guaranteed level of performance, as long as the systemuncertainties remain the assumed bounds.

To determine the potential impacts of fuel cells on future distribution system, dynamicmodels 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 orderlinear system models of Solid Oxide Fuel Cells (SOFCs) from time domain simulations. Themethod is the Box-Jenkins algorithm for calculating the transfer function of a linear systemfrom samples of its input and output.

When connected in small amounts, the impact of distributed generation on distributionsystem 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 indistribution system stability studies. The model is applied to a distributed utility grid that usesa solid oxide fuel cell plant as distributed resource. Examinations include transient stabilityand 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 modelingof SOFC plants on the distribution system using identification algorithms. Effect of a SOFCplant on distribution system stability is discussed in Section 5, and finally, conclusions areprovided in Section 6.

2. Study of Molten Carbonate Fuel Cell-Microturbine HybridPower Cycles

Fuel cells are particularly well suited for power generation in cogeneration plants becausethey convert energy directly into electricity in an electrochemical process whilesimultaneously producing heat (Silveira et al., 2001; Akai et al., 1997; Leal and Silveira,

A Potential Alternative in the Electric Utility 163

2002). However, it is first necessary to reform fuels such as natural gas, converting it into agas 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 itinternally 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). Thelargest demonstration of MCFC technology has been California’s 2-MW Santa ClaraDemonstration Project (Fuel Cell Engineering Corporation, 1997).

Microturbines, which are typically fueled with natural gas, generate between 25 and 200kW of electricity. Their small size and relatively low cost allow them to be located nearwhere they are needed. They can operate at very low emission levels and reduce theefficiency losses and environmental impact of large transmission and distribution systems. Inthis 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 andrelatively high emissions of small gas turbines, and the high cost of small fuel cell powerplants (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 2.3describes 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.

2.1. Molten Carbonate Fuel Cells

MCFC can reach fuel-to-electricity efficiencies approaching 60%, considerably higher thanthe 37-42% efficiencies of a phosphoric acid fuel cell (PAFC) plant. When the waste heat iscaptured and used, overall fuel efficiencies can be as high as 85%.

Improved efficiencies are one reason why MCFC offers significant cost reductions overPAFC technology. Another is that the electrodes of a MCFC be made of nickel catalystsrather than the more costly platinum of PAFC systems.

Natural gas is internally reformed, partially in an internal reforming unit and partially atthe 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 internalreforming (DIR) which provides for better thermal management. At the cathode, oxygenreacts with carbon dioxide and electrons to form carbonate ions:

−− →++ 2322 22/1 COeCOO (1)

The carbonate ions flow through the electrolyte matrix from cathode to anode. At theanode, the carbonate ions are consumed by the oxidation of hydrogen to form steam andcarbon dioxide, releasing electrons to the external circuit:

−− ++→+ eCOOHCOH 222232 (2)

Francisco Jurado164

Figure 1. IIR/DIR structure of MCFC stack.

Cell voltage under load current (A/cm) is expressed as (Hirschenhofer et al., 1998):

izVV concactoc −−−= ηη (3)

Activation polarization is caused by electrode kinetics while concentration polarization iscaused by concentration gradients in the electrode. Equilibrium potential is described by theNernst equation (Hirschenhofer et al., 1998):

aCOaOH

cCOcOaHoo pp

pppF

RTEV,,

,2/1,,

22

222ln2

+= (4)

Partial pressures (normalized to atmospheric pressure) depend on anode/cathode gaspressure and composition while standard potential and ohmic impedance are both temperaturedependent. Fuel cell polarization losses are generally dependent on partial pressures,temperature, and current density, and are spatially distributed in an actual cell. In this paperthe dynamic model is lumped parameter, where outlet properties are equal to averageproperties (Ding et al., 1997; Lukas et al., 1999; Lukas et al., 2001).

2.2. Gas Turbine

A thorough introduction to gas turbine theory is provided in (Cohen et al., 1987). There alsoexists a large literature on the modeling of gas turbines. Model complexity varies according to


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 dividingthe gas turbine into a number of sections. Throughout each section, the thermodynamic stateis 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 ofordinary differential equations, which facilitate easier application within a computersimulation program. For a detailed model, a section might consist of a single compressor orturbine stage. Much simpler models result if the gas turbine is decomposed into just threesections corresponding to the main turbine components i.e. compressor, combustor andturbine, as in (Hussain and Seifi, 1992).

Instead of applying the fundamental conservation equations, as described above, anothermodeling approach is to characterize gas turbine performance by utilizing real steady stateengine performance data, as in (Hung, 1991). It is assumed that transient thermodynamic andflow processes are characterized by a continuous progression along the steady stateperformance curves. This is known as the quasi-static assumption. The dynamics of the gasturbine, e.g. combustion delay, motor inertia, fuel pump lag etc. are then represented aslumped quantities separate from the steady-state performance curves. Very simple modelsresult 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. Aftercompression of the air to achieve the most favorable conditions for combustion, fuel gas ismixed with the air in the combustion chamber. Combustion takes place and the hot exhaustgases are expanded through the turbine to produce mechanical power. In terms of energyconversion, the chemical energy present in the combustion reactants is transferred to the gasstream during combustion. This energy - measured in terms of gas enthalpy- is then convertedinto mechanical work by expanding the gas through the turbine. Thus, the excess mechanicalpower available for application elsewhere, after accounting for the power required to drive thecompressor, is derived ultimately from the combustion process.

Compressor power consumption equation:

transc

ICac

hwPηηΔ

= (5)

Combustion energy equation:

( ) ( )( ) 0298

298298 25

=−+

+−+Δ+−

ispsis

coutpaafTinpgg

Tcw

TcwhwTcw (6)

Power delivery equation:

ITgTT hwP Δ=η (7)

Francisco Jurado166

cTmec PPP −= (8)

Fig. 2 shows the block diagram of the gas turbine.

Figure 2. Gas turbine.

2.3. Gas Turbine Control Configuration

The concept of the gas turbine control system, which is applied in this chapter, is based on theSpeedtronic Mark 4 description as presented in (Rowen, 1988). Some considerationsconcerning 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 twointerconnected subsystems. The subsystems are the fuel system (fuel valve with actuator), andthe turbine. The fuel flow out from the fuel systems results from the inertia of the fuel systemactuator and of the valve positioner.

The fuel system actuator equation is:

11e

sk

wf

ff +=τ

(9)

The valve positioner equation is:

dFcbs

ae+

=1 (10)

where the input variable to the fuel system is Fd. The output variable from the fuel systemmodel is wf . A single gas turbine does not require the digital setpoint feature. The kLHV factordepends on the LHV. The kLHV and 0.23 factors cater for the typical turbine power/fuel rate


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:

)(.).w(kT fLHV ωΔ50230 +−= (11)

Equation (11) allows the turbine torque to be calculated algebraically. This torque is usedin the equations which model the mechanical system:

ωTPmec = (12)

Input variables to the turbine are wf , Δω and ω. Output variable from the turbine is Pmec .For the purpose of this chapter only modulating control of mechanical side of the gas turbineis of interest. The simplified model of the gas turbine controller in this work consists of twoinputs and one output. Inputs to the controller are Pmec and ω. The output from the controllersis Fd .

The block diagram of the gas turbine control system is presented in Fig. 3 and describedby 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 incomingsignals.

Figure 3. Gas turbine control system.

Table 1. Gas turbine. Constants.

a b c kf τb

1 0.05 1 1 0.4

2.4. Fuel Cell/ Turbine Hybrid

The coupling of high temperature fuel cells with gas turbines is by no means new, and hasbeen 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

Francisco Jurado168

and Lundberg, 1999); (2) dynamic modeling, performance, technical issues, and integrationconsiderations (Liese et al., 1999); (3) microturbine specific studies (Campanari, 1999); and(4) details of experimental programs to demonstrate the coupling of a microturbine with asolid-oxide fuel cell SOFC (Leeper, 1999).

When used in combination with turbines, fuel cells can produce from 55 – 90% of theelectricity of the system while turbines produce the remainder. Several cycle configurationshave been proposed and the terminology is still evolving, but one useful way of looking at thedifferences is to divide the configurations into those that include “directly fired turbines” andthose 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 acombustor. In the “indirectly fired turbine” system, an atmospheric fuel cell is used and apressurized heat exchanger provides input to the turbine. When utilizing the fuel cell/gasturbine combination the combined efficiency of the system is raised to greater than 60 % andcriteria pollutant emissions are essentially eliminated. An example of a hybrid system with an“indirectly fired turbine” is shown in Fig. 4.

Figure 4. Fuel cell/ turbine hybrid system.

This type of integration could utilize any gas turbine system. The indirect-fired approachplaces the fuel cell in the exhaust of an indirect fired gas turbine. The air and residualproducts of the fuel cell are then fed to an atmospheric combustor. This combustor heats theair leaving the compressor via a heat recovery unit (HRU) and delivers it to the turbine. Thecombination of the atmospheric combustor and heat exchanger, replaces the normal internalpressurized combustor.

The turbine exhaust flows to the fuel cell anode exhaust oxidizer. Exhaust from the anodeexhaust 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 oxygenand 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 theturbine pressure, instead it operates at the preferred ambient pressure and its independent ofgas turbine cycle pressure ratio. The system works efficiently with a wide range of turbinecompression ratio. This allows taking a system developed for integration at the multi-MW


scale with industrial size turbines, and configuring a small MW class system using amicroturbine at a lower pressure ratio.

2.5. Adaptive Control

A real-world plant can be usually characterized by time-varying dynamical properties, whichaffect the plant behavior. Stochastic models are used to represent the disturbances acting atthe plant output because of the large number and different nature of the factors disturbing thenormal plant operation.

Robustness properties can usually be ensured by the feedback structure of the controlsystem. The feedback compensates for the deviation of the plant output signal value from itssetpoint: disturbances affecting the plant (load power variation) or change in the plant modelparameters (HHV), such a change, is convenient to observe parameter tracking.

It may be possible to identify the parameters of the controller that we are seeking. Thisscheme is called direct adaptive control, because we are going to obtain directly the requiredcontroller parameters through their estimation in an appropriately redefined plant model.

Adaptive control is usually used to cope with an unknown or/and changing plant to becontrolled (Aström and Wittenmark, 1995). Analysis and synthesis of such a control system ispossible only under some assumptions concerning the nature of the plant and its dynamics. Inthis chapter only linear, discrete-time plants disturbed in a stochastic manner will beconsidered. The following plant model will be used (Moscinski and Ogonowski, 1995):

( ) ( )( ) ( ) ( )ieiuzAzBziy k += −

−−

1

1

(13)

Equation (13) is one of the most typical in the field of adaptive control and non-standarddiscrete-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 ΔPL.

Generally, most control algorithms can be described by the structure and parameters ofthe difference equation:

( ) ( ) ( ) ( ) ( ) ( ) 0111 =+−+ −−− hizTiyzSiuzR ω (14)

The coefficients of the R(z-1), S(z-1) and T(z-1) polynomials and the h term are chosenbefore the simulation experiment and stay constant during the experiment.

The aim of the minimum-variance control algorithm is the minimization of the followingperformance index (Aström and Wittenmark, 1995; Aström and Wittenmark, 1990).

( ) ( )( ) ( ) ( ) ( ) ( ) ( )[ ]

⎪⎭

⎪⎬

⎫

⎪⎩

⎪⎨

⎧

+⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡−+= −−

−−

−−− 21

2

11

11 iuzQqizVKiy

zBzBzPEJ r ω (15)

Francisco Jurado170

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 offuel flow setpoint. This measurement of the stack current determines the plant output signaly(i). Stack current is continuously adjusted by the inverter control to maintain power. Thiscurrent is the setpoint of the plant output signal ω(i).

2.6. Results and Discussion

A plant consisting of a load is fed from the fuel cell/ turbine hybrid system. The selectedsystem comprises a 250 kW fuel cell and a 30 kW gas microturbine. The plant and the fuelcell/ microturbine system are modeled using MATLAB™.

All parameters correspond to a 2-stack equivalent. The fuel cell stacks used in thischapter are rated at 125 kW. Stack voltage is taken across a parallel connection of 2 stacks,each stack consisting of 258 cells.

Figure 5. Fuel cell control system.

To investigate transient behavior, the plant is assumed to be at steady state correspondingto rated power and subjected to a sudden variation in power demand. The HRU, powerconditioning system (PCS), and plant control system are included in the simulation (WorkingGroup on Prime Mover and Energy Supply Models, 1994; Hannett and Feltes, 2001; Tolbertet al. 1999; Lukas et al. 2000). The inverter is assumed to regulate load voltage perfectly, andsimply draws stack current proportional to load current and inversely proportional to stackvoltage using a power demand setpoint. The time constants for changes in power output forthe microturbines and fuel cell range from 5 ms to 50 s, and the PCS dynamics are notimportant (Lasseter, 2001).


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). Theplant parameters are: polynomials A(z-1), B(z-1) and standard deviation sd of white noise ofnormal distribution. Parameters of RLS are: initial values of diagonal of covariance matrixP_init and forgetting factor λ.

The disturbance considered is a load power variation ΔPL. This is zero-mean, white noiseand has constant variance σ2.

The controller has the form (Aström and Wittenmark, 1990):

( ) ( )( ) ( )iyzRzS

iuc

c1

1

−

−

−= (16)

where Sc of order 2 and Rc of order 2 . The structure of the controller is calculated accordingto the plant structure. To show the ability of the controller to adapt to varying operatingconditions, a new value of gas composition is introduced at 100 s what provides time-varyingplant. 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 valueof the output signal standard deviation is obtained in the steady state. The plant is assumed tobe of ARX type (AutoRegressive with eXogenous input).

Figure 6 a. Control signal U.

Figure 6 b. Load current deviation ΔIdc.

Francisco Jurado172

The results with the derived model are summarized. Fig. 6 a presents the control signalU. Referring to the results shown in Fig. 6 b, it can be seen the stack current deviation ΔIdc.Stack voltage starts at its initial value, corresponding to regulated power, and drops quickly toa sudden increase in stack current, in turn, controlled by the inverter to maintain the newpower setpoint. Stack voltage varies according to the temperature dependence, while stackcurrent 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 doesnot cover the entire plant uncertainty.

Figure 8 a. Polynomial R parameters (r1, r2).


Figure 8 b. Polynomial S parameters (s1, s2).

3. Robust Control for Fuel Cell-Microturbine Hybrid Power PlantUsing Biomass

Currently the design of control systems relies heavily upon an explicit mathematical model ofthe system. However, such a model is often very difficult and sometimes impossible to find.For this reason, conventional control systems are usually based on a linearized and highlysimplified mathematical model of the dynamics of the system. From another point of view,robust control incorporates the varying or uncertain parameters of the model. This workconsiders the application of a robust control methodology that allows the controller toautomatically adjust to changing process variables and thereby provide a uniform responseover a wide range of operating conditions.

The main contribution of this chapter is the modeling of both microturbines and fuel cellsin a manner suitable to study the robust control.

MCFC is a class of “second generation” fuel cells. Designed to operate at highertemperatures than phosphoric acid fuel cell (PAFC), this technology can achieve higher fuel-to-electricity and overall energy use efficiencies than lower temperature cells. In a MCFC, theelectrolyte is made up of lithium-potassium carbonate salts heated to about 1200ºF (650º C).At these temperatures, the salts melt into a molten state that can conduct charged particles,called ions, between two porous electrodes. The MCFC can reach fuel-to-electricityefficiencies approaching 60%, considerably higher than the 37-42% efficiencies of a PAFCplant. When the waste heat is captured and used, overall fuel efficiencies can be as high as85% (Layne and Holcombe, 2000).

This Section is structured as follows. Subsection 3.1 describes the biomass gasificationand Subsection 3.2 outlines the robust control. Finally, Section 3.3 depicts some simulationresults and discussion.

3.1. Biomass Gasification

A power plant can generate electric power using biomass from the trees. A gasifier is capableof converting tons of wood chips per day into a gaseous fuel that is fed into a fuel cell. Thecomposition of gas depend on biomass gasification technology.

Francisco Jurado174

Gasification is the term often used to describe the thermochemical processes that takeplace between a solid fuel and a gaseous reacting media, like air, oxygen or steam in order toproduce a combustible gas or synthesis gas rich in carbon monoxide and hydrogen. Thedifference compared to combustion is that the amount of oxidant, i.e. oxygen in air or pureoxygen or steam, is not high enough to enable complete oxidation, i.e. combustion. Instead, apartial oxidation, i.e. gasification, will take place. The overall gasification process can bedivided 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 calleddevolatilization. This is followed by pyrolysis, the step where the major part of the carboncontent 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 thelast step, called gasification, the char is partly oxidized into gaseous products and partly burntto generate the heat needed for the overall gasification process. Even though the reactionstaking place during gasification are not very complicated and complex themselves, thedifficult part is to establish the kinetic relations between the different species.

“Hydrogen products” are defined as any hydrogen (H2) and other gases, whose usefulnessor value is defined by its hydrogen or hydrogen-plus-methane (CH4) content.

Biomass has a low fixed carbon/volatile matter ratio, favoring processes that maximizepyrolysis and cracking, producing higher CH4/H2 ratios, desirable for use in the most nearlycommercial fuel cells.

The reactions taking place during the different steps of gasification are:

( ) 22 HCOOHsC +→+ (17)

222 HCOOHCO +→+ (18)

( ) 422 CHHsC →+ (19)

224 3HCOOHCH +→+ (20)

In this work, the fuel cell model uses a constant average cell voltage and a constantaverage fuel utilization to calculate the production of the fuel cells according to Faraday'slaw, Equation (21):

( ) cafueliidcel FVNUynP ∑=, (21)

where ni is the number of electrons transferred during the electrochemical conversion of i tocarbon dioxide and water. This number equals two for hydrogen and carbon monoxide, andwould be eight for direct electrochemical oxidation of methane. In Equation (21), CH4 isincluded as a fuel only in the case of internal reforming.

When biomass is gasified, relatively large quantities of tar are produced, e.g., from 10 %of the fuel in an updraft gasifier to 2 to 4 % in a fluidized bed. The amount and compositionof the tars are dependent on the fuel, the pyrolysis conditions and the secondary gas phase


reactions. If the tar is allowed to condense when the gas is cooled (condensation temperatureof tar from woody biomass is 200 to 500°C) considerable problems with equipmentcontamination (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 aminimum LHV of the gas and are sensitive to dust. In addition, alkali compounds aredetrimental to gas turbines, causing corrosion and deposits at the high working temperaturesof turbines.

The requirement for a higher gas pressure has led to the development of two distinctschemes for integrating the gasification system to the gas turbine; pressurized gasification-hotgas filtering and cleaning and; atmospheric gasification-cold gas filtering and cleaning.

A variety of relatively large-scale biomass gasification technologies are at variousadvanced stages of development. Three gasifier/gas cleanup designs are considered here: (i)atmospheric-pressure air-blown fluidized-bed gasification with wet scrubbing (Waldheim andCarpentieri, 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-pressureindirectly-heated gasification with wet scrubbing (Paisley and Anson, 1997) the HHV is 18.1MJ/kg, % H2 = 29.5, and % CH4 =13.05.

Gumz is the earliest reference found describing the concept of combining a pressurizedgasifier with a gas turbine engine (Gumz, 1950).

3.2. Robust Control

There are many ways to design a control system. Optimal control is a method that has beenaround since the 1950s and has seen a lot of successful applications, especially in theaerospace field. Adaptive control is a way of designing a control system to adjust to changesin the system that is being controlled. Fuzzy and neural control allow a designer to control anonlinear system without having a mathematical model of the system.

In many practical cases, the designer has a model of the system, but the systemparameters are subject to uncertainty. In this case robust control can be used to design acontrol system with a guaranteed level of performance, as long as the system uncertaintiesremain the assumed bounds.

Traditional control design approaches, consider a fixed operating point in the hope thatthe resulting controller is robust enough to stabilize the plant for different operatingconditions. These approaches definitely yield good results if the parameter variations aresmall or the system dynamics is not too sensitive with respect to these parameters(Ackermann et al., 2002). For significant parameter variations these control design methodsreach their performance limits. New design approaches, which already incorporate the plantuncertainty in the design step are then required.

Eigenvalue assignment plays an important role in control system theory because of itsmany applications, such as stabilization and optimization (Lee et al., 2001). Schemes toassign the closed-loop eigenvalues of the controllable linear time-varying system have beenproposed in (Wolovich, 1968; Nguyen, 1987; Valasek and Olgac, 1995 a; Valasek and Olgac,1995 b).

Francisco Jurado176

To stabilize this plant it would suffice to shift the poles only a little into the left complexhalf -plane. In view of a practical realization this is certainly not sufficient since the dampingstill could be arbitrarily small. The notion of Γ-stability is therefore introduced to assuresufficient stability margins. Considering the operating domain it cannot be the aim of thedesign to specify precisely the location of the poles in dependency of the operating point. Thedesign goal is reached if the damping is sufficiently large. This is the case if the roots of theclosed-loop system for all possible operating points have a degree of damping larger then avalue D0. Furthermore, a maximal settling time might be required. This corresponds to amaximal real part of the roots not larger than a certain value σ0. Both conditions can bevisualized graphically in the complex s-plane.

The admissible region for closed-loop eigenvalues is denoted as Γ, a system is called Γ-stable if all its eigenvalues are located in this region and an uncertain system is called robustlyΓ-stable if all eigenvalues for all operating conditions are contained in Γ. The definition of Γ-stability permits arbitrary regions in the complex s-plane and does not underlie anyrestrictions. It also includes the special cases of the left half-plane for Hurwitz stability andthe unit circle for Schur stability.

The region Γ is specified by its boundary ∂Γ. Besides lines, adequate elements todescribe the boundary of a Γ-region are circles, hyperbolas, and ellipses. Lines parallel to theimaginary axis restrict the settling time, circles the bandwidth. Circles can be used to limit thebandwidth, or – in its inverted version – to guarantee a minimal bandwidth.

The field of application of this method is not only robustness analysis but also controllerdesign. In this case the set of stabilizing controller parameters is determined. All controllersfrom this set stabilize the plant, thus, allowing to incorporate further design criteria to selectthe final controller. The task is to determine a controller which robustly Γ-stabilizes thesystem for the entire operating domain.

3.3. Results

A plant consisting of a load is fed from the fuel cell/ turbine hybrid system. The selectedsystem comprises a 250 kW fuel cell and a 30 kW gas microturbine. The plant and the fuelcell/ microturbine system are modeled using MATLAB (Math Works, 2006).

All parameters correspond to a 2-stack equivalent. The fuel cell stacks used in thischapter 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 correspondingto rated power and subjected to a sudden variation in power demand; this is zero-mean, whitenoise, 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 andEnergy 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 stackcurrent proportional to load current and inversely proportional to stack voltage using a powerdemand setpoint. Stack current is measured and used in calculation of fuel flow setpoint.


Fig. 9 presents the eigenvalues. Referring to the results shown in Fig. 10, it can be seenthe Γ-region, which is a hyperbola with a = -1, b = 1.73205, that corresponds to a maximalreal part of σ =-1 (maximal settling time), that guarantees a minimal degree of dampingequals to 0.5.

Figure 9. Eigenvalues.

Figure 10. Γ-region.

Many utilities prefer to use the PI controller for better system dynamic response and inthe 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. Thisfigure displays the set of simultaneously Γ-stabilizing controller parameters for the vertices ofthe 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 Γ-stable the point is marked (Paradise, 2004). The controller is selected from the set with

Francisco Jurado178

maximal distance from the stability boundaries. This guarantees some additional robustnessmargins in the case that the model does not cover the entire plant uncertainty.

Figure 11. Parameter space of the controller.

Fig. 12 depicts the parameter space where the varying parameters considered are %hydrogen and % methane. The controller is robustly Γ-stable since the vertices are Γ-stableand the operating domain is not intersected by stability boundaries.

Figure 12. Parameter space of the operating domain.


Fig. 13 displays the stack current deviation ΔIdc where the simulation time is 200 s, thisfigure demonstrates that the response of the PI controller with varying parameters is betterthan that of PI controller with fixed parameters.

Stack voltage starts at its initial value, corresponding to regulated power, and dropsquickly to a sudden increase in stack current, in turn, controlled by the inverter to maintainthe new power setpoint. Stack voltage varies according to the temperature dependence, whilestack current is continuously adjusted by the inverter control to maintain power.

Figure 13. Load current deviation ΔIdc.

4. Modeling SOFC Plants on the Distribution System UsingIdentification Algorithms

Exhibiting the dynamic influences of Solid Oxide Fuel Cell (SOFC) on the distribution gridrequires the use of a large dynamic model (Jurado, 2003). Since SOFCs will be proliferated, itis necessary to reduce the model order of each SOFC system to enable computationalanalysis.

The computation of linear system models of power systems from time domainsimulations is a topic of considerable practical interest. This interest is motivated by theinsight into the dynamic interactions among power system components that can be obtainedfrom a linear representation. Linear models allow for the application of linear analysistechniques to complement the information obtained from nonlinear time domain simulationsand often allow for a better understanding of the system dynamic characteristics than thatobtained from the inspection of time simulations alone. Although the nonlinear nature of aSOFC must be recognized, in many cases a linearized system representation allows for amore efficient means of analysis.

Several techniques for computing state space matrices and transfer function realizationsof power systems from time domain data have been proposed in recent years. Thesetechniques include the Prony method which is based on fitting a weighted sum of exponential

Francisco Jurado180

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). Inaddition, methods designed to estimate the frequency response of a system have also beenproposed (Trudnowski et al, 1994) The Prony method is perhaps the method for which moreextensive results and applications have been documented. Its application to the analysis andcontrol of electromechanical oscillations has shown the value of deriving linear models fromtime 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 andcontrol design (Ljung and Glad, 1994; Ljung, 1999; Söderström and Stoica, 1989). Thisalgorithm consists of a simple procedure for calculating the transfer function of a linearsystem from samples of its input and output.

Using MATLAB/Simulink (Mathworks, 2006), a dynamic model of a SOFC-penetrateddistribution 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 ARXmodels are described in Subsection 4.3. Subsection 4.4 compares the response of identifiedsystem versus the response of the actual system. In the end, Subsection 4.5 depicts somesimulation results.

4.1. Solid Oxide Fuel Cell

There are several types of fuel cells being developed for a variety of applications and thesehave been extensively discussed in the open literature. Unlike other variants, the SOFC isentirely solid state with no liquid components. Operation at elevated temperature is needed toachieve the necessary level of conductivity in the cell’s solid electrolyte for it to operateefficiently. With an outlet temperature in the range of 900–1000°C, the efficiency of the cellalone is about 50 percent.

Typically the fuel cell system consists of SOFC generator modules in a parallel flowarrangement, with the number of standard modules being determined by the plant powerrequirement. The SOFC generator module embodies a number of tubular cells, which arecombined to form cell bundle rows, several of which are arranged side by side to make up thecomplete assembly.

Most likely, fuel cell will be a dominant distributed energy resource. The SOFCs aredynamic devices and when connected to the distribution system they will affect its dynamicbehavior. Hence, researchers have developed dynamic models for these components(Massardo and Lubelli, 2000; Padullés 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 timeto change the chemical reaction parameters after a change in the flow of reactants. Thisdynamic response function is modeled as a first-order transfer function with a 5-s timeconstant.

The electrical response time in the fuel cells is generally fast and mainly associated withthe speed at which the chemical reaction is capable of restoring the charge that has been


drained by the load. This dynamic response function is also modeled as a first-order transferfunction but with a 0.8-s time constant.

With aid of the inverter, the fuel cell system can supply not only real power but alsoreactive power. Usually, power factor can be in the range of 0.8-1.0. The SOFC systemdynamic model is given in Fig. 14.

Figure 14. SOFC system dynamic model.

The fuel utilization (U) is the ratio between the fuel flow that reacts and the input fuelflow, such as the follows:

inH

rH

inH

oH

inH

qq

qqq

U2

2

2

22 =−

= (22)

Typically, an 80-90% fuel utilization is used.Every individual gas will be considered separately, and the perfect gas equation will be

applied to it. Hydrogen will be considered as an example.

RTnVp HanH 22= (23)

It is possible to isolate the pressure and to take the time derivative of the previousexpression, obtaining:

Francisco Jurado182

2

2H

an

H qVRT

dtdp

= (24)

There are three relevant contributions to the hydrogen molar flow: the input flow, theflow that takes part in the reaction and the output flow, thus:

( )rH

oH

inH

an

H qqqVRT

dtdp

222

2 −−= (25)

According to the basic electrochemical relationships, the molar flow of hydrogen thatreacts can be calculated as:

rfcr

rH IK

FINq 2

20

2== (26)

Returning to the calculation of the hydrogen partial pressure, it is possible to write:

( )rfcr

oH

inH

an

H IKqqVRT

dtdp

222

2 −−= (27)

If it could be considered that the molar flow of any gas through the valve is proportionalto its partial pressure inside the channel, according to the expressions:

2

22

2H

H

an

H

H KMK

pq

== (28)

Replacing the output flow by (28), taking the Laplace transform of both sides andisolating the hydrogen partial pressure, yields the following expression:

( )rfcr

inH

H

HH IKq

sK/

p 211

2

2

2

2−

+=

τ (29)

where ( )RTK/V HanH 22=τ .

A similar operation can be made for all the reactants and products. Applying Nernst’sequation and Ohm’s law (to consider ohmic losses), the stack output voltage is represented bythe following expression:

rfc

OH

OHc rI

ppp

FRTENV −

⎟⎟⎠

⎞

⎜⎜⎝

⎛

⎥⎥⎦

⎤

⎢⎢⎣

⎡+=

2

22

5.0

00 ln2

(30)


4.2. Utility-Connected Inverter Control

Control of the flux vector has been shown to have good dynamic and steady-stateperformance (Chandorkar, 2001). It also provides a convenient means to define the powerangle since the inverter voltage vector switches position in the d-q plane, whereas there is nodiscontinuity in the inverter flux vector.

For a six-pulse Voltage Source Inverter (VSI), the inverter output voltage space vectorcan take any of seven positions in the plane specified by the d-q coordinates. The timeintegral of the inverter output voltage space vector is called the “inverter flux vector” forshort. The d and q axis components of the inverter flux vector ψv are defined as

ψ ∫∞−

=t

ddv dv τ ψ ∫∞−

=t

qdq dv τ (31)

The magnitude and the angle of ψv with respect to the q axis are determined as

22dvqvv ψψψ +=

⎟⎟⎠

⎞⎜⎜⎝

⎛ −= −

qv

dvv ψ

ψδ 1tan (32)

The d and q axis components of the ac system voltage flux vector ψe , its magnitude, andangle are defined in a similar manner. The angle between ψv and ψe is defined as

evp δδδ −= (33)

It is useful to develop the power transfer relationships in terms of the flux vectors. Thebasic real power transfer relationship for the control system of Fig. 15 in the d-q referenceframe is

( )ddqq ieieP +=23

(34)

In (34), eq and ed are the q- and d-axis components, respectively, of the ac system voltagevector E. In addition, iq and id are the components of the current vector I. When iq and id areexpressed in terms of the fluxes, taking into account the spatial relationships between the twoflux vectors and assuming the ac system voltage to be sinusoidal, (34) can be expressed as

pveTL

P δψωψ sin2

3= (35)

In this expression, ψe and ψv, are the magnitudes of the ac system and the inverter fluxvectors, respectively, and δp is the spatial angle between the two flux vectors. ω is the

Francisco Jurado184

frequency of rotation of the two flux vectors. The expression for reactive power transfer canbe derived in a similar manner. This is

[ ]2cos23

epveTL

Q ψδψψω−= (36)

Figure 15. Control system for the inverter.

Equations (35) and (36) indicate that P can be controlled by controlling δp, which can bedefined as the power angle, and Q can be controlled by controlling ψv.

The two variables that are controlled directly by the inverter are ψv and δp. The vector ψv

is controlled to have a specified magnitude and a specified position relative to the ac systemflux vector ψe.

The errors between actual and desired amounts activate the remainder of the firingscheme only if they exceed a threshold value. If the error is larger than the hysteresis band(whose widths are Δδp and Δψv ) 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 transientvariations and fuel flow input control for slow variations.

2. Voltage control: done by adjusting the set point E* of the inverter, which effects themagnitude of the converter output voltage.


Using the fast response of the inverter, the maximum power can be used to alleviatetransients when fault occurs. However, the inverters for fuel cells have the followinglimitations:

•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).

4.3. Identification Algorithms

4.3.1. ARX Models

The most used model structure is the simple linear difference equation

( ) ( ) ( ) ( )( ) ( )tenbnktub

...nktubnatya...tyaty

nb

na

++−−++−=−++−+

11 11

(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 nb–1 is the number of zeros, while nk is the pure time-delay (thedead-time) in the system. For a system under sampled-data control, typically nk is equal to 1if there is no dead-time.

For multi-input systems nb and nk are row vectors, where the i-th element gives theorder/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-handside 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.

4.3.2. ARMAX, Output-Error and Box-Jenkins Models

There are several elaborations of the basic ARX model, where different disturbance modelsare introduced. These include well known model types, such as ARMAX, Output-Error, andBox-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 ycan be written:

( ) ( ) ( ) ( )[ ] ( ) ( ) ( )[ ] ( )∑=

+−=nu

iiiii teqD/qCnktuqFqBtyqA

1

(38)

Francisco Jurado186

Here ui denotes input #i, and A, Bi , C, D, and Fi , are polynomials in the shift operator (zor 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 thesepolynomials (i.e., the number of poles and zeros of the dynamic model from u to y, as well asof the disturbance model from e to y).

Most often the choices are confined to one of the following special cases:

ARX: ( ) ( ) ( ) ( ) ( )tenktuqBtyqA +−= (39)

ARMAX: ( ) ( ) ( ) ( ) ( ) ( )teqCnktuqBtyqA +−= (40)

Output-Error: ( ) ( ) ( )[ ] ( ) ( )tenktuqF/qBty +−= (41)

Box-Jenkins: ( ) ( ) ( )[ ] ( ) ( ) ( )[ ] ( )teqD/qCnktuqF/qBty +−= (42)

Note that A(q) corresponds to poles that are common between the dynamic model and thedisturbance model. Likewise Fi(q) determines the poles that are unique for the dynamics frominput # 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 analystcan adjust, one requirement an identified system must meet is that its response to a giveninput should match the response of the actual system. This practical criterion was used as aguideline to adjust the order of the identified systems.

4.4. Performance

SOFC models used in the distribution system analysis were constructed as shown in thefollowing :

• There is one 4.16 kV/ 480 V transformer.• All SOFCs were connected at the end of their respective feeders at 480 V.

In this Section SOFC modeled is denoted as “the actual system”. Once an identifiedsystem was obtained, its time domain response and transfer function were compared againstthe corresponding quantities of the actual system. To accomplish this, the actual system waslinearized around an operating point.

The results presented here correspond to a 0.02 p.u. by 0.1 s probing pulse into the realpower block of SOFC in Fig.14. The sampling time was 0.01 s and 600 points were used toperform 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 timeresponse of identified system versus the response of the actual system. The identified systemwas obtained using Box-Jenkins algorithm, and is of 4th order. This method estimatesparameters of the Box-Jenkins model structure using a prediction error method. The order ofthe identified system is the minimum order required to obtain a good time domain match.


Figure 16. Model output comparison.

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 thefrequency range.

Figure 17. Transfer function magnitude and phase comparison. Actual system and identified system.

Francisco Jurado188

4.5. Simulation Results

The IEEE test feeders (Kersting, 2001) are used as the test system to investigate the dynamiccharacteristics of the distribution system with fuel cells. Figs. 18, 21 and 23 show the testsystems with the fuel cells. In this chapter, all loads are balanced, and characterized byconstant 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 thegeneration equally. The IEEE 13 node test feeder is very small, short and relatively highlyloaded for a 4.16 kV feeder. Penetration means the proportion of the distribution feeder loadbeing supplied by SOFCs associated with the distribution feeder (Donnelly et al., 1996). Inthis model, an initial load of Pi is assumed and the penetration is thus,

iPP

PnPenetratio+

= (43)

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 ofthe 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 connectionof the distribution system at 0.5 s, while voltage was held constant


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 feeder’s nominal voltage is 24.9kV. It is characterized by very long and lightly loaded.

Francisco Jurado190


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 providevoltage drop problems that must be solved with the application of voltage regulators andshunt capacitors.



Figure 24. Fuel cell response to a response for a voltage step transient at node 46. IEEE 123 nodefeeder.

Francisco Jurado192

Fig. 25 shows the response to a 10 % step of system voltage.

Figure 25. Fuel cell response to a voltage step transient at node 634 for the IEEE 13 node feeder.

5. Effect of a SOFC Plant on Distribution System Stability

Distributed generation (DG) is electricity generation sited close to the load it serves, typicallyin the same building or complex. The DG embraces a palette of technologies in varying stagesof availability, from entrenched to pilot. It is sometimes called a “disruptive” technologybecause of its potential to upset the utility industry’s apple cart.

Most likely, fuel cells will be a dominant DGs (Moore, 1997; Scientific American 1999).These DGs are dynamic devices and when connected to the distribution system they willaffect its dynamic behavior. Hence, several researchers are working to develop dynamicmodels for these components (Bessette 1994; Haynes, 1999; Padullés et al., 2000; Massardoand Lubelli, 2000; Campanari, 2001; Lasseter, 2001; Rao and Samuelsen, 2002; Jurado,2003).

This Section develops a generic dynamic model for a grid-connected SOFC plant. Themodel is defined by a small number of parameters and is suitable for planning studies.

The steady state power generation characteristics of the plant are derived and analyzed.Understanding the transient behavior of SOFC is important for control of stationary utilitygenerators during power system faults, surges and switchings.

Voltage regulation is one of the main problems in the distribution systems, especially atthe much far-end load and in the rural areas. Voltage regulation and maintaining the voltagelevel are well known problems in the radial distribution network. Several techniques havebeen applied by implementing many devices in the distribution network to solve theseproblems. The most common devices and techniques used are transformer equipped by load


tap changer, supplementary line regulators installed on distribution feeders, shunt capacitorswitched on distribution feeders (Barker and De Mello, 2000) and shifting transformerstowards the load center (Ijumba et al. 1999). A multiple line drop compensation voltageregulation method that determines tap positions of under-load tap changer transformers isproposed in (Joon-Ho Choi, and Jae-Chul Kim, 2001) to maintain the customers’ voltageswithin the permissible limits.

The model derived is based on the main equations. It is developed in the Laplace domainand 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 transientstability. Some basic concepts of voltage stability are introduced in Subsection 5.3. andSubsection 4.4 depicts some simulation results.

5.1. Transient Stability

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 insynchronism, or in step each other. Conversely, instability denotes a condition involving lossof synchronism, or falling out of step (Kimbark, 1995).

Hence, for a simplified intuitive description of transient stability, a power system may beregarded as a set of synchronous machines and of loads interconnected through thetransmission network. Under normal operating condition, all the system machines run at thesynchronous speed. If a large disturbance occurs the machines start swinging with respect toeach other, their motion being governed by differential equations. Depending upon the powersystem modeling, the number of such first-order differential equations is lower bounded bytwice 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 alsobe 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 knockDG off the system, requiring it to be resynchronized with the grid. There are various means toenhance 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 stabilityof a power system under large and sudden disturbances. Fast valving schemes involve rapidclosing and opening of thermal turbine valves in a prescribed manner to reduce the generatoracceleration following a severe fault (Patel et al., 2001).

During steady state operation of a power system, there is equilibrium between themechanical input power of each unit and the sum of losses and electrical power output of thatunit. The problem arises when there is a sudden change in the electrical power output due to asevere and sudden disturbance. The severity is measured by drop of this power to a very lowor to zero value and a consequential sudden acceleration of the machines governed by theswing equation:

Francisco Jurado194

ePmPdt

dH−=

2

22

0

δ

ω (44)

Where,δ = rotor angle, in electrical radian Pm = mechanical power, in p.u.Pe = electrical power output, in p.u.H = inertia constant, in MW-s/MVAω0 = nominal speed, in electrical radian/s

From Equation (44), it is apparent that the decrease in the mechanical power has the sameimpact on the rotor angle swings as that of increase in the electrical power output. Fastvalving has a function of reducing the mechanical power input to the turbine and so thegenerated power.

A change in load demand (ΔPL, ΔQL) causes corresponding change in both bus voltagemagnitude and the power output of the generator. The voltage magnitude value is calculatedin the power flow analysis. The change in the generator’s active power output results insystem frequency variations through the swing equations.

5.2. Voltage Stability

Voltage instability spans a range in time from a fraction of second to tens of minutes. Timeresponse charts have been used to describe dynamic phenomena (Taylor, 1994; Cañizares,2002). Many electric system components and controls play a role in voltage stability. Onlysome, however, will significantly participate in a particular incident or scenario. The systemcharacteristics and the disturbance will determine which phenomena are important.

Mid-term voltage stability.The time frame is several minutes, typically two-threeminutes. Operator intervention is often not possible. This scenario involves high loads, and asudden large disturbance. The system is transiently stable because of the voltage sensitivity ofloads. 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 andis driven by a very large load buildup, or a large rapid power transfer increase. The loadbuildup, measured in MW/min, may be quite rapid.

Several papers have shown the direct relation between saddle-node bifurcations andvoltage collapse problems, e.g., (Cañizares and Alvarado, 1993; Cañizares, 1995). Saddle-node bifurcations, also known as turning points, are generic codimension one localbifurcations of nonlinear dynamical systems of the form:

( )λ,xfdtdx

= (45)


where x ∈ Rn are the state variables, λ ∈ R is a particular scalar parameter that drives thesystem 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, λ0), i.e., f (x0, λ0) = 0,if the corresponding system Jacobian Dxf|0=Dxf (x0, λ0) has a unique zero eigenvalue, andsome particular transversality conditions hold at that equilibrium point.

The state variables x and nonlinear function f (.) are typically defined in terms of thequasi-steady state phasor models used in transient stability studies. Thus, generator angularspeed deviations and phasor bus voltages, magnitudes and angles, are usually an importantpart of vector x. The parameter λ 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 theconstant power loads by a factor λ, until the singular point of the linearization of the powerflow equations is reached. The loads can be defined as,

( )nLL PPP λ+= 10 (46)

where P0 is the active power base load, PnL represents the load distribution factor and PL is theactive load at a bus L for the current operating point.

5.3. Instantaneous Power Issue

The introduction of a micro-source within an existing power system assumes that there maybe a cluster of small generators located in electrical proximity, interconnected through on-sitefeeders 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 kindof internal form of energy storage. These inertia-less systems are not well suited to handlestep changes in the requested output power. It must be remembered that the current powersystems have storage in the mechanical energy of the inertia of the generators. When a newload is applied the initial energy balance is satisfied by the inertia of the system, this results ina slight reduction in system frequency.

The step sized power demand in the distribution system should be instantaneouslymatched by an identical supply of power from the micro source power source.

Load coming on-line with grid connection. SOFCs have a problem in instantaneouspower tracking. SOFC shows a poor load tracking, as a load is applied and the micro-sourceramps up to pick up the whole quota of extra power request. The missing transient power thatthe micro-source is not fast enough to provide is taken from the connection with the grid thatsupplies only a part of the local requested power during steady state. Voltage is maintained bythe power grid.

Load coming on-line without grid connection. The requested power from the load comingon-line is a step function, while the inertia-less micro source always takes a finite amount oftime to ramp up to the newly requested value. Since the unit cannot change its output powerinstantaneously, the power is balanced by voltage reduction. As the power injected from themicro source increases to supply the needed load the voltage is restored.

Francisco Jurado196

There is a missing quota of energy from the difference of the power that the load isrequesting and the power that the unit is able to provide. If the connection to the main grid isactive, then the missing quota can be provided by the utility. It will be seen from the gridterminals as a temporary, pulse-like power request.

5.4. Results

The analysis is tested on a system consists of 13 buses and is representative of a medium-sized industrial plant. The system is shown in Fig.26 and described by the data in Table 2.

The SOFC plant is connected at bus 4 and consists of 20 SOFCs, as depicted in Fig. 26.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. Each SOFC has the sameparameters, yielding a rated power of 2 MW.

Suppose at a certain time, the total load in this distribution system is as shown in Table 2.The SOFC plant mainly provide some peak shaving capability and ancillary services for thefeeder.

G1 1

2

3

SOFC plant 4

5

6 9 10

7 8

13

12 11

Figure 26. Test system.

It is assumed bilateral contracts to perform load-following among SOFCs at a givenfeeder is not allowed. That is, the ancillary service must be coordinated at the substation ortransmission level.

It is desirable to minimize the number of dynamic equations to be solved in thesimulations. Here, all SOFCs in the plant are regarded as coherent devices, which have thesame dynamic response and share the generation equally.


Table 2. Generation, load, and bus voltage data. Base case.

Bus Vmag

(%)Pgen

kWQgen

kvarPload

kWQload

kvar1 100.00 7450 540 - -2 99.89 - - - -3 99.59 - - 2240 20004 99.64 2000 1910 - -5 101.88 - - 600 5306 99.56 - - - -7 101.94 - - 1150 2908 104.06 1310 11309 99.56 - - - -

10 99.52 - - - -11 99.31 - - 370 33012 103.30 - - 2800 250013 101.64 - - 810 800

5.4.1. Step Change Response

A power measurement is fed in the fuel flow regulator, whose output determines the positionof a fuel valve. A step change in the input reactant element will not be noticed as a suddenincrease in charges. Typically, there is a sudden but contained rise in output power that takesplace in 1-3 s, while the newly desired output power level is reached limited until the SOFCreaches its thermal equilibrium, which may tens of seconds. This characteristic makes theSOFC a poor candidate for isolated systems, where loads require instantaneous power.

Figure 27. SOFC plant response to a step change in the power demand.

Francisco Jurado198

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 isincreased, 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 isdue to the slow chemical response time in the fuel processor. The total response time of Pfrom 0.6 to 0.9 p.u. is about 30 s. Fig. 27 shows the response of the SOFC to a step change inthe fuel valve.

Normally the fuel cell supports base load thus, the fuel flow regulator in Fig. 14 is set toprovide a constant input flow-rate for the cell fuel. The response of the fuel flow regulator isslow. The plant, however, can respond continuously to small instantaneous power adjustmentby the control of the angle δp

* . Any steady state power change must be followed by acorresponding change in the fuel flow-rate. If the fuel cell power increases while the fuel flowrate is maintained constant, the steady state cell utilization will increase and the cell voltagewill drop.

5.4.2. Transient Stability

The penetration level of this SOFC plant is set at 20 %. A faults occurs that disconnects theutility supply from the distribution system. The fault clearing is after 0.4 s. The comparisonbetween Figs. 28 and 29 shows the advantage of a fast power controller. Since the SOFCinverter has the capability to adjust its angle δp

* quickly, the inverter maintains constantpower output under fast transient disturbances. In addition, SOFC inverter has powermodulation capabilities that enhance transient stability of the system.

Figure 28. Fast transient disturbances without inverter.


Figure 29. Fast transient disturbances with inverter.

5.4.3. Voltage Stability

This Section discusses the nature of the response of system to a linear increase in loaddemand. 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.

Francisco Jurado200

The power load PL increases by a factor until the voltage collapse is reached. The analysisis tested on the distribution system. Continuation method is used to identify the location ofbifurcation point (λ = 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 rampfunction equal to the linear increase in load demand.

The plant can contribute to the utility reactive power supply by continuously adjustingthe set points ψv* 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 magnitudeof 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 todrive a gas turbine. A dynamic model for this MCFC-microturbine system has beenelaborated.

A MCFC stack interacts with other system components and causes an interdependencybetween them. To investigate transient behavior, the plant is assumed to be at steady statecorresponding to rated power and subjected to a sudden variation in power demand. Thischapter has developed the control system with an adaptive minimum variance controller andbased on the simulation study, the resulting controller is robust enough to stabilize the systemfor different operating conditions.


A simple, biogas-fed, hybrid MCFC power-generation system has been developed. Thecharacteristics of gas from biomass gasification may vary significantly. Energy, measured interms of HHV, and the composition of gas depend on biomass gasification technology.

Traditional control design approaches, consider a fixed operating point in the hope thatthe resulting controller is robust enough to stabilize the system for different operatingconditions. On the other hand, robust control incorporates the varying parameters of themodel.

This chapter establishes the possibilities for biomass-based fuel cell and gas microturbineintegration. Referring to the results obtained, it can be seen that the derived model is quitevalid for robust control studies.

In this work it was felt to be premature, based on the current state of technologyreadiness, to identify an optimum hybrid plant in terms of parameter selection, efficiency, andmajor feature selection. A better understanding of these will be realized with the tests of acoupled system, and this will facilitate more accurate system modeling in the future.

The capability to calculate low-order equivalent linear systems from time domainsimulations of SOFC models using the ARX algorithm has been established. After the SOFCmodel was created, it was reduced to transfer functions using the ARX algorithm; thus, thetransfer function (reduced-order model) exhibited the same dynamic response as the originalSOFC model.

A significant reduction in the model order was achieved. The time domain response ofthe identified system matches the response of the actual system.

Therefore, each SOFC reduced-order model influences the grid in the same manner asSOFC and loads would, modulating real and reactive power in response to voltage andfrequency changes on the grid.

The introduction of SOFC within the context of an already existing grid can providecontrol of local bus voltage, control of base power flow thus reducing the power demand fromthe main grid feeder, and ultimately frequency control associated with load sharing withinunits located in the micro-grid.

The SOFC is capable of providing effective load-following service in the distributionsystem. However, the results also show that the SOFC system is not an uninterruptible powersupply and does not protect the load from voltage stability while in grid-connect mode. WhenSOFC 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

A(z-1), B(z-1) polynomials of the plantBr

-(z-1), B-(z-1) polynomials of the performance indexP(z-1), Q(z-1), V(z-1) filter polynomialsR(z-1), S(z-1), T(z-1) polynomials of the difference equationSc (z-1), Rc(z-1) polynomials of the controllere(i) disturbance of the planth constant term

Francisco Jurado202

J performance indexK prediction horizon in the performance indexP_init covariance matrixq weighting coefficientr1, r2 varying parameters of the polynomial Rc(z-1)s1, s2 varying parameters of the polynomial Sc(z-1)sd standard deviationu(i) plant control signaly(i) plant output signalΔIdc stack current deviation (A)ΔPL load power variation (W) forgetting factorσ2 constant varianceω(i) setpoint of the plant output signal

Distribution network stability

Dxf system JacobianH inertia constant (s)Pe electrical power output (p.u.)Pi initial load (p.u.)PL active load at a bus (p.u.)Pm mechanical power (p.u.)PnL load distribution factor (p.u.)P0 active power base load (p.u.)x state variablesΔPL, ΔQL changes in load demand (p.u.)δ rotor angle (electrical rad) λ changes in the system loadingω0 nominal speed (electrical rad/s)

Fuel cell

E0 ideal standard potential (V)F Faraday’s constant (C/ kmol)

rfcI fuel cell current (A)

i cell load current (A/cm2)Kan anode valve constant

2HK valve molar constant for hydrogen

OHK2

valve molar constant for water

2OK valve molar constant for oxygen

Kr constant (kmol/s A)

OHM2

molecular mass of hydrogen


Na reactant flow to the anode (kmol/s)N0 number of cells in series in the stack

2Hn number of hydrogen moles in the anode channel (kmol/s)

ni number of electrons transferred during the electrochemical conversionP real power (W)P* set point for the real power (W)pi partial pressure (Pa)Pel,dc electrical power produced by the fuel cell stacks (W)

inHq

2input fuel flow (mol/s)

oHq

2output fuel flow (mol/s)

rHq

2fuel flow that reacts (mol/s)

R universal gas constant (8.31 J/(mol K))r ohmic loss (Ω)rH-O ratio of hydrogen to oxygenT absolute temperature (K)Te electrical response timeTf fuel processor response timeU fuel utilizationUfuel average fuel utilizationVan volume of the anode (m3)Vc cell voltage under load current (V)Vo equilibrium potential (V)yi mole fraction of fuel species i in the anode gasz cell ohmic impedance (Ωcm2)ηact activation polarization (V)ηconc concentration polarization (V)

2Hτ response time for hydrogen flow (s)

OH2τ response time for water flow (s)

2Oτ response time for oxygen flow (s)

Gas turbine

a, b, c valve parameterscpa 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 positionFd fuel demand signalHHV higher heating value (MJ/kg)K I PID parameterKP PID parameterkf fuel system gain constant

Francisco Jurado204

kHHV factor which depends on HHVkLHV factor which depends on LHVLHV lower heating value (MJ/kg)Pc compressor power consumption (W)pcin air pressure at compressor inlet (Pa)pcout air pressure at compressor outlet (Pa)Pmec mechanical power delivered by turbine(W)PT total mechanical power delivered by turbine (W)pTin pressure of combustion gases at turbine inlet (Pa)pTout pressure of combustion gases at turbine outlet (Pa)s Laplace transformt time (s)T mechanical torque delivered by turbine (Nm)Tcout outlet air temperature (K)Tis temperature of injected steam (K)TTin turbine inlet gas temperature (K)U(t) control signalwa air mass flow into the compressor (kg/s)wf fuel mass flow (kg/s)wg turbine gas mass flow (kg/s)wis injection steam mass flow (kg/s)Δh25 specific enthalpy of reaction at reference temperature of 25ºC (J/kg)ΔhIC isentropic enthalpy change for a compression from pcin to pcout (J/kg)ΔhIT isentropic enthalpy change for a gas expansion from pTin to pTout (J/kg)Δω rotation speed deviation of the turbine (rad/s)ηc overall compressor efficiencyηT overall turbine efficiencyηtrans transmission efficiency from turbine to compressorτf fuel system time constant (s)ω rotation speed of the turbine (rad/s)

Inverter

E load bus voltage (V)E* set point for the load bus voltage (V)LT inductance (H)Q reactive power (var)Q* set point for the reactive power (var)XT reactance (Ω)V inverter output voltage space vector (V)δp angle between ψv and ψe (rad)δp

* angle reference (rad)ψe flux vector associated with E (Vs)ψv flux vector associated with V (Vs)ψv* flux-vector reference (Vs)


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 powerplant with CO2 recovery and a sequestering system. Energy Vol. 22, No. (2-3), pp.249-255.

Aström K.J., Wittenmark B., 1990. Computer-controlled systems: theory and design. 2nd ed.Englewood Cliffs: Prentice Hall, cop.

Aström 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 powersystem 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 andmicroturbine systems. ASME paper GT-65.

Campanari S., 2001. Thermodynamic model and parametric analysis of a tubular SOFCmodule, Journal of Power Sources, Vol. 92, No. (1-2), pp. 26-34.

Cañizares C.A., 1995. On bifurcations, voltage collapse and load modelling, IEEE Trans.Power Systems, Vol. 10, No.1, pp. 512-522.

Cañizares C.A., editor, 2002. Voltage Stability Assessment: Concepts, Practices and Tools,Final report, IEEE/PES Power Systems Stability Subcommittee Special Publication.

Cañizares C.A., Alvarado F.L., 1993. Point of collapse and continuation methods for largeac/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 IndustryApplications, Vol. 37, No. 3, pp. 880–887.

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 carbonatefuel cells. In: Proceedings of the Fourth International Symposium on Carbonate FuelCell Technology, Montreal, Quebec.

Donnelly M.K., Dagle J.E., Trudnowski D.J., Rogers G.J., 1996. Impacts of the distributedutility on transmission system stability, IEEE Trans. on Power Systems, Vol. 11, No. 2,pp. 741–746.

Farooque M., 1991. Development of internal reforming carbonate fuel cell technology, finalreport, 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 digitalcomputer simulation method for predicting gas turbine dynamic behaviour. Institution ofMechanical Engineers Proc. Vol. 186, No. 27.

Fuel Cell Engineering Corporation, Santa Clara, 1997. 2 MW Fuel cell demonstration powerplant test, report TR-108 252. Electric Power Research Institute.

Francisco Jurado206

Ghezel-Ayagh H., Leo A.J., Maru H.C., 2002. Direct Fuelcell®/turbine hybrid system forultra 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-shaftcombustion 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 powerplants, 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 IntegrationInto 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 PowerPlants and Power Systems, Munich, Germany, Pergamon Press, pp. 43-48.

Ijumba N.M., Jimoh A.A., Nkabinde M., 1999. Influence of distribution generation ondistribution network performance, Africon IEEE, Vol. 2, pp. 961–964.

Joon-Ho Choi, Jae-Chul Kim, 2001. Advanced voltage regulation method of powerdistribution systems interconnected with dispersed storage and generation systems, IEEETrans. on Power Delivery Vol. 16, No. 2, pp. 329–334.

Jurado F., Ortega M., Cano A., Carpio J., 2002. Neuro-fuzzy controller for gas turbine inbiomass-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 GasTurbine News, No.2.

Leal E.M., Silveira J.L., 2002. Study of fuel cell cogeneration systems applied to a dairyindustry, Journal of Power Sources Vol. 106, No. (1-2), pp. 102-108.


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 IndustrialElectronics, Vol. 3, pp. 1696-1701.

Leeper J.D., 1999. The hybrid cycle: integration of a fuel cell with a gas turbine. ASME PaperGT-430.

Liese E.A., et al., 1999. Technical development issues and dynamic modeling of gas turbineand fuel cell hybrid systems. ASME paper GT-360.

Lukas M,D,, Lee K,Y,, Ghezel-Ayagh H., 1999. Development of a stack simulation model forcontrol study on direct reforming molten carbonate fuel cell power plant. IEEE Trans.Energy Conversion Vol. 14, No. 4, pp. 1651–1657.

Lukas M.D., Lee K.Y., Ghezel-Ayagh H., 2000. Operation and control of direct reformingfuel cell power plant. in Proceedings of the IEEE Power Engineering Society WinterMeeting.

Lukas M,D,, Lee K,Y,, Ghezel-Ayagh H., 2001. An explicit dynamic model for directreforming 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 SummerMeeting, Vol. 2, pp. 761-766.

Ljung L., Glad T., 1994. Modeling of Dynamic Systems, Englewood Cliffs, N.J., PrenticeHall.

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 turbinecombined 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, TheMathworks Inc., Natick, MA.

Matsumoto S., Sasaki A., Urushibata H., Tanaka T., 1990. Performance model of moltencarbonate 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. 76–79.

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.Padullés 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.

Francisco Jurado208

Phillips C.L., Nagle. H.T., 1995. Digital control system: analysis and design 3rd ed. NewJersey: Prentice Hall, cop.

Rao A.D., Samuelsen G.S., 2002. Analysis strategies for tubular solid oxide fuel cell basedhybrid 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 ReferenceLibrary, AGTR 880.

Salo K., Horvath A., Patel J., 1998. Pressurized gasification of biomass, ASME paperGT-349.

Sanchez-Gasca J.J., Chow J.H., 1997. Computation of power system low-order models fromtime 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 identificationmethods for the analysis of electromechanical oscillations, IEEE Trans. Power Systems,Vol. 14, No. 3, pp. 995–1002.

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, NewJersey.

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.Söderström 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/advancedturbine power cycles. In: Fuel Cells Conference.

Stephenson D., Ritchey I., 1997. Parametric study of fuel cell and gas turbine combined cycleperformance. ASME paper GT-340.

Stoica P., Soderstrom T., Friedlander B., 1985. Optimal instrumental variable estimates of theAR-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.


Valasek M., Olgac N., 1995 a. Efficient eigenvalue assignments for general linear mimosystems. Automatica, Vol. 31 b, No. 11, pp. 1605-1617.

Valasek M., Olgac N., 1995 b. Efficient pole placement technique for linear time-variant sisosystems. 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 paperGT-419.

Waldheim L., Carpentieri E., 1998. Update on the progress of the Brazilian wood BIG-GTdemonstration project, ASME paper GT-472.

Wolovich W.A., 1968, On the stabilization of controllable systems. IEEE Trans. onAutomatic Control, Vol. 13, No. 5, pp. 569-572.

Working Group on Prime Mover and Energy Supply Models, 1994. Dynamic models forcombined 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-followingperformance of microturbines and fuel cells, Electric Power Systems Research, Vol. 62,No. 1, pp. 1-11.

In: Fuel Cell Research TrendsEditor: L.O. Vasquez, pp. 211-245

ISBN 1-60021-669-2© 2007 Nova Science Publishers, Inc.

Chapter 5

M ICROFABRICATION TECHNIQUES : USEFUL

TOOLS FOR M INIATURIZING FUEL CELLS

Tristan Pichonat∗

Microsystems Group, IEMN / CNRS UMR 8520, Cité ScientifiqueAv. Poincaré, BP 60069

F-59652 VILLENEUVE D’ASCQ Cedex, France

Abstract

This chapter will introduce the reader to the reasons for miniaturizing fuel cellsand to the specifications required by this miniaturization. It will then show what kindsof fuel cells can fit to these specifications and which fuels can be employed to supplythem. The techniques presently used for the realization of miniature fuel cells willbe described, underlining particularly the growing part of the microfabrication tech-niques inherited from microelectronics. It will present an overview on the applicationsof 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 poroussilicon membrane as the proton-exchange membrane instead of a common ionomersuch as Nafion®.

1. Introduction

Numerous electronic devices such as computers, cellular phones, camcorders, have becomeportable with the miniaturization of electronic components. Such equipments require ap-propriate power supply, which have to be more and more effective especially in terms ofpower density and lifetime, in order to be able to provide for the increasing number offunctionalities of these portable applications (like Internet on cellular phones, for example).

In this strong growth market, fuel cells represent promising power sources for theseapplications. Indeed even if they are generally associated with much larger applicationssuch as vehicles and power plant, they can be of any size, given their modularity. It enables

∗E-mail address: [email protected]. Tél. +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 competingbatteries,after [2,4].

to conceive very small size fuel cells, involving only a few thin cells with a small effectivearea.

30 years ago, only the Nickel Cadmium (NiCd) technology was available for the supplyof 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 1980’s, 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 effortbatteries remain heavy, bulky, of limited charge and above all polluting. Actually, the majorproblem for batteries is that their energy is stored in heavy and voluminous sources such asmetal oxides and graphitic materials [1–3].

Compared with batteries fuel cells can produce electrical energy from much lighter ma-terials such as gases, alcohols, and hydrides by catalytic electrochemical processes. Theytheoretically 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 topresent 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 andthe refilling is immediate by changing the fuel tank. This technology is also theoreticallynon-polluting. As many arguments that presently put research in effervescence.

Yet miniaturizing fuel cells is still a challenge, with many issues to consider such asthe fabrication techniques, the cost and the customer safety [5]. Actually fuel cells aretwo-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 whichhave to work as one. If the ability to store and manage efficiently the fuel willmostly influence the complete volume and the autonomy of the fuel cell, the capability totransform efficiently the specific energy of the fuels into electrical energy will depend onthe effectiveness of catalytic reactions on the electrodes of the fuel cell. The cost is thenbound, among other factors, to the catalyst itself, mostly platinum, even if in very smallamounts. Moreover using fuel cells in portable applications implies methods to provide ahigh level of safety. This means, given that fuel cells are not generally sealed since theyrequire oxygen from air, that the fuel has to be isolated to avoid any risk of accidental directoxidation with air or user contact if toxic [2]. Other issues such as the management of theheat 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 globalconception 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 thefirst solution seems to be the easiest, it involves the miniaturization of all the auxiliarycomponents and the difficulty to adapt "macrofabrication" techniques to miniature sizes. Soresearch 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 standardmicrofabrication techniques mainly used in microelectronics and more especially the fabri-cation of micro- and nano-electro-mechanical systems (MEMS/NEMS). Actually more andmore papers show the interest in developing MEMS-based fuel cells, either directly withsilicon substrates (fig. 2), or adapting the methods to other substrates such as metals orpolymers. These techniques enable notably mass fabrication at low cost (very large numberof devices on a very small area) and then could lead to the reduction of the global cost ofthe 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 electronicsand 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 understandablethe review on microfabricated fuel cells. Finally the complete fabrication process of a mi-


cromachined fuel cell based on a silane-grafted porous silicon membrane will be describedasan example.

2. Miniaturization

2.1. Microelectronics and MEMS

Historically microfabrication techniques have first been developed to meet the requirementsof microelectronics, but they also have allowed the emergence and the development of a newresearch field where mechanical elementscan be manufactured and actuated with electricalsignals 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 MEMSin the late 1980’s [14]. MEMS is not the only term used to describe this field which is alsocalled as micromachines, for instance in Japan, or more broadly referenced as microsystemtechnology (MST) in Europe.

These submillimeter devices are machined using specific techniques globally calledmicrofabrication technology. This definition also includes microelectronic devices, but inaddition 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 tointroduce special steps to free the mechanical structures. Moreover, many MEMS are nownot only based on silicon but are also manufactured with polymer, glass, quartz or thin metalfilms.

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 theonly all-digital display chip by Texas Instruments) but also pressure sensors for automotiveand medical applications. Further applications to come could be lab-on-chip for medicalanalysis and radiofrequency MEMS for telecommunications (switches, filters, etc.). Twoexamples of MEMS fabricated at the IEMN (Villeneuve d’Ascq, France), a microgripperand a micromotor, are shown in figures 3 and 4.

Besides the many applications previously cited, there are three main reasons that prompt tothe use of MEMS technology today:

- Miniaturization of existing devices, like for example the production of silicon basedgyroscope reducing previously existing voluminous devices to a microchip;

- Development of new devicesbased on principles that do not work at larger scale, likefor instance the electro-osmotic effect in microchannels;

- Development of new toolsto interact with the micro-world.

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 massfabrication, not denigrating the performances of the system. Furthermore the materials and


Figure 3. SEM view of an electrostatic microgripper,see[15].

Figure 4. SEM view of an electrostatic micromotor,see[16].


techniques used for fabricating MEMS are compatible with the fabrication of fuel cells, andprovide, for instance, easy ways to conceive microchannels to bring efficiently the fuel andthe oxidant to the membrane-electrode assembly, but mostly can offer to the fuel cell anappropriate and dedicated micromachined support which may eventually be integrated withother microelectronic or MEMS devices.

2.2. Miniaturizable Fuel Cells

If it has been shown in section 1. that miniaturization can bring several limitations in theconception of small fuel cells (materials, flows, techniques, etc.) additional restrictionscome from the devices which they are intended to. Actually portable electronics also havetheir own requirements regarding to the power to supply, the packaging, the working tem-perature and the fuels to use. Thus the manufacturing of miniature fuel cells will be deter-mined by the final applications.

So as it can not be considered to provide a fuel cell functioning at higher temperaturesthan 80 or even 100°C for the user’s safety, the choice in the type of fuel cell to use inportable devices is limited to low temperature fuel cells such as PEMFC (for Proton Ex-change Membrane Fuel Cell or sometimes Polymer Electrolyte Membrane Fuel Cell) andDMFC (for Direct Methanol Fuel Cell).

Miniature SOFCs (for Solid-Oxid Fuel Cells) functioning at temperatures as low as500°C (low temperature for SOFCs usually functioning at 800°C or more) have also beenreported [6–10]. They are based on thin films technology and use techniques similar asthose which will be described later on (photolithography, etching, physical vapour depo-sition; see 3.1.). However decreasing the operating temperature involves the decrease ofperformances of the fuel cell. Besides these temperatures remain inadequate to portabledevices because they imply important heat transfer and thus would require an additionalcooling system which would increase the overall volume of the fuel cell.

3. Microfabrication

3.1. Introduction

The designation microfabrication techniques gathers the collection of technologies used formaking microdevices [17]. To fabricate a microdevice, many processes must be performed,one after the other, many times repeatedly. For instance the fabrication of memory chiprequires 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 ionimplantation), 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 amicrofluidic device or a laser diode.


This section will briefly introduce the reader to the microfabrication techniques thatwill be more generally used for applications in miniature fuel cells, i.e. photolithography,etching, deposition, and sealing. These techniques mainly concern silicon as the basicsubstrate, 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 onmicrofabricated fuel cells (section 4.). More informations on techniques and MEMS ingeneral can be found in [18–21].

3.2. Photolithography

Photolithography or optical lithography is the transfer of a pattern by selective exposure to aradiation source such as UV light from a photomask to a photosensitive material depositedon 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 butglass 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 thinand 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 becomesoluble in a developing solution while a negative photoresist will become insoluble uponillumination. Exposed areas (when using a positive photoresist) or unexposed areas (whenusing a negative photoresist) are removed with the adapted developing solution, leaving thesubstrate 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 underlyingsubstrate (fig. 5).

3.3. Deposition of Materials

One of the basic building blocks in MEMS processing is the ability to deposit thin filmsof materials in order for instance to make local zones conducting or insulating, or to useas masking layers. Film thicknesses from atomic layers up to several tens ofµm can beachieved. The deposited film can subsequently be locally etched using the techniques de-scribed in sections 3.2. and 3.4.. MEMS deposition techniques can be classified in twogroups [21]:

- Depositions thanks to achemical reaction: Chemical Vapour Deposition (CVD), elec-trodeposition, thermal oxidation, and epitaxy. These processes exploit the creation ofsolid materials directly from chemical reactions in gas and/ or liquid compositions oreven with the substrate material. Byproducts can also be created during these reactions.

- Depositions thanks to aphysical reaction: Physical vapour deposition (PVD), and cast-ing. In this latter kind of processes, the material to be deposited is moved from a sourceto the substrate.


Figure 5. Principle of the photolithography for a positive photoresist.

3.3.1. Deposition Thanksto Chemical Reactions

Chemical Vapour Deposition (CVD)

In this process, the substrate is placed inside a reactor supplied by different gases [21]. Theprinciple 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 widelyused 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-kdielectrics.

The two most important CVD technologies in MEMS are the Low Pressure CVD(LPCVD) and Plasma Enhanced CVD (PECVD). The LPCVD process produces layerswith excellent uniformity of thickness and material characteristics. The main problemswith the process are the high deposition temperature (higher than 600 °C) and the relativelyslow 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 oneside of the wafers on 1 to 4 wafers at a time whereas LPCVD systems deposit films on bothsides of at least 25 wafers at a time.

Electrodeposition

It involves the coating of an electrically conductive sample with a layer of metal usingelectrical current. The part to be plated is the cathode of the circuit while anode is made of


the metal to be plated. Both of these components are immersed into a solution containingone or moremetal salts as well as other ions that permit the flow of electricity. A rectifiersupplies a direct current to the cathode causing the metal ions in solution to lose their chargeand plate out on the cathode. As the electrical current flows through the circuit, the anodeslowly dissolves and replenishes the ions in the bath. Other electroplating processes use anon-consumable anode such as lead.

Thermal Oxidation

This process allows the deposition of silicon dioxide on silicon. Silicon oxidation is madeat high temperature (typically between 800 and 1100 °C) in an oven. It is performed withthe reaction of an oxidizing gas (generally O2) on the silicon surface. It can be made moreeffective by the use of water in the form of vapour. This process is particular because thematerial for the deposition (atoms of Si) is taken directly from the substrate. For instancewhen a thickness of 1µm is grown on the Si substrate, the thickness of the substrate isdecreased of a few hundreds of nm.

Epitaxy

This technology is quite similar to what happens in CVD processes, though here with thisprocess 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 orientationwith the substrate acting as a seed for the deposition. If an amorphous/polycrystallinesubstrate surface is used, the film will also be amorphous or polycrystalline.

There are several technologies for creating the conditions inside a reactor needed tosupport epitaxial growth, of which the most important is Vapor Phase Epitaxy (VPE). Inthis process, a number of gases are introduced in an induction heated reactor where onlythe substrate is heated. The temperature of the substrate typically must be at least 50 % ofthe 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 (>100µm). Epitaxy is a widely used technologyfor producing silicon on insulator (SOI) substrates. The technology is primarily used fordeposition of silicon.

3.3.2. Deposition Thanks to Chemical Reactions

Physical Vapour Deposition (PVD)

Two techniques are mainly used:sputtering andevaporation, especially for metal deposi-tion. In both cases the formation of a deposit on a substrate away from the source consistsof three steps: (1) converting the condensed phase (generally a solid) into a gaseous orvapor phase, (2) transporting the gaseous phase from the source to the substrate, and (3)condensing the gaseous source on the substrate, followed by the nucleation and growth ofa film [18].

In the sputter deposition, atoms from a source material (target) are ejected from a sourcematerial (target) by energetic ions. These atoms are then condensed on the substrate to form


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.

Figure 6. Principle of the sputtering deposition process.

In the evaporation method, a film is deposited by the condensation of the vapor on asubstrate, which is maintained at a lower temperature than that of the vapor. All metals va-porize when heated to sufficient temperatures. Several methods, such as resistive, inductive(or radio frequency), electron bombardment (e-beam), or laser heating, can be used to attainthese temperatures, according to the metal to deposit.

These techniques are performed under high vacuum (around 10−4-10−5 Pa).

More specifically, metals can be deposited either by electroplating, evaporation, or sputter-ing processes. Commonly used metals include gold, nickel, aluminum, chromium, titanium,tungsten, platinum and silver.

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 thematerial is dissolved when immersed in a chemical solution. The same thing happens withelectroetching 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 materialsto etch, either metal of polymer, wet etching processes are generally used.

Wet Chemical Etching

It consists in a selective removal of material by dipping a substrate into a solution that candissolve it. Due to the chemical nature of this etching process, a good selectivity can oftenbe obtained, which means that the etching rate of the target material is considerably higherthan that of the mask material if selected carefully. For instance, solutions such as potassiumhydroxide (KOH) or tetramethylammonium hydroxide (TMAH) are generally used to etchsilicon with photoresist and silicon dioxide or nitride masking whereas hydrofluoric acid(HF) solutions etch preferentially silicon dioxide without etching silicon. Solutions forselective metal or polymer etching are also available.


Electroetching of Silicon

Discoveredin the late 1950’s 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 hydrofluoricacid (HF) bath between two platinum electrodes. When a current is established between theelectrodes, the silicon is etched on the anode side. According to the current, it can eitherlead to porous silicon with pores of controlled diameter (from the nm to severalµm, stillaccording to the applied current but also to the type of silicon and the concentration of theHF bath used) or electropolishing (complete etching of the silicon). Masking of the materialcan be achieved with a precious metal layer (gold, platinum) or with silicon nitride but fora 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 reactorin which several gases are introduced. A plasma is struck in the gas mixture using a radiofrequency (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 andforming another gaseous material. The ions can also physically etch the material due totheir high kinetic energy. By changing the balance between the chemical and physical partsit is possible to influence the anisotropy of the etching, since the chemical part is isotropicand the physical part is highly anisotropic. Then the combination can form sidewalls thathave shapes from rounded to vertical. Etch conditions in an RIE system are very muchdependent 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 ofµm can be achieved withalmost vertical sidewalls. The main technology for etching silicon is based on the Boschprocess where two different gases are alternated in the reactor: C4F8 for creating a Teflon-like passivation and protective layer on the surface of the substrate, and SF6 for etching thesubstrate. The passivation layer is immediately sputtered away by the physical part of theetching, but only on the horizontal surfaces and not on the sidewalls. Since the passivationlayer only dissolves very slowly in the chemical part of the etching, it builds up on thesidewalls and protects them from etching. As a result, etching aspect ratios of 50 to 1 canbe 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, usuallypoly(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 toachieve conformal and high resolution (nm scale) patterning with flat or curval substratesand enables the use of materials incompatible with optical lithography or other conven-tional microfabrication techniques. Different replication techniques are possible such as


theembossingwith a rigid master (replication by mechanical deformation of a polymer),the molding where structures are formed inside the void spaces of an elastomeric mold,the printing to transfer patterned ink onto a substrate with an elastomeric stamp or themicrofluidic patterning where the channels of a microfluidic network are used to deliverreactants from a reservoir to the surface of a substrate. Far more documentation on pattern-ing processes can be found in [28].

3.6. Sealing

MEMS devices have movable mechanical parts which require free space to move. Giventheir 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 forinstance anodic bonding. It allows the bonding of glass and silicon by applying a hightemperature between 320 and 400 °C and a high voltage (400 V up to 1 kV). The silicondevices are encapsulated by the glass cap and thus remain protected from the atmosphereat a low vacuum.

In the following section, a review of MEMS-based fuel cells is going to show that all theseprevious techniques are used to manufacture miniature fuel cells and to make them moreeffective. Actually, photolithography / etching and micromolding will be useful to transferpatterns such as microchannels for fuel flows or membranes on a substrate and depositiontechniques 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, andlikewise the safety of consumers.

4. Microfabricated Fuel Cells

In the domain of MEMS-based miniature fuel cells (FCs), various solutions have beenreported. Actually it can be noticed in the literature that the power densities of theminiature FCs range from a few tens ofµW cm−2 up to several hundreds of mW cm−2.Two basic design approaches are currently employed: the classic bipolar design whereall the components of the micro fuel cell are stacked together (fig. 7) and where fuel andoxidant are separated by the membrane electrodes assembly (MEA), and the planar design(fig. 8) where the fuel and oxidant channels are interdigitated and both electrodes are onthe same single side [29]. Bipolar design ensures the separation of fuel and oxidant butrequires all components to be fabricated separately and then assembled together. In thecontrary, the planar design is more suitable for a monolithic integration but requires alarger surface area to deliver similar performance.

The real micromachined parts of the micro fuel cell are generally the electrode platesand the fuel delivery system. The latter is often achieved by microchannels. Nguyenetal made a review of the different designs for the fuel delivery systems [29], illustrated in


Figure 7. Bipolar design of micro fuel cell,after [29].

Figure 8. Planar design of micro fuel cell,after [29].

figure 9.

Concerning theusable fuels for the supply of miniature fuel cells, the more investigatedare presently hydrogen and methanol. Methanol is a low cost liquide fuel with a highenergy density and can be directly oxidized in the fuel cell. Though due to the poorkinetics of the charge transfer reaction at the heart of the fuel cell, direct methanol fuelcells need substantial amounts of noble metals as catalyst, increasing the cost. Moreoverconcentrated methanol can not be used as it leaks through the membranes of fuel cells, sodiluted methanol has to be used which drastically reduced energy density. On the otherhand hydrogen is the basic fuel but is far too dangerous to be stored directly as compressedgas 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 theoverall fuel cell [2]. Fuels such as ethanol [11] or formic acid [12, 13] are also currentlyconsidered as interesting alternatives.

In the following sections the indexation of the different FCs will be made according tothe 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


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 eachFC described.

4.1. Silicon-Based Fuel Cells

As the basic material for MEMS technologies, silicon (Si) is also the most common mate-rial encountered in MEMS-based FCs. Its properties and the microfabrication techniquesassociated to it and previously described are now well-known and mastered. Another ad-vantage of Si-based FCs may also be to facilitate the possible integration of the FCs withother electronic devices on the same chip.

Since 2000, Meyerset al (Lucent Technologies) have proposed two alternative designsusing Si: a classical bipolar using separate Si wafers for the cathode and the anode and aless effective monolithic design that integrated the two electrodes onto the same Si surface[30,31]. In the bipolar design, both electrodes were constructed from conductive Si wafers.The reactants were distributed through a series of tunnels created by first forming a poroussilicon (PS) layer and then electropolishing away the Si beneath the porous film. The FCwas completed by adding a catalyst film on top of the tunnels and finally by casting aNafion®solution. Two of these membrane-electrode structures were fabricated and thensandwiched together. A power density of 60 mW cm−2 was announced for the bipolardesign with H2 supply.

Unlike Meyers and Maynard, Leeet al (Stanford Univ.) proposed a "flip-flop"µ-FCdesign (fig. 10) where both electrodes were present on both sides [32]. If this design does


provide ease of manufacturing by allowing in-plane electrical connectivity, it complicatesthe gas management. Instead of electrons being routed from front to back, gassesmust be routed in crossing patterns, significantly complicating the fabrication process andsealing. Peak power in a four-cell assembly achieved was still 40 mW cm−2 with H2 as fuel.

Figure 10. "Flip-flop" fuel cell design,after [32].

A variant of this design was reported by Minet al (Tohoky Univ, Japan) who pro-posed two structures ofµ-PEFC using microfabrication techniques [33], the "alternativelyinverted structure" and the "coplanar structure". These structures used Si substrates withporous SiO2 layers with platinum (Pt)-based catalytic electrodes and gas feed holes, glasssubstrates with micro-gas channels, and a polymer membrane (Flemion®S). In spite of areported enhancement [34], the FC reached poor results (only 0.8 mW cm−2 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, Yuet al (Hong KongUniv.) were able to decrease the internal resistance of the thin-film current collectors. Itinvolved an increase in the performance of the FC, achieving a peak power density of 193mW cm−2 with H2 and O2 [35].

Yen et al [36] (Univ. of California / Pennsylvania State Univ.) presented a bipolarSi-based micro DMFC with a MEA (Nafion®-112 membrane) integrated and micro chan-nels 750µm wide and 400µm deep fabricated using Si micromachining (Deep RIE). ThisµDMFC with an active area of 1.625 cm2 was characterized at near room temperature (RT),showing a maximum power density of 47.2 mW cm−2 at 60 °C then 1M methanol was fedbut only 14.3 mW cm−2 at RT. Since then, Luet al [37] have enhanced the performance oftheµDMFC to a maximum power density of 16 mW cm−2 at RT and 50 mW cm−2 at 60°C with 2 M and 4 M methanol supply with a modified anode backing structure enablingto reduce methanol crossover (fig. 11). A further evolution of their technology is shown inthe 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. Theelectrodes 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 andreached 35 mW cm−2, 0.38 mW cm−2 and 17 mW cm−2 respectively at RT with forced


Figure 11. Picture of a silicon wafer with flow channels,after [37].

O2. More recently, performances with formic acid as fuel were increased up to 28 mWcm−2 with electrodeposited palladium (Pd)-containing catalyst at the anode [39]. An all-passive system (passive 10 M formic acid and quiescent air) was also recently announcedwith interesting first results (12.3 mW cm−2) [40].

Through the works of Gold and Chuet al, the University of Illinois also proposedanother solution consisting in using nanoporous silicon as the basic material for the proton-conducting membrane [41–43]. A first study by Goldet alhad already shown the relevanceto use porous silicon loaded with sulfuric acid as the membrane of the fuel cell, though theassembled fuel cell exhibited poor results (a few hundredµW cm−2) [41]. Further evo-lutions proved to be more effective. Starting from a standard silicon wafer, 100-µm-thicksilicon membranes were micromachined using deep RIE (with silicon nitride masking) andthen anodized in an ethanoic-HF bath to convert them to porous silicon membranes. Insula-tors (made of TiO2) and catalyst layers (ink made of Pd black (anode), Pt black (cathode),5 % Nafion solution and deionized water) were finally painted onto the porous structuresto complete the membrane electrodes assembly. Supplied with a 5 M formic acid and 0.5M sulfuric acid fuel solution, this fuel cell achieved a current density of 34 mW cm−2 at120 mA cm−2 [42] with the sulfuric acid acting as the proton-conductor. An improvedfabrication process (similar to the one described in section 5.) has been recently proposedwith a membrane wet etched on both sides with a KOH solution, silicon nitride insula-tion layers on both sides grown by LPCVD and a backside RIE after anodization to openup the nanopores [43]. In order to partially oxidize the pore walls, the porous membranewas immersed in a 37 % hydrochloric acid solution during one night. Actually the porousmembrane only acts here as a separator that limits fuel crossover and does not need to beproton-conducting. A 5 nm thick gold palladium alloy film was deposited on top of the cat-alyst layer of both anode and cathode sides uisng sputtering. Current collectors were thenformed by painting a gold ink around and on top of the catalyst films. The performanceswere also improved up to 94 mW cm−2 at 314 mA cm−2 with the same fuel supplying


conditions than previously.Aravamudhanet al (Univ. of South Florida) presented a FC powered by ethanol at

RT [11]. The electrodes were fabricated using macroporous Si technology. The poresdeveloped acted both as micro-capillaries/wicking structures and as built-in fuel reservoir,reducing the size of the FC. The pore sizes dictated the pumping/priming pressure in the FC.The PS electrode thus eliminated the need for an active external fuel pump. The structureof the MEA consisted of two PS electrodes sandwiching a Nafion®-115 membrane. Pt wasdeposited on both the electrodes micro-columns to act both as an electrocatalyst and as acurrent collector. The FC reached a maximum power density of 8.1 mW cm−2 by supplying8.5 M ethanol solution at RT.

Recently, Yaoet al (Carnegie Mellon Univ.) were currently working on a RT DMFCto produce a net output of 10 mW for continuous power generation [44]. Their worksfocused on the design of the complete system including water management at the cathode,micro pumps and valves, CO2 gas separation and other fluidic devices. A passive gasbubble separator removed CO2 from the methanol chamber at the anode side. The backplanes of both electrodes were made of Si wafers with an array of etched micro-sized holes.Nano-tubes catalysts were fabricated on the planes. A 3 % methanol solution at the anodeand the air at the cathode were driven by natural convection instead of being pumped.A micro pump sent water back to the anode side. With 25 mW cm−2, the total MEAarea around 1 cm2 could provide enough power to a 10 mW microsensor along with theextra power needed for internal use, such as water pumping, electronic controls and processconditioning.

Xiao et al proposed a silicon / glass micro fuel cell with catalyst layers formed by Ptsputtering on ICP etched silicon columns (fig. 12) and integrated gold-based micro currentcollectors patterned on both silicon and glass surfaces [45]. The complete process on siliconincluded notably the use of thermal oxide patterned by RIE, dry etching of feedholes andmicrocolumns, wet etching of fuel reservoirs and metallization for current collectors. Theglass substrate (Pyrex Corning 7740) was wet etched by HF to form the chamber and inlet/ outlet. The two substrates were then bonded by anodic bonding at 320 °C and 400 V.The bonded pair was sputtered with TiW (adhesion layer) and Pt to produce the catalystlayer on the micro columns. This process enabled to increase the catalyst area with 3Dcolumns. To complete the fuel cell, the fabricated electrodes were bonded with a Nafion®-117 membrane at 135 °C. Two prototypes were achieved providing current densities of7.1 and 0.1 mW cm−2 with H2 and methanol respectively for the first one and improvedperformances of 13.7 and 0.21 mw cm−2 for the second one which benefited from a newetching process [46].

Liu et alalso applied the MEMS technology to the fabrication of micro DMFCs [47,48].The process included the use of two silicon substrates (respectively anode and cathode)sandwiching a MEA using a modified Nafion®. First both wafers were patterned by pho-tolithography and wet etched. Then, gold was deposited on both to form the electrodesusing a PVD method. The fuel channels through the anode silicon substrate were made bylaser beam and silicon-glass anodic bonding was employed to complete the anode side.After the assembly of the silicon substrates and the MEA, epoxy resin was used as asealant around the edges of the MEA and electrodes, which completed theµDMFC. TheNafion®membrane was modified byγ-ray radiation and palladium deposition to reduce the


Figure 12. Platinum particles on ICP etched rings,after [46].

methanol crossover. This modification allowed performances of 4.9 mW cm−2, whereaswith an unmodified membrane, power density was only 2.5 mW cm−2.

Figure 13. Photograph and schematic views of the finalµDMFC, after [48].

Chen et alrecently reported the fabrication and characterization of a high powerself-breathing PEMFC with optimally designed wet KOH etched flow-fields and elec-trodes [49]. Ni/Cu/Au layers are used for current collecting, the 1.5µm-thick in-betweenlayer instead of a thick Au layer allowing to reduce the fabrication cost. The MEA iscomposed of a classical Nafion®-1135 membrane with Pt-alloy sprayed carbon paper onboth sides. The two silicon electrodes are sandwiched and pressed at RT with the MEA,and then sealed with epoxy. A base-chip formed by drilled Pyrex glass anodically bondedwith KOH eched silicon acted both as H2 inlet/outlet manifolds and as a support for a


Figure 14. From the wafer to the phone, the fuel cells by the CEA,after [50].

six-cell pack. Anaverage power density of 144 mW cm−2 was measured with H2 and airbreathing, which was one of the highest performance currently reported.

Concerning near commercial applications, only public announcements allow us to knowa few details on FCs about to come on the market. Toshiba, Hitachi, Casio, NEC, Motorola,Samsung, Neah Power Systems, to quote only some of the most notorious, have presentedFC prototypes supplying portable equipments and regularly announce forthcoming com-mercialization but as far as we know, none of the main actors have already made theirproducts really available for the masses. In return, more information from laboratories con-nected to the industry can be known. For instance, the French Nuclear Research CenterCEA announced the successful fabrication of high performances prototypes (fig. 1) fueledby H2 and based on thin films type structures on Si substrate obtained by microelectronicsfabrication techniques (RIE for fuel microchannels, PVD for anode collector, CVD, serig-raphy, inkjet for Pt catalyst, lithography) with a Nafion®membrane [50]. An impressivepower density of 300 mW cm−2 with a stabilization around 150 mW cm−2 during hundredhours was reported.

Since many years now, Morseet al (Lawrence Livermore National Laboratory) havebeen 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 forFCs based on either PEM or SO membrane. In 2002, the PEM cell yielded a computedpeak power of 37 mW cm−2 at 0.45 V and 40 °C whereas the SOFC reached 145 mWcm−2 at 0.35 V and 600 °C. Since 2002, the UltraCell company has an exclusive licensewith Lawrence Livermore National Laboratory to exploit their technology to develop com-mercial FCs.


Figure 15. Fraünhofer IZM prototype of miniature fuel cell powering a LED,after [53].

4.2. Metal-Based FuelCells

Less expensive than silicon and easier to assemble with other FCs components, microma-chined foil materials are also one possible choice to obtain low-cost functional FCs.

Using the same structure as previously quoted [36, 37], Luet al have recently replacedthe silicon judged too fragile for compressing and good sealing with the MEA by verythin stainless steel plates as bipolar plates with the flow field machined by photochemicaletching technology [52]. A gold layer was deposited on the stainless steel plates to preventcorrosion. This enhanced fuel cell reached 34 mW cm−2 at RT and 100 mW cm−2 at 60°C.

R. Hahn and the Fraünhofer IZM have also chosen foil materials for their prototypesof self-breathing planar PEMFCs (one example in fig. 2) [53]. The dimensions were 1cm2 with 200µm thickness. Stable long term operation was achieved at 80 mW cm−2 atvarying ambient conditions with dry H2 fuel. They used a commercially available MEAfurther processed in their laboratory. Micro patterning technologies were employed onstainless steel foils for current collectors and flow fields and the assembled technology wasadapted from microelectronics packaging. In their design, the interconnection between cellswas performed outside the membrane area which reduced sealing problems. The completeFC consisted of only 3 layers: the current collectors with integrated flow fields on top andbottom and the patterned MEA in-between. The current collector foils with integrated flowfield of the electrodes consisted of Au-electroplated and microstructured (micro patterning,wet etching, laser cutting, RIE) sandwiched metal-polymer foils. The commercial MEAwas structured using RIE and laser ablation to avoid possible internal bypasses.

Müller et al (IMTEK) used micro-machined metal foils to form the flow fields of theirµ-FC [54]. Using Gore MEA, they were able to form very thin, high power density stacks.They demonstrated both uncompressed and compressed FC designs. The uncompressed


Figure 16. Structure of the passive air breathing direct formic acid fuel cell,after [13].

one had apeak power density of 20 mW cm−2 and the 5-cell compressed stack achieved250 mW cm−2.

Though they do not clearly refer to microfabrication techniques, another interestingmetallic miniature FC structure can also be reported, once again coming from the Universityof Illinois, and proposed by Haet al [12]. They described the design and performanceof a passive air breathing direct formic acid FC (DFAFC). The MEA was fabricated inhouse with catalyst inks (Pt black powder for cathode and Pt-Ru black for anode, Milliporewater and 5 % Nafion®solution) directly painted on a Nafion®-117 membrane. The currentcollectors were fabricated from titanium (Ti) foils electrochemically coated with gold. Theminiature cell at 8.8 M formic acid produced a maximum power density of 33 mW cm−2

with pieces of gold mesh inserted between the current collector and the MEA on both sidesof the MEA. With Pd black used as the catalyst at the anode side [13], they recently obtaineda maximum power density of 177 mW cm−2 at 0.53 V for their passive DFAFC with 10 Mformic acid.

4.3. Polymer-Based Fuel Cells

A third solution currently developed speculates on MEMS-polymers such as polydimethyl-siloxane (PDMS) or polymethyl methacrylate (PMMA). These polymers can be microma-chined by molding or by laser ablation.

Chanet al (Nanyang Univ.) reported the fabrication of a polymericµPEMFC (fig. 17)developed on the basis of micromachining of PMMA by laser [55]. The microchannels forfuel flow and oxidant were ablated with a CO2-laser. The energy of the laser beam has aGaussian distribution thus the cross section of the channel also has a Gaussian shape. A40 nm gold layer was then sputtered over the substrate surface to act both as the current


Figure 17. Assembled polymeric micro fuel cell,after [55]. (a) openview and (b) assem-bled view.

collector and corrosion protection layer. The Gaussian shape allows gold to cover all sidesof the channel. In thisµ-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 lineto 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 adiffusion layer (carbon powder and PTFE) on one side and coated with a catalyst layer (inkwith Nafion®-112 solution and Pt carbon) on the other side. Silver conductive paint wasprinted on the other side of the carbon paper to increase its conductivity and to contributeto collect current. In the final step, the two PMMA substrates and the MEA were bondedtogether using an adhesive gasket. H2 was supplied by hydride storage, the air by an airpump with a constant flow rate of 50 mL min−1, O2 flow with a constant flow rate of 20mL min−1, all tests were performed at RT. A high power density of 315 mW cm−2 at 0.35V has been achieved when O2 is supplied at the cathode side, 82 mw cm−2 with air.

Another PMMA-based micro PEMFC has been developed by Hsiehet al. The technol-ogy included excimer laser lithography for patterning the flow field in the PMMA struc-ture [56]. The channels were about 400µm wide, and 200µm deep with a rib spacing of 50µm for both anode and cathode. Electrically conductive regions acting as current collectorfor 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 bysurface mount technology or hot-pressing. A power density of 25 mW cm−2 at 0.65 V wasobtained at room temperature, when supplied by H2 and O2. Further developments usinghigh aspect ratio UV-curable epoxy SU-8 instead of PMMA and silver as current collectorwere reported with similar performances (30 mW cm−2 with H2 and forced air) [57].

SU-8 epoxy was also the solution adopted by Chaet al to create microchannels bysimple UV photolithography and lift-off from a glass substrate [60]. Nafion-115 was usedas the PEM and sandwiched between two polymer chips with microchannels and two layers


of photosensitive polymers. Microchannels were 300µm wide and100µm deep. Currentcollectors 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. Themaximum power density observed with this all-polymerµ-DMFC was 8 mW cm−2 with 2M methanol as fuel and oxygen feeding.

A complete PEMFC system consisting in a PDMS substrate with micro flow channelsupon which the MEA was vertically stacked has been developped by Shahet al [58, 59].PDMS microreactors were fabricated by employing micromolding with a dry-etched siliconmaster. The PDMS spincoated on micromachined Si was then cured and peeled off from themaster. The MEA consisted in a Nafion®-112 membrane where Pt was sputtered through aMylar®mask. Despite an interesting method, this FC gave poor results (0.8 mW cm−2).

4.4. Hybrid and Emerging Solutions

An original approach was recently proposed by Park and Madou. It used carbonized ma-chined polyimide (Cirlex®from Dupont, made from Kapton®) as their microfluidic plates[61]. The technique they currently called C-MEMS (for carbon-MEMS) consists in mi-cromachining the polyimide and then pyrolyzed it at high temperature (900 °C) to makemicromachined carbon (fig. 18). The goal is to reproduce the advantages of graphite usedin large fuel cells (such as high electrcal conductivity, resistance to corrosion, high thermalconductivity, chemical compatibility, etc.) but uneasy to report on micro fuel cells due tothe brittleness of the material and its high cost. The structure of the complete fuel cell wasas follows: carbon microfluidic flow channels on each side of a commercial MEA madeof a Nafion®-115 membrane and 20 % Pt loaded carbon paper from Electrochem. The FCwas sealed with epoxy and the membrane was larger than the microfluidic plates in orderto hydrate the inner membrane from the external part. A power density of 1.21 mW cm−2

was achieved. The major drawback of this technology is the cost of the polyimide whichremains expensive compared to other materials.

Wanet alpresented another composite solution which used porous polytetrafluoroethy-lene (PTFE) as the support for the proton-conducting electrolyte [62]. An expanded PTFEsubstrate was sandwhiched on both sides with very thin porous metal sheets made of tita-nium coated with Ru layer and fabricated by photolithography and wet etching techniques.An electrolyte dispersion (DE 2021 from DuPont) was then impregnated into the pores ofthe PTFE membrane and the metal sheets, dried and protonized by immersing in boiling 5% sulfuric acid solution. To complete the MEA (fig. 19), a catalyst ink made of Pt-loadedcarbon, alcohol, and 5 % Nafion®solution, was directly brush-coated onto both surfaces ofelectrolyte membrane and dried. Performances of 80 mW cm−2 were reached with a H2/airsupply. The metal sheets served here both as flow fields and current collectors.

This brief overview shows that it is not the material that really matters to achieve highperformances but mostly the way the fluids are managed and the stack is sealed. If Si is stillthe material the most commonly used, the promising performances and technology obtainedby the Fraunhöfer Institute with stainless steel foils, for instance, or the ones reached withPMMA by Chan et al [55], have shown that interesting alternative technologies can alsolead to functional FCs. The table 1 summarizes the main characteristics of the previouslycited fuel cells.


Figure 18. Carbonized fluidic plates shown to the left of 1 cm2 Cirlex®square,after [61].

Figure 19. View of the integrated composite MEA,after [62].

Table 1. Main characteristics of the reviewed micromachined fuel cells, with Pmax in mW cm−2, OCV the open circuit voltage in V,FA for formic acid, SA for sulfuric acid, RT for room temperature and T for working temperature in °C. NR is for non reported

information.

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] Müller 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


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®orNafion®-like ionomer as the proton exchange membrane. However these proton-conductingpolymers are not really suitable for microfabrication: they can not be patterned with stan-dard photolithography and their volumetric changes with humidification are a majorproblem to deal with at a micrometer scale. Yet alternate solutions are currently underactive development. Actually, there have been several papers on membraneless laminarflow-based microfluidic fuel cells [63–65], which do not require any proton-exchangemembrane. These micro fuel cells use the liquid-liquid interface between the fuel and theoxidant. An aqueous stream containing a liquid fuel, such as formic acid, methanol, ordissolved hydrogen, and an aqueous stream containing a liquid oxidant, such as dissolvedhydrogen, permanganate, or hydrogen peroxide, are introduced into a single microfluidicchannel in which the opposite sidewalls are the anode and cathode [68]. Thanks to thedominant laminar convective transport, fuel crossover can be avoided, and so the anodedry-out and cathode flooding. Furthermore these fuel cells are media flexible, i.e. thecomposition 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 newdual electrolyte H2/O2 fuel cell system, Cohenet al exhibited thus an OCV of 1.4 V and1.5 mW was generated from a single planar device with Si microchannels [66]. With theintegration of a gas-diffusion electrode enabling oxygen delivery directly from air, thelaminar flow-based fuel cell proposed by Jayashreeet al achieved performances as highas 26 mW cm−2 with 1 M formic acid as fuel [68]. This novel technology is thus alreadycompeting with proton-exchange membrane fuel cells and allows to consider furtherminiaturization 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 asan example of a miniature fuel cell manufactured thanks to microfabrication technology.

5. Example: Grafted Porous Silicon-Based Miniature Fuel Cells

This section focuses on the fabrication of a particular proton-exchange membrane (PEM)for small FCs using microtechnology. The solution has been developed as an alternative toclassical ionomeric membranes like Nafion®and is based on a porous silicon (PS) membraneon which is grafted a proton-conducting silane. Like previously cited Chuet al [41–43],previous works have reported the relevance to use PS directly as the membrane of the FC,either as the support for a Nafion®solution [69–72] or with appropriate proton-conductingmolecules grafted on the pore walls [70,73–75]. This second solution involves the chemicalgrafting of silane molecules containing sulfonate (SO−

3 ) or carboxylate (COO−) groups onthe pores walls in order to mimic the structure of an ionomer.


5.1. Manufacturing Process

5.1.1. Membranes

The completeprocess for the fabrication of the proton-conducting membranes, previouslyreported in [69,73], can be described as follows. A 4-inch 520µm thick n+ (100)-orienteddouble-side polished silicon wafer is first thermally oxidized in an oven at 1000 °C underO2 and water steam flows to obtain a 2µm 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 Crlayers are used as adherence layers for the Au layers and are relatively thin (30 nm). The Aulayers are 800 nm thick and will serve as current collector layers for the fuel cells. They arealso useful as masking layers for the next different etchings since Au is not etched neitherby 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 50µm thick double-sidedmembranes. These membranes are then made porous by anodization in an ethanoic HF bathwith an average pore diameter of 10 nm (size controlled with the input current density). Inorder to be sure that all the pores are open on the rear side of the membranes, a short reactiveion 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 andthe figures 21 and 22 show the result of the process.

5.1.2. Proton Conduction and Membrane Electrode Assembly

The previous manufactured membranes are not usable as they are. Actually, PS is notnaturally proton-conducting. In order to make the membranes proton-conducting, a solu-tion consists in grafting silane molecules containing proton-conducting groups on the poreswalls. A commercially available silane salt from United Chemical Technologies Inc (UCT)containing three carboxylate groups have been used for the first investigations. The first stepconsists in creating silanol functions (Si-OH) at the surface of PS. A soft process involvingUV-ozone cleaner has been successfully implemented. The grafting of silane molecules isthen realized by immersing the hydrophilic porous membranes into a 1 % silane solution inethanol for 1 h at RT and ambient air. In order to replace -Na endings from the silane saltby -H endings to get the real carboxylic behaviour for the grafted function, membranes areimmersed for 12 hours in a 20 % solution of sulfuric acid, then carefully rinsed in deionizedwater. 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 (inblack 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


Figure 20. Sketch of a complete process for manufacturing porous silicon membranes forminiature fuelcells.


Figure 21. Photograph of the silicon wafer after the membrane etching process. 69 mem-branesof 7 mm2 are micromachined on a 4-inch wafer.

Figure 22. SEM (scanning electron microscope) cross section view of a typical poroussilicon membranemade with the process in figure 20.


Figure 23. Membrane-electrode assembly (8 x 8 mm), scale comparison with a 1 cent eurocoin.

gasto the membrane, a home-made test cell was used in which up to 4 single cells can betested separately or together with a serial or parallel connexion. A maximum power densityof 58 mW cm−2 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 shouldbe reduced at least as small as that of assumed pores in Nafion®(3 to 5 nm). Moreover, toincrease the power density, the grafting density should be controlled to be sure that all theinternal surface of the pores is grafted. The use of a catalyst ink instead of a Pt-loadedcarbon cloth should create a more intimate link between the membrane and the catalystand further increase the performances. A study on the thickness of current collecting metallayers [49] had also recently shown that the thickness of the Au layer was probably too thinto provide the lowest electrical resistance.

6. Conclusion

Though no commercial product is available yet, microfabrication techniques have nowproved 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 everycomponent of a fuel cell: flow channels for the proper circulation of fuel and waste withphotolithography and reactive ion etching or micromolding, catalyst support by etching anddeposition, current collector by metal deposition and proton-conducting membrane withinsulator deposition, photolithography, wet etching, electroetching and post-fabricationchemical treatment.

Several examples of miniature fuel cells using microfabrication techniques have beenpresented. Among the numerous solutions developed today, the basic structure of fuel cells


Figure 24. Performances of the miniature fuel cell obtained with a mesoporous siliconmembrane graftedwith proton-conducting silane (voltage vs current density in continuousline, power density vs current density in dashed line).

remains the same: thin film planar stack (generally silicon, foils, polymer or glass) withcommercial ionomer, most often Nafion®, the reported layers being micromachined (mi-crochannels or porous media) for gas/liquid management and coated with gold for currentcollecting. Performances range from the tenth ofµW cm−2 to several hundreds of mWcm−2 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-basedfuel cells, but foils and polymers have shown interesting perspectives for future commercialdevelopments. 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 usingreal 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 toget rid of ionomer membranes that have notably the disadvantages to change in size withthe humidification and to be inappropriate with microtechnology.

As a detailed example of a micromachined fuel cell, an alternative solution which doesnot use an ionomer for the proton-exchange membrane has also been reported. It consistsin a porous silicon membrane with a proton-conducting silane grafted on the pores walls.With this membrane, performances as high as 58 mW cm−2 have been achieved. Thispromising membrane can still be improved. Future work should focus on the reduction ofthe pores diameter to decrease the gas crossover, the control of the grafting density intothe porous silicon, the replacement of the electrode carbon cloth by an ink and the use of amore proton-conducting silane (with SO−3 terminations).


Acknowledgement

All previouslypublished figures were reproduced with kind permission from the editors and/or the authors.

References

[1] Dyer, C.K.Scientific American1999, 281 (1), 88-93.

[2] Dyer, C.K. In: 2004 IEEE Symposium on VLSI Circuits, Digest of Technical Papers2004, 124-127.

[3] Dyer, C.K.J. Power Sources2002, 106, 31-34.

[4] Boisard, V.Etats Unis Microélectronique2000, 22, 4-16.

[5] Hebling, C.; Heinzel, A.Fuel Cells Bulletin2002, 7, 8-16.

[6] Morse, J.D.; Graff, R.T.; Hayes, J.P.; Jankowski, A.F. In:Proceedings of the MaterialsResearch Society Symposium, 2002; Vol. 575, pp 321-324.

[7] Srikar, V.T.; Turner, K.T.; Ie, T.Y.A.; Spearing, S.M.J. Power Sources2004, 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 A2005, 123-124,497-504.

[12] Ha, S.; Adams, B.; Masel, R.I.J. Power Sources2004, 128, 119-124.

[13] Ha, S.; Dunbar, Z.; Masel, R.I.J. Power Sources2006, 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.; Régnier, S.; Bidaud, P.; Tsitsiris, E.; Collard, D.; Buchail-lot, L. Sens. Actuators A2004, 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-ers’05, Séoul, Korea, 2005; pp.57-60.

[17] http://en.wikipedia.org/wiki/Microfabrication


[18] Sze, S.M.VLSI technology, 2nd edition, McGraw-Hill, New York, 1988.

[19] Senturia, S.D.Microsystem Design, Springer Science, New York , 2001.

[20] Madou, M.J.Fundamentals of Microfabrication: The Science of Miniaturization2ndedition, CRC PRESS, New York, 2002.

[21] http://www.memsnet.org/mems/

[22] http://en.wikibooks.org/wiki/Microwiki

[23] Ulhir, A. Bell Syst. Tech. J.1956, 35, 333.

[24] Canham, L.T.Appl. Phys. Lett.1990, 57, 1046-1048.

[25] Lehmann, V.; Gösele, U.Appl. Phys. Lett.1991, 58, 856-858.

[26] Halimaoui, A. In:Properties of Porous Silicon; Canham, L. (Ed.), Emis DatareviewsSeries, INSPEC, 1997; pp 12-22.

[27] Lehmann, V.Electrochemistry of silicon, Wiley-VCH, Weinheim, Germany, 2002.

[28] Geissler, M.; Xia, Y.Adv. Mater.2004, 16 (15), 1249-1269.

[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 andBattery Technologies for Portable Power Applications International Symposium, NewOrleans, LA, U.S.A., 2000.

[31] Meyers, J.P.; Maynard, H.L.J. Power Sources2002, 109, 76-88.

[32] Lee, S.J.; Chang-Chien, A.; Cha, S.W.; O’Hayre,R.; Park, Y.I.; Saito, Y.; Prinz, F.B.J. Power Sources2002, 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 Sources2003, 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. Acta2004, 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 B2005, 107 (2), 882-891.

[39] Jayashree, R.S.; Spendelow, J.S.; Yeom, J.; Rastogi, C.; Shannon, M.A.; Kenis, P.J.A.Electrochim. Acta2005, 50 (24), 4674-4682.


[40] Yeom, J.; Jayashree, R.S.; Rastogi, C.; Shannon, M.A.; Kenis, P.J.A.J. Power Sources2006, 160 (2), 1058-1064.

[41] Gold, S.; Chu, K.-L.; Lu, C.; Shannon, M.A.; Masel, R.I.J. Power Sources2004, 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. Energy2006, 31, 636-649.

[45] Xiao, Z.; Yan, G.; Feng, C.; Chan, P.C.H.; Hsing, I-M. In:Proceedings of the 13thInternational 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 the8th 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 CEA2004, 50-51.

[51] Jankowski, A.F.; Hayes, J.P.; Graff, R.T.; Morse, J.D. In:Proceedings of the MaterialsResearch Society Symposium, San Francisco, CA, U.S.A., 2002; Vol. 730, pp 93-98.

[52] Lu, G.Q.; Wang, C.Y.J. Power Sources2005, 144, 141-145.

[53] Hahn, R.; Wagner, S.; Schmitz, A.; Reichl, H.J. Power Sources2004, 131, 73-78.

[54] Müller, M.; Müller, C.; Gromball, F.; Wölfe, M.; Menz, W.Microsyst.Technol.2003,9, 159-162.

[55] Chan, S.H.; Nguyen, N.-T.; Xia, Z.; Wu, Z.J. Micromech. Microeng2005, 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.


[58] Shah, K.; Shin, W.C.; Besser, R.S.J. Power Sources2003, 123, 172-181.

[59] Shah, K.; Shin, W.C.; Besser, R.S.Sens. Actuators B2004, 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. Acta2004, 50, 795-799.

[61] Park, B.Y.; Madou, M.J.J. Power Sources2006, 162 (1), 369-379.

[62] Wan, W.; Wang, G.J. Power Sources2006, 159, 951-955.

[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 Sources2004,128, 54-60.

[65] Choban, E.R.; Spendelow, J.S.; Gancs, L.; Wieckowski, A.; Kenis P.J.A.Electrochim.Acta2005, 50, 5390-8.

[66] Cohen, J.L.; Westly, D.A.; Pechenik, A.; Abruna, H.D.J. Power Sources2005, 139,96-105.

[67] Cohen, J.L.; Volpe, D.J.; Westly, D.A.; Pechenik, A.; Abruna, H.D.Langmuir2005,21, 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 portableapplications, Ph.D. Thesis, 2004; available at http://tel.ccsd.cnrs.fr/tel-00010710.

[71] Pichonat, T.; Gauthier-Manuel, B.J. Memb. Sci.2006, 280, 494-500.

[72] Pichonat, T.; Gauthier-Manuel, B.Microsyst. Technol.2006, 12 (4), 330-334.

[73] Pichonat, T.; Gauthier-Manuel, B.J. Micromech. Microeng.2005, 15, S179-S184.

[74] Pichonat, T.; Gauthier-Manuel, B.J. Power Sources2006, 154, 198-201.

[75] Pichonat, T.; Gauthier-Manuel, B. In:Proceedings of the 8th Symposium on Design,Test, Integration and Packaging of MEMS/MOEMS (DTIP’06), Stresa, Italy, 2006; pp354-359.

[76] Pichonat, T.; Gauthier-Manuel, B.Microsyst. Technol.2007, DOI: 10.1007/s00542-006-0342-5.

Owner

Note


Chapter 6

TECHNICAL H2 ELECTRODES FOR LOWTEMPERATURE FUEL CELLS

Francisco Alcaide-Monterrubio1*, Pere L. Cabot2 and Enric Brillas2

1CIDETEC, Centre for Electrochemical Technologies,Pº Miramón, 196, E-20009 San Sebastián, Spain.

2Laboratori d’Electroquímica dels Materials i del Medi Ambient.Departament de Química Física, Facultat de Química,

Universitat de Barcelona, C/ Martí i Franquès 1-11, E-08028 Barcelona, Spain.

Abstract

Hydrogen electrode reaction has been widely studied regarding to its applications in fuelcell technology. In the last years, a lot of studies concerning kinetic and mechanistic aspectsabout smooth and well defined surfaces have been recently published, but less is known aboutthe electrochemical behavior of technical electrodes, in half-cells and complete cells. In thispaper, the main manufacture methods of technical hydrogen electrodes reported in theliterature 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 hydrogenelectrode oxidation reaction both, in liquid and solid electrolytes is also reviewed. Previouswork constitutes a significant background that can help to develop technical hydrogendiffusion anodes for application in practical fuel cells. In particular, the electrochemicalbehavior 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 energysources has been well demonstrated. However, high cost, durability, high system complexityand lack of fuel infrastructure are drawbaks that still keep in standby their large-scalecommercialization [1, 2]. Table 1 summarizes some technical characteristics of the main


Francisco Alcaide-Monterrubio, Pere L. Cabot and Enric Brillas248

types of FCs, classified according to the electrolyte used and the operation temperature,which, are:

Table 1. Summary of some features of the main fuel cell types

PEFC AFC PAFC MCFC SOFC

Electrolyte ion exchangemembrane

potassiumhidroxyde

phosphoric acid liquid moltencarbonate

ceramic

Operating temp. / °C 60-120 < 100-240 150-220 600-700 600-1000

Charge carrier inelectrolyte H+ OH- H+ CO3

2- O2-

Anode catalyst Pt, Pt/Ru Pt/Au, Pt, Ag Pt Ni, Ni/Cr Ni/ZrO2

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 exchangemembrane, which has to be hydrated for a proper operation. The working temperature isbetween 60-120 °C, limited by the polymer membrane.

Alkaline Fuel Cells (AFCs). The electrolyte is 35-50 or 85 wt. % KOH in aqueoussolution, 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 therange 150-220 °C, but it can be increased to about 300 °C, because of the stability of theconcentrated acid solution.

Molten Carbonate Fuel Cell (MCFCs). The electrolyte is a mixture of sodium, potassiumand lithium carbonates, retained in a ceramic matrix of LiAlO2. The working temperature isaround 600-700 °C.

Solid Oxide Fuel Cells (SOFCs). Usually, Y2O3-stabilized ZrO2 (yttria-stabilized zirconiaYSZ) is used as electrolyte. The cell operates at 600-1000 °C.

Nowadays, such FCs are in different stages of development and have differentapplications. The only practical application of low temperature fuel cells considered in thischapter are AFCs in the American space shuttles. On the other hand, PAFCs have been usedin stationary power generation plants since the 70s. Finally, PEFCs have experimented aresurgence in the 90s due to their performance improvement, as a consequence of the use of anew proton exchange membrane (Nafion) and new techniques that enhanced the efficiency ofplatinum catalyst in the electrodes. This resurgence has been mainly directed towards portableand 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 hydrogenas fuel, because platinum is used as electrocatalyst in the anode. Pure hydrogen is expensiveand 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 someimpurities such as CO2 and CO or diluents (e.g. N2). The consequence of the use of this fuelis a loss in the fuel cell performance because the anode becomes poisoned. One way to avoidthis 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 hydrogenoxidation reaction (HOR) taking place in the anode.

Hydrogen diffusion anodes utilized in practical fuel cells are technical electrodes. In thiskind of electrodes, complications arise from the undefined structure of the catalyst.Furthermore, most of the surface analytical methods used with smooth surfaces can not beapplied with highly dispersed technical electrodes. That is the reason because, traditionally,HOR has been studied in model electrodes, used to reduce complexity and to simplifyinterpretation of electrochemical experiments. For example, in model electrodes there is noinfluence of gas flow nor effect of the ionomer, as would be the case for a technical electrodemeasurement. Besides, another motivation to work with model electrodes lies in thepreparation of defined structures, like well-defined surfaces. In this scenario, the relationshipbetween electronic property and electrochemical reactivity may be evident in theexperimental measurements.

It seems that still, the results obtained with model electrodes can not be applied in astraightforward way to technical electrodes. However, they are very interesting studies carriedout with technical electrodes in experimental conditions that reproduce as faithfully aspossible those found in real FCs.

In this chapter, we will describe first the main methods utilized to fabricate technicalelectrodes, 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 willdiscuss later. Finally, we will try to correlate the electrochemical behavior of modelelectrodes with those shown for technical electrodes on the basis of their chemicalcomposition.

2. Fabrication Methods of H2 Technical Electrodes

2.1. Gas Diffusion Electrodes

The development of hydrogen electrodes for fuel cells (FCs) is directly related to thedevelopment of the different kinds of FCs. In this sense, it has been established that a goodelectrode structure has to verify the following conditions: (i) to allow a thin film of electrolyteover the electrode with easy access of reactants and avoid drying and drowning, (ii) to have ahigh surface area for reaction and (iii) to keep the electrolyte with constant thickness andcomposition.

Hydrogen technical electrodes for low temperature fuel cells such as AFCs, PAFCs andPEFCs are porous gas diffusion electrodes (GDEs) [4]. These electrodes have a large areareaction zone with minimum mass transport hindrances, thus allowing the easy access of


reactants and removal of products. Then, the goal of GDEs fabrication is to provide astructure that maximizes the catalyst utilization and avoids large mass transport losses in theelectrode.

GDEs typically consist, at least, of two layers: a gas diffusion layer (GDL) and an activelayer (AL). The GDL should provide mechanical support and electrical contact (currentcollector), optimal distribution of reactant gases and a pore structure suitable for the removaof liquid or vapor phase water (water management). The AL contains the catalyst, where theelectrochemical reaction takes place, but only in those sites where reactant, electrolyte andcatalyst meet, that is, in the three-phase zone or boundary (TPB).

2.2. Preparation of Gas Diffusion Electrodes

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 presentin them. We will report below general procedures for their preparation, mainly focused to themanufacture of hydrogen diffusion electrodes [5]. The main steps in the fabrication of GDEsare the preparation of the gas diffusion medium and the preparation and application of thecatalyst layer.

2.2.1. Gas Diffusion Layer Preparation

The preparation of a GDL involves the use of a substrate, carbon cloth or paper [6-8], whichare in general commercially available. They are usually treated to havehydrophobic/hydrophilic properties, typically using polytetrafluoroethylene (PTFE) [9]. Amicroporous carbon layer, made with carbon and PTFE with controlled porosity is applied tothe substrate in the catalyst layer side or to both sides [10]. This improves the gas and watertransport properties.

2.2.2. Catalyst Layer Preparation and Application

The PTFE-bonded electrode was introduced by Neidrach and Alford [11]. It consists of metalblacks or carbon supported metal catalysts that are hydrophilic, blended with fine particles ofhydrophobic PTFE, that flow and bind the structure as a result of heat treatment duringfabrication. Thanks to its physical properties, PTFE flows and penetrates the pores, thusallowing a good interfacial contact between catalyst and carbon and providing hydrophobicgas pores for reactants.

During the electrode operation, some pores are filled with the electrolyte while others areempty. This produces a large vapor/liquid interfacial area and increases the interfacialreaction rate due to better catalyst utilization and electrode performance. The interface is quitestable by the existence of high capillary forces in small pores. The pressure control system ofthis 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 witha Pt load of about 0.3 mg cm-2 [12-14]. PAFCs employ H2-diffusion PTFE-bonded electrodeswith 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


premixed before the catalyst layer is deposited (thin film method), being the catalyst loadingbetween 0.1-0.35 mg cm-2 [16-18].

The most important methods for catalyst layer preparation and application are:

i) Spray [19, 20]An ink is first prepared by stirring a mixture of the catalyst and Nafion® solution in an

alcohol (e.g., isopropanol)-water solution for several hours. A PTFE emulsion is normallyadded immediately before the spray process. The ink is then repeatedly sprayed using airbrush, air-less spray or electrostatic spray, depending on the size of the electrodes, onto theGDL. Between each spraying, the electrode is sintered. The last step can be the electroderolling to produce a thin layer of uniform thickness and low porosity.

ii) Screen printing [21, 22]This method is the same as that used commercially for screen printing logos on clothing.

In this case, the ink is fairly viscous. For example, a paste like ink can be prepared [23] bymixing 1 g of Pt/C (Pt 20% wt.) with 3 mL of Nafion® (5% wt.), 5 mL of an alcoholic solvent(such as ethanol or isopropanol), and 2 g of camphor as the pore forming agent. Sonication isapplied to homogeneously disperse the contents. Another example of formulation to preparean ink to make a H2 electrode for a PAFC [24] is Pt/C, 10 wt. % PTFE, 5 wt. % polyvinylalcohol, and 3wt.% t-butyl phosphate solution, the Pt catalyst loading being 0.5mg cm−2. Agelling agent to increase the viscosity is sometimes added [25].

iii) Gel-roll process [26]This method consists of preparing a gel by mixing the desired quantities of the catalyst

powder, PTFE emulsion and toluene or xylene. This gel can be rolled in a pinch roller toobtain electrode sheets of desired thickness.

iv) Dry-ice spread [26]The process involves mixing appropriate amounts of catalyst and PTFE (in powder form)

with dry ice, spreading the resulting mix on a GDL and letting it dry. The loose structure isthen pressed in a hydraulic press to compact the catalyst onto the substrate.

v) Dry-roll process [27]A mixture of catalyst powder, PTFE and a pore former (e.g., sodium carbamate) is first

prepared by blending them with isopropanol. Afterwards, the blend mix is filtered and driedto remove the solvents and water. The mix is then fed to a pair of heated vertical rollers. Theshear force applied between the rollers converts the powder mix into a sheet of uniformthickness. This sheet is further pressed onto the GDL in a hydraulic press and finally, theresulting electrode is soaked in warm water to remove the pore former.

vi) Filter casting [26]The first step is the preparation of a gas diffusion medium, which usually consists of a

thin layer of uncatalyzed carbon and PTFE on a hydrophobized carbon cloth or paper. Thefollowing step involves the preparation of catalyst-PTFE suspension. The catalyst is firstdispersed in distilled water, using a combination of stirring and sonication. The PTFEemulsion is added and the PTFE flocculated onto the catalyst powder. The pH of the catalyst


has to be between 2.5-3.0 to avoid improper flocculation, thus resulting in a colloidalsuspension of the catalyst-PTFE emulsion that can pass through the filter or substrate.

vii) Other processesAnother method for electrode preparation is a direct filtration. The catalyst-PTFE

suspension can be directly filtered onto the carbon paper or filtered onto a thin porous paperand subsequently transferred to the carbon paper (or cloth) by rolling. This step is usuallyfollowed by cold and hot pressing in a hydraulic press to ensure good adhesion of the catalystlayer to the substrate. Finally, a sintering process is applied to allow a cohesive interactionbetween the carbon/catalyst particles and PTFE and to confer good mechanical strength to theelectrodes. It consists of two steps, a backing process for 10-15 minutes at 270°C (to removemost of the surfactants) and a sintering step for 10-20 minutes at 340-350°C.

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 usedin PEFCs, PAFCs and AFCs, respectively. Nowadays, hybrid techniques involving


combination of two processes explained above, for example, spray and screen-printing, arealso tested to fabricate PEFCs electrodes [28]. On the other hand, vacuum deposition methodssuch 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 obtaincatalyst layers because its scaling-up is difficult.

Figure 3. Production steps for the manufacturing of dual-layer H2 technical electrodes using PTFE as apowder for AFCs.

Figure 4. Production steps for the manufacturing of dual-layer H2 technical electrodes using PTFE as apowder and in emulsion for AFCs.


3. Electrocatalysis of Hydrogen Oxidation Reaction

Electrocatalysis plays a crucial role in the development of efficient and high performance lowtemperature fuel cells. To enhance the electrode reactions of these systems, it is needed thestudy of different electrocatalysts that have a direct influence on the rate of theelectrochemical reactions considered in order to identify the most active one and optimize it.In this sense, geometric and electronic factors are determinant to choose a suitableelectrocatalyst [30], which should have a high electronic conductivity, be stable in the fuelcell environment and display suitable adsorption characteristics for reactants andintermediates.

Most of the electrocatalysts used in H2 technical electrodes for HOR in low temperaturefuel cells are noble metals such as platinum, ruthenium, and palladium. Some non-noblemetals with acceptable catalytic activity, alloyed or mixed with noble metals, are cobalt, iron,molybdenum, nickel, tin, and tungsten [31]. Some organic materials, like metal phtalocianineshave also been satisfactory for some reactions [32].

Unsupported materials such as metal blacks were initially used as catalytic electrodes, butthey were rapidly disregarded due to the need of using high loadings of expensive noblemetals. This led to the development of supported systems, in which the fine dispersion of thecatalyst in the very porous support allows increasing the active surface where theelectrochemical reaction takes place.

The main supports used for fuel cell catalysts are carbonaceous materials such as carbonblacks [33-35]. Carbons have sufficient electrical conductivity (in the range 1-103 S cm-1),good thermal conductivity, and low thermal expansion coefficient. They also exhibit areasonable corrosion resistance, which can be improved by suitable heat treatments [36].From a technical point of view, carbons are lightweight and have a porous structure and alarge surface area (25-2000 m2 g-1, according to BET measurements), that provide many sitesfor deposition of catalyst particles. On the other hand, different carbon types arecommercially available and they are not expensive. The most used are Vulcan XC-72, VulcanXC-72R (extra conductive carbon black), Black Pearls 2000 and Shawinigan acetylene black.The choice of a particular carbon support will affect the performance and lifetime of theelectrodes, and consequently, of the fuel cell considered.

3.1. Hydrogen Oxidation in Alkaline Media

The use of a particular catalyst for HOR in alkaline media depends on the operatingconditions 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 Nidoped with Co or alloyed with Fe, Ti or Mo to improve its performance) approaches that ofnoble 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 beenemployed to catalyze HOR in alkaline solutions.


3.2. Hydrogen Oxidation in Acid Media

HOR has a very high exchange current density and platinum is generally considered to be themost effective electrocatalyst amongst the stable noble metals [40]. When pure hydrogen isfed, the reaction is highly reversible and proceeds at high current densities with lowpolarization loss, i.e. the overpotential associated with the hydrogen electrode is very loweven at very high current densities. However, when hydrogen contains impurities such as COor CO2, the anode overpotential increases and the cell performance is reduced. Tungstentrioxide (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 practicalchallenge in the HOR research is to find electrocatalysts with high exchange currents per unitcost and resistance to poisoning, i.e., low cost and stable catalysts able to oxidize hydrogen ata very high rate in the presence of impurities. Of course, carbon supported electrocatalysts arenowadays generally used, since their performance depends on the overall active surface area.

Commercial application of H2 is generally based on the gas obtained by hydrocarbonsreforming and therefore, it is necessary to dispose of an anode electrocatalyst able to toleratea limited amount of CO, either under steady state operation or under transient conditions ofhigh CO content. In addition, CO2 adversely affects the catalyst performance, because areverse 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 forHOR in the presence of CO include Pt and several Pt alloys or Pt mixtures with other noble ornon-noble metals [44]. The most used include ruthenium, molybdenum and tin.

At temperatures below about 125°C, CO adsorption on platinum is very strong. Even fewppms in the H2 stream cause substantial performance losses on the anode. Therefore, the useof 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 anodeelectrocatalysts for hydrogen oxidation in the presence of CO because they tolerate low ppmsof CO without excessive polarization losses. Tungsten carbide (WC) also shows high CO-tolerance [38, 44].

4. Mechanistic Aspects of Hydrogen Oxidation Reaction

In aqueous acid solution, the hydrogen oxidation reaction (HOR) proceeds according to thefollowing equation:

H2 → 2 H+ + 2 e- (1)

whereas in alkaline solution, the overall anodic process can be written as:

H2 + 2 OH- → 2 H2O + 2 e- (2)

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 theelectrolyte to the electrode surface:


H2 → H2,electrode surface (3)

b) Hydration and ionization of molecular hydrogen, which may take place through one orboth of the two following parallel reaction paths:

i) dissociative adsorption of molecular hydrogen into hydrogen atoms (Had) on theactive sites (Tafel reaction):

H2,electrode surface → 2 Had (4)

followed by its ionization and hydration (Volmer reaction):

Hads → H+ + e- (5)or

Hads + OH- → H2O + e- (6)

ii) hydration and partial ionization in a quasi-single step (Heyrovsky reaction):

H2,ad → Had + H+ + e- (7)or

H2,ad + OH- → Had + H2O + e- (8)

c) Transport of H+ or water from the electrode surface.

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 assumingthat adsorption of hydrogen atoms on the electrode surface follows a Langmuir isotherm, withinsignificant adsorption of molecular H2. It is assumed that such reactions occur on ahomogeneous surface. Then, current densities for Tafel, Volmer and Heyrovsky reactions canbe written as [45]:

( ) 2T-H

2TT --1

2θθ kakj = (9)

( ) ( )⎥⎦⎤

⎢⎣⎡

⎥⎦⎤

⎢⎣⎡= + RT

Fak

RTF

kj -1-

exp -1-

exp VHV-

VVV

ηαθ

ηαθ (10)

( ) ( )⎥⎦⎤

⎢⎣⎡

⎥⎦⎤

⎢⎣⎡= + RT

FakRT

Fakj -1-exp - exp -1 HHH-

HHHH 2

ηαθ

ηαθ (11)

where η is the hydrogen overpotential, θ is the surface coverage by Had, ki’s and k-i’s (i = T,V, H) are rate constants of the forward and backward reactions, respectively; αi are the


transfer coefficients and the a’s with subcripts mean the thermodynamic activities of theindicated 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 partialcathodic current density of the same reaction. Tafel, Volmer and Heyrovsky exchange currentdensities are given by equations (12), (13) and (14), respectively:

( ) 2T-H

2TT0, -1 ) (

2θθθ kakj == (12)

( ) ( )=⎥⎦

⎤⎢⎣⎡=⎥⎦

⎤⎢⎣⎡= + RT

Fak

RTF

kj -1-

exp -1

exp ) ( VHV-

VVV0,

ηαθ

ηαθθ

( ) ααααα θθ +=H

-1V-

-1V -1 akk (13)

( ) ( )⎥⎦⎤

⎢⎣⎡=⎥⎦

⎤⎢⎣⎡= + RT

FakRT

Fakj -1-exp exp -1 ) ( HHH-

HHHH0, 2

ηαθ

ηαθθ

( ) αααααα θθ +

−=H

-1H

1H-

-1H 2

-1 aakk (14)

where i0,j (θ ) (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:

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎟⎠

⎞⎜⎜⎝

⎛= ∗

2

0H

H2

0T0,T -

-1-1

2

2

θθ

θθ

pp

jj (15)

( )⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛= ∗

+

+

RTF-

aa

RTF

jj -1

exp -1-1-

exp V

H

H

0

V

0V0,V

ηαθθηα

θθ

(16)

( )⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛= ∗∗

+

+

RTF-

aa

RTF

pp

jj -1exp - exp -1-1 H

H

H

0

H

H

H

0H0,H

2

2 ηαθθηα

θθ

(17)

where, i0,j (i = T, V, H) are the exchange current densities corresponding to 2Ha = 1 (

2Hp ≈

1 atm) at the equilibrium coverage θ0, and represent a measure of the reactivity of theelectrode material for the Tafel ,Volmer or Heyrovsky reactions. The superscript ∗ indicatebulk values. Here, activities for H2 were replaced by partial pressures, approaching the gasbehavior to ideality.

The electrode works as a H2 electrode at 2Hp = 1 atm if either T0,j , V0,j ≥ 10 Rj or

V0,j , H0,j ≥ 10 Rj , Rj being the total residual current density due to oxidation and


reduction of impurities, to anodic metal dissolution and oxide formation, and to cathodicreduction of oxygen.

4.1. Tafel-Volmer Mechanism

The T-V mechanism is involved when T0,j > 10 H0,j and the steady-state condition of

hydrogen oxidation (or evolution) is given by

j = Tj = Vj (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 particularexperimental conditions. The following cases can be considered:

i) When T0,j , V0,j ≥ 10 j , the overpotential is caused only by transport processes

(concentration changes of H2 and H+ in the diffusion layer) and it is independent of theelectrode material.

The dependence of the hydrogen overpotential upon current density is given by,

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛+⎟

⎟⎠

⎞⎜⎜⎝

⎛=

+Hd,Hd,

-1ln -1ln 2

- 2

jj

FRT

jj

FRTη (19)

In equation (19) 2Hd,j and +Hd,j are the limiting current densities due to convective

diffusion:

∗=2

2

2

2 HH

HHd,

2c

FDj

δ (20a)

∗+

+

+

+ = HH

HHd,

- cFD

jδ

(20b)

where Di’s are the diffusion coefficients and δi’s are the thicknesses of the diffusion layers ofthe indicated species.

If the hydrogen diffusion alone determines the j–η curve, then T0,j , V0,j >> 2Hd,j , and

taking into account that at 25 °C the second term in equation (19) is negligible for solutionswith values of ∗

+Hc ≥ 0.1 M, one obtains

2Hd,j j≈ at η = 0.06 V.


ii) In the case T0,j ≥ 10 V0,j ; 0.1 T0,j > j > 10 V0,j , the Volmer reaction is the rate-

determining step, so that no limiting current density is expected, and the Tafel reaction is atquasi-equilibrium. Then, equation (16) for anodic current density yields:

⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛=

RTF

jj

exp V

0V0,V

ηαθθ

(21)

The following Tafel relation is obtained when the ratio of coverages in equation (21)remains close to the unity in the studied range:

VVV0, logln 2ln 2 jbajFRTj

FRT

×+=+=η (22)

The slope of the straight line is 2 RT / F ( Vα = ½). At 25 °C it has a value of 0.0513 V or0.1183 V dec-1 (taking R = 8.314 J K mol-1 and F = 96,485.3 C).

iii) In the case that V0,j ≥ 10 T0,j ; 0.1 V0,j > j > 10 T0,j , the Tafel reaction is a pure

chemical reaction. Considering that all partial reactions that occur before and after the Tafelreaction are fast, the following expression for the hydrogen overpotential is established [46]:

⎥⎥⎦

⎤

⎢⎢⎣

⎡=

Tl,

-1ln2

-jj

FRTη (23)

where Tl,j 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 ( dη )

appears. When the Tafel reaction is also the rate-determining step, the overall potential willhave then another contribution, Tη , coming from equation (23) and the following equationcan be derived:

⎥⎦

⎤⎢⎣

⎡=+

lTd -1ln

2-

jj

FRTηη (24)

where Tl,Hd,l 1112

jjj += .

The limiting current density for H2 oxidation, determined by the Tafel reaction, Tl,j , is

proportional to the concentration of hydrogen. Since, ∗∝22 HHd, cj (see equation (20)), lj is


expected to be always proportional to hydrogen concentration, independently of whether thediffusion or the Tafel reaction is the rate-determining step.

On the other hand, for the equilibrium of the Volmer reaction, one has:

⎟⎟⎠

⎞⎜⎜⎝

⎛= ∗

+

+

H

H

0

0

-1-1ln

aa

FRT

θθ

θθη (25a)

⎥⎦⎤

⎢⎣⎡

−−

= ∗+

+

ηθθ

θθ

RTF

aa

-exp11

H

H

00

(25b)

Combining equation (15) with equation (25a), and assuming that polarization due toprotons concentration is insignificant, 1

HH≈∗

++ aa , which is equivalent to consider that

only the concentration of molecular hydrogen decreases in the plane of reaction with respectto the bulk concentration, one can deduce that:

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

⎟⎟⎠

⎞⎜⎜⎝

⎛= ∗ 2

0T0,

H

H

-1-1

-ln2

-2

2

θθ

η

j

jpp

FRT

(26)

From equation (26) a Tafel line with slope b = RT / 2F equal to 0.0128 V or 0.0296 Vdec-1 at 25 °C is expected for small values of θ.

On the other hand, if equation (25b) is put into equation (15) and the overpotential isassumed to be a large positive number (tending towards the limiting current density), thefollowing equation can be obtained:

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛= ∗

2

2

H

H2

0T0,T

-1-1

pp

jjθθ

(27)

When, 1-1-1 0 ≈θθ , equation (27) predicts a maximum anodic limiting currentdensity independent of stirring, given by:

∗=2

2

H

HT0,Tl, p

pjj (28)

In contrast, when θ << 1, the maximum anodic limiting current density will be:


( ) ∗=2

2

H

H2

0

T0,Tl, -1 p

pjj

θ (29)

Equations (28) and (29) hold for 2Hd,j → ∞, being expected that 1

22 HH =∗pp in this

case [47].If the Tafel reaction alone is the rate-controlling step, then,

2Hd,j >> Tl,j for 0θ ≤ 0.1

and Tl,l jj = is reached at V 0.05≤η , whereas for 0θ = 0.9, anl,j is attained at

V 0.120≈η [46]. Only in these conditions, the comparison of theory and experiment issimple, 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 isgiven by VT ηηη += , that is, it contains reaction and charge-transfer contributions. Theoverall overpotential can be expressed through the following generalized equation, whichtakes into account the effect of coverage, θ [49]:

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜

⎝

⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎟⎠

⎞⎜⎜⎝

⎛+

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

⎟⎟⎠

⎞⎜⎜⎝

⎛=

41-

2

0T0,

21-

0V0,

1-2

0T0, -1

-1-1

-1-1

2sinh2

-1-1

-1ln2

-

θθ

η

j

jθθ

jj

FRT

θθj

jF

RT (30)

Figure 5. Theoretical anodic hydrogen overpotential for the Tafel-Volmer mechanism, with θ <<1 and

α = 1/2, according to equation (30) at different V0,T0, jj ratios: (a) 0.1, (b) 1 and (c) 10.

Temperature 298 K.


Fig. 5 shows theoretical T0,jj vs. η curves for several V0,T0, jj ratios according to

equation (30), with 0≅θ and α = 0.5. As can be seen, curve a represents the contributionof the reaction overpotential ( Tη ), whereas the remaining part corresponds to the charge

transfer overpotential ( Vη ).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

( ) ( )⎥⎥⎦

⎤

⎢⎢⎣

⎡

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛+= ηαα

ηRTF

jjj

FRTj

FRT -exp--1ln-ln

ln

-

1/2

l0 (31)

Equation (31) is valid for θ << 1 and represents a Tafel equation with a correction term,which takes into account the lowering of concentration due to the rate control by diffusionand reaction, and the reverse reaction [44]. The slope and the extrapolation to 0=η of thestraight lines obtained from equation (31) allows the determination of the charge transfercoefficient, α , and the exchange current density, 0j , respectively, without diffusion andreaction rate control.

iv) When T0,j , V0,j < Rj , the potential difference metal/solution is determined by two

different electrode reactions, which occur simultaneously. The anodic current density of onereaction 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.

4.2. Volmer- Heyrovsky Mechanism

The V-H mechanism is involved if H0,j > 10 T0,j and the steady-state condition for

hydrogen oxidation is given by

j = Vj = Hj (32)

Under these conditions, the following approaches can be considered:

v) In the case that V0,j , H0,j > 0.1 j , the situation is identical to case i) of Tafel-

Volmer mechanism.

vi) When H0,j ≥ 10 V0,j ; 0.1 H0,j > j >10 V0,j , the Volmer reaction is the rate-

determinig step and the following equation is obtained:


⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛=

RTF

jj

exp 2 V

0V0,V

ηαθθ

(33)

vii) For V0,j ≥ 10 H0,j ; 0.1 V0,j > j >10 H0,j , 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:

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛= ∗ RT

Fpp

jj exp -1-1 H

H

H

0H0,H

2

2 ηαθθ

(34)

Under these conditions, the anodic current becomes negligible at overpotentials where theratio of pressures for molecular hydrogen decreases up tpclose to 0.

viii) When V0,j , H0,j < Rj , 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-Vmechanisms can only be discriminated in the case that Tafel reaction is the rate-determiningstep [49, 50].

5. Hydrogen Oxidation Reaction Studies on Bulk and ModelElectrodes

HOR studies on smooth noble metal surfaces in acidic media show that platinum is the mostactive. The reaction mechanism on bulk polycrystalline platinum electrode is usually assumedto proceed through a Tafel–Volmer mechanism via reactions (4) and (5), the dissociativeadsorption of hydrogen being the rate-determining step (r.d.s.) [51]. In alkaline electrolytes,HOR on polycrystalline Pt follows a Tafel–Volmer sequence and the r.d.s. is the dissociativehydrogen adsorption given by reactions (4) and (6) [52].

Since HOR is a fast reaction on platinum, diffusion effects lead to difficulties inobtaining kinetic parameters of the process such as Tafel slopes, even at low overpotentials,which make necessary the use of rotating or high surface area electrodes.

In the Rotating Disk Electrode (RDE) technique, the current-potential curves on smoothplatinum exhibit an anodic limiting current density, which depends on rotation rate in bothacidic and alkaline media [46]. These plots are well described by equation (19), which holdsfor a diffusion overpotential alone. Similar relationships have been observed in acidicsolutions for Ir, Rh, and Pd, and well-characterized Pt-Ru, Pt-Rh, Pt-Sn [53], and Pt-Au [51]alloys, and also for Ni in alkaline solutions. In the case of platinum, a evolution of thelimiting diffusion current density to a limiting reaction current density ( Tl,j ), independent of

rotation rate, is observed as a consequence of the rate-determining H2 adsorption.


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

274 K 333 KPt(hkl)acid alkaline acid alkaline

Pt(111) 0.21 0.01 0.83 0.3Pt(110) 0.65 0.125 1.35 0.675Pt(100) 0.36 0.05 0.76 -

aObtained from the micropolarization region

HOR at low-index Pt single-crystal surfaces, Pt(hkl), is a structure sensitive reaction,since its kinetics is a function of the cristallographic symmetry of the Pt surface [54]. Usingthe RDE technique, Marković et al. [55] pointed out the differences between the exchangecurrent densities, Tafel slopes and apparent activation energies for Pt(111), Pt(100), andPt(110) at different temperatures in acidic and in alkaline media. These results collected inTable 2 show that the kinetics of HOR on Pt(hkl) surfaces in alkaline solutions is slower thanin acidic solutions, increasing the absolute kinetic activities for Pt(hkl) in the sequencePt(111) < Pt(100) < Pt(110). In addition, from mass-transfer corrected j-E curves, single Tafelslopes of 2.3RT/2F and 2.3RT/F were found for Pt(110) and Pt(111), respectively, at 274 K inacidic electrolyte. This allows concluding that on Pt(110) the reaction follows the Tafel(r.d.s.)–Volmer mechanism, whereas on Pt(111) it verifies the Heyrovsky (r.d.s.)–Volmermechanism. Nevertheless, in alkaline solutions Tafel slopes cannot be determinedunambigously.

The kinetics of HOR on platinum and platinum-based alloys have been typicallyexamined in aqueous solutions using the RDE technique. However, the mass transport and theelectrode kinetics are different in a PEFC environment because of the presence of a solidpolymer electrolyte. Some studies have been carried out on polymer-covered (e.g. Nafion®)bulk and model Pt electrodes with and without supporting electrolyte [51, 56]. The resultsshow that dissociative adsorption of H2 by equation (4), is the rate-determining step in acidicsolutions. Since Nafion® also behaves as an acid, the kinetic analogy between the Pt|acidicsolution and Pt|Nafion® interfaces is reasonable.

A dual-pathway kinetic equation has been proposed to describe the HOR current behavioron Pt over the whole relevant overpotential region [57]. This equation is based on the Tafel–Volmer–Heyrovsky mechanism and has been applied to analyse the polarisation curvesmeasured with platinum microelectrodes and RDE, as well as, with high surface area catalystsoperating under PEFCs conditions.

6. Hydrogen Oxidation Reaction Studies on Technical Electrodes

Electrochemical studies of technical electrodes can be carried out in liquid media using astandard three-electrode cell by means of the RDE technique [58] and also with stationaryelectrolyte. In the former case, the electrolyte is saturated with hydrogen and in the latter one,the electrode is mounted in a holder where H2 gas is fed through the back of the electrode


[59]. When using a solid electrolyte such as an ionic exchange membrane is used, the study isthen performed in a fuel cell environment.

The hydrogen oxidation kinetics on technical carbon-supported Pt-Ru catalysts wasinvestigated using the thin-film rotating electrode technique [60]. The electrode was preparedplacing a certain volume of catalyst suspension onto a glassy carbon rotating electrode. In thiscase, the kinetic parameters were determined by nonlinear fitting of the polarisation curve inthe mixed kinetic- and diffusion-controlled region, assuming a Tafel slope of RT / 2F, whichcorresponds to a Tafel–Volmer mechanism with the dissociative adsorption of hydrogen asthe r.d.s.

Vogel et al. [61] studied the electrochemical oxidation of hydrogen on carbon-supportedPt electrodes in acidic electrolyte. The electrodes were made by pressing the catalyst with aminimum amount of PTFE into a small gold disk.

Experimental current density vs. overpotential curves in hydrogen-saturated H3PO4 wereanalysed in terms of theoretical equations derived from the model of porous floodedelectrodes considering a Tafel–Volmer 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 toelectrodes with relatively thick catalyst layers, high concentration of active catalyst, and highexchange current density and ii) equation (36) related to a pure reaction control ocurring inthe case of a very thin catalyst layer, low concentration of active catalyst or a slow reactionrate:

( )⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠⎞

⎜⎝⎛

⎟⎠⎞

⎜⎝⎛

=

RT2F-exp-1-1

2-exp-1-1

0

0l ηθ

η

θ RTF

jj

(35)

and

( ) 2

0

0l

RT2F-exp-1-1

2-exp-1-1

⎥⎦

⎤⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠⎞

⎜⎝⎛

⎟⎠⎞

⎜⎝⎛

=ηθ

η

θ RTF

jj

(36)

Figure 6 shows the comparative shapes for the theoretically normalised η-j curvesdecribed by equations (35) and (36). As can be seen, the shape of such curves normalised bythe limiting current density, jl, is function only of the hydrogen equilibrium coverage, 0θ . Forthe carbon-supported Pt electrodes used in the work of Vogel et al. [61], the experimentaldata fitted better to equation (35), suggesting that the electrode is in a mixed region ofactivation and diffusion control. The poisoning effect of CO on HOR in the Pt electrodes wasalso studied by the same authors using similar equations to (35) and (36), but replacing 2H

0j ,


the exchange current density for hydrogen oxidation, by CO0j , the exchange current density

for hydrogen oxidation in the presence of CO, expressed as:

( )COH0

CO0 -12 θjj = (37)

where COθ is the coverage of platinum by CO. Results obtained from anodic polarisation forCO-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 thesequence proposed.

Figure 6. Comparison between the shapes of the normalised polarisation curves in the two limitingcases: (a) and (c) diffusion-reaction control; (b) and (d) reaction control. 0θ = 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 aPt/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 carbonpaper, being tested as a floating electrode in a half-cell assembly [62]. The porous platinumanode was mechanically floated on top of the acid electrolyte, while pure or mixed hydrogenwere supplied through the electrode surface opposite to the active layer This configurationminimizes diffusion resistances [63]. From the current density vs. overpotential plots for H2

and H2/CO mixtures, COθ , 2H0j and CO

0j , can be calculated and used to measure thecatalytic 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 activecatalyst for hydrogen oxidation in acidic solution, it is quickly deactivated by CO. Thisrepresents a trouble for PEFCs and PAFCs running with industrial hydrogen or reformatefeed streams. To solve this problem, some binary platinum alloys tolerant to CO weredeveloped [51].

An electrode kinetics for the evaluation of hydrogen oxidation on carbon-supportedbinary Pt-M alloys (M = Cr, Mn, Fe, Co and Ni) was conducted at a solid polymer electrolytemembrane interface [65]. The gas diffusion electrodes consisted of a carbon cloth substrate


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 ofequation (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 diffusionelectrodes catalysed by Pt/C, Pt-Ru/C and Pt-Mo/C in aqueous H2SO4. The substrate was acarbon paper and the GDL was composed of a mixture of carbon and PTFE. Electrochemicalmeasurements were carried out using a conventional three-electrode cell, where the GDE wasmounted 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 anodicgalvanostatic polarisation data, obtained with pure H2 at room temperature:

( )jfj RjREE ×++= Ω exp0pocp (38)

where Eocp is the anode open circuit potential (OCP); RΩ is the ohmic resistance; 0pR is the

polarisation resistance at OCP, and f is a mass transfer related factor ( limiting1 jf ∝ ). RΩ and0pR were obtained from electrochemical impedance spectroscopy measurements.

The analysis of the 0pR and jlimiting values, which can be ascribed to H2 oxidation on

catalytic sites, revealed to the authors some characteristics of the mechanism of hydrogenoxidation on Pt-Ru/C and Pt-Mo/C catalysts. In particular, the decay of limiting currentdensity in these catalysts was partially related to the change of the HOR rate constant. Then,they proposed the Heyrovsky–Volmer mechanism given by equations (5) and (7), wasproposed for HOR in Pt-Ru and Pt-Mo supported electrodes. This proposal was reinforced bythe same authors investigating the H2 oxidation mechanism on Pt and Pt-Ru electrocatalystssupported on a high surface area carbon powder, by electrochemical impedance spectroscopymeasurements at OCP, using the same experimental setup described above. In the light of theanalyisis 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-Volmermechanism was proposed for Pt-Ru/C [67].

Janssen et al. [68] investigated the mechanism of hydrogen oxidation on a commercialfuel cell grade electrode on Toray paper with a platinum loading of 0.50 mg cm-2 [68]. Thesolution used was 0.5 mol dm-3 H2SO4(aq) at 25 °C. Experimental j-η curves in the lowoverpotential region up to 200 mV were obtained for pure and diluted H2. By comparingexperimental and theoretical relations, they concluded that HOR occurs according to theTafel–Volmer mechanism. To obtain the kinetic parameters of HOR, they further applied anequation similar to (31), the corrected Tafel slope depending on T0,V0, αα , T0,ld, αα and

0θ , assuming that the gas compartment behaves like a continously fed and stirred tankreactor [69]. They found that the reactivity of the electrode, which depends on theelectrochemical pretreatment applied, had a large effect on the kinetic parameters of H2

oxidation. The transfer coefficient of the Volmer reaction, Vα , was ½, with 1T0,V0, ≤αα ,

and the limiting current was determined by hydrogen diffusion for a very reactive gas


diffusion electrode. In contrast, the Tafel reaction was rate determining for a low reactive gasdiffusion electrode, whereas Vα increased and T0,V0, αα decreased with lowering

reactivity of the gas diffusion electrode. The effect of hydrogen concentration (partialpressure 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 polymerelectrolyte 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 sandwichedbetween 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 referenceelectrode. The analysis of the anodic polarisation scans together with the dependence of theexchange current density on the partial pressure of hydrogen allowed concluding that the Pt-Ru catalysed anode follows the Tafel–Volmer 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 Tafel–Volmermechanism was proposed to quantitatively explain the current vs. potential curves obtained ina variety of experimental conditions:

( ) anl

anan -1ln -ln OCP jR

jj

CjBAE Ω+⎟⎟⎠

⎞⎜⎜⎝

⎛++= (39)

where OCP is the open circuit potential; A and B are related to the charge transfer; C isrelated to diffusion or reaction overpotential, and ΩR is an ohmic-type resistance. Equation(39) accounts for diffusion, adsorption, inhibition, charge transfer and ohmic-type drop in thewetted electrode pores.

In addition, the effect of H2 partial pressure was modelized modifying equation (35) forhigh overpotentials as follows:

an'

anl

an

0

0 -

ln -1

ln - jRjj

jF

RTF

RTΩ+⎟⎟

⎠

⎞⎜⎜⎝

⎛⎟⎠⎞

⎜⎝⎛+⎟⎟

⎠

⎞⎜⎜⎝

⎛⎟⎠⎞

⎜⎝⎛=

θθ

η (40)

where 0θ and 'ΩR are the only fitting parameters.

7. Comparison between Model and Technical Electrodes

The electrocatalytic studies performed with model electrodes permit a more direct assignmentof the kinetic results than measurements made with gas diffusion electrodes in practical fuelcells, even using a half-cell mode (three electrode cell experiments).

Model electrodes are bulk, single-crystal or polycrystalline. Electrochemicalmeasurements usually are carried out under defined mass transport conditions in liquid media,as in the RDE technique configuration. In this scenario, model electrodes are free from


complications due to water management and mass transport limitations for fuel transport, asituation very far from the realistic operating conditions in a fuel cell, and it may lead to someloss of information.

The transition to more realistic conditions, near fuel cell environment, is achieved using athin-film rotating disk electrode with a practical supported catalyst. It allows obtaining kineticparameters such as the the exchange current density and Tafel slopes, avoiding the use ofcomplex models. This is of fundamental importance, because small catalyst particles having asignificant 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 orsolid electrolyte, but in this case, the complexity is related to the evaluation of diffusion andohmic overpotentials, closely connected to the charge transfer kinetics. However, choosingappropiate models, it is possible to calculate different contributions from the simulation ofcurrent vs. potential data.

HOR is so fast that it becomes very difficult to determine experimentally the kineticparameters 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 inmodel and technical electrodes, the r.d.s. being the dissociative adsorption of H2 (Tafelreaction). On the other hand, significant differences in activity are reported because of thestructure, 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 reportedto be equal to 1.9 mA cm-2 [51]. On highly dispersed platinum supported on carbon, it couldbe one order of magnitude greater due to its bigger specific area and particular surfacestructure of the catalytic particles. In alkaline media, j0 reported for smooth and highlydispersed 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 andhighly dispersed Pt in alkaline electrolyte.

From the above considerations, it seems that the extrapolation of the results obtained withmodel 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 totechnical diffusion electrodes in fuel cells.

8. Conclusion

The main preparation methods for H2 technical electrodes for low temperature fuel cells havebeen examined. It has been demonstrated that the electrochemical behavior of the electrodesdepends on their fabrication, thus affecting the fuel cell operation. The preparation of thecatalyst of the active layer also influences its physical properties and electrochemicalperformance. Different electrochemical approaches to study HOR on model, as a firstapproximation, and technical electrodes, are exhaustively analyzed and their kineticparameters are discussed to evaluate their performance and system modelling. The existenceof a gap between the knowledge obtained from studies on model electrodes and technicalelectrodes is emphasized. To optimize the performance of practical fuel cell electrodes, thepreparation of high surface area catalysts with the same characteristics as those shown at theatomic level then seems necessary. In this sense, mechanistic studies are fundamental to


provide kinetics parameters, which can help us to design catalysts to operate in a real fuel cellenvironment.

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 Nº.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] Gouérec, 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 Nº. 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.; Spinacé 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.; Fueβ, H. J. Appl. Electrochem. 2005, 35, 85-90.[33] Dicks, A. L.; J. Power Sources. 2006, 156, 128-141.


[34] McCreery, R. L. In Interfacial Electrochemistry, Theory, Experiments, andApplications; Wieckowski, A.; Ed.; Mercel Dekker, Inc.: New York (NY), 1999;pp 631-647

[35] Kinoshita, K. Carbon, electrochemical and physicochemical properties; John Wiley &Sons: New York (NY), 1988; pp 388-468.

[36] Landsman, D. A.; Luczak, F. J. In Handbook of Fuel Cells. Fundamentals Technologyand Applications; Vielstich, W.; Lamm, A.; Gasteiger, H. A.; Eds.; John Wiley & Sons:Chichester (UK), 2003; Vol. 4, pp 811-831.

[37] Gulzow, E. J. Power Sources. 1996, 61, 99-104.[38] Appleby, A. J.; Foulkes, F. R.; Fuel Cell Handbook; Van Nostrand Reinhold: New York

(NY), 1989; pp 313-335.[39] Cifrain M.; Kordesh K. J. Power Sources. 2004, 127, 234-242.[40] Antolini, E. J. Mat. Sci. 2003, 38, 2995-3005.[41] Gu T.; Lee W. K; Van Zee J. W. Appl. Catal. B. 2005, 56, 43-49.[42] de Bruijn F. A; Papageorgopoulos D. C.; Sitters E. F.; Janssen G. J. Power Sources.

2002, 110, 117-124.[43] Wee, J-H.; Lee K-Y. J. Power Sources. 2006, 157, 128-135.[44] Pereira, L G. S.; dos Santos, F. R.; Pereira, M. E.; Paganin, V. A. Ticianelli, E. A.

Electrochim. Acta, 2006, 51, 4061-4066.[45] Breiter, M. W. In Handbook of Fuel Cells. Fundamentals Technology and Applications;

Vielstich, W.; Lamm, A.; Gasteiger, H. A.; Eds.; John Wiley & Sons: Chichester (UK),2003; Vol. 2, pp 362-367.

[46] Vetter, K. J., Electrochemical Kinetics; Academic Press: New York (NY), 1967;pp 516-554.

[47] Gennero de Chialvo, M. R.; Chialvo, A. C. Phys. Chem. Chem. Phys. 2004, 6,4009-4017.

[48] Vielstich, W. Fuel Cells; John Wiley & Sons: London (UK), 1970; pp 47-50.[49] Breiter, M.; Clamroth, R., Z. Elektrochem. 1954, 58, 493-505.[50] Austin, L. G. In Handbook of Fuel Cell Technology; Berger C.; Ed.; Prentice-Hall, Inc:

Englewood Cliffs, NJ, 1968; pp 92-112.[51] Conway, B. E.; Tilak, B. V. Electrochim. Acta. 2002, 47, 3571-3594.[52] Bagotzky, V. S.; Osterova, V. J. Electroanal. Chem. 1973, 43, 233-249.[53] Gasteiger, H. A.; Marković, N. M.; Ross, Jr. P. N. J. Phys. Chem. 1995, 99, 8290-8301.[54] Schmidt, T. J.; Ross, P. N.; Marković, N. M. J. Electroanal. Chem. 2002, 524-525,

252-260.[55] Marković, N. M.; Sarraf, S. T.; Gasteiger, H. A.; Ross, Jr. P. N. J. Chem. Soc. Faraday

Trans. 1996, 92, 3719-3725.[56] Jiang, J.; Kucernak, A. J. Electroanal. Chem. 2004, 567, 123-137.[57] Wang, J. X.; Springer, T. E; Adžić, R. R. J. Electrochem. Soc. 2006, 153, A1732-

A1740.[58] Schmidt, T. J.; Gasteiger, H. A. In Handbook of Fuel Cells. Fundamentals Technology

and Applications; Vielstich, W.; Lamm, A.; Gasteiger, H. A.; Eds.; John Wiley & Sons:Chichester (UK), 2003; Vol. 2, pp 316-333.

[59] Srinivasan, S. Fuel Cells. From Fundamentals to Applications; Springer: New York(NY), 2006; pp 267-285.


[60] Wang, J. X.; Brankovic, S. R., Zhu, Y.; Hanson, J. C.; Adžić, 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.; Turtù, 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.


Chapter 7

PEM FUEL CELL MODELING

Maher A.R. Sadiq Al-Baghdadi*

Department of Mechanical Engineering, International Technological University,115 Dollis Hill Lane, London NW2 6HS, UK

Abstract

Fuel Cells are growing in importance as sources of sustainable energy and will doubtlessform part of the changing program of energy resources in the future. Two key issues limitingthe widespread commercialization of fuel cell technology are better performance and lowercost. The development of physically representative models that allow reliable simulation ofthe processes under realistic conditions is essential to the development and optimization offuel cells, the introduction of cheaper materials and fabrication techniques, and the design anddevelopment of novel architectures. The difficult experimental environment of fuel cellsystems has stimulated efforts to develop models that could simulate and predict multi-dimensional coupled transport of reactants, heat and charged species using computationalfluid dynamic (CFD) methods. The strength of the CFD numerical approach is in providingdetailed insight into the various transport mechanisms and their interaction, and in thepossibility of performing parameters sensitivity analyses.

Among all kinds of fuel cells, proton exchange membrane (PEM) fuel cells are compactand lightweight, work at low temperatures with a high output power density and lowenvironmental impact, and offer superior system startup and shutdown performance. Theseadvantages have sparked development efforts in various quarters of industry to open up newfield of applications for PEM fuel cells, including transportation power supplies, compactcogeneration stationary power supplies, portable power supplies, and emergency and disasterbackup power supplies.

This chapter of "PEM Fuel Cell Modeling" looks at how engineers can model PEM fuelcells to get optimal results for any application. A review of recent literature on PEM fuel cellmodeling was presented. A full three-dimensional, non-isothermal CFD model of a PEM fuelcell with straight flow field channels has been developed in this chapter. The model wasdeveloped to improve fundamental understanding of transport phenomena in PEM fuel cellsand to investigate the impact of various operation parameters on performance. Thiscomprehensive 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: [email protected]. Present address: Currently at the Higher Institute of Mechanical

Engineering, Head of Mechanical & Energy Engineering Department, Yefren, P.O. Box 65943, Libya.

Maher A.R. Sadiq Al-Baghdadi274

mechanism of water, and potential fields. In addition, the hygro and thermal stresses inmembrane, which developed during the cell operation, were modeled and investigated. Thenew feature of the algorithm developed in this model is its capability for accurate calculationof the local activation overpotentials, which in turn results in improved prediction of the localcurrent density distribution. Fully three-dimensional results of the velocity flow field, speciesprofiles, liquid water saturation, temperature distribution, potential distribution, water contentin the membrane, stresses distribution in the membrane, and local current density distributionare presented and analyzed with a focus on the physical insight and fundamentalunderstanding. The model is shown to be able to understand the many interacting, complexelectrochemical, and transport phenomena that cannot be studied experimentally. This chapteris 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 providedduring 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 NovaScience Publishers, for his invitation to the publishing program "Fuel Cell Research andDevelopment."

My gratitude and appreciation is due to my wife for her patience, care, and supportduring the period of preparing this work.

Most of all, I want to thank my parents for their unconditional support throughout all theyears of my education. This study would not have been possible without them. Many thanksalso are due to my brother and sister.

Thanks to all who have helped me in carrying out this research.

Maher A.R. Sadiq AL-Baghdadi

Nomenclature

Symbol Description Unitsa Water activity -

MEAA Area of the MEA 2m

chA Cross sectional area of flow channel 2m

Wc Water content -

fC Fixed charge concentration 3. −mmol

2HC Local hydrogen concentration 3. −mmolrefHC

2Reference hydrogen concentration 3. −mmol

PEM Fuel Cell Modeling 275

Symbol Description Units

2OC Local oxygen concentration 3. −mmolrefOC

2Reference oxygen concentration 3. −mmol

Cp Specific heat capacity 11.. −− KkgJD Diffusion coefficients 12 . −sm

+HD Protonic diffusion coefficient 12 . −sm

dropD Diameter of droplet water m

E Reversible cell potential volts

cellE Cell operating potential volts

fcE Thermodynamic efficiency of the cell -

F Faraday's constant 1.96487 −molCG Gibb's free energy Jg Specific Gibb's free energy 1. −molJh Specific enthalpy 1. −molJI Cell operating (nominal) current density 2. −mA

ai Anode local current density 2. −mA

ci Cathode local current density 2. −mAref

coi ,Anode reference exchange current density 2. −mA

refaoi ,

Cathode reference exchange current density 2. −mA

aLi ,Anode local limiting current density 2. −mA

cLi ,Cathode local limiting current density 2. −mA

k Gases thermal conductivity 11.. −− KmW

effk Effective electrode thermal conductivity 11.. −− KmW

grk Graphite thermal conductivity 11.. −− KmW

memk Membrane thermal conductivity 11.. −− KmW

xmk Mass transfer coefficient 12 .. −− smmolKp Hydraulic permeability 2m

2HLHV Lower heating value of hydrogen 1. −kgJ

phasem Mass transfer in the form of: evaporation

evapphase mm = and condensation condphase mm =1. −skg



M Molecular weight of mixture gases 1. −molkg

2HM Molecular weight of hydrogen 1. −molkg

OHM2

Molecular weight of water 1. −molkg

2OM Molecular weight of oxygen 1. −molkg

WN Net water flux across the membrane 12 .. −− smkgnd Electro-osmotic drag coefficient -

en Number of electrons transfer -

P Pressure Pa

cP Capillary pressure Pa

satP Water saturation pressure Pa

q Heat generation 2. −mW

gr Volume fraction of the gas phase -

lr Volume fraction of the liquid phase -

R Universal gas constant 11..314.8 −− KmolJs Specific entropy 11.. −− KmolJsat Saturation -

T Temperature Ku Velocity vector 1. −sm

∞v Free-stream velocity 1. −sm

cellW Cell power density 2. −mW

ix Molar fraction -

iy Mass fraction -

fz Fixed-site charge -

Greek Symbols Description


aα Charge transfer coefficient, anode side -

cα Charge transfer coefficient, cathode side -

β Modified heat transfer coefficient 13 .. −− KmW



τ Chemical potential 1. −molJ

τ Electrochemical potential 1. −molJ

2Hγ Hydrogen concentration parameter -

2Oγ Oxygen concentration parameter -

evapHΔ Enthalpy of evaporation 1. −kgJSΔ Entropy change of cathode side reaction 11 −− KmoleJ

CLδ Catalyst layer thickness m

GDLδ Gas diffusion layer thickness m

memδ Membrane thickness mε Porosity -

ξ Stoichiometric flow ratio -

η Overpotential volts

eλ Electrode electronic conductivity 1. −mS

mλ Membrane ionic conductivity 1. −mSμ Viscosity 11.. −− smkgρ Density 3. −mkg

GDLφ Electric potential inside the gas diffusion layer volts

memφ Electric potential inside the membrane voltsℜ Relative humidity of inlet fuel and air %ψ Inlet Oxygen/Nitrogen ratio -

ϑ Local relative humidity of the gas phase -

σ Surface Tension 1. −mNπ Strain -

℘ Thermal expansion coefficient -

Swelling expansion coefficient -

Ψ Young's modulus Pa

Ω Stress Pa


Abbreviations Definition

CFD Computational Fluid DynamicsCL Catalyst LayerGDL Gas Diffusion LayerMEA Membrane Electrode AssemblyMEM Membrane

Subscript Definition

a Anodec Cathodeg Gas phasel Liquid phasew Water

1. Introduction

1.1. Background

Fuel Cells are electrochemical devices that directly convert the chemical energy of a fuel intoelectricity. In general, fuel cells offer many advantages over conventional energy conversiondevices. 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 inFigure 1.1. The figure provides a comparison between the exergy efficiencies for electricitygeneration by the principal energy conversion system in use today [2]. The solid barsrepresent what currently is possible. The figure demonstrates that the exergy efficiency of fuelcell systems compares favorably to all other energy conversion systems using hydrocarbonsfor fuel. In fact, fuel cell systems have the highest overall average efficiency of all systemsexcept hydroelectric plants. The non-solid extensions above each efficiency column indicatethe theoretical improvements that are predicted for each system. With these predictedimprovements in fuel cell system efficiency, fuel cell systems would gain an even largerefficiency advantage over other systems.

Figure 1.1 also indicates fuel cell systems can be used to generate power over a largerange 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 stillbeing efficient at lower power requirements. This offers a distinct advantage over internalcombustion (IC) engines, whose efficiencies drop off drastically the further they operate fromtheir peak (i.e. design) power point [3]. In fact, the vast majority of time, a vehicle, whoseprimary power source is an IC engine, operates far below peak power. This is also true ofmost other energy conversion systems, depending of course, on the application.


Fuel cell systems also are easily scaled, allowing them to be used for small applicationssuch as the power source for a personal computer as well as large applications like astationary power plant. Large or small, all these applications harness the fuel cells inherentefficiency advantages to make better use of a fuel's chemical energy.

As the world becomes more industrialized, a second key advantage of fuel cells, loweremissions, becomes essential for controlling global pollution. With fuel cells, the conversionof chemical energy to electrical energy is accomplished electrochemically only water iscreated. Although some emissions do occur during the fuel reforming process, temperaturesare not high enough for NOx, sulfur components are removed prior to reforming so that SOxformation does not occur, and much less CO2 is released due to the higher averageefficiencies of fuel cell systems. In addition, CO concentrations are too small by necessity andun-reacted hydrocarbons are re-circulated in the system [2, 4]. Furthermore since there are nomoving parts in a fuel cell system, such systems operate much more quietly, resulting in lessnoise 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 stationaryresidential power applications [5].

Figure 1.1. Exergy efficiencies of the principal types of energy conversion systems [2].

1.2. Proton Exchange Membrane Fuel Cells

According to their electrolytes, fuel cells are classified into four types; PEMFC (ProtonExchange Membrane Fuel Cell), PAFC (Phosphoric Acid Fuel Cell), MCFC (Molten


Carbonate Fuel Cell), and SOFC (Solid Oxide Fuel Cell). An exception to this classificationis the DMFC (Direct Methanol Fuel Cell) which is a fuel cell in which methanol is directlyfed to the anode. The electrolyte of this cell is not determining for the class. Table 1.1compares the different types of fuel cell systems [2, 5-8]. A schematic representation of a fuelcell with reactant and product, and ions flow directions for these types of fuel cells are shownin Figure 1.2 [6].

From the Table 1.1 can conclude that the PEMFC is the only fuel cell that excels in allthe characteristics essential for private vehicle applications. Additional advantages that thePEMFC offers over some of the other fuel cells are that the PEMFC is a less complicatedsystem 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 lowtemperatures with a high output power density, and offer superior system startup andshutdown performance [9, 10]. These advantages have sparked development efforts in variousquarters of industry to open up new field of applications for PEM fuel cells, includingtransportation power supplies, compact cogeneration stationary power supplies, portablepower supplies, and emergency and disaster backup power supplies [10, 11].

Figure 1.2. Operating principle of various types of fuel cells.


It is believed that there will be a time in the future when global energy demands will bemet by some source other than fossil fuels. It is believed that hydrogen will play a major rolein such a future [11]. The concept of a hydrogen economy describes an economy where theprincipal source of energy is hydrogen related. Fuel cells, in particular proton exchangemembrane fuel cells (PEMFC), are expected to play a major role in a future hydrogeneconomy [12]. Fuel cells are particularly attractive for use in vehicles as a replacement to thecombustion engine [13]. The low operating temperatures of a PEMFC allows for easy start upand quick response to changes in load and operating conditions.

Table 1.1. The different types of fuel cells.

AFC PEMFC DMFC PAFC MCFC SOFC

Electrolyte NaOH /KOH Polymer membrane H3PO4

LiCO3–K2CO3

ZrO2 withY2O3

ElectrodeMaterial

Metal orcarbon Pt-on-carbon Ni +Cr Ni/Y2O3–

ZrO2

OperatingTemp. (Co) 60-100 50-100 50-200 160-210 650-800 800-1000

Power Density(kW/m2) 0.7 – 8.1 3.8 – 6.5 > 1.5 0.8 – 1.9 0.1 – 1.5 1.5 – 2.6

PracticalEfficiency (%) 60 60 60 55 55-65 60-65

Applications Transportation, Space, Military,Energy storage systems

Combine heat andpower fordecentralizedstationary powersystems

Combine heat and powerfor stationarydecentralized systemsand for transportation(trains, boats, …)

Start-Up Time min sec sec-min hrs hrs hrs

1.3. Operation Principle of a PEM Fuel Cell

Figure 1.3 shows the operation principle of a PEM Fuel Cell. Humidified air enters thecathode channel, and a hydrogen gas enters the anode channel. The hydrogen diffuses throughthe anode diffusion layer towards the catalyst layer, where each hydrogen molecule splits upinto two hydrogen protons and two electrons on catalyst surface according to [14]:

−+ +→ eHH 442 2 (1.1)

The protons migrate through the membrane and the electrons travel through theconductive diffusion layer and an external circuit where they produce electric work. On thecathode side the oxygen diffuses through the diffusion layer, splits up at the catalyst layersurface and reacts with the protons and the electrons to form water:


OHeHO 22 244 →++ −+ (1.2)

Reaction 1 is slightly endothermic, and reaction 2 is heavily exothermic, so that overallheat is created [14]. From above it can be seen that the overall reaction in a PEM Fuel Cellcan be written as:

OHOH 222 22 →+ (1.3)

Based on its physical dimensions, a single cell produces a total amount of current, whichis related to the geometrical cell area by the current density of the cell in [A/cm2]. The cellcurrent density is related to the cell voltage via the polarization curve, and the product of thecurrent density and the cell voltage gives the power density in [W/cm2] of a single cell.

Figure 1.3. Operating scheme of a PEM Fuel Cell.

1.4. Fuel Cell Components

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 polymermembrane (Figure 1.4).


Reactants enter the cell through gas channels, which are embedded in the currentcollectors (bipolar plate). The gas diffusion layers (GDL) are used to uniformly distribute thereactants across the surface of the catalyst layers (CL), as well as to provide an electricalconnection between the catalyst layers and the current collectors.

Figure 1.4. PEM fuel cell components.

The electrochemical reactions that drive a fuel cell occur in the catalyst layers which areattached to both sides of the membrane. The catalyst layers must be designed in such amanner 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 conductingpolymer within the catalyst layers and the membrane. Electrons produced at the anode aretransported through the electrically conductive portion of the catalyst layers to the gasdiffusion 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 thecatalyst layers. Protons are conducted across the polymer membrane from the anode, wherethey are produced, to the cathode where they combine with oxygen and electrons to formwater, 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


mechanism that may be similar to capillary flow. Upon reaching the gas channels, liquidwater is transported out of the cell along with the bulk gas flow [15]. Water may also betransported, in dissolved form, through the polymer portion of the catalyst layers and throughthe membrane [16]. The mechanisms of dissolved water transport are diffusion, due to aconcentration 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 andconvection by a coolant in contact with the collector plates.

Figure 1.5. Transport of gases, protons, and electrons in a PEM fuel cell electrode.

Figure 1.6. Water transport processes in a PEM fuel cell.


1.4.1. Polymer Electrolyte MembraneAn important part of the fuel cell is the electrolyte, which gives every fuel cell its name. Atthe core of a PEMFC is the polymer electrolyte membrane that separates the anode from thecathode. The desired characteristics of PEMs are high proton conductivity, good electronicinsulation, good separation of fuel in the anode side from oxygen in the cathode side, highchemical and thermal stability, and low production cost [18]. One type of PEMs that meetsmost of these requirements is Nafion. This is why Nafion is the most commonly used andinvestigated PEM in fuel cells.

In the Proton-Exchange Membrane Fuel Cell (or Polymer-Electrolyte Membrane FuelCell) the electrolyte consists of an acidic polymeric membrane that conducts protons butrepels electrons, which have to travel through the outer circuit providing the electric work. Acommon electrolyte material is Nafion® from DuPont™, which consists of a fluoro-carbonbackbone, similar to Teflon, with attached sulfonic acid −

3SO groups (Figure 1.7). Themembrane is characterized by the fixed-charge concentration (the acidic groups): the higherthe concentration of fixed-charges, the higher is the protonic conductivity of the membrane.Alternatively, the term “equivalent weight” is used to express the mass of electrolyte per unitcharge.

For optimum fuel cell performance it is crucial to keep the membrane fully humidified atall times, since the conductivity depends directly on water content [19]. The thickness of themembrane is also important, since a thinner membrane reduces the ohmic losses in a cell.However, if the membrane is too thin, hydrogen, which is much more diffusive than oxygen,will be allowed to cross-over to the cathode side and recombine with the oxygen withoutproviding electrons for the external circuit. Typically, the thickness of a membrane is in therange of 5-300 mμ [20].

1.4.2. Catalyst LayerThe best catalyst material for both anode and cathode PEM fuel cell is platinum. Since thecatalytic activity occurs on the surface of the platinum particles, it is desirable to maximizethe surface area of the platinum particles. A common procedure for surface maximization isto deposit the platinum particles on larger carbon black particles [21].

Therefore, the catalyst is characterized by the surface area of platinum by mass of carbonsupport. The electrochemical half-cell reactions can only occur, where all the necessaryreactants have access to the catalyst surface. This means that the carbon particles have to bemixed with some electrolyte material in order to ensure that the hydrogen protons can migratetowards the catalyst surface. This "coating" of electrolyte must be sufficiently thin to allowthe reactant gases to dissolve and diffuse towards the catalyst surface (Figure 1.8). Since theelectrons travel through the solid matrix of the electrodes, these have to be connected to thecatalyst material, i.e. an isolated carbon particle with platinum surrounded by electrolytematerial will not contribute to the chemical reaction.

Several methods of applying the catalyst layer to the gas diffusion electrode have beenreported. These methods are spreading, spraying, and catalyst power deposition. For thespreading method, a mixture of carbon support catalyst and electrolyte is spread on the GDLsurface by rolling a metal cylinder on its surface [22]. In the spraying method, the catalyst andelectrolyte mixture is repeatedly sprayed onto the GDL surface until a desired thickness isachieved.


Although the catalyst layer thickness can be up to 50 mμ thick, it has been found thatalmost all of the electrochemical reaction occurs in a 10 mμ thick layer closest to themembrane [22].

1.4.3. Gas Diffusion LayerThe typical materials for gas diffusion layers are carbon paper and carbon cloth. These areporous materials with typical thickness of 100-300 mμ [22]. The functions of the gasdiffusion layers are to provide structural support for the catalyst layers, passages for reactantgases to reach the catalyst layers and transport of water to or from the catalyst layers, electrontransport from the catalyst layer to the bipolar plate in the anode side and from the bipolarplate to the catalyst layer in the cathode side, and heat removal from the catalyst layers. Gasdiffusion layers are usually coated with Teflon to reduce flooding which can significantlyreduce fuel cell performance due to poor reactant gas transport.

The gas diffusion layers are characterized mainly by their thickness and porosity (Figure1.9). The hot-pressed assembly of the membrane and the gas-diffusion layer including thecatalyst is called the Membrane-Electrode-Assembly (MEA).

1.4.4. Bipolar PlateThe functions of the bipolar plate are to provide the pathways for reactant gas transport andelectron conduction paths from one cell to another in the fuel cell stack, separate theindividual cells in the stack, carry water away from the cells, and provide cooling passages.Plate material and topologies facilitate these functions. Common plate topologies are straight,serpentine, or inter-digitated flow fields. The area of the channels is important, since in somecases a lot of gas has to be pumped through them, but on the other hand there has to be a goodelectrical connection between the bipolar plates and the gas-diffusion layers to minimize thecontact resistance and hence ohmic losses [22]. A judicious choice of the land to openchannel width ratio is necessary to balance these requirements.

Figure 1.7. Membrane structure.

Desirable material characteristics of bipolar plates are high electrical conductivity,impermeability to gases, good thermal conductivity, light weight, high corrosion resistance,and easy to manufacture. The common materials used in bipolar plates are graphite, metalssuch as stainless steel, aluminum, or composite materials [22]. Graphite plates meet most ofthe requirements for optimal fuel cell performance; however the disadvantage of graphite


plates is the high cost of machining the flow fields. Metallic plates are cheap and easy tomanufacture, but they have high contact resistance due to the metal oxide layer formingbetween the plate and the gas diffusion layer. Metallic plates also suffer high degradationfrom the corrosive fuel cell environment that leads to short life cycles. However, some coatedmetallic plates have been shown to produce performance comparable to graphite plates.Finally, composite plates can offer the combined advantages of high electrical and thermalconductivity of graphite plates and low manufacturing cost of metallic plates [23].

Figure 1.8. Catalyst layer structure.

Figure 1.9. The gas diffusion layer.


1.5. Fuel Cell Thermodynamics

1.5.1. Gibbs Free Energy Change in Fuel Cell ReactionsFuel cell electrochemical reactions convert free energy change associated with the chemicalreaction into electrical energy directly. The Gibbs free energy change in a chemical reactionis a measure of the maximum net work obtainable from a chemical reaction [24].

sThg Δ−Δ=Δ (1.4)

The basic thermodynamic functions are internal energy U, enthalpy H, entropy S, andGibbs free energy G. These are extensive properties of a thermodynamic system and they arefirst order homogenous functions of the components of the system. Pressure and temperatureare intensive properties of the system and they are zero-order homogenous functions of thecomponents of the system. Electrochemical potentials are the driving force in anelectrochemical system. The electrochemical potential comprises chemical potential andelectrostatic potential in the following relation.

Chemical potential:

jNPTii N

G

,,⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

=τ (1.5)

Electrochemical potential:

φττ zFii += (1.6)

Where iτ is the chemical potential, z is the charge number of the ion, φ is the potentialat the location of the particles i.

1.5.2. Electrode PotentialThe reversible electrode potential of a chemical reaction can be obtained from the Nernstequation;

⎟⎟⎟

⎠

⎞

⎜⎜⎜

⎝

⎛

−=∑∑

j

vjj

i

vii

e

o

a

a

FnRTEE ln (1.7)

Where oE is the reversible potential at standard state, ia and ja are activity

coefficients of the products and reactants respectively, iv and jv are the stoichiometric

coefficients respectively. The standard state reversible electrode potential can be calculatedfrom the thermodynamic property of standard Gibbs free energy change of the reaction as[24];


FnGEe

oo Δ

−= (1.8)

It can be seen from these relations that the reversible potential is dependent ontemperature and pressure since the Gibbs free energy is a function of temperature and theactivity coefficients are dependent on temperature, pressure for gases and ionic strengths forionic electrolytes. The Nernst equation (1.7) is used to derive a formula for calculating thereversible cell potential as follows: Anode electrode potential;

⎟⎟⎠

⎞⎜⎜⎝

⎛+=

+

2

2

ln2 H

Hoaa a

aF

RTEE (1.9)

Cathode electrode potential;

⎟⎟⎠

⎞⎜⎜⎝

⎛+=

+212)(

2

2ln2 OH

gOHocc aa

aF

RTEE (1.10)

The absolute electrode potential of the fuel cell is difficult to measure. However, only theelectrode potential difference between the cathode and anode is important in fuel cells. Thereversible cell potential can be obtained from the difference between the reversible electrodepotentials at the cathode and anode.

⎟⎟⎠

⎞⎜⎜⎝

⎛+= 21

22

2ln2 OH

OHo

aaa

FRTEE (1.11)

1.5.3. Electrode KineticsElectrochemical kinetics is a complex process and only a short summary is provided in thissection. The first step in understanding the kinetics of the electrode is to determine thegoverning electrochemical reaction mechanism. The reaction mechanism can be single step ormulti-step with electron transfer. The operation of an electrochemical system is a highly non-equilibrium process that involves the transfer of electrons and reactant species at the electrodesurface. The reaction rates are directly related to the Faradic current flows through theelectrode. This rate depends on three important parameters: exchange current density which isrelated to catalytic activity of the electrode surface, concentration of oxidizing and reducingspecies at the electrode surface, and surface activation overpotential. Mathematically, thesephysical quantities are used to derive the Butler-Volmer equation for calculating the current[25].


1.6. Fuel Cell Performance

The reversible potential obtained from the Nernst equation corresponds to the thermodynamicequilibrium state of the electrochemical system. However, when current starts flowingthrough the cell, the cell potential drops below the reversible potential due to several types ofoverpotential including activation overpotential, ohmic overpotential, and concentrationoverpotential. In addition to these typical electrochemical overpotentials, PEM fuel cells alsosuffer 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 dividedinto four regions, which are governed by different losses. Fuel crossover loss occurs when theouter circuit is disconnected in region 1. Activation loss dominates at low current densities inregion 2. Region 3 is governed by the ohmic loss and the bending down of the polarizationcurve 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 variouscurrent density regions.

The details of these potential losses can be summarized as follows:

1- Fuel Crossover and Internal Currents LossesFuel crossover is the amount of fuel that crosses the membrane from the anode to the

cathode without being oxidized at the anode catalyst layer, which results in a loss of fuel.Internal current is the flow of electron from the anode to the cathode through the membraneinstead of going through the external circuit. The combination of these two losses is typically


small compared to other types of losses. This loss can be reduced by increasing the thicknessof the electrolyte at the cost of a higher ohmic loss.

2- Activation OverpotentialsThe 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 currentproduction from the electrode, i.e. a net reaction rate. In PEM fuel cell, the activationoverpotential at the anode is negligible compared to that of the cathode. Activationpolarization depends on factors such as the properties of the electrode material, ion-ioninteractions, ion-solvent interactions and characteristics of the electric double layer at theelectrode-electrolyte interface. Activation polarization may be reduced by increasingoperating temperature and by increasing the active surface area of the catalyst.

3- Ohmic OverpotentialsTwo 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 toproton transport through the membrane. The magnitudes of these potential losses depend onthe materials used in the construction of the fuel cells and its operating conditions [27].Membrane conductivity increases with membrane water content. Reduction in the thicknessof the membrane between anode and cathode may be thought of as an expedient way toeliminate ohmic overpotential. However, “thin” membrane may cause the problem ofcrossover or intermixing of anodic and cathodic reactants [27].

4- Mass Transport OverpotentialsMass transport loss becomes significant when the fuel cell is operated at high current

density. This is created by the concentration gradient due to the consumption of oxygen orfuel at the electrodes. The mass transport loss at the anode can be negligible compared to thecathode. At the limiting current density, oxygen at the catalyst layer is depleted and no morecurrent can be obtained from the fuel cell. This is responsible for the sharp decline inpotential at high current densities. To reduce mass transport loss, the cathode is usually run athigh pressure.

2. Literature Review

2.1. Introduction

Fuel cell modeling has received much attention over the past 20 years in an attempt to betterunderstand the phenomena occurring within the cell. Parametric models allow engineers anddesigners to predict the performance of the fuel cell given geometric parameters, materialproperties and operating conditions. Such models are advantageous because experimentationis costly and time consuming. Furthermore, experimentation is limited to designs, whichalready exist, thus does not facilitate innovative design. Given the highly reactiveenvironment within the fuel cell, it is often impossible to measure critical parameters, such astemperature, pressure and potential gradients, or species concentration within the cell. Thus,


detailed transport models, which accurately predict the flux and concentration of multiplespecies, 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 20years, discusses contemporary trends and compares various approaches to modeling in recenttimes.

2.2. Categories of Fuel Cell Models

A fuel cell model may fall into one of three categories: analytical, semi-empirical, andmechanistic (theoretical) [28]. Table 2.1 categorizes the models reviewed in this chapteraccording to their areas of investigation and dimension of study.

2.2.1. Analytical ModelsExamples of analytical modeling are those reported by Standaert et al. [29, 30]. Manysimplifying assumptions were made concerning variable profiles within the cell in order todevelop an approximate analytical voltage versus current density relationship. This modelalso predicted water management requirements. This was done in the case of isothermal andnon-isothermal cells. However, analytical models are only approximate and do not give anaccurate picture of transport processes occurring within the cell. They are limited topredicting voltage losses and water management requirements for simple designs. They maybe useful if quick calculations are required for simple systems.

2.2.2. Semi-Empirical ModelsSemi-empirical modeling combines theoretically derived differential and algebraic equationswith empirically determined relationships. Empirical relationships are employed when thephysical phenomena are difficult to be modeled or the theory governing the phenomena is notwell understood.

Springer et al. [31] developed a semi-empirical model for use in a fuel cell with apartially hydrated membrane (as opposed to a fully hydrated membrane). Empiricallydetermined relationships were developed correlating membrane conductivity and electrodeporosity with water content in the Nafion® membrane. Most of the models subsequentlydeveloped used these correlations to determine the conductivity of the Nafion® membrane.

Amphlett et al. [32] used semi-empirical relationships to estimate the potential losses andto fit coefficients in a formula used to predict the cell voltage given the operating currentdensity. This model accounted for activation and ohmic overpotentials. The partial pressuresand dissolved concentrations of hydrogen and oxygen were determined empirically asfunctions of temperature, current density and gas channel mole fractions. Subsequently, thereversible cell voltage, activation overpotentials and cell resistance were correlated withtemperature, partial pressures, dissolved concentrations and operating current density.

Pisani et al. [33] also used a semi-empirical approach to study the activation and ohmiclosses as well as transport limitations at the cathode reactive region.

Table 2.1. PEM fuel cell model categorization based on areas of investigation


Maggio et al. [34] studied the water transport in a fuel cell using a semi-empiricalapproach. They modeled the concentration overpotential effect using an empirical functionbetween the cathode gas porosity and current density (since current density is related to waterproduction). The effective gas porosity was assumed to decrease linearly with increasingcurrent density. This is due to the increasing percentage of gas pores occupied by liquidwater. Their results indicate that dehydration of the membrane is likely to occur on the anodeside 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 automobiledriving cycles. They combined the model of Amphlett et al. [32] with the commercialsoftware ADVISOR, which was used to simulate vehicular driving conditions. They studiedthe effects of the number of cells per stack, electrode kinetics and water concentration in themembrane on the fuel consumption. They concluded that a larger number of cells per stackresult in greater stack efficiency resulting in better fuel economy.

Lee et al. [36] developed an artificial neural network model for use to design and analysisPEM fuel cell power systems. The artificial neural network model can simulate theexperimental data for different operating conditions and hence can be used to investigate theinfluence of process variables.

Al-Baghdadi [37] developed a semi-empirical model to provide a tool for the design andanalysis of fuel cell total systems. The model take into account the process variations, such asthe gas pressure, temperature, humidity, and used to cover operating processes, which areimportant 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 investigatingthe performance optimization of a PEM fuel cell. Their results indicate that operatingtemperature and pressure can be optimized, based on cell performance, for given design andoperating point (cell voltage and related current density).

Semi-empirical models are, however, limited to a narrow corridor of operatingconditions. They cannot accurately predict performance outside of that range. They are veryuseful for making quick predictions for designs that already exists. They cannot be used topredict the performance of innovative designs, or the response of the fuel cell to parameterchanges outside of the conditions under which the empirical relationships were developed.Empirical relationships also do not provide an adequate physical understanding of thephenomena inside the cell. They only correlate output with input.

2.2.3. Mechanistic ModelsMechanistic modeling has received the most attention in the literature. In mechanisticmodeling, differential and algebraic equations are derived based on the physical and electro-chemical principals governing the phenomena internal to the cell. These equations are solvedusing some sort of computational method. Figure 2.1 gives a chronology of the developmentof mechanistic modeling. It shows the evolution of PEM fuel cell modeling as it increased incomplexity.

Mechanistic models can generally be characterized by the scope of the model. In manycases, modeling efforts focus on a specific part or parts of the fuel cell, like the cathodecatalyst layer [39], the cathode electrode (gas diffusion layer plus catalyst layer) [40-42], orthe membrane electrode assembly (MEA) [43, 44]. These models are very useful in that they


may include a large portion of the relevant fuel cell physics while at the same time havingrelatively short solution times. However, these narrowly focused models neglect importantparts of the fuel cell making it impossible to get a complete picture of the phenomenagoverning fuel cell behavior.

One of the early mechanistic models for a PEM fuel cell was the pioneering work ofBernardi and Verbrugge [45, 46]. They developed a one-dimensional, steady state, isothermalmodel which described water transport, reactant species transport, as well as ohmic andactivation overpotentials. Their model assumed a fully hydrated membrane at all times, andthus calculated the water input and removal requirements to maintain full hydration of themembrane. The model was based on the Stefan Maxwell equations to describe gas phasediffusion in the electrode regions, the Nernst–Planck equation to describe dissolved speciesfluxes in the membrane and catalyst layers, the Butler Volmer equation to describe electroderate kinetics and Schlogl’s equation for liquid water transport.

This model was used primarily to predict the polarization effects (due to ohmic andactivation overpotentials) and the water management requirements. The model computed therequired water input at the anode side and required water removal rate at the cathode sidenecessary to maintain full hydration of the Nafion® membrane at all times.

Their model also predicted the dissolved hydrogen and oxygen concentrations within thecatalyst layers and it was found that at sufficiently high current densities most of theelectrochemical reactions occurred on the outer surface of the catalyst layer. This informationis vital to designers of fuel cells. It allows them to economically distribute the catalyst whereit is most needed. Considering that the platinum catalyst is one of the largest expenses in afuel cell, this could help reduce the cost. Although their model was basic and manyimprovements have been made since, this work served as a foundation for PEM fuel cellmodeling.

Baschuk and Li [47] developed a one-dimensional model, which accounted for cathodemass limitation effects by allowing variable degrees of flooding at the cathode catalystlayer/backing region. They account for concentration overpotential as a result of thedecreased concentration of dissolved oxygen in the catalyst region due to the excessive watercontent. Darcy’s law is used to obtain the drop in partial pressure of the oxygen at the cathodecatalyst layer, and Henry’s law is used to determine the dissolved oxygen concentration.Their results showed excellent agreement with experimental results. The model also predictedthat increasing the cell pressure lowers the limiting current density. High pressures result inmaximum flooding occurring at lower current densities, and this effect is more significantthan the increase in partial pressure of the oxygen. The results also showed, predictably, thatincreasing the temperature increases the limiting current density.

Models that include all parts of a fuel cell are typically two- or three-dimensional andreflect many of the physical processes occurring within the fuel cell. In a real PEM fuel cellgeometry, the gas diffusion layers are used to enhance the reaction area accessible by thereactants. The effect of using these diffusion layers is to allow a spatial distribution in thecurrent density on the membrane in both the direction of bulk flow and the directionorthogonal to the flow but parallel to the membrane. This two-dimensional distribution cannotbe modeled with the well-used two-dimensional models, (like models that developed by theresearchers in Ref. [48-52]), where the mass-transport limitation is absent in the thirddirection.


LargeComputationalDomains

67

Heat GenerationMechanism 66

Multi PhaseInterdigitatedFlow Field

6564

InterdigitatedFlow Field 63

Serpentine FlowField 54 62

Multi Phase 60 51

3D nonisothermal 55 58

57 56

3Disothermal 59

656461

Catalyst LayerMode 52

Gas ChannelFlow Mode 50

Anode MassLimitation 49

CFD methods 53

Cathode MassLimitation 47

2D 48 71

1D 45 46

91 92 93 94 95 96 97 98 99 00 01 02 03 04 05Year

Figure 2.1. PEM fuel cell mechanistic modeling evolution.


Fuller and Newman [48] developed a pseudo two-dimensional model, which predictedwater and thermal management as well as fuel utilization for a fuel cell operating withreformed methanol as the fuel.

Um et al. [49] developed a two-dimensional transient model based on the single domainapproach, which studied the effects of hydrogen dilution along the anode gas channel. Thetwo-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 hydrogendiffuses from the gas channel into the gas diffusion region, its concentration along the gaschannel decreases resulting in a two-dimensional concentration gradient in the gas diffusionelectrodes. The result is that mass transport limitations are seen on the anode side especiallyat high current densities, and when reformed fuel is used instead of pure hydrogen. At highcurrent densities, hydrogen is extracted from the flow channels at a much fast rate than at lowcurrent densities. With reformed fuel, the partial pressure of the hydrogen is already loweredby the presence of carbon dioxide in the gas feed, so as it is used up toward the end of the gaschannel, the partial pressure of the hydrogen may be too low and it may not be able to diffusefast 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 instraight gas channels, i.e., counter flow versus co-flow. It was found that the flow mode onlymade a difference when dry or low humidity inlet gases were used. For such cases, counterflow operation produced better results since by so doing the reactant gases were sufficientlyhumidified internally. If the inlet gases are already humidified, the flow mode makes littledifference. The reason for this is that for high humidity gases, the increase in membraneconductivity due to the high humidity is counteracted by the increase in cathode concentrationoverpotential 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 humidificationof the gas streams. For co-flow low humidity gases, the membrane dehydrates. Thisinformation 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 inthe 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 interfaceunderestimates the transport limitations due to water build-up. Hence, they modeled thecatalyst layer as a finite region. Their model showed that 20–40% of the water building up atthe cathode catalyst layer comes from water which is transported across the membrane. Thisproblem may be counteracted by applying a pressure differential to force back diffusion ofwater, 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 entiredomain of interest, with source and sink terms accounting for species consumption andgeneration within the cell. The single-domain is less cumbersome in that no internal boundaryconditions and conditions of continuity need to be specified. It is also easier to incorporateinto 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 fuelcell, namely the anode and cathode gas diffusion regions, anode and cathode gas flow


channels, membrane and catalyst layers. These equations are solved separately andsimultaneously.

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 governingdifferential equations are solved over a single computational domain, which consists of a gaschannel, gas diffusion layer, and catalyst layer for both the anode and cathode sides of thefuel cell as well as the solid polymer membrane. The model for the catalyst regions is basedon an agglomerate geometry, which requires water species to exist in both dissolved andgaseous 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 computationaleffect is that all regions can be considered as one domain where no internal boundaryconditions or statements of continuity need be defined. The only difference is that materialproperties and source terms assume different values for the all regions. Conservationequations which govern the entire domain of interest, typically the entire fuel cell (gas flowregions and the membrane electrode assembly) are defined. In each region, the differences areaccounted for by source and sink terms. All equations are written in the form of a genericconvection–diffusion equation, and all terms, which do not fit that format are dumped into thesource or sink term. This formulation allows for solution using known computational fluiddynamics (CFD) methods. The strength of the CFD numerical approach is in providingdetailed insight into the various transport mechanisms and their interaction, and in thepossibility of performing parameters sensitivity analyses.

Two-dimensional models can accurately predict spatial variations in speciesconcentration and fluxes for simple flow regimes, e.g., straight flow channels (co-flow andcounter-flow arrangements). However, for more complex flow regimes, such as interdigitatedflow 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 isrequired. Comprehensive three-dimensional models can study the effect of a large number ofproperties and operating parameters and therefore much more computationally intensive,leading to longer solution times. However, these disadvantages are typically outweighed bythe benefit of being able to assess the influence of a greater number of design parameters andtheir associated physical processes.

Dutta et al. [54] used the unified approach to study mass transport between the channelsof a PEM fuel cell with a serpentine flow field. Their model is three-dimensional and allowsfor multi-species transport. They studied the effect of flow channel width in the serpentineflow 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 resultsshowed that for low humidity conditions, water transport is dominated by electro-osmoticeffects, i.e., water flows from anode to cathode at the side of the cell closer to the gas channelinlet. At the outlet side of the cell, water transport is dominated by back diffusion, and itflows in the opposite direction. Thus the serpentine flow field allows for circulation of thewater 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 densitydistributions and temperature gradients within the cell as well as water flux and species


transport. For gas flow fields separated by current collecting plates, three-dimensional effectswere observed due to the unevenness of the hydrogen and oxygen supply. These effects werepronounced under the collector plate land areas. These effects may result in transport limitingconditions at high current densities. The development of a non-isothermal model wasintended to study the heat transfers within the cell. It was observed that a temperaturedifference of 2–3K existed within the cell. Yan et al. [56] performed a similar study andfound a temperature variation within the cell of the same magnitude. However, the magnitudeof the heat transfer was not reported so it is difficult to compare the magnitude of theconductive heat transfer relative to the total heat transfer. This information would have helpedto justify the need for non-isothermal modeling. If the heat transfer by conduction were smallcompared to other heat transfers, then a temperature difference of 2–3K could hardly besignificant.

Wang et al. [57] developed a single-phase, three-dimensional parametric model,considering the effects of temperature, humidity and pressure. It was found that theperformance of the fuel cell is improved with increasing temperature if the inlet gases arefully humidified. If the gases are not fully humidified, dehydration of the membrane is likelyto occur resulting in reduced conductivity values, hence reduced cell performance. They alsofound that at low current densities anode humidification is required, but not at higher currentdensities. This is because at high current densities, sufficient water is produced at the cathodeto keep the membrane hydrated. Their results further showed that cathode humidification isnot significant at all, especially at high current densities. This is because dehydration is likelyto occur on the anode side and flooding on the cathode side. Therefore, humidifying thecathode gas stream adds no benefit. Finally, increasing the pressure of the inlet gases wasseen to improve performance by increasing the activation currents and the partial pressures ofreactant gases. The authors report that at higher current densities, their model overestimatesthe cell current density compared to experimental results. The reason for this is that the modeldid not take into account mass transport effects. The impact of liquid water on transport in thegas-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 thetemperature and pressure on the fuel cell performance was investigated. In addition,geometrical and material parameters such as the gas diffusion electrode thickness andporosity as well as the ratio between the channel width and the land area were investigated. Itwas found that in order to obtain physically realistic results experimental measurements ofvarious modeling parameters were needed. In addition, the contact resistance inside the cellwas found to play an important role for the evaluation of impact of such parameters on thefuel cell performance. The impact of liquid water on transport in the gas-diffusion electrodewas, however, not account for.

Using the unified approach, Kumar and Reddy [59] studied the effects of having metalfoam in the flow field of the bipolar plates. Their three-dimensional steady state model showsthat decreasing the permeability of the gas flow field improves performance. This is becauseat low flow field permeability reactant gases are transported by forced convection rather thandiffusion. Having many tiny gas channels results in a lower permeability than having fewlarge channels. However, due to limitations in machining processes, the flow channels cannotbe made too small. Placing metal foam in the flow field allows the flow field permeability tobe lowered without resorting to precise machining processes. They found that decreasing the


permeability from 10−6 to 10−12 m2 increases the “average current density” of their systemfrom 5943 to 8425Am−2.

Berning and Djilali [60] developed a computational fluid dynamics multiphase model ofa PEM fuel cell. Their model accounts for three-dimensional transport processes includingphase change and heat transfer, and including the gas diffusion layers and gas flow channelsfor both anode and cathode, as well as a cooling channel. Transfer of liquid water inside thegas diffusion layers was modeled using viscous force and capillary pressure terms. Thephysics of phase change was accounted for by prescribing local evaporation as a function ofthe undersaturation and liquid water concentration. Simulation has been performed for fullyhumidified gases entering the cell. The results showed that different competing mechanismslead to phase change at both anode and cathode sides of the fuel cell. The predicted amount ofliquid water depends strongly on the prescribed material properties, particularly the hydraulicpermeability of the gas diffusion layer. Analysis of the simulations at a current density of 1.2Acm-2 showed that both condensation and evaporation take place within the cathode gasdiffusion 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. Inparticular, the impact of dimensional scales on the order of 100 micrometers and below hasbeen investigated. A model based on isothermal three-dimensional computational fluiddynamics has been developed which predicts that very small channels result in significantlyhigher peak power densities compared to their larger counterparts. The experimental resultsconfirm the predicted outcome at relatively large scale. At especially small scale ( mμ100⟨ ),the model (which dose not consider two-phase flow) disagree with the measured data. Liquidwater flooding at the small channel scale is hypothesized as a primary cause for thisdiscrepancy.

Nguyen et al. [62] developed a three-dimensional model which accounts for mass andheat 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 directionof flow. Also, in the gas diffusion layer, the oxygen concentration is a minimum under theland area. At high current densities the oxygen is almost completely depleted under the landareas. The result is an uneven distribution of oxygen concentration along the catalyst layerresulting in local overpotentials, which vary spatially. A unique feature of this model is avoltage-to-current (VTC) algorithm, which allows for the solution of the potential field andthe local activation overpotential. Since the reactant concentration is not constant across thecatalyst layer, the activation overpotential will not be constant. Their simulations show avariation in local activation overpotential from 0.31 to 0.37 V at a current density of 1.2Acm−2. This VTC algorithm however, comes with a computational cost. It slows down thesolution requiring 6000–8000 iterations for convergence. In addition, the predicteddistribution of current densities show profiles which are fundamentally different from thedistribution 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 interdigitatedflow field. The model accounted for mass transport, electrochemical kinetics, species profilesand current density distribution within the cell. Interdigitated flow fields result in forcedconvection of gases, which aids in liquid water removal at the cathode. This would helpimprove performance at high current densities when transport limitations due to excessive


water production are expected. The model shows that there is little to no difference at low tomedium 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 limitingcurrent, 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 beused 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-phasemixture. Water properties such as specific volume change depending on the degree ofmixture. They used a CFD algorithm to solve for the flow field in the gas flow channels anddiffusion regions, and the fourth order Runge–Kutta method together with a shootingtechnique to solve for the flow field in the catalyst layers and the membrane. For theinterdigitated flow field, results show that the oxygen concentration is higher and liquid watersaturation is lower than those for a conventional straight channel flow field. The higheroxygen concentration results in fast reaction rates and the lower liquid water saturation resultsin less concentration overpotential. It is also shown that the local current densities are muchmore 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 thanthat with a conventional flow field if the inlet gases are well humidified. This is because theinterdigitated 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 modelto account rigorously for various heat generation mechanisms, including irreversible heat dueto electrochemical reactions, entropic heat, and Joule heating arising from the electrolyteionic resistance. Their results show that the irreversible reaction heat and entropic heat in thecathode catalyst layer are major contributors to heat generation in PEM fuel cells, accountingfor about 90% of total waste heat released. Further, it was revealed that the gas diffusion layerthermal conductivity strongly impacts on the membrane temperature rise. There exist vastlydifferent thermal behaviors under different humidification conditions. In the medium range ofcurrent density (i.e. the ohmic control regime of IV curve), the low inlet humidity case show asignificant decrease in fuel cell performance as the membrane temperature rise, indicatingthat the fuel cell performance is primarily controlled by membrane hydration. Thus, efficientcooling through the current collector ribs becomes critical in low-humidity operation in orderto maintain good membrane proton conductivity. On the other hand, for fully humidifiedcases, the performance is dominated by oxygen reduction reaction kinetics and thereforebecome higher with larger membrane temperature rises. In addition, severe flooding ofelectrodes is more likely in the fully humidified operation, particularly in the cold region nearribs. Therefore relatively higher membrane temperature rise enabled by low gas diffusionlayer conductivity should be helpful to alleviate electrode flooding and enhance the oxygenreduction reaction kinetics.

Sivertsen and Djilali [67] developed a single-phase, non-isothermal 3D model which isimplemented into a computational fluid dynamic code. The model allows parallel computing,thus making it practical to perform well-resolved simulations for large computationaldomains. The parallel solver allows them to use a large computational grid (total of 546000


cells). The maximum number of processors used was eight. The model solves for the electricand ionic potentials in the electrodes and membrane. The local activation overpotentialdistribution is resolved, rather than assumed uniform, using a voltage-to-current (VTC)algorithm. Their results showed that the predicted distribution of current densities showprofiles which are fundamentally different from the distribution obtained in all other previousmodels that used current-to-voltage algorithm. The maximum current density occurs underthe land areas as a result of the dominant influence of ohmic losses over concentration losseson the activity at the catalyst layer.

The need for improved lifetime of PEM fuel cells necessitates that the failuremechanisms be clearly understood and life prediction models be developed, so that newdesigns can be introduced to improve long-term performance.

Weber and Newman [68] developed one-dimensional model to study the stressesdevelopment in the fuel cell. They showed that hygro-thermal stresses might be an importantreason for membrane failure, and the mechanical stresses might be particularly important insystems that are non-isothermal. However, their model is one-dimensional and does notinclude 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 alsostudied, 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 eachupper 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 powergeneration (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 ofhygro-thermal stresses into actual fuel cell models to study the effect of these conditions onthe stresses developed in the membrane. In addition, the transport phenomena in a fuel cellare 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 fordistributed power generation. In an ongoing effort to meet increasing energy demand and topreserve the global environment, the development of energy systems with readily availablefuels, high efficiency and minimal environmental impact is urgently required. A fuel cellsystem 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 electricalenergy. Still a maturing technology, fuel cell technology has already indicated its advantages,such as its high-energy conversion efficiency, modular design and very low environmentalintrusion, over conventional power generation equipment.

Two key issues limiting the widespread commercialization of fuel cell technology whichare better performance and lower cost. Recent years have seen significant increase in powerdensities, reliability and overall performance of PEM fuel cells but the underlying physics ofthe transport processes in a fuel cell, (which involve coupled fluid flow, heat and mass


transport with electrochemistry), remain poorly understood. The development of physicallyrepresentative models that allow a reliable simulation of the processes under realisticconditions is essential to the development and optimization of fuel cells, the introduction ofcheaper materials and fabrication techniques, and the design and development of novelarchitectures. Such models are advantageous because experimentation is costly, difficult, andtime consuming. Furthermore, experimentation is limited to designs, which already exist, thusdoes not facilitate innovative design. Given the highly reactive environment within the fuelcell, it is often impossible to measure critical parameters, such as temperature, pressure andpotential gradients, or species concentration. Thus, detailed transport models, whichaccurately predict the flux and concentration of multiple species, are required. The difficultexperimental environment of fuel cell systems has stimulated efforts to develop models thatcould simulate and predict multi-dimensional coupled transport of reactants, heat and chargedspecies using computational fluid dynamic (CFD) methods. The strength of the CFDnumerical approach is in providing detailed insight into the various transport mechanisms andtheir interaction, and in the possibility of performing parameters sensitivity analyses. Thesemodels allow engineers and designers to predict the performance of the fuel cell given designparameters, material properties and operating conditions.

In the present work, a CFD model of PEM fuel cell with straight flow field channels isdeveloped. The model developed in this chapter is different from those in the literature in thatit is fully three-dimensional as opposed to two-dimensional (e.g. [51]). It is accounts forliquid water transport through the gas diffusion layer as well as transport across themembrane, rather than restricted to transport of liquid water through GDL only (e.g. [60]).Furthermore, it is non-isothermal, rather than assumed constant cell temperature (e.g. [61],[64], and [65]). Also, the present model incorporates the effect of hygro and thermal stressesinto actual three-dimension fuel cell model, rather than assumed simplified temperature andhumidity profile, with no internal heat generation (e.g. [68] and [69]).

In addition, the local activation overpotential distribution is resolved using the current-to-voltage algorithm, rather than assumed uniform (as all research presented in the literature),making it possible to predict the local current density distribution more accurately, andtherefore, high accuracy prediction of temperature distribution in the cell and then thermalstresses.

The model accounts for detailed species mass transport, heat transfer in the solids as wellas in the reactants, potential losses in the gas diffusion layers and membrane, electrochemicalkinetics, and the transport of water through the membrane. In addition, the model accounts forthe physics of phase change in that the rate of evaporation is a function of the amount ofliquid water present and the level of undersaturation. Finally, the model is not limited torelatively low humidity reactants, as was the case in prior studies, and can be used to simulateconditions representative of actual fuel cell operation.


3. A Three-Dimensional, Single-Phase CFD Model of a PEM FuelCell

3.1. Introduction

The model presented here is a comprehensive full three–dimensional, non-isothermal, single–phase, steady–state model that resolves coupled transport processes in the membrane, catalystlayer, gas diffusion electrodes and reactant flow channels of a PEM fuel cell. This modelaccounts for a distributed over potential at the catalyst layer as well as in the membrane andgas diffusion electrodes. The model features an algorithm that allows for a more realisticrepresentation of the local activation overpotentials which leads to improved prediction of thelocal current density distribution. This model also takes into account convection and diffusionof different species in the channels as well as in the porous gas diffusion layer, heat transfer inthe solids as well as in the gases, electrochemical reactions and the transport of water throughthe membrane.

3.2. Computational Domain

A computational model of an entire cell would require very large computing resources andexcessively long simulation times. The computational domain in this chapter is thereforelimited to one straight flow channel with the land areas. The full computational domainconsists of cathode and anode gas flow channels, and the membrane electrode assembly asshown in Figure 3.1.

Figure 3.1. Three-dimensional computational domain.


3.3. Assumption

The model that is presented here is based on the following assumptions:

1- The fuel cell operates under study–state conditions.2- All gases are assumed to behave as ideal gases and fully saturated with water vapor, i.e.

fully humidified conditions is assumed.3- The gas mixture is assumed well mixed at the molecular level, with all components

sharing the same velocity, pressure, and temperature fields.4- The flow in the channels is considered as laminar.5- The membrane is assumed to be fully humidified so that the ionic conductivity is

constant.6- Anode and cathode gases are not permitted to crossover, i.e. the membrane is

impermeable.7- The model is restricted to single–phase water transport in the gas diffusion electrodes and

gas flow channels.8- Water is assumed to be product in the vapor phase, which is consistent with the

assumption of single–phase water transport.9- The potential drop in the bipolar plate is negligible since graphite is a good conductor.10- The gas diffusion layer is assumed to be homogeneous and isotropic.

3.4. Modeling Equations

3.4.1. Flow in Gas ChannelsIn the fuel cell channels, the gas-flow field is obtained by solving the steady-state Navier-Stokes equations, i.e. the continuity equation;

( ) 0=⋅∇ uρ (3.1)

and momentum equation;

( ) ( )[ ]TP uuuuu ∇⋅∇+⎟⎠⎞

⎜⎝⎛ ⋅∇+−∇=∇−⊗⋅∇ μμμρ

32

(3.2)

The mass balance is described by the divergence of the mass flux through diffusion andconvection. The steady state mass transport equation can be written in the followingexpression for species i;

( )01 =

⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢

⎣

⎡

∇+

⋅+⎥⎦

⎤⎢⎣

⎡ ∇−+⎟

⎠⎞

⎜⎝⎛ ∇

+∇−⋅∇

∑=

TTD

yPPyx

MMyy

MMDy

Ti

i

N

jjjjj

jiji uρρ

(3.3)


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;

( ) 0=∇−⋅∇ TkTCpuρ (3.4)

The Maxwell-Stefan diffusion coefficients of any two species are dependent ontemperature and pressure. They can be calculated according to the empirical relation based onkinetic gas theory [70];

21

23131

75.18 111016.3

⎥⎥⎦

⎤

⎢⎢⎣

⎡+

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎠

⎞⎜⎝

⎛+⎟

⎠

⎞⎜⎝

⎛×=

∑∑

−

ji

kkj

kki

ij MMVVP

TD (3.5)

The values of the molar diffusion volumes, ( )∑ kiV , are given by Fuller et al. [70] as

shown in Table 3.1.

Table 3.1. Molar diffusion volumes of reactants ( )∑ kiV .

Species Diffusion Volumes [m3/mole]

2O 6106.16 −×

2N 6109.17 −×

2H 6107.12 −×OH 2

61007.7 −×

In order to determine the inlet gas composition to the fuel cell, the following equation ofthe saturation pressure of water vapor has been used [49];

( ) ( )( )37

2510

15.273104454.1

15.273101837.915.27302953.01794.2log

−×+

−×−−+−=−

−

T

TTPsat (3.6)

The above calculated saturated pressure is in bars. The molar fraction of water vapor inthe fully humidified inlet gas stream at the cathode gas flow channel is simply the ratio of thesaturation pressure and the total pressure;

C

satcinOH P

Px =,,2

(3.7)


Since the ratio of nitrogen and oxygen in dry air is known to be 79:21, the inlet oxygenfraction can be found via;

21791

1 ,,,

2

2

+

−= cinOH

inO

xx (3.8)

Since the sum of all molar fractions is unity, the molar fraction of nitrogen can beobtained;

inOcinOHinN xxx ,,,, 2221 −−= (3.9)

The gas composition of the anode inlet stream is much simpler since there are only twogas species, hydrogen and water vapor. The molar fraction of water vapor in the anodeincoming gas stream can be determined as follow;

a

satainOH P

Px =,,2

(3.10)

The molar fraction of hydrogen is determined from the unity relation as follow;

ainOHinH xx ,,, 221−= (3.11)

The mass fraction is related to the molar fraction by [55];

∑=

= n

jjj

iii

Mx

Mxy

1

.

. (3.12)

The gas mixture density is calculated as follow [51];

∑=

= n

i i

i

My

TR

P

1

.ρ (3.13)

The molar fraction of each species is related to the mass fraction by [55];


∑=

=n

j j

j

i

i

i

My

My

x

1

(3.14)

The gas mixture molecular weight is calculated as follow;

∑=

=n

iii MxM

1

. (3.15)

The correlation for the gas viscosity of each species can be expressed as [71];

( ) 72 10.. −×++= TCTBAiμ (3.16)

where A, B, and C are correlation constants of species i as shown in Table 3.2 [71].

Table 3.2. Constants for gas viscosity equation.

Species A B C

2O 11.18 21032.66 −× 6109.187 −×−

2N 43.30 21089.49 −× 6103.109 −×−

2H 87.21 21020.22 −× 61051.37 −×−OH 2 89.31− 21045.41 −× 610272.8 −×−

The Herning-Zipperer correlation [71] for the calculation of gas mixture viscosity is usedin this work;

( )

( )∑

∑

=

== n

iii

n

iiii

Mx

Mx

1

5.0

1

5.0

.

..μμ (3.17)

The correlation for the gas thermal conductivity of each species can be expressed as [71];

( )324 ...10184.4 TDTCTBAki +++×= − (3.18)

where A, B, C, and D are correlation constants of species i as shown in Table 3.3 [71].


Table 3.3. Constants for gas thermal conductivity equation.

Species A B C D

2O 7816.0− 21080.23 −× 4108939.0 −×− 810324.2 −×

2N 9359.0 21044.23 −× 41021.1 −×− 810591.3 −×

2H 34.19 21074.159 −× 41093.9 −×− 81029.37 −×OH 2 53.17 21042.2 −×− 4103.4 −× 81073.21 −×−

The gas mixture thermal conductivity is calculated as follow;

∑=

=n

iii ykk

1

. (3.19)

The correlation for the gas specific heat capacity of each species can be expressed as[71];

( )323

...101868.4 TDTCTBAM

Cpi

i +++×

=−

(3.20)

where A, B, C, and D are correlation constant of species i as shown in Table 3.4 [71];

Table 3.4. Constants for gas specific heat capacity equation.

Species A B C D

2O 22.6 310710.2 −× 61037.0 −×− 91022.0 −×−

2N 07.7 310320.1 −×− 61031.3 −× 91026.1 −×−

2H 88.6 310022.0 −×− 61021.0 −× 91013.0 −×OH 2 10.8 310720.0 −×− 61063.3 −× 91016.1 −×−

The specific heat capacity of the gas mixture is calculated as follow;

∑=

=n

iii yCpCp

1

. (3.21)

3.4.2. Gas Diffusion LayersTransport in the gas diffusion layer is modeled as transport in a porous media. The continuityequation in the gas diffusion layers becomes:

( ) 0=⋅∇ uρε (3.22)


The momentum equation reduces to Darcy’s law:

PKp∇=

μu (3.23)

The mass transport equation in porous media becomes;

( ) 01

=⎥⎥⎦

⎤

⎢⎢⎣

⎡ ∇+⋅+⎥

⎦

⎤⎢⎣

⎡ ∇−+⎟

⎠⎞

⎜⎝⎛ ∇

+∇−⋅∇ ∑= T

TDyPPyx

MMyy

MMDy T

ii

N

jjjjj

jiji ερερε u (3.24)

In order to account for geometric constraints of the porous media, the diffusivities arecorrected using the Bruggemann correction formula [72];

5.1ε×= ijeffij DD (3.25)

The heat transfer in the gas diffusion layers is governed by;

( ) ( )TTTkTCp solideff −=∇−⋅∇ εβερε u (3.26)

Where the term on the right hand side accounts for the heat exchange to and from thesolid matrix of the GDL.

β 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 ofβ is [W/m3]. The heat transfer coefficient between the gas phase and the solid matrix of theelectrodes, β , has been found by trial-and error. It has been adjusted so that the temperaturedifference between the solid and the gas-phase is minimal, i.e. below 0.1 K throughout thewhole domain. This is equivalent to assuming thermal equilibrium between the phases. Thelow velocity of the gas-phase inside the porous medium and the high specific surface areawhich accommodates the heat transfer justify this assumption.

The effective thermal conductivity can be calculated from the thermal conductivities ofgas and graphite by an expression given by Gurau et al. [73];

grgr

greff

kkk

kk

31

2

12εε −

++

+−= (3.27)

However, the effective thermal conductivity of Ballard Arvcarb® was used in this model.The potential distribution in the gas diffusion layers is governed by;

( ) 0=∇⋅∇ GDLe φλ (3.28)


3.4.3. Catalyst LayersThe catalyst layer is treated as a thin interface, where sink and source terms for the reactantsare implemented. Due to the infinitesimal thickness, the source terms are actuallyimplemented in the last grid cell of the porous medium. At the cathode side, the sink term foroxygen is given by;

cO

O iF

MS

42

2−= (3.29)

Whereas the sink term for hydrogen is specified as;

aH

H iF

MS

22

2−= (3.30)

The production of water is modeled as a source terms, and hence can be written as;

cOH

OH iF

MS

22

2= (3.31)

The generation of heat in the cell is due to entropy changes as well as irreversibilities dueto the activation overpotential;

( )ccact

e

iFn

sTq , ⎥⎦

⎤⎢⎣

⎡+

Δ−= η (3.32)

The local current density distribution in the catalyst layers is modeled by the Butler-Volmer equation;

⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−+⎟

⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛= cact

ccact

arefO

Orefcoc RT

FRT

FCC

iiO

,,, expexp2

2

2 ηα

ηα

γ

(3.33)

⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−+⎟

⎠⎞

⎜⎝⎛

⎟⎟⎠

⎞⎜⎜⎝

⎛= aact

caact

arefH

Hrefaoa RT

FRT

FCC

iiH

,,, expexp2

2

2 ηα

ηα

γ

(3.34)

The variation of the reference exchange current density with the operating cell pressureand temperature was computed using the procedure given by Parthasarathy et al. [74];

K15.353

,4, 101325

06.2exp10028.6 2

=

−

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛ ××=

T

inOcrefco

xPi (3.35)


K15.353

,, 101325

06.2exp822 2

=⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛ ×=

T

inHarefao

xPi (3.36)

atm 3

21, 6295.8

exp1005.3=

−⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛×=

CP

refco

Ti (3.37)

atm 3, 3306.64

exp1815.10=

⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛=

aP

refao

Ti (3.38)

3.4.4. MembraneThe balance between the electro-osmotic drag of water from anode to cathode and backdiffusion from cathode to anode yields the net water flux through the membrane [44];

( )WWOHdW yDFiMnN ∇⋅∇−= ρ

2 (3.39)

The water diffusivity in the polymer can be calculated as follow [51];

⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛ −×= −

TDW

130312416exp103.1 10 (3.40)

Heat transfer in the membrane is governed by;

( ) 0=∇⋅⋅∇ Tkmem (3.41)

The potential loss in the membrane is due to resistance to proton transport acrossmembrane, and is governed by;

( ) 0=∇⋅∇ memm φλ (3.42)

The membrane type is Nafion 117®. Bernardi and Verbrugge [46] developed thefollowing theoretical expression for the electric conductivity of the fully humidifiedmembrane;

fHfm CDZRTF ...

2+=λ (3.43)


3.5. Hygrothermal Stresses

As a result of in the changes in temperature and moisture, the membrane, GDL and bipolarplates will all experience expansion and contraction. Because of the different thermalexpansion and swelling coefficients between these materials, hygrothermal stresses areexpected to be introduced into the unit cell during operation. In addition, the non-uniformcurrent and reactant flow distributions in the cell result in non-uniform temperature andmoisture content of the cell which could in turn, potentially causing localized increases in thestress magnitudes.

Using hygrothermoelasticity theory, the effects of temperature and moisture as well asthe mechanical forces on the behavior of elastic bodies have been addressed.

The total strain tensor is determined using Tang et al. [69] formula;

STM ππππ ++= (3.44)

where, Mπ is the contribution from the mechanical forces and Tπ , Sπ are the thermal andswelling induced strains, respectively.

The thermal strains resulting from a change in temperature of an unconstrained isotropicvolume are given by;

( )fmemT TT Re−℘=π (3.45)

Similarly, the swelling strains caused by moisture uptake are given by;

( )fmemS

Reℜ−ℜ=π (3.46)

Hooke's law is used to determine the stress tensor;

πΨ=Ω∇ (3.47)

The initial conditions corresponding to zero stress-state are defined; all components ofthe cell stack are set to reference temperature 20 C, and relative humidity 35% (correspondingto the assembly conditions) [69, 75]. In addition, a constant pressure of (1 MPa) is applied onthe surface of lower graphite plate, corresponding to a case where the fuel cell stack isequipped with springs to control the clamping force.

3.6. Potential Drop Across the Cell and Cell Performance

3.6.1. Cell PotentialUseful work (electrical energy) is obtained from a fuel cell only when a current is drawn, butthe actual cell potential, Ecell, is decreased from its equilibrium thermodynamic potential, E,because of irreversible losses. The various irreversible loss mechanisms which are oftencalled overpotentials, η , are defined as the deviation of the cell potential, Ecell, from the


equilibrium thermodynamic potential E. The cell potential is obtained by subtracting alloverpotentials (losses) from the equilibrium thermodynamic potential as the followingexpression;

Diffmemohmactcell EE ηηηη −−−−= (3.48)

The equilibrium potential is determined using the Nernst equation [37];

( ) ( ) ( )⎥⎦⎤

⎢⎣⎡ +×+−×−= −−

22ln

21ln103085.415.2981083.0229.1 53

OH PPTTE (3.49)

3.6.2. Activation OverpotentialActivation overpotential arises from the kinetics of charge transfer reaction across theelectrode-electrolyte interface. In other words, a portion of the electrode potential is lost indriving the electron transfer reaction. Activation overpotential is directly related to the natureof the electrochemical reactions and represents the magnitude of activation energy, when thereaction propagates at the rate demanded by the current. The activation overpotential can bedivided into the anode and cathode overpotentials. The anode and cathode activationoverpotentials are calculated from Butler-Volmer equation (3.33 and 3.34).

3.6.3. Ohmic Overpotential in Gas Diffusion LayersThe potential loss due to current conduction through the anode and cathode gas diffusionlayers can be modeled by equation (3.28).

3.6.4. Membrane OverpotentialThe membrane overpotential is related to the fact that an electric field is necessary in order tomaintain the motion of the hydrogen protons through the membrane. This field is provided bythe existence of a potential gradient across the cell, which is directed in the opposite directionfrom the outer field that gives us the cell potential, and thus has to be subtracted. Theoverpotential in membrane is calculated from the potential equation (3.42).

3.6.5 Diffusion OverpotentialDiffusion overpotential is caused by mass transfer limitations on the availability of thereactants near the electrodes. The electrode reactions require a constant supply of reactants inorder to sustain the current flow. When the diffusion limitations reduce the availability of areactant, part of the available reaction energy is used to drive the mass transfer, thus creatinga corresponding loss in output voltage. Similar problems can develop if a reaction productaccumulates near the electrode surface and obstructs the diffusion paths or dilutes thereactants. Mass transport loss becomes significant when the fuel cell is operated at highcurrent density. This is created by the concentration gradient due to the consumption ofoxygen or fuel at the electrodes. The mass transport loss at the anode is negligible comparedto that at the cathode. At the limiting current density, oxygen at the catalyst layer is depletedand no more current increase can be obtained from the fuel cell. This is responsible for thesharp decline in potential at high current densities. To reduce mass transport loss, the cathode


is usually run at high pressure. The anode and cathode diffusion overpotentials are calculatedfrom the following equations [26];

⎟⎟⎠

⎞⎜⎜⎝

⎛−=

cL

ccDiff i

iF

RT

,, 1ln

2η (3.50)

⎟⎟⎠

⎞⎜⎜⎝

⎛−=

aL

aaDiff i

iF

RT

,, 1ln

2η (3.51)

GDL

OOcL

CFDi

δ22

2, = (3.52)

GDL

HHaL

CFDi

δ22

2, = (3.53)

The diffusivity of oxygen and hydrogen are calculated from the following equations [63];

⎟⎠⎞

⎜⎝⎛

⎟⎠⎞

⎜⎝⎛×= −

PTDO

101325353

102.323

52

(3.54)

⎟⎠⎞

⎜⎝⎛

⎟⎠⎞

⎜⎝⎛×= −

PTDH

101325353

101.123

42

(3.55)

3.6.6. Cell Power and EfficiencyOnce the cell potential is determined for a given current density, the output power density isfound as;

cellcell EIW .= (3.56)

The thermodynamic efficiency of the cell can be determined as [38];

22.

2

HH

cellfc LHVM

FEE = (3.57)

3.7. Boundary Conditions

Boundary conditions are specified at all external boundaries of the computational domain aswell as boundaries for various mass and scalar equations inside the computational domain.


3.7.1. InletsThe inlet values at the anode and cathode are prescribed for the velocity, temperature andspecies concentrations (Dirichlet boundary conditions). The inlet velocity is a function of thedesired 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]:

chc

cin

inOMEAccin AP

RTx

AFIu 11

4,

,,

2

ζ= (3.58)

cha

ain

inHMEAaain AP

RTx

AFIu 11

2,

,,

2

ζ= (3.59)

3.7.2. OutletsAt the outlets of the gas-flow channels, only the pressure is being prescribed as the desiredelectrode pressure; for all other variables, the gradient in the flow direction (x) is assumed tobe zero (Neumann boundary conditions).

3.7.3. External SurfacesIn order to reduce computational cost, the advantage of the geometric periodicity of the cell istaken. Hence symmetry is assumed in the y-direction, i.e. all gradients in the y-direction areset to zero at the x-z plane boundaries of the domain. All variable are mathematicallysymmetric, with no flux across the boundary. At the external surfaces in the z-direction, (topand bottom surfaces of the cell), temperature is specified and zero heat flux is applied at thex-y plane of the conducting boundary surfaces.

3.7.4. Interfaces Inside the Computational DomainCombinations of Dirichlet and Neumann boundary conditions are used to solve the electronicand protonic potential equations. Dirichlet boundary conditions are applied at the land area(interface between the bipolar plates and the gas diffusion layers). Neumann boundaryconditions are applied at the interface between the gas channels and the gas diffusion layers togive zero potential flux into the gas channels. Similarly, the protonic potential field requires aset of potential boundary condition and zero flux boundary condition at the anode catalystlayer interface and cathode catalyst layer interface respectively.

3.8. Computational Procedure

3.8.1. Computational GridThe governing equations were discretized using a finite volume method and solved using ageneral-purpose computational fluid dynamic code. The computational domains are dividedinto a finite number of control volumes (cells). All variables are stored at the centroid of eachcell. Interpolation is used to express variable values at the control volume surface in terms ofthe control volume center values. Stringent numerical tests were performed to ensure that the


solutions were independent of the grid size. The coupled set of equations was solvediteratively, and the solution was considered to be convergent when the relative error in eachfield between two consecutive iterations was less than 10-6. The number of iterations requiredto obtain converged solutions dependent on the nominal current density of the cell; the higherthe load the slower the convergence.

3.8.2. Solution AlgorithmThe solution begins by specifying a desired current density of the cell to be used forcalculating the inlet flow rates at the anode and cathode sides. An initial guess of theactivation overpotential is obtained from the desired current density using the Butler-Volmerequation. Then follows by computing the flow fields for velocities u,v,w, and pressure P.

Figure 3.2. Flow diagram of the solution procedure used.


Once the flow field is obtained, the mass fraction equations are solved for the mass fractionsof oxygen, hydrogen, water vapor, and nitrogen. Scalar equations are solved last in thesequence of the transport equations for the temperature field in the cell and potential fields inthe gas diffusion layers and the membrane. The local current densities are solved based on theButler-Volmer equation. After the local current densities are obtained, the local activationoverpotentials can be readily calculated from the Butler-Volmer equation. The localactivation overpotentials are updated after each global iterative loop. Convergence criteria arethen performed on each variable and the procedure is repeated until convergence. Theproperties and then source terms are updated after each global iterative loop based on the newlocal gas composition and temperature. The strength of the current model is clearly toperform parametric studies and explore the impact of various parameters on the transportmechanisms and on fuel cell performance. The new feature of the algorithm developed in thiswork is its capability for accurate calculation of the local activation overpotentials, which inturn results in improved prediction of the local current density distribution. The flow diagramof the algorithm is shown in Figure 3.2.

3.9. Modeling Parameters

Choosing the right modeling parameters is important in establishing the base case validationof the model against experimental results. Since the fuel cell model that is presented in thischapter 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 obtaingood agreement with experimental results obtained from a real fuel cell. It is important tonote that because this model accounts for all major transport processes and the modelingdomain comprises all the elements of a complete cell, no parameters needed to be adjusted inorder to obtain physical results.

Table 3.5 shows the dimensions of the computational domain. All parameters listed inTable 3.5 refer to both side, anode and cathode. The membrane refers to fully wetted Nafion117® membrane.

Table 3.5. Cell dimensions for base case.

Parameter Symbol Value UnitChannel length L 31050 −× mChannel height H 3101 −× mChannel width W 3101 −× mLand area width

landW 3101 −× m

Electrode thickness (GDL)GDLδ 31026.0 −× m

Catalyst layer thicknessCLδ 3100287.0 −× m

Membrane thicknessmemδ 31023.0 −× m

The operational parameters are based on the experimental operating conditions used byWang et al. [57]. These values are listed in Table 3.6.


Table 3.6. Operating conditions for base case.

Parameter Symbol Value UnitAir pressure (Cathode pressure) cP 3 atm

Fuel pressure (Anode pressure) aP 3 atm

Air stoichiometric flow ratio cξ 2 -

Fuel stoichiometric flow ratio aξ 2 -

Relative humidity of inlet air cℜ 100 %

Relative humidity of inlet fuel aℜ 100 %

Air inlet temperature cellT 353.15 K

Fuel inlet temperature cellT 353.15 KInlet Oxygen/Nitrogen ratio ψ 0.79/0.21 -

Electrode material properties have important impact on fuel cell performance. Theimportant electrode material properties are thickness, porosity, and thermal and electronicconductivities. The base case values are listed in Table 3.7.

Table 3.7. Electrode properties at base case conditions.

Parameter Symbol Value Unit Ref.Electrode porosity ε 0.4 - [46]Electrode hydraulic permeability ekp 11107.1 −× 2m [57]Electrode therm. conductivity(Ballard AvCarb®-P150) effk 17.1223 KmW ./ [62]

Electrode electronic conductivity eλ 180 mS / [67]

Transfer coefficient, anode side aα 0.5 - [73]

Transfer coefficient, cathode side cα 1 - [74]

Anode ref. exchange current density refaoi , 2465.598 2/ mA [58]

Cathode ref. exchange currentdensity

refcoi ,

3108081.1 × 2/ mA [58]

Oxygen concentration parameter2Oγ 1 - [74]

Hydrogen concentration parameter2Hγ .5 - [73]

Entropy change of cathode reaction SΔ -326.36 KmoleJ ./ [24]Heat transfer coefficient β 6104 × 3/ mW Assumed

Thermal expansion coefficient GDL℘ 6108.0 −×− K/1 [69]

Poisson's ratio GDLℑ 0.25 - [75]

Young's modulus GDLΨ 10101× Pa [75]


The membrane properties are required to model various transport phenomena across themembrane. Table 3.8 lists the membrane properties taken for the base case. The membranetype is fully humidified Nafion 117®.

Table 3.8. Membrane properties.

Parameter Symbol Value Unit RefProtonic diffusion coefficient +H

D 9105.4 −× sm /2 [46]

Fixed-charge concentration fc 1200 3/ mmole [46]

Fixed-site charge fz -1 - [46]

Electro-osmotic drag coefficient dn 2.5 - [31]

Membrane thermal conductivity memk 0.455 KmW ./ [62]Membrane ionic conductivity(humidified Nafion® 117) mλ 17.1223 mS / [46]

Membrane hydraulic permeability mkp 111004.7 −× 2m [57]

Thermal expansion coefficient mem℘ 610123 −× K/1 [69]

Swelling expansion coefficient mem41023 −× - [69]

Poisson's ratio memℑ 0.25 - [75]

Young's modulus memΨ 610249× Pa [75]

3.10. Model Accuracy Validation

3.10.1. Grid Refinement StudyThe convergence behavior and accuracy of a numerical solution depends on the discretizationscheme, equation solver algorithm, and grid quality. Stringent numerical tests were performedto ensure that the solutions were independent of the grid size.

In general, a finer grid provides solution that is more accurate. However, larger grid sizewill increase the computational power required to carry out the simulation. Therefore, it isimportant to find the grid size that is just adequate to provide the accuracy needed for thesimulation. In order to investigate this, a grid of 50% finer than the base case grid was used.The computations at base case conditions ware repeated on this refined grid and the solutionsare compared. The polarization curves obtained with the refined grid and the base case areshown in Figure 3.3. The polarization curves for the two cases show that there is nosignificant difference between the two results. It is impossible to distinguish the two differentlines. This indicates that in terms of the fuel cell performance the base case grid providesadequate resolution. Therefore, the base case mesh is sufficient and no further refinement isnecessary, considering the increase in computational cost (time) for finer mesh.

The computational cost (time of the solution) increases with the number of grid cells.Given the essentially grid-independent solution obtained with the base case grid and the


impracticality of performing a large number of parametric simulations with the finer grid, thebase 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.

3.10.2. Comparing with Experimental DataThe results for base case operating conditions were verified with experimental resultsprovided by Wang et al. [57]. The computed polarization curve shown in Figure 3.4 is ingood agreement with the experimental polarization curve. However, the predicted currentdensities in the mass transport limited region (>1.3A/cm2) are higher than the experimentalvalues. This discrepancy is a common feature of single-phase models where the effect ofreduced oxygen transport due to water flooding at the cathode at high current density cannotbe accounted-for [60]. In addition to this flooding effect, anode drying can also be acontributing factor to the reduced performance at high current density [62].

In general, it is possible to obtain good agreement between the predicted and theexperimental polarization curves with most models. Even the earlier one-dimensional modelof the MEA developed by Bernadi and Verbrugge [45, 46] or the two-dimensional modeldeveloped by Siegel et al. [51], resulted in excellent agreement between each model andexperiment with the adjustment of some parameters. In the model presented here, all theparameters are within physical limits and since no parameters needed to be adjusted, this willhelp to conduct a systematic study on the importance of each single parameter on the fuel cellperformance.


Figure 3.4. Comparison of three-dimensional simulation with an experimental polarization curve.

3.11. Results

In addition to the polarization curve, the comprehensive three-dimensional model also allowsfor the assessment of important information about the detail of transport phenomena insidethe fuel cell. These transport phenomena are the velocity flow field, variation of localconcentration of gas reactants, temperature field, local current densities, and potential field.

3.11.1. Velocity ProfileThe velocity profiles in the mid-plane of the cathode and anode gas flow channels for threedifferent nominal current densities are shown in Figure 3.5. The velocity exhibits a laminarfully developed profile where the highest velocity is located at the center of the channels andthis profile is maintained until the end of the channels. However, the velocity scale in theanode gas flow channel is several orders of magnitude smaller than that in the cathode gasflow channel. This is because the molar fraction of the hydrogen of the incoming humid gas isgreater than that of oxygen, and hence, the inlet velocity at the anode gas flow channel issmaller than that at cathode gas flow channel (equations 3.54 and 3.55). In all loadingconditions, the velocity profiles are similar. However, the velocities are higher for highnominal current density. This is because the nominal current density is a function of inletvelocity. In other words, the inlet mass flow must be sufficient to supply the amount ofreacting species required for a specific current density.


The velocity profiles in the cathode GDL at nominal current densities of 1.2 A/cm2 areshown in Figure 3.6. The figure shows the 2D velocity vectors (cone) in the cross plane ofGDL (y-z) at x=25 mm (lower figure) and 3D velocity vectors arrows (upper figure). Thevectors are pointed upward from the cathode GDL, spreading into the catalyst layer, andpointed up to the membrane surface. In the cathode, oxygen is consumed on the cathodecatalyst layer surface due to the electrochemical reactions while the water vapor is producedon the surface.

Figure 3.5. Velocity profiles in the mid-plane of the cathode and anode gas flow channels for threedifferent nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).


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 areshown in Figure 3.7. The figure shows the 2D velocity vectors (cone) in the cross plane ofGDL (y-z) at x=25 mm (lower figure) and 3D velocity vectors arrows (upper figure). Thevectors are pointed downward from the anode GDL, spreading into the catalyst layer, andpointed down to the membrane surface because both species (H2 and water vapor) are


consumed on the anode catalyst layer surface due to the electrochemical reactions and thewater 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.

3.11.2. Oxygen DistributionThe detailed distribution of oxygen molar fraction for three different nominal currentdensities is shown in Figure 3.8. In all cases, oxygen concentration decreases gradually fromthe inflow channel to the outflow channel due to the consumption of oxygen at the catalystlayer. In the GDL, oxygen concentration under the land area is smaller than that under the


channel area. The concentration of oxygen at the catalyst layer is balanced by the oxygen thatis being consumed and the amount of oxygen that diffuses towards the catalyst layer, drivenby the concentration gradient. The lower diffusivity of the oxygen along with the lowconcentration of oxygen in ambient air results in noticeable oxygen depletion under the landareas. At a low current density, the oxygen consumption rate is low enough not to causediffusive limitations, whereas at a high current density the concentration of oxygen under theland areas has already reached near-zero values. Since, in addition, the local current densityof the cathode side reaction depends directly on the oxygen concentration; this means that thelocal current density distribution under the land areas is much smaller than under the channelareas, especially near the outlet. It becomes clear that the diffusion of the oxygen towards thecatalyst layer is the main impediment for reaching high current densities.

Due to the relatively low diffusivity of the oxygen compared to the hydrogen, the cathodeoperation conditions usually determine the limiting current density when the fuel cell is runon humidified air. This is because an increase in current density corresponds to require higheroxygen consumption, which is limited by diffusivity of oxygen.

3.11.3. Hydrogen DistributionThe hydrogen molar fraction distribution in the anode side is shown in Figure 3.9 for threedifferent nominal current densities. In general, the hydrogen concentration decreases frominlet to outlet as it is being consumed. However, the decrease is quite small along the channeland the decrease in molar concentration of the hydrogen under the land areas is smaller thanfor the oxygen in cathode side due to the higher diffusivity of the hydrogen.

Figure 3.8. Continued on next page.


Figure 3.8. Oxygen molar fraction distribution in the cathode side for three different nominal currentdensities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).


Figure 3.9. Hydrogen molar fraction distribution in the anode side for three different nominal currentdensities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).

3.11.4. Water DistributionThe present model does not account for phase change and two-phase flow. This single-phaseassumption then allows for super-saturation of water in the gas phase, namely the wateractivity greater than unity [31, 63]. The water molar fraction distribution in the cell is shownin Figure 3.10 for three different nominal current densities. Significant condensation isexpected to occur under the cathode land area for all nominal current densities. Note that themaximum mole fraction of water vapor in all cases exceeds the saturated value, indicatingthat vapor condensation takes place. However, the magnitude of water mole fractions ishigher for high nominal current density than for low current density. In addition, under lownominal current density, the back diffusion is sufficient to counteract the electro-osmoticdrag, but under high nominal current density, the electro-osmotic effect dominates backdiffusion. These phenomena result in a drier anode at higher nominal current density.

The results shown in Figure 3.10 also represent the profiles of water content in themembrane. The influence of electro-osmotic drag is readily apparent from these results. Atlow current density, there is very little change in water content across the membrane. This isdue to a relatively low amount of drag and to the fact that the vapor activity at the anode andcathode is nearly identical. As current density is increased, the water content profile becomessteeper as the anode dehydrates and the cathode water content increases.


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).

3.11.5. Cell Temperature DistributionThermal management is required to remove the heat produced by the electrochemical reaction(up to ~50% of the energy produced during high power density operation) in order to preventdrying out of the membrane and excessive thermal stresses that may result in rupture of themembrane. The small temperature differential between the fuel cell stack and the operatingenvironment make thermal management a challenging problem in PEM fuel cells. Thetemperature distribution inside the fuel cell has important effects on nearly all transportphenomena, and knowledge of the magnitude of temperature increases due to irreversibilitiesmight help preventing failure. Temperature distribution for the low, intermediate and highload conditions are demonstrates in the following three figures. Figs. 3.11 and 3.12 show thedistribution of the temperature inside the cell and at the cathode side catalyst layerrespectively. In addition, Figure 3.13 shows the temperature profiles in the through-planedirection where data taken at x=10 mm length of the cell cutting across the middle of the flowchannel. In general, the temperature at the cathode side is higher than at the anode side, due tothe reversible and irreversible entropy production. Naturally, the maximum temperatureoccurs, where the electrochemical activity is highest, which is near the cathode side inlet area.The temperature peak appears in the cathode catalyst layer, implying that major heatgeneration takes place in this region. In all loading conditions, the distributions oftemperature are similar. However, the temperature increase for low load condition of 0.3A/cm2 is small, only 1.537 K. This is different for high nominal current density (1.2 A/cm2).A much larger fraction of the current is being generated near the inlet of the cathode side at


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 occursinside 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 nominalcurrent densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).


Figure 3.13. Temperature profiles along through-plane direction for three different nominal currentdensities of 0.3 A/cm2, 0.7 A/cm2, and 1.2 A/cm2.

3.11.6. Current Density DistributionFigure 3.14 shows the local current density distribution at the cathode side catalyst layer forthree different nominal current densities. The local current densities have been normalized bydivided through the nominal current density in each case (i.e. Iic ). It can be seen that for alow nominal current density, (0.3 A/cm2), the local current is evenly distributed, themaximum value is about 20% higher and the minimum value is 15% lower than the nominalcurrent density. The result is an evenly distributed heat generation, as have seen before.

An increase in the nominal current density to 0.7 A/cm2 leads to a more pronounceddistribution of the local current, and the maximum value can exceed the nominal currentdensity by more than 25% at the cathode side inlet, the minimum value being 25% below thenominal current. Further increase in the current leads to a more current distribution inside thecell.

For a nominal current density of 1.2 A/cm2, a high fraction of the current is generated atthe catalyst layer that lies beneath the channels, leading to an under-utilization of the catalystunder the land areas. The maximum current fraction being just about 40% higher and theminimum 35% lower than the nominal current density.

For optimal fuel cell performance, a uniform current density generation is desirable, andthis could only be achieved with a non-uniform catalyst distribution, possibly in conjunctionwith non-homogeneous gas diffusion electrodes [55, 58].


Figure 3.14. Dimensionless local current density distribution ( Iic ), 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).

3.11.7. Activation Overpotential DistributionThe variation of the cathode activation overpotentials is shown in Figure 3.15. For all nominalcurrent densities, the distribution patterns of activation overpotentials are similar, with highervalues under the channel area. It can be seen that the activation overpotential profile correlatewith the local current density, where the current densities are highest in the center of thechannel and coincide with the highest reactant concentrations.

The variation of the anode activation overpotentials is shown in Figure 3.16. It can beseen that the anode activation overpotentials is evenly distributed for all nominal currentdensities and several orders of magnitude smaller than that of cathode activationoverpotentials.

It can be seen from the figures 3.15 and 3.16 that the activation overpotential is directlyrelated to the nature of the electrochemical reactions and represents the magnitude ofactivation energy, when the reaction propagates at the rate demanded by the current.

3.11.8. Ohmic Overpotential DistributionOhmic overpotential is the loss associated with resistance to electron transport in the gasdiffusion layers. For a given nominal current density, the magnitude of this overpotential isdependent on the path of the electrons. The potential field in the cathodic and the anodic gasdiffusion electrodes are shown in Figure 3.17. The potential distributions are normal to theflow channel and the sidewalls, while there is a gradient into the land areas where electronsflow into the bipolar plate. The distributions exhibit gradients in both x and y direction due to


the non-uniform local current production and show that ohmic losses are larger in the area ofthe catalyst layer under the flow channels.

3.11.9. Membrane Overpotential DistributionThe potential loss in the membrane is due to resistance to proton transport across themembrane from anode catalyst layer to cathode catalyst layer. The distribution pattern of theprotonic overpotential is dependent on the path traveled by the protons and the activities inthe catalyst layers. Figure 3.18 shows the potential loss distribution in the membrane for threedifferent nominal current densities. It can be seen that the potential drop is more uniformlydistributed across the membrane. This is because of the smaller gradient of the hydrogenconcentration distribution under the channel and land areas at the anode catalyst layer due tothe higher diffusivity of the hydrogen.

3.11.10. Diffusion Overpotential DistributionThe variation of the cathode diffusion overpotentials is shown in Figure 3.19. For all nominalcurrent densities, the distribution patterns of diffusion overpotentials are similar, with highervalues under the channel area. In addition, it can be seen that the diffusion overpotentialprofile correlate with the local current density.

The variation of the anode diffusion overpotentials is shown in Figure 3.20. It can be seenthat the anode diffusion overpotentials is evenly distributed for all nominal current densitiesand several orders of magnitude smaller than that of cathode diffusion overpotentials.

In addition, it can be seen from Figure 3.19 and Figure 3.20 that the mass transport lossbecomes significant when the fuel cell is operated at high current density. This is created bythe concentration gradient due to the consumption of oxygen or fuel at the electrodes.



Figure 3.15. Activation overpotential distribution at the cathode side catalyst layer for three differentnominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).


Figure 3.16. Activation overpotential distribution at the anode side catalyst layer for three differentnominal 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 threedifferent nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).


Figure 3.18. Membrane overpotential distribution across the membrane due to proton transport for threedifferent 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 differentnominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).


Figure 3.20. Diffusion overpotential distribution at the anode side catalyst layer for three differentnominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2 (middle); 1.2 A/cm2 (lower).

3.11.11. Stresses Distribution in MembraneThe durability of proton exchange membranes used in fuel cells is a major factor in theoperating lifetime of fuel cell systems. The Mises stresses distribution in membrane thatdeveloped during the cell operating can be seen in Figure 3.21, for three different nominalcurrent densities. It can be seen that the maximum stress occurs, where the temperature ishighest, which is near the cathode side inlet area. The maximum stress appears in the lowersurface of membrane (cathode side), implying that major heat generation takes place near thisregion.

In all loading conditions, the distributions of Mises stresses are similar. However, themaximum stress for low load condition of 0.3 A/cm2 is small, only 2.4 MPa. This is differentfor high nominal current density (1.2 A/cm2). A much larger fraction of the current is beinggenerated near the inlet of the cathode side at the catalyst layer and this leads to asignificantly larger amount of heat being generated here. The maximum stress is about 2.62MPa and it occurs inside the lower surface of the membrane (cathode side).

Figure 3.22 shows the total displacement values that occur in membrane (slice contourplots) and the deformation shape of the membrane (scale enlarged 180 times) for threedifferent nominal current densities. It can be seen that the total displacement and the degree ofthe deformation in membrane are directly related to the increasing of current density, due toincreasing of heat generation. In addition, the total displacement profile correlate with theMises stresses. The deformation that occurs in membrane under the land areas is muchsmaller than under the channel areas due to the clamping force effect.

The MEA is the core component of PEMFC and consists of membrane with the gas-diffusion layers including the catalyst attached to each side. Figure 3.23 shows total


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 thermalexpansion and swelling coefficients between gas diffusion layers and membrane materialswith non-uniform temperature distributions in the cell during operation, hygrothermal stressesand deformation are introduced. The non-uniform distribution of stresses, caused by thetemperature gradient in the MEA, induces localized bending stresses, which can contribute todelaminating between the membrane and the GDLs.

3.11.12. Analysis of Different Loss MechanismsOne of the advantages of a comprehensive fuel cell model is that it allows for the assessmentof the different loss mechanisms. This is shown in Figure 3.24. The most important lossmechanism is the activation overpotential at the cathode side, which has to be addressed withimproved catalyst deposition techniques. Due to the transfer coefficient of ( 5.0=aα ) for theanode side reaction, the anodic activation loss increases relatively fast once the cell currentdensity exceeds the exchange current density of the anodic reaction. However, it also shouldbe possible to alleviate anodic activation losses with improved catalyst deposition. At highcurrent densities, the membrane loss becomes significant. It can be seen that due to its ohmicnature, it increases linearly with increasing current density. The diffusion loss at the cathodeside is quantitatively small, until the oxygen concentration approaches zero at the limitingcurrent density.

In addition to the polarization curve, power and efficiency of the cell can be determinedfrom the resultant cell potential, which is shown in Figure 3.25. The figure shows that theefficiency at maximum power is much lower than the efficiency at partial loads, which makesthe fuel cells very attractive alternative for applications with highly variable loads where mostof the time the fuel cell is operated at low load and high efficiency.

3.12. Summary

A three-dimensional, single-phase CFD model of a PEM fuel cell has been presented in thissection. The complete set of equations was given, and the computational procedure wasoutlined. The results of the base case show good agreement with experimentally obtaineddata, taken from the literature. A detailed distribution of the reactants and the temperaturefield inside the fuel cell for different current densities were presented. In addition, potentialdistribution 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 hygroand thermal stresses in membrane, which developed during the cell operation, were modeledand the behavior of the membrane was investigated. A grid refinement study revealed thatalready 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 transportphenomena that occur in a fuel cell, and furthermore provide guidelines for fuel cell designand prototyping.

One of the simplifications of the current model is the assumption that the volume of theliquid water inside the gas diffusion layer is negligible (single-phase). In order to eliminate


this shortcoming, a multi-phase model has been developed, which will be presented in thefollowing 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 180times) for the membrane for three different nominal current densities: 0.3 A/cm2 (upper); 0.7 A/cm2

(middle); 1.2 A/cm2 (lower).


Figure 3.23. Total displacement (slice contour plots) and deformed shape plots (scale enlarged 300times) for the GDLs (upper), membrane (middle), and MEA (lower) at a nominal current density of 1.2A/cm2 on the y-z plane at x=10 mm.

Figure 3.24. The break-up of different loss mechanisms at base case conditions.


Figure 3.25. Efficiency and power density of the cell at base case conditions.

4. A Three-Dimensional, Multi-Phase CFD Model of a PEM FuelCell

4.1. Introduction

Water management is one of the critical operation issues in proton exchange membrane fuelcells. Spatially varying concentrations of water in both vapor and liquid form are expectedthroughout the cell because of varying rates of production and transport. Water emanatesfrom two sources: the product water from the oxygen-reduction reaction in the cathodecatalyst layer and the humidification water carried by the inlet streams or injected into thefuel cell.

One of the main difficulties in managing water in a PEM fuel cell is the conflictingrequirements 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 diffusionlayer or even the gas channel, thereby inhibiting or even completely blocking oxygen masstransfer. On the anode side, as water is dragged toward the cathode via electro-osmotictransport, dehumidification of the membrane may occur, resulting in deterioration of protonicconductivity. In the extreme case of complete drying, local burnout of the membrane mayresult.


Devising better water management is therefore a key issue in PEMFC design, and thisrequires 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 hasbeen developed that accounts for both the gas and liquid phase in the same computationaldomain and thus allows for the implementation of phase change inside the gas diffusionlayers. The model includes the transport of liquid water within the porous electrodes as wellas the transport of gaseous species, protons, energy, and water dissolved in the ion conductingpolymer.

Water is assumed to be exchanged among three phases; liquid, vapor, and dissolved, andequilibrium among these phases is assumed. Water transport inside the porous gas diffusionlayer is described by two physical mechanisms: viscous drag and capillary pressure forcesand is described by advection within the gas channels. Liquid water, created by theelectrochemical reaction and condensation, is dragged along with the gas phase. At thecathode side, the humidity level of the incoming air determines whether this drag is directedinto or out of the gas diffusion layer, whereas at the anode side this drag is always directedinto the GDL. The capillary pressure gradient drives the liquid water out of the gas diffusionlayers 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 waterfrom anode to cathode and back diffusion from cathode to anode yields the net water contentthrough the membrane. The present multi-phase model is capable of identifying importantparameters for the wetting behavior of the gas diffusion layers and can be used to identifyconditions that might lead to the onset of pore plugging, which has a detrimental effect of thefuel cell performance.

4.2. Assumption

The assumptions made in this multi-phase model are basically identical to the ones stated insection 3. In order to implement the phase change of water, the following additionalassumptions 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.

4.3. Modeling Equations

4.3.1. Flow in Gas ChannelsThe mass conservation equation for each phase yields the volume fraction ( )r and along withthe momentum equations the pressure distribution inside the channels. The continuityequation for the gases (gas phase) inside the channel is given by;


( ) 0=⋅∇ gggr uρ (4.1)

and for the liquid water (liquid phase) inside the channel becomes;

( ) 0=⋅∇ lllr uρ (4.2)

Two sets of momentum equations are solved in the channels, and they share the samepressure field;

PPP lg == (4.3)

Under these conditions, it can be shown that the momentum equations becomes;

( ) ( )[ ]Tgggggggggg Pr uuuuu ∇⋅∇+⎟

⎠⎞

⎜⎝⎛ ⋅∇+−∇=∇−⊗⋅∇ μμμρ

32

(4.4)

and

( ) ( )[ ]Tllllllllll Pr uuuuu ∇⋅∇+⎟

⎠⎞

⎜⎝⎛ ⋅∇+−∇=∇−⊗⋅∇ μμμρ

32

(4.5)

Multiple species are considered in the gas phase only, and the species conservationequation in multi-component, multi-phase flow is the same as in the single-phasecomputations in section 3, except for the consideration of the volume fraction of the gasphase;

( )0

1 =

⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢

⎣

⎡

∇+⋅

+⎥⎦

⎤⎢⎣

⎡ ∇−+⎟

⎠⎞

⎜⎝⎛ ∇

+∇−⋅∇

∑=

TTDyr

PPyx

MMyy

MMDyr

Tigigg

N

jjjjj

jijigg

uρ

ρ (4.6)

The temperature field is obtained by solving the convective energy equation;

( )( ) 0=∇−⋅∇ TkTCpr ggggg uρ (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 phasethe following properties are considered;

The liquid water density is calculated as follow [71];


( )( )cTTl BA 15.27313 ..101 −−−×=ρ (4.8)

where C 2.374 274.0 3471.0 === cTBA .The correlation for the liquid water viscosity can be expressed as [71];

⎟⎠⎞

⎜⎝⎛ +++

−×=2

10.101 3DTCT

TB

A

lμ (4.9)

where 62 1066.14 10966.1 1828 73.10 −− ×−=×==−= DCBA .

4.3.2. Gas Diffusion LayersThe physics of multiple phases through a porous medium is further complicated here withphase change and the sources and sinks associated with the electrochemical reaction. Theequations used to describe transport in the gas diffusion layers are given below.

Mass transfer in the form of evaporation ( )0>phasem and condensation ( )0<phasem is

assumed, so that the mass balance equations for both phases are;

( )( ) phasegg msat =−⋅∇ uερ1 (4.10)

and

( ) phasell msat =⋅∇ uερ. (4.11)

The momentum equation for the gas phase reduces to Darcy’s law, which is, however,based on the relative permeability for the gas phase ( )KP . The relative permeability accountsfor the reduction in pore space available for one phase due to the existence of the secondphase. The relative permeability for the gas phase is given by;

( )KPsatKPg −= 1 (4.12)

and for the liquid phase;

KPsatKPl .= (4.13)

The momentum equation for the gas phase inside the gas diffusion layer becomes;

( ) PKpsatg

g ∇−−=μ

1u (4.14)


Two liquid water transport mechanisms are considered; shear, which drags the liquidphase 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 Darcy’s law,the following equation can write;

ll

ll P

Kp∇−=

μu (4.15)

where the liquid water pressure stems from the gas-phase pressure and the capillary pressureaccording to [60];

satsatP

PPPP ccl ∇

∂∂

−∇=∇−∇=∇ (4.16)

Introducing this expression into Equation (4.15) yields a liquid water velocity fieldequation;

satsatPKP

PKP c

l

l

l

ll ∇

∂∂

+∇−=μμ

u (4.17)

The functional variation of capillary pressure with saturation is prescribed followingLeverett [60] who has shown that;

( )satfKP

Pc

21

⎟⎠⎞

⎜⎝⎛=εσ (4.18)

( ) ( ) ( ) ( )32 1263.1112.21417.1 satsatsatsatf −+−−−= (4.19)

The correlation for interfacial liquid/gas tension can be expressed as [71];

( )3

11 101 −×⎟⎟

⎠

⎞⎜⎜⎝

⎛−−

=Z

c

c

TTTT

σσ (4.20)

where 8105.0 15.298 35.647 97.71 11 ==== ZTTcσ .The liquid phase consists of pure water, while the gas phase has multi components. The

transport of each species in the gas phase is governed by a general convection-diffusionequation in conjunction which the Stefan-Maxwell equations to account for multi speciesdiffusion, as described in section 3, with the addition of a source term accounting for phasechange;


( ) ( )

( )phase

Tigig

N

jjjjj

jijig

m

TTDysat

PPyx

MMyy

MMDysat

=

⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢

⎣

⎡

∇+⋅−

+⎥⎦

⎤⎢⎣

⎡ ∇−+⎟

⎠⎞

⎜⎝⎛ ∇

+∇−−⋅∇

∑=

εερ

ερ

u1

11 (4.21)

The heat transfer in the gas diffusion layers is governed by the energy equation asfollows;

( )( )( ) ( ) evapphasesolidgeffggg HmTTTkTCpsat Δ−−=∇−−⋅∇ εεβεερ ,1 u (4.22)

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 evaporationfor water is calculated as follow [71];

80.41861

, ×⎟⎟⎠

⎞⎜⎜⎝

⎛−−

Δ=ΔZ

c

clevapevap TT

TTHH (4.23)

where 38.0 15.373 35.647 7.538 1, ====Δ ZTTH clevap .

Implementation of phase changeIn order to account for the magnitude of phase change inside the GDL, expressions are

required to relate the level of over- and undersaturation as well as the amount of liquid waterpresent to the amount of water undergoing phase change.

In the case of evaporation, such relations must be dependent on (i) the level ofundersaturation of the gas phase in each control volume and on (ii) the surface area of theliquid water in the control volume. The surface area can be assumed proportional to thevolume fraction of the liquid water in each cell. A plausible choice for the shape of the liquidwater is droplets, especially since the catalyst area is Teflonated [60].

The evaporation rate of a droplet in a convective stream depends on the rate ofundersaturation, the surface area of the liquid droplet, and a (diffusivity dependent) mass-transfer coefficient. The mass flux of water undergoing evaporation in each control volumecan be represented by [60];

0

0

1

2w

wwdropxmDOHevap x

xxDkNMm

−−

= ∞πϖ (4.24)

The bulk concentration ∞wx is known by solving the continuity equation of water

vapor. To obtain the concentration at the surface 0wx , it is reasonable to assumethermodynamic equilibrium between the liquid phase and the gas phase at the interface, i.e.,the relative humidity of the gas in the immediate vicinity of the liquid is 100%. Under that


condition, the surface concentration can be calculated based on the saturation pressure and isonly a function of temperature.

The heat-transfer coefficient for convection around a sphere is well established, and byinvoking the analogy between convective heat and mass transfer, the following mass-transfercoefficient is obtained [60];

⎥⎥

⎦

⎤

⎢⎢

⎣

⎡

⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎟⎠

⎞⎜⎜⎝

⎛+= ∞

3121

2

2 6.02OHg

g

g

gdrop

drop

OHairxm D

vDD

Dck

ρμ

μρ

(4.25)

It is further assumed that all droplets have a specified diameter dropD , and the number of

droplets in each control volume is found by dividing the total volume of the liquid phase ineach control volume by the volume of one droplet;

3

61

.

drop

cvD

D

VsatN

π= (4.26)

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 surfacearea, but the rate of condensation can be different when it takes place on a wetted surface. Inaddition, the overall surface area in each control volume available for condensation shrinkswith an increasing amount of liquid water present. Berning and Djilali [60] assumed that therate of condensation depends only on the level of oversaturation of the gas phase multipliedby a condensation constant. Thus, the mass flux of water undergoing condensation in eachcontrol volume can be represented by;

0

0

1

w

wwcond x

xxCm

−−

= ∞ϖ (4.27)

4.3.3. Catalyst LayersThe catalyst layers are treated as thin interfaces, where the oxygen and hydrogen are depletedand liquid water and heat are produced. The depletion and production rates depend on thelocal current density, which is described by the Butler-Volmer equation. In addition, the localactivation overpotential distribution is resolved making it possible to predict the local currentdensity distribution more accurately. The sink and source terms applied at the catalyst layerare the same as those in section (3.4.3).

4.3.4. MembraneThe balance between the electro-osmotic drag of water from anode to cathode and backdiffusion from cathode to anode yields the net water flux through the membrane [44];


( )WWOHdW cDFiMnN ∇⋅∇−= ρ

2 (4.28)

The water diffusivity in the polymer can be calculated as follow [51];

⎥⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛ −×= −

TDW

130312416exp103.1 10 (4.29)

The variable Wc represents the number of water molecules per sulfonic acid group

(i.e. 132 equivalent mol −SOOH ).The water content in the electrolyte phase is related to

water vapor activity via [64, 65];

( )( ) ( )

( )3 8.1631 14.10.1410 0.3685.3981.17043.0 32

≥=≤<−+=≤<+−+=

acaacaaaac

W

W

W

(4.30)

The water vapor activity given by;

sat

W

PPx

a = (4.31)

4.4. Boundary Conditions

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 thesize of the computational domain and computational costs. In the x direction, zero fluxconditions 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 densityand stoichiometric flow ratio, as described in section 3.7. The gas streams entering the cell arefully humidified, but no liquid water is contained in the gas stream. At the outlets, thepressure is prescribed for the momentum equation and a zero gradient conditions are imposedfor all scalar equations.

4.5. Computational Procedure

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 competingtransport and phase-change mechanisms, the computational time required was about threetimes greater than in single-phase.


4.6. Modeling Parameters

The parameters introduced to account for the multi-phase flow and phase change phenomenaare listed in Table 4.1. With the exception of the parameters listed in Table 4.1, thephysicochemical and geometric parameters are identical to those used in the single-phasesimulations (section 3).

Table 4.1. Multi-phase parameters at base case conditions.

Parameter Symbol Value Unit Ref.Droplet diameter

dropD 8100.1 −× m [60]

Condensation constant C 5100.1 −× - [60]Scaling parameter for evaporation ϖ 01.0 - [60]

4.7. Model Validation

The multi-phase model is validated by comparing model results to experimental data providedby Wang et al. [57]. The importance of phase change to the accurate modeling of fuel cellperformance is also illustrated. Performance curves with and without phase change as well asexperimental data are shown in Figure 4.1 for the base case conditions. Comparison of thetwo curves demonstrates that the effects of liquid water accumulation become apparent evenat relatively low values of current density. Furthermore, when liquid water effects are notincluded in the model, the cell voltage dose not exhibit an increasingly steep drop as the cellapproaches its limiting current density. This drop off in performance is clearly demonstrated

Figure 4.1. Effect of liquid water buildup on cell performance.


by experimental data, but cannot be accurately modeled without the incorporation of phasechange. By including the effects of phase change, the current model is able to more closelysimulate performance, especially in the region where mass transport effects begin todominate.

4.8. Results

Results with and without phase change for the cell operates at nominal current density of 1.4A/cm2 are discussed in this section. The selection of relatively high current density is due toillustrate the phase change effects, where it becomes clearly apparent between single andmulti-phase model in the mass transport limited region.

Prior to proceeding with a detailed analysis of the results, it is useful to discuss some ofthe phase-change mechanisms. The central parameter for determining the direction of phasechange is the relative humidity of the gas phase;

sat

OH

PP

2=ϕ (4.32)

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 Dalton’s law the partial pressure of aspecies is equal to its molar fraction multiplied with the total pressure of the gas phase, i.e.;

satOH P

Px2

=ϕ (4.33)

When the relative humidity is below 100% in the presence of liquid water, this give riseto evaporation. Condensation, on the other hand, occurs when the relative humidity exceeds100% 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 OHx2

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 avery interesting effect and, depending on the incoming gas condition, it can lead toeither evaporation or condensation. In the first place, there is a pressure drop insidethe GDL due again to reactants consumption. This pressure drop depends strongly onthe permeability of the gas diffusion layer, i.e., for the same amount of consumedreactants, the pressure drop will be higher for a lower permeability. The bulk velocityof the gas phase is directed into the GDL, and is governed by Darcy’s law. Thus, thepressure drop inside the GDL depends strongly on the permeability. The partial


pressure of the water vapor decreases with the gas-phase pressure. Hence, this effectalone would lead to undersaturation, causing evaporation.

3. The saturation pressure satP increases with an increase in temperature caused by theheat production term due to the electrochemical reaction. The order of magnitude ofthe temperature increase depends primarily on the thermal conductivity of the gasdiffusion layer. It was found with the single-phase model that the temperature canrise by a few degrees Kelvin and this effect alone would lead to evaporation of liquidwater.

The net phase change is a result of the balance between these competing, coupled, andspatially varying mechanisms. It should be noted that the first two effects are also ofimportance inside the gas flow channels; the oxygen depletion from inlet to outlet results inoversaturation and condensation at the walls and channel/GDL interface, whereas the overallpressure drop along the channel would alone cause evaporation. For the straight channelsection considered here, the total pressure drop is very small and hence the oxygen depletioneffect dominates, causing condensation. The dominant mechanisms highlighted in thisdiscussion are relevant to cases where the incoming air is at a high humidification level, as isthe case in practical fuel cell operation.

4.8.1. Velocity ProfileThe velocity fields inside the cathodic and anodic gas diffusion layers are shown in Figures4.2 and 4.3 for both gas and liquid phase. The pressure gradient induces bulk gas flow from



Figure 4.2. Gas phase velocity vectors (upper) and liquid water velocity vectors (lower) inside thecathode 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 theanode 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 ofthe gas diffusion layers into the flow channels. Therefore, the liquid water flux is directedfrom the GDL into the channel, i.e., in the opposite direction of the gas-phase velocity, whereit can leave the cell. The velocity of the liquid phase, however, is lower than for the gasphase, which is due to the higher viscosity, and the highest liquid water velocity occurs at thecorners of the channel/GDL interface. The liquid water oozes out of the GDL, mainly at thecorners of the GDL/channel interface.

4.8.2. Liquid Water SaturationThe liquid water saturation inside the cathodic and anodic gas diffusion layers are shown inFigure 4.4. Condensation occurs mainly in two areas inside the cathodic GDL: at the catalystlayer where the molar water vapor fraction increases due to the oxygen depletion, and at thechannel/GDL interface, where the oversaturated bulk flow condenses out. This term isrelatively small compared to the other effects.

Condensation occurs throughout the anodic GDL due to hydrogen depletion. Similar tothe cathode side, the liquid water can only leave the GDL through the build-up of a capillarypressure gradient to overcome the viscous drag, because at steady state operation, all thecondensed water has to leave the cell.

The liquid water saturation is distributed through the entire cathodic and anodic GDLwith maximum saturations found under the land areas. The reason for this is clear: once liquidwater is being created by condensation, it is dragged into the GDL by the gas phase. The high


spatial variation of the saturation demonstrates again the three-dimensional nature of transportprocesses in PEM fuel cells.

Clearly, liquid water saturation depends strongly on the specified capillary pressure, andagain, the permeability of the gas diffusion layer becomes the central parameter.

4.8.3. Liquid Water Content in the MembraneSeveral transport mechanisms in the cell affect water distribution. In the membrane, primarytransport is through (i) electro-osmotic drag associated with the protonic current in theelectrolyte, which results in water transport from anode to cathode; and (ii) diffusionassociated with water-content gradients in the membrane. One of the main difficulties inmanaging water in a PEM fuel cell is the conflicting requirements of the membrane and of thecatalyst gas diffusion layer. On the cathode side, excessive liquid water may block or floodthe pores of the catalyst layer, the gas diffusion layer or even the gas channel, therebyinhibiting or even completely blocking oxygen mass transfer. On the anode side, as water isdragged toward the cathode via electro-osmotic transport, dehumidification of the membranemay occur, resulting in deterioration of protonic conductivity. In the extreme case ofcomplete drying, local burnout of the membrane can result.

Figure 4.5 shows profiles for polymer water content in the membrane for the base caseconditions. The influence of electro-osmotic drag and back diffusion are readily apparentfrom this result.

Figure 4.4. Liquid water saturation inside the cathode and anode GDLs at a nominal current density of1.4 A/cm2.


Figure 4.5. Water content profiles through the MEA at a nominal current density of 1.4 A/cm2.

4.8.4. Oxygen DistributionFigure 4.6 shows the oxygen distribution inside the cathode. Similar to the single phasecomputations, the oxygen concentration is highest under the channel, and exhibits a three-dimensional behavior with a fairly significant drop under the land areas, and a more gradualdepletion towards the outlet. The lower diffusivity of the oxygen along with the lowconcentration of oxygen in ambient air results in noticeable oxygen depletion under the landareas.

The concentration of oxygen at the catalyst layer is balanced by the oxygen that is beingconsumed and the amount of oxygen that diffuses towards the catalyst layer, driven by theconcentration gradient. It can be seen from the figure that, the oxygen concentration at thecatalyst layer predicted in multi-phase model is quantitatively smaller than that predicted insingle-phase simulations, due to the effects of liquid water inside GDL. This water reducesthe performance of the cell by increasing the accumulation of liquid water at the cathodeGDL, which decreases its permeability to reactant gas flow and lead to the onset of poreplugging by liquid water.

4.8.5. Hydrogen DistributionThe underlying mechanisms of the effect of phase change are the same for cathode and anodesides, but their magnitudes differ. In addition, the phase change process takes place in abinary rather than a ternary mixture.

Figure 4.7 shows the hydrogen distribution inside the anode. The molar hydrogen fractionis almost constant inside the GDL due to the higher diffusivity of the hydrogen. However, thedecrease under the land areas is much less than in the absence of phase change. This can beexplained by the fact that consumption of hydrogen leads to a direct increase of the molarwater vapor fraction since the anode gas steam is a binary mixture, and hence a high level ofoversaturation results in a strong condensation potential inside the gas diffusion layer,


condensation of liquid water reduces the molar water vapor fraction in return. As a result, themolar 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 theaverage oxygen molar fraction at the cathode CL.


The virtually constant hydrogen molar fraction is a particularly interesting feature whenconsidering 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 theaverage hydrogen molar fraction at the anode CL.


4.8.6. Water Vapor DistributionThe water vapor molar fraction distribution in the cell is shown in Figure 4.7. The molarwater vapor fraction, however, remains almost constant throughout the gas diffusion layer inmulti-phase model. In the absence of phase change, this is not being the case, since thenitrogen and water vapor fraction would increase as the oxygen fraction decreases.

4.8.7. Diffusion Overpotential DistributionThe variations of the cathode and anode diffusion overpotentials are shown in Figures 4.8 and4.9 respectively. For both single and multi-phase models, the distribution patterns of diffusionoverpotentials have similar profiles, but their magnitudes differ.

It can be seen from the figures that, the diffusion overpotentials predicted in multi-phasemodel is quantitatively higher than that predicted in single-phase simulations, due to thepresence of liquid water inside GDL. This water reduces the limiting current density in thecell by increasing the accumulation of liquid water at the GDL, which decreases itspermeability to reactant gas flow and lead to the onset of pore plugging by liquid water.

4.8.8. Current Density DistributionThe local current density distribution is shown in Figure 4.10. The local current densities havebeen normalized relating to the nominal current density (i.e. Iic ).



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 catalystlayer predicted in multi-phase model has a much higher fraction of the total current, generatedunder the channel area. This is due to the effects of liquid water inside the GDL, whichdecreases its permeability to reactant gas flow (oxygen) and lead to the onset of pore pluggingby liquid water. This can lead to local hot-spots inside the membrane electrode assembly, andleads to a further drying out of the membrane, thus increasing the electric resistance, which inturn leads to more heat generation and can lead to a failure of the membrane. Thus, it isimportant 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 thegas diffusion layers and can be used to identify conditions that might lead to the onset of poreplugging, which has a detrimental effect of the fuel cell performance, especially in the masstransport limited region.


Figure 4.8. Diffusion overpotential distribution at the cathode catalyst layer predicted in single-phasemodel (upper) and multi-phase model (lower) at a nominal current density of 1.4 A/cm2.


Figure 4.9. Diffusion overpotential distribution at the anode catalyst layer predicted in single-phasemodel (upper) and multi-phase model (lower) at a nominal current density of 1.4 A/cm2.


Figure 4.10. Dimensionless local current density distribution ( Iic ) at the cathode catalyst layer

predicted in single-phase model (upper) and multi-phase model (lower) at a nominal current density of1.4 A/cm2.


4.9. Summary

This section presented a three-dimensional, two-phase model of the cathode and anode of aPEM Fuel Cell. The mathematical model accounts for the liquid water flux inside the gasdiffusion layers by viscous and capillary forces and hence is capable of predicting the amountof liquid water inside the gas diffusion layers. The physics of phase change are included inthis model by prescribing the local evaporation term as a function of the amount of liquidwater present and the level of undersaturation, whereas the condensation has been simplifiedto be a function of the level of oversaturation only.

Three different physical mechanisms that lead to phase change inside the gas diffusionlayers were identified. A rise in temperature because of the electrochemical reaction leads toevaporation, mainly at the cathode side. If the gases entering the cell are fully humidified, thedepletion of the reactants leads to an increase in the partial pressure of the water vapor, andhence to condensation along the channel and inside the gas diffusion layers. Finally, adecrease in the gas phase pressure inside the gas diffusion layers leads to a decrease in thewater vapor pressure, and hence causes evaporation.

The results show that the multi-phase model is capable of identifying importantparameters for the wetting behavior of the gas diffusion layers and can be used to identifyconditions that might lead to the onset of pore plugging, which has a detrimental effect of thefuel 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 interpretresults. 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 flowchannels has been developed. This model provides valuable information about the transportphenomena inside the fuel cell such as reactant gas concentration distribution, liquid watersaturation distribution, temperature distribution, potential distribution in the membrane andgas diffusion layers, activation overpotential distribution, diffusion overpotential distribution,and local current density distribution. In addition, the hygro and thermal stresses inmembrane, 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 forthe calculation of the electrochemical kinetics without simplifications. This calculationinvolves the coupling of the potential field with the reactant species concentration field,which results in an accurate prediction of local current density distribution. Results arephysically consistent and in good agreement with available experimental data.

The model is shown to be able to: (1) understand the many interacting, complexelectrochemical, phase change mechanism, and transport phenomena that cannot be studiedexperimentally; (2) identify limiting steps and components; and (3) provide a computer-aidedtool 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 PEMfuel cells to get optimal results for any application.


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), 2684–2699.[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, 193–201.[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), 3537–3546.[25] Hyunchul, J.; Chao-Yang, W. J. Electrochem. Soc. 2004, 151(11), 1954–1960.[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.


[29] Standaert, F.; Hemmes, K.; Woudstra, N. J. Power Sources. 1996, 63(2), 212–234.[30] Standaert, F.; Hemmes, K.; Woudstra, N. J. Power Sources. 1998, 70(2), 181–189.[31] Springer, T.E.; Zawodzinski, T.A.; Gottesfeld, S. J. Electrochem. Soc. 1991, 138(8),

2334–2342.[32] Amphlett, J.C.; Baumert, R.M.; Mann, R.F.; Peppley, B.A.; Roberge, P.R. J.

Electrochem. Soc. 1995, 142(1), 1–8.[33] Pisani, L.; Murgia, G.; Valent'ın, M.; D'Aguanno, B. J. Power Sources. 2002, 108(1-2),

192–203.[34] Maggio, G.; Recupero, V.; Pino, L. J. Power Sources. 2001, 101(2), 275–286.[35] Maxoulis, C.N.; Tsinoglou, D.N.; Koltsakis, G.C. Energy Conversion Manage. 2004,

45(4), 559–573.[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); 145–151.[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),5883–5895.[42] Grujicic, M.; Zhao, C.L.; Chittajallu, K.M.; Ochterbeck, J.M. Materials Science

Engineering B. 2004, 108(3), 241–252.[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), 181–196.[48] Fuller, T.F.; Newman, J. J. Electrochem. Soc. 1993, 140(5), 1218–1225.[49] Um, S.; Wang, C.Y.; Chen, K.S. J. Electrochem. Soc. 2000, 147(12), 4485–4493.[50] Ge, S.; Yi, B. J. Power Sources. 2003, 124(1), 1–11.[51] Siegel, N.P.; Ellis, M.W.; Nelson, D.J.; von Spakovsky, M.R. J. Power Sources. 2004,

128(2), 173–184.[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), 177–198.[54] Dutta, S.; Shimpalee, S.; Van Zee, J.W. Int. J. Heat Mass Transfer. 2001, 44(11),

2029–2042.[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), 127–137.[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), 54–62.[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.


[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), 40–51.[64] Hu, M.; Gu, A.; Wang, M.; Zhu, X.; Yu, L. Energy Conversion Manage. 2004, 45(11-

12), 1861–1882.[65] Hu, M.; Gu, A.; Wang, M.; Zhu, X.; Yu, L. Energy Conversion Manage. 2004, 45(11-

12), 1883–1916.[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), 3215–3226.[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), 827–836.[73] Gurau, V.; Liu, H.; Kakac, S. AIChE J. 1998, 44(11), 2410–2422.[74] Parthasarathy, A.; Srinivasan, S.; Appleby, J.A.; Martin, C.R. J. Electrochem. Soc.

1992, 139(10), 2856–2862.[75] Product Information, 2005, DuPont™ Nafion® PFSA Membranes N-112, NE-1135, N-

115, N-117, NE-1110 Perfluorosulfonic Acid Polymer. NAE101.


Chapter 8

SURFACE FUNCTIONALIZATION OF CARBON CATALYST-SUPPORT FOR PEM FUEL CELLS:

A REVIEW

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: [email protected], Tel: 1-518-738-0229 (work), Fax: 1-518-782-7914

Zhigang Qi 382

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.

1.1. 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.

2H2 → 4H+ + 4e- (1) O2 + 4H+ + 4e- → 2H2O (2) 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 DuPont’s 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

1.2. Carbon-Supported Catalysts

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

Zhigang Qi 384

methods such as thermal treatment, chemical oxidation, electrochemical oxidation, plasma treatment [10], ion bombardment, and chemical linkage of functional groups. Vulcan XC-72 is probably the most commonly used carbon support for PEM fuel cells.

Some surface groups can act as the nucleation sites during the catalyst deposition process, which could result in smaller particles and higher thermal stability against sintering. For example, Roy et al. found that by creating nitrogen (or sulfur)-containing groups on the surface of carbon, the mean size of the Pt particles generated was 1.5 (or 1.0) nm compared to 2.5 nm generated on the untreated counterpart [11]. Kim and Park reported that increasing the basicity of the carbon support resulted in the formation of Pt particles with smaller size and higher activity, while increasing the acidity of the carbon support led to the opposite effects [12]. They suggested that well-distributed basic functional groups increased the effectiveness of the Pt reduction process.

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.


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.

2. Fuel Cell Performance

According to thermodynamics a fuel cell should perform much better (e.g., providing higher electrical efficiency) at temperatures lower than about 750 oC than a combustion engine whose maximum efficiency is determined by the Carnot limit [= (Thigh – Tlow)/Thigh]. In reality, the electrical efficiency of a PEM fuel cell is much less than the theoretical prediction for several reasons, including activation overpotential loss, concentration overpotential loss, and losses due to the electric and the ionic resistances. The activation overpotential is the dominating loss for PEM fuel cells due to the relatively low operating temperature. In addition, the following stringent requirements for the reaction to proceed further limit the fuel cell performance.

2.1. Three-Phase Boundary

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

Zhigang Qi 386

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.

H+

Electrolyte

Catalyste- 3-phase

regionH+

Electrolyte

Catalyste- 3-phase

region

Figure 2. Illustration of catalyst-electrolyte-reactant three-phase boundary.

In order to increase the three-phase boundaries, an ionomer is typically incorporated into the catalyst layer in the state-of-the-art electrodes [16,17]. The ionomer can be added into a PTFE-bonded electrode by a variety of methods such as brushing, spraying, and dipping. More recently, an ionomer is incorporated into electrodes by directly mixing it with Pt/C [18]. This approach significantly increases the number of the three-phase boundaries, but the overall catalyst utilization is still far from satisfactory. Park et al. suggested that even in the best performing electrodes the Pt utilization was around 10-25% [19]. However, Janssen and Sitters reported catalyst utilization ranging from 41 to 64% based on the electrochemical surface area measured by cyclic voltammetry [20]. Rajalakshmi et al. showed through cyclic voltammetry that electrodes made with Pt/CNT had a Pt utilization of around 44% [21]. Fig. 3 depicts a mixture of Pt/C and an ionomer, and it is clear that the three-phase boundaries are not easy to create.

Carbon

Bare Pt:Inactive

Pt at 3-phaseregions:Active

Pt covered by ionomer:Less active or inactive

Carbon

Bare Pt:Inactive

Pt at 3-phaseregions:Active

Pt covered by ionomer:Less active or inactive

Figure 3. Illustration of three-phase boundaries in a mixture of Pt/C and ionomer.

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.


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.

3.1. Dual Conducting Support

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.

Zhigang Qi 388

3.2. Conducting Polymer/Polyelectrolyte Composite

The concept was demonstrated by depositing Pt on composites of conductive polymers and polyelectrolytes. Fig. 5 shows the chemical structures of several representative conducting polymers. The conductive polymer transports electrons, and the polyelectrolyte in the acid form transports protons. However, conventional chemical polymerization of a conductive polymer typically leads to a precipitate that is not processable and it is difficult to use as a catalyst support. By using a polyelectrolyte such as sodium polystyrenesulfonate (PSS) as a molecular template that can pre-concentrate the monomer such as pyrrole and an oxidant such as Fe3+, Qi and Pickup successfully synthesized polypyrrole/PSS composite particles with variable sizes from ca. 30 nm to 1 μm [23]. They found that smaller composite particles with higher electronic conductivity were obtained when higher concentrations of pyrrole and Fe3+ were used. The composite particles did not dissolve in either water or organic solvents and did not form colloidal dispersion. This made them excellent candidates for use as supports for catalyst particles. Washing the composite using water might not be able to completely remove Fe3+ that was associated with the polyelectrolyte as counter ions (typically < 2% wt.). These counter ions could be completely removed by washing using a dilute acid such as HNO3.

* HC CH *npolyacetylene

NH

* *n

Polypyrrole

S* *n

Polythiophene

* NH *n

Polyaniline

S* *n

Poly(3,4-ethylenedioxythiophene)

O O

N* *n

Poly(3-methylpyrrole)CH3

Figure 5. Chemical structures of polyacetylene, polypyrrole, poly(3-methylpyrrole), polyaniline, polythiophene, and 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.


Unfortunately, the electronic conductivity of PPy/PSS was reduced by 3 orders of magnitude after the Pt deposition. Impedance measurement indicated that the proton conductivity of the electrodes made using Pt/PPy/PSS and PTFE was about 6 mS cm-1 at around 0.0 V, and it decreased to about 2 mS cm-1 when the voltage was increased to 0.4 V. This reduction in proton conductivity was believed to be due to the expulsion of protons from PSSH when PPy became more oxidized at higher potentials. The proton conductivity of the composite was not affected by the catalyst deposition.

Figure 6. TEM of Pt nanoparticles dispersed on PPy/PSS composite particles [24]. Reproduced by permission of The Royal Society of Chemistry.

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

Zhigang Qi 390

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.

Pt alloys such as Pt-Ru were also deposited onto conductive polymer/polyelectrolyte composite particles by reducing a mixture of H2PtCl6 and RuCl3 using either formaldehyde or hydrazine at reflux [26]. Yields for Pt-Ru alloys were somehow lower than those for Pt. Also, when poly(vinylsulfate) was used as the polyelectrolyte, the yields were lower than when polystyrenesulfonate was used. The loss of the electronic conductivity of the polymer composite was more severe when hydrazine was used as the reducing agent than formaldehyde during the catalyst deposition step.

Disappointing results were obtained when Pt/PEDOT/PSS was used to catalyze HOR [26]. Impedance spectroscopy showed a significant increase in electronic resistance of the polymer composite at potentials lower than 0.25 V. This is typical for organic conducting polymers because they are reduced to the non-conducting neutral form at such lower potentials. In addition, the presence of H2 was expected to accelerate the reduction process because H2 could chemically reduce the polymer. So, the low performance of the catalyzed composite was due to the loss of the electronic conductivity.

In summary, the following obstacles must be overcome before conducting polymer can be used as Pt supports for commercial applications. First, the electronic conductivity of the composite is often seriously reduced during the Pt deposition process. Second, the resulting Pt/composite gradually loses its electronic conductivity during storage. Third, the irreversible


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. Surface Functionalization of Carbon

Since carbon does not have issues associated with the loss of the electronic conductivity, it should be a more practical support. The question is how to make it conduct protons.

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).

Zhigang Qi 392

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.

Unfortunately, carbon oxidation is a corrosion process where the morphology of the carbon particles can be changed and some carbon can be consumed to form gases such as carbon dioxide (CO2) under strong oxidizing conditions. This can adversely affect the fuel cell performance.

Carbon oxidation can also happen at the cathode via an electrochemical route due to the voltage when a fuel cell is under operation. The early stage of carbon oxidation may contribute to the increased performance during the fuel cell incubation stage. Continuous oxidation, however, can lower the resistance to flooding, and severe corrosion of the carbon support will cause other problems as discussed later.

4.2. Chemical Linkage

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


trichlorosilane in anhydrous dichloromethane under the protection of nitrogen. Unreacted silane was washed away by dichloromethane during filtration. The silane molecules that attached to the surface of the carbon went through a hydrolysis process by absorbing moisture in air to become crosslinked. Pt deposition was performed for silane treated Vulcan XC-72 by the reduction of H2PtCl6 using formaldehyde at reflux.

Optimal Nafion® content in electrodes made with sulfonated silane-treated Pt/C was around 10% wt., while the optimal Nafion® content for electrodes made with untreated Pt/C was 30% wt. The performance of the former electrode with 10% Nafion® was only slightly lower than that of the latter with 30% Nafion®. When 10% Nafion® was used with the untreated Pt/C to make the electrode, its performance was much lower than that obtained by silane-treated Pt/C (Fig. 9). Clearly, the presence of sulfonated silane contributed significantly to the proton conductivity in the electrodes. It was found that modification of the carbon support prior to the Pt deposition was more effective than modification of Pt-catalyzed carbon, presumably due to the blocking of the active Pt sites by the silane in the latter case. Also, estimates indicated that the sulfonate loading was similar at the optimal Nafion® loadings for untreated and silane-treated Pt/C electrodes. Fuel cells showed no performance loss in 12 hours of operation, indicating that the attached sulfonated silane groups were stable in the fuel cell environment.

Figure 9. Polarization curves of (, ) untreated; and (Δ, ) sulfonated silane treated catalysts [29]. Reproduced by permission of The Electrochemical Society, Inc.

4.2.2. Diazonium Salt Although the sulfonated silane-treated Pt/C electrode reduced the optimal Nafion® content to one third of the untreated Pt/C electrode, the best performance of the former was slightly

Zhigang Qi 394

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.

NH2(CH2)2PO3H2 + (CH3)2CHCH2ONO → +N2(CH2)2PO3H- + (CH3)2CHCH2OH + H2O (3)

Pt/Carbon-H + +N2(CH2)2PO3H- → Pt/Carbon-CH2CH2PO3H2 + N2 (4) NH2(CH2)2SO3H + (CH3)2CHCH2ONO → +N2(CH2)2SO3

- + (CH3)2CHCH2OH + H2O (5) Pt/Carbon-H + +N2(CH2)2SO3

- → Pt/Carbon-CH2CH2SO3H + N2 (6)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Current density (A cm-2)

Cel

l Vol

tage

(V)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Pow

er Density (W

cm-2)

Figure 10. Performance of (, ) phosphonated; and (, ) unphosphonated 20% Pt/Vulcan XC-72 [30]. Reproduced by permission of The Royal Society of Chemistry.

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%


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].

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

Current Density (A cm-2)

Cel

l Vol

tage

(V)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Pow

er D

ensi

ty (W

cm

-2)

V-I, Untreated, 30.0% NafionV-I, Sulfonated, 17.5% NafionW-I, Untreated, 30.0% NafionW-I, Sulfonated, 17.5% Nafion

,

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)nΦmSO3H with m and n ranging from 0 (m and n do not equal 0 at the same time) and up (Φ 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

Zhigang Qi 396

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.

C6F5NH2 + (CH3)2CHCH2ONO + HBF4 → +N2C6F5 BF4

- + (CH3)2CHCH2OH + H2O (7) Pt/Carbon-H + +N2C6F5 BF4

- → Pt/Carbon-C6F5 + HBF4 + N2 (8)

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0


Cel

l Vol

tage

(V)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Pow

er D

ensi

ty (W

cm

-2)

TreatedUntreated

Figure 12. Performance of () untreated; and (Δ) –C6F5 treated 20% Pt/Vulcan XC-72 [32]. Reproduced by permission of The Electrochemical Society, Inc.


4.2.3. Inorganic Salt Xu el al. also discovered a simpler method that enabled the direct linkage of –SO3H groups onto the surface of carbon by thermally decomposing a sulfur-containing inorganic salt such as (NH4)2SO4 at 235 oC in the presence of carbon particles [33]. They proposed that (NH4)2SO4 decomposes according to Eq. 9, and the SO3 formed thereby reacts with carbon via its surface hydrogen atoms to have –SO3H groups linked onto the surface according to Eq. 10:

(NH4)2SO4 → 2NH3 + 2H2O + SO3 (9) Carbon-H + SO3 → Carbon-SO3H (10) 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.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8


Cel

l Vol

tage

(V)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Pow

er Density (W

cm-2)

Treated with 20% ammonium sulfate, 17.5% NafionTreated with 10% ammonium sulfate, 17.5% NafionTreated with 5% ammonium sulfate, 17.5% NafionUntreated, 30% Nafion

Figure 13. Polarization curves of 20% Pt/Vulcan XC-72 treated with () 0; (Δ) 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

Zhigang Qi 398

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.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6


Cel

l Vol

tage

(V)

Heated at 135 C, no ammonium sulfateHeated at 235 C, no ammonium sulfateHeated at 235 C, 20% ammonium sulfate

Figure 14. Effect of drying temperatures on fuel cell performance in the (◊, ) absence; and (Δ) 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


(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.

Carbon O Si CH2 CH2 S

Carbon CH2 CH2 P

Carbon CH2 CH2

Carbon

F F

F

FF

O

O

OH

O

OH

OH

S

O

O

OH

Carbon S

O

O

OH

Sulfonated silane

Ethylphosphonic acid

Ethylsulfonic acid

2,3,4,5,6-pentafluorophenylSulfonic acid

Figure 15. Chemical structures of functional groups.

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.

Zhigang Qi 400

4.3. Polymer Grafting

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 16. Illustration of a polymer modified Pt/C.

Figure 17. Polarization curves of (, ) untreated; and (, Δ) 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.


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

Zhigang Qi 402

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.

5.1. Carbon Corrosion

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.

Membrane

Cathode

Anode

Air (D) Air (B)

Hydrogen (A)Air (C)

H2 → 2H++2e-

O2+4H++4e- → 2H2O

O2+4H++4e-→ 2H2O

C+2H2O → CO2+4H++4e-

2H2O → O2+4H++4e-

~0.0 V

~1.0 V~2.0 V

H+ H+e-

e-

~1.0 V

Figure 19. Illustration of localized high voltage at cathode due to formation of air/H2 boundary at anode [39]. Reprinted from J. Power Sources 158, Tang, H.; Qi, Z.; Ramani, M., & Elter, J., PEM fuel cell cathode carbon corrosion due to the formation of air/fuel boundary at the anode, 1306-1312, Copyright (2004), with permission from Elsevier.

Thermodynamically, carbon can go through the oxidation process shown by Eq. 11 at around 0.2 V. However, the reaction kinetics is slow and carbon typically shows little corrosion at such a voltage. Unfortunately, when carbon is used at the cathode, its potential can be 1.0 V at OCV, although during the operation of the fuel cell, the voltage is typically lower than 0.8 V. In addition, during the fuel cell startup and showdown processes, a fuel/air boundary may form at the anode when air gradually gets into the anode side that still contains some unreacted hydrogen. Such a boundary results in a localized cathode voltage that can be


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].

C + 2H2O → CO2 + 4H+ + 4e- (11) 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 Faraday’s 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 μA 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 μA cm-2.

Table 1. Corrosion current density vs. complete carbon support consumption

Corrosion current density (μA/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

5.2. Carbon Graphitization

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.

Zhigang Qi 404

Unfortunately, the graphitization of a Pt catalyzed carbon causes the Pt particles to grow due to the high temperature. At the same time, when Pt is deposited onto the surface of a graphic carbon, the resulting Pt particles are much larger than those depositing on an amorphous carbon under the same condition. This is because a graphitic carbon contains fewer edge defects that allow the anchor of the Pt particles. So, although a graphitic carbon support provides a higher stability, the dispersion and the utilization of the catalyst particles are normally lower.

Carbon nanotubes (CNTs) have definitive graphitic surfaces with no basal plane edges and thus they are expected to be highly stable against corrosion. Li and Xing studied the electrochemical stability of electrodes made of CNTs and carbon black, respectively, by holding the potential at 1.2 V for different number of hours [42]. They found that the electrochemical oxidation of the noncatalyzed and catalyzed CNTs was basically stopped after an initial period of corroding time, although the catalyzed CNTs showed higher degree of surface oxidation. In contrast, both the noncatalyzed and catalyzed carbon black showed continuous oxidation, and the amounts of the surface oxide groups per unit mass were several times higher than those on the CNTs (Fig. 20). They proposed that the oxidation of the CNTs were limited to the top one or two graphene layers that had been damaged by a prior acid treatment step (Fig. 21). Shao et al. showed that the electrochemical active surface area of Pt/C degraded by 50% during 192 hours of accelerated testing, while that of Pt/CNTs only lost 26% [43]. They found that the loss was mainly due to the Pt particle growth, and proposed that the higher stability of Pt/CNTs were probably due to a stronger interaction between Pt and CNTs and the higher resistance of the CNTs to the electrochemical oxidation. Studies by Wang et al. showed that Pt/Vulcan XC-72 lost almost 80% of the Pt surface area, while Pt/MWNT lost only 37% in 168 hours of accelerated testing. They also found that the presence of Pt enhanced the corrosion current by 70% for Vulcan XC-72, but only by 40% for MWNT [44].

Figure 20. Surface oxidization with time for () 20% Pt/carbon black; and (Δ) 20% Pt/CNT [42]. Reproduced by permission of The Electrochemical Society, Inc.


Multiwalled CNT

Oxidized CB

Graphene Layers

Pre-treated CNT

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.

5.3. Carbon Functionalization

Hypothetically, carbon corrosion could be slowed down by a proper surface functionalization for several speculated reasons. First, the functionlization process may use up the active sites (e.g., the more easily corroding sites) on the surface of carbon, leaving the sites with more corrosion resistance unoccupied. Second, the functional groups that are attached to the surface of carbon may be able to act as a shield that protects the carbon surface from being corroded. Third, the functional group may go through an oxidation-reduction process that consumes the charge that is otherwise to be used to oxidize the carbon. However, there is no reported study on the long term durability of surface functionalized carbon for PEM fuel cell applications up to date.

Carbon surface functionalization could also be used to help solve some other problems. For example, CO at part-per-million (ppm) levels can seriously poison a PEM fuel cell anode. A functional group that can assist CO oxidation should be able to significantly enhance the CO-tolerance of a PEM fuel cell, and thus dramatically simplify the fuel cell system. For another example, carbon surface functionalization may be able to reduce the generation of hydrogen peroxide. It has been found that the presence of carbon in the catalyst layer promotes the formation of hydrogen peroxide, whose radicals can attack the ionomer to cause it to decay gradually. This has been accepted as one of the major reasons that cause the failure of the membrane in a PEM fuel cell. The change of the physical and the chemical properties of the surface of carbon after being modified by suitable functional groups could lead to the reduction of hydrogen peroxide formation. Alternatively, if the functional groups can not

Zhigang Qi 406

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.

5.4. Carbonless Electrodes

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 μm) 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].

Gullá 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


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.

Zhigang Qi 408

[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.


[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.

Zhigang Qi 410

[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] Gullá, 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 oxidation–resistant indium tin oxide catalyst support for proton exchange membrane fuel cells. J. Power Sources, 161, 893-900.


Chapter 9

A STUDY ON THE ROLE OF CARBON SUPPORT MATERIALS FOR FUEL-CELL CATALYSTS

Seok Kim1* and Soo-Jin Park2** 1Advanced Material Division, Korea Research Institute of Chemical Technology,

P.O. Box 107, Yusong, Daejeon 305-600, Korea (South) 2Dept. of Chemistry, Inha University, 253, Nam-gu, Incheon 402-751, Korea (South)

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: [email protected], Tel: +82-42-860-7232, Fax: +82-42-861-4151 (First Author) ** E-mail address: [email protected], Tel: +82-42-860-7234, Fax: +82-42-861-4151 (Corresponding Author)

Seok Kim and Soo-Jin Park 412

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-Teller’s (BET’s) surface area. However, the effect of the surface characteristics of the various carbon materials had not been fully studied to our best knowledge.

2. Literature Survey on Carbon Supports

A. Preparation Methods

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

step of Pt and Ru precursors (e.g., H2PtCl6, RuCl3, Pt(NH3)2(OH)2, Ru3CO12, Pt(NH3)2(NO)2) followed by a reduction step of Pt and Ru procedures. The first is characterized by a deposition step. This can be the chemical reduction of the electrocatalyst slurry in aqueous solution by using N2H4, NaS2O5, NaS2O3, NaBH4 (liquid-phase reduction) or the gas-phase reduction of the impregnated carbon black using a flowing hydrogen stream. It has been shown that the impregnation method can be used for the synthesis of multifunctional systems, e.g. from a bimetallic to a quaternary electrocatalyst. However, it requires the use of high surface area carbon black such as Ketjen black whose limitations for the operation of a methanol fuel cell have been pointed out above. Furthermore, this procedure does not allow one to obtain high dispersions in the presence of high metal loadings.

There are various colloidal deposition routes, as used by Jalan [21], Bőnnemann et al. [22], Petrow and Allen [23]. The advantages of these preparation routes stem from the fact that significantly higher surface areas can be attained in the presence of a high metal loading on carbon. The main disadvantages are the complexity of the preparation steps in the overall synthesis the use of organic compounds/solvents and the higher production costs.

The choice of the most appropriate preparation procedure relies on the following considerations. It is well known that the preparation procedure for electrocatalysts influences their physico-chemical properties and thus their activity. The performance characteristics of an electrocatalyst depend on its chemical composition (surface and bulk), structure and morphology. Accordingly, the selected methodology of electrocatalyst synthesis should allow one to address the process for the attainment of a proper structure (crystalline or amorphous) and with a chemical composition on the surface as close as possible to the nominal or bulk composition.

Since the rate of all electrocatalytic reactions is strictly related to the active surface area, besides the surface chemistry, the morphology of the electrocatalyst needs to be tailored. Morphology is not only related to the metal-phase area but also to the presence of micro- and macro pores in the electrocatalyst support that could facilitate or hinder the mass transport properties. All these characteristics determine the cell performance even if the relative influence of each parameter is still not known in detail. It is thus necessary to select appropriate procedures for the optimization of these characteristics, i.e. composition, structure, particle size, porosity, etc. Generally a combination of physico-chemical and electrochemical analyses carried out on different electrocatalysts indicates the system that best suits the scope of application in a DMFC.

B. Role of the Carbon Support

In general preparation procedures, such as impregnation, colloidal deposition and surface reaction involve the adsorption of active compounds on a carbon black surface. The synthesis of a highly dispersed electrocatalyst phase in conjunction with a high metal loading on carbon support is one of the present goals in the field of DMFCs. In this regard it is of interest to determine which carbon black is most suitable as a support. In recent reports [24-26], the most used carbon blacks were: Acetylene Black (BET Area: 50 m2/g), Vulcan XC-72(BET Area: 250m2/g) and KETJEN Black (BET Area: 1000m2 /g).

All these materials have optimal electronic conductivity but, as denoted above, they differ in their BET surface area and thus very probably in morphology. High surface area carbon


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.

C. Various Carbon Supports

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


and then sulfonation of the resulting compounds. Membrane-electrode assemblies prepared using these electrode materials showed an improved power density compared with a commercial 60% PtRu/Vulcan XC-72. This performance is ascribed to the higher ionic conductivity of the anode layer and well-exposed catalytic active sites.

Highly dispersed Pt nanoparticles supported on sulfonated carbon nanotubes (Pt/sulfonated-CNTs) with a high density of sulfonic acid groups were recently prepared, and then a Pt/sulfonated-CNTs modified GC electrodes was constructed successfully [34]. Sulfonation showed a positive effect of electroactivity of catalysts.

PtRu catalysts with MCMB as support [35] showed lower polarization characteristics than that with CB as support. Pt-Ru nanoparticles (1.6 nm) were supported on carbon nanotubes (200nm diameter, 8-10 um length) obtained by carbonization of PPy on an alumina membrane [36]. The amount and morphology of Pt nanoparticles depend on the types of carbon nanomaterials, including GNFs or CNTs [37]. Surfactant-stabilized Pt and Pt/Ru electrocatalysts for PEMFC had been prepared and investigated by X. Wang [38].

3. Acid/Base Treatments of Carbon Blacks Supports

A. Introduction

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.


B. Preparation of Carbon-Supported Pt Nanoclusters

Carbon-supported Pt nanoclusters were prepared by a modified polyol synthesis method, as reported previously [45]. Carbon materials (500 mg) were suspended and stirred by ultrasonic treatment for 20min in ethylene glycol (EG) solution, and then a 104mg hexachloroplatinic acid (H2PtCl6) (Pt: 50 mg; Pt weight percent: 10 wt.% against carbon materials)-dissolved EG solution was added dropwise slowly to the EG solution, which was stirred mechanically for 4 h. NaOH solutions of various concentrations (0.1 M, 0.5 M, 1.0 M, and 3.0 M in EG solution) was added to adjust the pH of the solution, and then the solution was heated at 140°C for 3 h for complete reduction of the Pt. The whole preparation process was conducted under flowing Argon gas. The solid was filtered and washed with 2 L of deionized water and then dried at 70°C for 24 h.

Virgin carbon blacks (VCB) of 24 nm size, 153 (cc/100g) DBP adsorption and 112 (m2/g) specific surface area, supplied by Korea Carbon Black Co., were used in our experiments. Neutral-treated carbon blacks (NCB), base-treated carbon blacks (BCB), and acid-treated carbon blacks (ACB) were prepared by treating the VCB with 99.9% C6H6, 0.2 N KOH, and 0.2 N H3PO4, respectively. Prior to each analysis, the residual chemicals were removed by Soxhlet extraction, that is, by boiling with acetone at 80°C for 3 h. Finally, the CBs were washed several times with distilled water and dried in a vacuum oven at 90°C for 12 h.

To assess the acid-base characteristics of the CBs after chemical treatment, the pH of the CB slurry was determined by a boiling method in accordance with ASTM D1512. The boiling method was performed by placing 5g of the CBs into 50 ml of deionized water contained in a 100 ml beaker. The pH measurement was taken after 2 min equilibrium time.

The acid and base values of the surface functional groups of the samples were determined by Boehm’s titration method [50]. To determine the acid value, 0.1g of the sample was added to 100 ml of 0.1 M NaOH solution and the mixture was shaken for 24 h. The solution was then filtered through a membrane filter (pore size = 0.24 μm, nylon) and titrated with 0.1 M HCl. Likewise, the base value was determined by the reverse titration of the acid value. The specific surface areas (SBET, [51]) of the samples were determined by gas adsorption. Physical adsorption of gases was used to characterize the CBs support, and the adsorbate used was N2 at 77 K with automated adsorption apparatus (Micromeritics, ASAP 2400). Prior to adsorption measurements, the samples were outgassed at 298 K for 6 h to obtain a residual pressure of less than 10-3 torr in high vacuum. To analyze the functional groups of CBs, the treated CBs were subjected to infrared (IR) spectroscopy (FTS-165 spectrometer, Bio-Rad Co.).

All of the Pt-based carbon catalyst (Pt/C) samples were characterized by recording their X-ray diffraction (XRD) patterns on a Rigaku X-ray diffractometer (Model D/Max-III B) using Cu Kα radiation with an Ni filter. Pt loading efficiency was calculated by comparing the atomic ratio of Pt intensity against the carbon intensity using the Energy Dispersive X-ray Spectroscopy (EDS) method coupled with Scanning Electron Microscopy (SEM). Alternatively, the Pt loading efficiency was measured using Jobin-Yvon Ultima-C Inductively Coupled Plasma-Atomic Emission Spectrometer (ICP-AES). The samples were also analyzed by X-ray photoelectron spectroscopy (XPS) on a VG ESCALAB MKII spectrometer. The Pt4f


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

To determine the condition of the formation of the carbon-supported Pt nanoclusters, various NaOH concentrations of 3.0 M, 1.0 M, 0.5 M, and 0.1 M were added into H2PtCl6- and VCB-containing EG solution. Figure 1 shows the powder X-ray diffraction patterns of Pt/CBs, prepared in (a) 3.0 M, (b) 1.0 M, (c) 0.5 M, and (d) 0.1 M NaOH aqueous solutions. All of the samples show typical Pt crystalline peaks of Pt (111), Pt (200), and Pt (220), except for those prepared in 0.1 M concentration. This indicates that the Pt clusters prepared in the 0.1 M NaOH concentration show a rather low loading efficiency. From this result, it was confirmed that the base concentration in the Pt precursor- and carbon-dispersed solutions is a critical factor in the manufacture of carbon-supported Pt nanoclusters.

0 20 40 60 80

500

1000

Pt(2

20)

Pt(2

00)

(d)

(c)

(b)

(a)

Pt(1

11)

Inte

nsity

/ a.

u.

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.


Table 1. Size and loading efficiencies of platinum nanoclusters of Pt/CBs, prepared in different NaOH aqueous solutions

NaOH Conc.

Avg. Size (nm)

Loading Efficiency a (%)

Loading Efficiency b (%)

3.0 4.90 87 83 1.0 3.07 90 88 0.5 11.40 82 79 0.1 - 31 28

a: EDS measurements b: ICP-AES measurements

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].

θ

λcosBKLC = (1)

where Lc is the mean size of the Pt particles, K is the Scherrer constant (=0.89), λ=0.154 nm, B is the half-height width of the (220) diffraction line, and θ 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


Pt nanoclusters. However, the values of Pt loading efficiency for the samples show an identical tendency.

D. Surface Treatment Effects

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 2. Surface and adsorption parameter results for carbon blacks

Carbon Blacks pH Acid value

[meq/g] Base value

[meq/g] SBET

a [m2/g]

VCB 7.0 50 70 112 NCB 7.0 40 58 114 BCB 7.1 70 90 118 ACB 2.6 110 13 68

a: specific surface area measured by gas adsorption

Table 3. Structural parameter results for carbon blacks obtained by XRD analysis

Carbon Blacks

d002a

nm Bb Lcc (nm)

VCB 0.3648 2.04 3.94 NCB 0.3636 2.06 3.90 BCB 0.3657 1.98 4.06 ACB 0.3665 1.37 5.86

a: interlayer spacing b: half-height width of the (002) diffraction line c: crystalline height

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


CB aggregation, resulting in increased crystalline size and increased concentration of crystalline boundaries as a typical turbostatic CB structure. A similar result was shown in the case of the base treatment of VCB (BCB) in the XRD studies. The values of d002 and Lc for NCB, which was treated by a nonpolar solution, were slightly reduced to 0.3636 nm and 3.90 nm, respectively. Thus, it was confirmed that acid treatment is most effective for changing the crystalline micro-structure of CBs. Contrastingly, a base or neutral treatment shows only a slight change of CB micro-structure.

500 1000 1500 2000 2500 3000 3500 4000 4500

3430291016301240

(d)

(c)

(b)

(a)

Tran

smitt

ance

Wavenumber (cm-1)

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 800°C 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


from the Pt reduction reaction being incomplete due to the high-acidic characteristics of the CB surfaces.

0 10 20 30 40 50 60 70 800

1000

2000

3000

4000

Pt(2

20)

Pt(2

00)

Pt(1

11)

Inte

nsity

/ a.

u.

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.

Table 4. Size and loading efficiencies of platinum nanoclusters of Pt/VCB, Pt/NCB, Pt/BCB, and Pt/ACB

Sample Avg. Size (nm)

Loading Efficiency a (%)

Loading Efficiency b (%)

Pt/VCB 3.07 90 88 Pt/NCB 3.74 93 89 Pt/BCB 2.65 97 94 Pt/ACB - 38 36

a: EDS measurements b: ICP-AES measurements

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


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.

64 66 68 70 72 74 76 78 80 82 84

0

500

1000

1500

VI

I

VIII

IV

II

(a)

Inte

nsity

(cps

)

Binding Energy (eV)64 66 68 70 72 74 76 78 80 82 84

0

500

1000

1500

VIII VI

IV

III

(b)

Inte

nsity

(cps

)

Binding Energy (eV)

64 66 68 70 72 74 76 78 80 82 84

0

500

1000I

VIIII

IV

II

V

(c)

Inte

nsity

(cps

)

Binding Energy (eV)64 66 68 70 72 74 76 78 80 82 84

0

500

1000

(d)

VIV

VIIII

II

I

Inte

nsity

(cps

)

Binding Energy (eV)

Figure 4. XPS Pt4f spectra of (a) VCB, (b) NCB, (c) BCB, and (d) ACB.


276 278 280 282 284 286 288 290 292 294

0

40000

80000

I

IIIII

(a)In

tens

ity (c

ps)

Binding Energy (eV)

276 278 280 282 284 286 288 290 292 294

0

40000

80000

(d)

I

IIIII

Inte

nsity

(cps

)

Binding Energy (eV)

276 278 280 282 284 286 288 290 292 294

0

40000

80000

(b)

I

IIIII

Inte

nsity

(cps

)

Binding Energy (eV)

276 278 280 282 284 286 288 290 292 294

0

40000

80000

(c)

I

IIIII

Inte

nsity

(cps

)

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.


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)

Binding Energy (eV) Sample 4f7/2 4f5/2

Relative Peak Area (%)

72.1 75.5 67.78 73.0 77.1 19.45 Pt/VCB 74.1 78.4 12.77 71.6 74.8 59.20 72.7 76.4 26.69 Pt/NCB 73.5 78.2 14.11 71.3 74.8 70.85 72.4 76.5 19.12 Pt/BCB 73.3 77.9 10.03 71.9 75.1 59.58 72.5 76.4 19.40 Pt/ACB 73.6 77.8 21.02

(b)

Sample Binding Energy (eV) Relative Peak Area (%) 284.6 76.53 285.9 13.75 Pt/VCB 286.9 9.72 284.6 72.67 285.8 17.08 Pt/NCB 286.8 10.26 284.6 72.40 285.9 17.49 Pt/BCB 286.9 10.10 284.6 70.64 285.8 19.29 Pt/ACB 286.9 10.07

In comparing the Pt4f spectra of the various chemical-treated carbon-supported catalysts,

it was observed that the Pt/ACB and Pt/NCB samples, in comparison with Pt/VCB, had a higher content of oxidized Pt species (Pt (II) and Pt (IV)) and a lower content of Pt (0). It is thought that the Pt could be easily oxidized in the case of using ACB and NCB supports. The C1s spectrum appeared to be composed of graphitic carbon (284.6 eV) and –C=O- like species (285.8 eV) [58]. A small amount of surface functional groups with high oxygen contents was also noted in the spectrum (286.8 eV). Although there were no prominent differences between the samples, Pt/ACB showed the smallest relative peak I area (70.64%) compared with that (76.53%) of Pt/VCB. Pt/NCB and Pt/BCB also showed decreased relative areas of 72.67%


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)

20nm20nm20nm

(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


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.

200 400 600 800 1000

0

1

2

3

bc

III

c

d

b

a

Cur

rent

Den

sity

(mA/

mg)

Voltage (mV vs. Ag/AgCl)

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.

4. Mixed Carbon Materials Supports

A. Binary Carbon Supports

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,


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 µm, 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.


Carbon-supported Pt nanoparticles were prepared by using a modified polyol synthesis method as reported before [27,45]. Binary carbon materials (500 mg) with a different mixing ratio of CBs and GNFs were suspended and stirred with an ultrasonic treatment for 20min in ethylene glycol (EG) solution and then a 104mg hexachloroplatinic acid (H2PtCl6) (Pt: 50mg, Pt weight percent is 10% against the carbon materials) dissolved EG solution was added dropwise slowly to the above solution and stirred mechanically for 4h. A 1.0M NaOH dissolved EG solution was added to adjust the pH of the solution to about 11. A formaldehyde (37%, 1.0 ml) aqueous solution was added to the solution to reduce Pt at 85°C for 1.5 h, and then the solution was heated at 140°C for 3h for a complete reduction of Pt. The whole preparation process was conducted under flowing Argon gas. The solid was filtered and washed with 2L of deionized water and then dried at 70°C for 24h.

To check an electroactivity of catalysts, cyclic voltammetry method for a three-electrode cell system was performed. We had prepared a working electrode by coating the catalyst powder mixed with Nafion® polymer onto a glassy carbon electrode. The preparation of thin film electrodes followed the previous researchers’ method [74-76]. Electrochemical impedance spectroscopy (EIS) measurements were studied by means of above electrochemical device coupled with FRA2 module (Eco Chemie, Netherlands) in a frequency range of 1 MHz ~ 0.1 Hz.

C. Particle Size and Loading Level of Catalysts

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 2θ = 26° and a small peak at 2θ = 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


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.

Table 6. Average sizes for platinum nanoparticle catalysts deposited on different carbon materials.

Carbon Samples

Average Size (nm)

GNF0 3.58 GNF10 4.87 GNF30 5.62 GNF50 6.13

GNF100 6.41

0 20 40 60 800

2000

4000

6000

Pt(3

11)

Pt(2

20)

Pt(2

00)

Pt(1

11)

(e)

(d)(c)(b)(a)

Inte

nsity

/ a.

u.

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


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.

Table 7. Loading levels of platinum nanoparticles and sulfur contents of 10 wt.% Pt/GNF-CB catalysts. (A 100% loading level means that the impregnated Pt weight

percent is 10% against the carbon materials. Sulfur content was referenced to the total weight of catalysts.)

Carbon Samples

Pt Loading levela

(%)

Pt Loading levelb

(%)

Sulfur Contenta (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

The Pt loading level is also measured independently by using an ICP-AES method. It is

also demonstrated in Table 7 for comparison. The values are slightly different from those of EDS method due to the existence of a little inhomogeneous distribution of Pt nanoparticles in the case of EDS method. However, the relative value of Pt loading level between samples shows similar trends.

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.


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.

Table 8. Electrochemical peak parameters of catalysts measured by a cyclic voltammetry.

Carbon Samples

Peak Potentiala

(mV)

Peak Currenta (mA cm-2)

Peak Potentialb

(mV)

Peak Currentb (mA cm-2)

GNF0 642 0.489 392 0.388 GNF10 664 0.618 508 0.260 GNF30 667 0.989 461 0.630 GNF50 679 0.449 438 0.320

GNF100 734 0.288 477 0.158 a: peak in forward scan, b: peak in backward scan

200 400 600 800 1000 1200-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

(a)

(e)I (m

A/cm

2 )

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).

E. Morphology of Catalyst Electrodes

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


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)

(c)

(b)

(d)

Figure 10. SEM micrographs of Pt catalysts deposited on (a) GNF10, (b) GNF30, (c) GNF50, and (d) GNF100 (Each scale bar means 2μm length).

F. Impedance and Resistance of Catalyst Electrodes

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


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.

0 100 200 3000

100

200

300

8.5 Hz

0.53 Hz

-Im (O

hm c

m2 )

Re (Ohm cm2)

(a) (b) (c) (d) (e)

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.

0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

Res

ista

nce

(Ohm

cm

2 )

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.


5. Plasma Treatment on Carbon Supports

A. Introduction

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

3500 3000 2500 2000 1500 1000 500

P50

P30

P20

Wavenumber (cm-1)

Tran

smitt

ance

(%) P0

P10

Figure 13. FT-IR spectra of the N2-plasma treated carbon blacks as a function of N2-plasma treatment time.


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.

1000 800 600 400 200 0

Pt/P50

Pt/P30

Pt/P20

Inte

nsity

(a. u

.)

Binding energy (eV)

Pt4f

O1s

C1sPt/P0

Pt/P10

N1s

Figure 14. XPS survey scan spectra of the N2-plasma treated Pt/carbon blacks catalysts.

Table 9. EA results of N2-plasma treated carbon blacks

Sample

P50

P10

P20

P30

P0

Elemental analysis

87.8

91.8

83.4

80.6

98.1

C1s N1s

10.7

3.8

5.3

10.6

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.

400 600 800 10001.0x10-4

0.0

-1.0x10-4

-2.0x10-4

-3.0x10-4

-4.0x10-4

-5.0x10-4

-6.0x10-4

-7.0x10-4

I (A

)

mV (Ag/AgCl)

Pt/P0 Pt/P10 Pt/P20 Pt/P30 Pt/P50

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.


6. Oxyfluorination Treatment on Carbon Supports

A. Introduction

The surface modification of carbon materials is of great importance in a wide variety of fields, such as structural applications [81], biomedicine [82], electrochemistry [83], microelectronics [84], and thin-film technology [85]. Surface modification is accomplished through different types of treatment: thermal treatment, wet chemical or electrochemical oxidation, plasma treatment, ion or cluster bombardment, covalent linkage of biomolecules, etc. The ultimate goal of these treatments is to change the surface chemistry and microstructure of the material and thus modulate a number of properties: biocompatibility, roughness, reactivity, conductivity, etc.

In recent years, controlled surface oxyfluorination has been investigated as an interesting industrial alternative. In comparison with other surface modification techniques, fluorine atoms penetrate the material surfaces to relatively great depths, the extent of which depends upon treatment conditions. In this work, the effects of the oxyfluorination for MWNTs on the surface functional groups and Pt catalyst deposition are investigated.

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 (O–H) at 3450 cm-1 are observed for the oxyfluorinated MWNTs. Particularly, the intensity of the O–H 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.


4000 3500 3000 2500 2000 1500 1000 500

300-MWNTs

200-MWNTs

100-MWNTs

R.T-MWNTs

neat-MWNTsTran

smitt

ance

(%)

Wavenumber (cm-1)

Figure 16. FT-IR spectra of the oxyfluorination treated Pt/MWNT catalysts.

20 40 60 80

Pt(111)Pt(200) Pt(220)

Pt/300-MWNTs

Pt/200-MWNTs

Pt/100-MWNTs

Pt/R.T MWNTs

Pt/neat MWNTs

Inte

nsity

(a. u

)

2θ

Figure 17. XRD spectra of the oxyfluorination treated Pt/MWNT catalysts.

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 2θ = 39.7˚ (111), 46.2˚ (200), and 67.4˚ (220). The broader diffraction peaks for the catalysts also lead to a smaller alloy particle


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

Fig. 18 shows cyclic voltammograms (CV) of oxyfluorinated Pt/MWNT catalysts in 0.5 M H2SO4 containing 1.0 M CH3OH. Pt/100-MWNTs show the best electroactivity among the samples due to its highest current density of anodic peak. This shows the oxyfluorinated Pt/100-MWNTs has high activity towards methanol oxidation. Anodic peaks of methanol oxidation are observed at about 750 ~ 850 mV.

300 600 900 1200

0.000

-0.001

-0.002

-0.003

-0.004

-0.005

I (A

)

mV (Ag/AgCl)

Pt/neat MWNTs Pt/R.T MWNTs Pt/100-MWNTs Pt/200-MWNTs Pt/300-MWNTs

Figure 18. Cyclic voltammograms of the oxyfluorination treated Pt/MWNT catalysts.

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 .


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 Boehm’s 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


component on the MWNT surfaces and other functional groups, such as C–O, C=O, HO–C=O, C–Fx, 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)


[24] P.S. Kaurenan, E. Skou, J. Electroanal. Chem., 408 (1996) 189. [25] M.K. Ravikumar, A.K. Shukla, J. Electrochem. Soc., 143, (1996) 2601. [26] J.B. Goodenough, A. Hamnett, B.J. Kennedy, S.A. Weeks, Electrochim. Acta, 32,

(1987) 1233. [27] W. Li, C. Liang, J. Qiu, W. Zhou, J. Qiu, Z. Zhou, G. Sun, and Q. Xin, J. Phys. Chem.

B, 107, (2003) 6292. [28] M. Carmo, V.A. Pagania, J.M. Rosolen, E.R. Gonzalez, J. Power Sources, 142 (2005)

169. [29] E.S. Steigerwalt, G.A. Deluga, D.E. Cliffel, and C.M. Lukehart, J. Phys. Chem. B, 105,

(2001) 8097. [30] C.A. Bessel, K. Laubernds, N.M. Rodriguez, and R.T.K. Baker, J. Phys. Chem. B, 105,

(2001) 1115. [31] J.B. Joo, P. Kim, W. Kim, J. Yi, Catalysis, 111 (2006) 171 [32] J.L. Figueiredo, M.F.R. Pereira, P. Serp, P. Kalck, P.V. Samant, J.B. Fernandes,

Carbon, 44 (2006) 2516 [33] J. Lee, Y. Park, O. Yang, Y. Kang, K, Jun, Y. Lee, H.Y. Kim, K. Lee, W.C. Choi, J.

Power Sources, 158 (2006) 1251 [34] H.J. Wang, H. Yu, F. Peng, P. Lv, Electrochem. Commun., 8 (2006) 499 [35] Y. Liu, X. Qiu, Y. Huang, W.T. Zhu, J. Power Sources, 111 (2002) 160 [36] Rajesh, K. Thampi, J. Bonard, B. Viswanathan, J. Material Chem., 10 (2000) 1757 [37] Lee, Y. Ju, P. Chou, Y. Huang, L. Kuo, J.C. Oung, Electrochem Commun., 7 (2005) 453 [38] X. Wang, I. Hsing, Electrochim. Acta, 47 (2002) 2981 [39] Z. Zhou, S. Wang, W. Zhou, G. Wang, L. Jiang, W. Li, S. Song, J. Liu, G. Sun, and Q.

Xin, Chem. Commun., (2003) 394. [40] W.X. Chen, J.Y. Lee, and Z. Liu, Chem. Commun. (2002) 2588. [41] Z. Liu, X.Y. Ling, J.Y. Lee, X. Su, and L.M. Gan, J. Mater. Chem., 13, (2003) 3049. [42] T.C. Deivaraj, W. Chen, and J.Y. Lee, J. Mater. Chem., 13, (2003) 2555. [43] H. Tang, J.H. Chen, Z.P. Huang, D.Z. Wang, Z.F. Ren, L.H. Nie, Y.F. Kuang, and S.Z.

Yao, Carbon, 42, (2004) 191. [44] M.S. Loffler, B. Gros, H. Natter, R. Hempelmann, Th. Krajewski, and J. Divisek, Phys.

Chem. Chem. Phys., 3, (2001) 333. [45] W. Li, C. Liang, J. Qiu, W. Zhou, H. Han, Z. Wei, G. Sun, and Q. Xin, Carbon, 40,

(2002) 791. [46] D.A. Stevens and J.R. Dahn, J. Electrochem. Soc., 150, (2003) A770. [47] H. Shioyama, Y. Yamada, A. Ueda, T. Kobayashi, Carbon, 41, (2003) 579. [48] D.L. Boxall, G.A. Deluga, E.A. Kenik, W.D. King, and C.M. Lukehart, Chem. Mater.,

13, (2001) 891. [49] C.H. Jung, W.Z. Oh, Y.R. Lee, and S.J. Park, Carbon Science, 6, (2005) 25. [50] H.P. Boehm, Adv. Catal., 16, (1966) 179. [51] S. Brunauer, P.H. Emmett, and E. Teller, J. Am. Chem. Soc., 60, (1938) 309. [52] K. Kinoshita, Carbon: Electrochemical and Physicochemical Properties, John Wiley,

New York (1988) p. 31. [53] V. Radmilovic, H.A. Gasteiger, and P.N. Ross, J. Catal., 154, (1995) 98. [54] K. Park, J. Choi, B. Kwon, S. Lee, Y. Sung, H. Ha, S. Hong, H. Kim, A. Wieckowski, J.

Phys. Chem. B, 106, (2002) 1869.


[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


Chapter 10

DEVELOPMENTS OF ALKALINE SOLID POLYMERELECTROLYTE MEMBRANES BASED ON POLYVINYL

ALCOHOL AND THEIR APPLICATIONSIN ELECTROCHEMICAL CELLS

G.M. Wu* and S.J. LinInstitute of Electro-Optical Engineering, Dept of Chemical and Materials Engineering,

Chang Gung University, Taoyuan 333, Taiwan R.O.C.C.C. Yang

Dept of Chemical Engineering, Ming Chi University of Technology,Taipei 243, Taiwan R.O.C.

Abstract

Alkaline solid polymer electrolyte membranes (ASPEM) have been extensively studiedfor the varied electrochemical device applications due to the thinner thickness, lowerelectrolyte permeability, higher ionic conductivity and ease in processibility. We prepared aseries 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, andPVA/PAA/PP/PE composites. These new material systems were introduced with uniqueimprovements for fuel cell applications. They have not only the potential to lower theprocessing cost but also can provide high ionic conductivity. The PVA/PAA polymer blendsystem exhibited high ionic conductivity of 0.301 S cm-1 and the anionic transport numbercould reach 0.99, both at room temperature. For Zn/air and Al/air battery application, thepower density was as high as 90-110 mW cm-2. The PVA-based composite SPE has greatpotential for use in alkaline battery systems. The progressive advancements in the science andtechnology of solid polymer electrolyte membranes are presented in this chapter.

We will demonstrate the preparation techniques for ASPEM and the characterizationresults. 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 cellfabrication. The alkaline solid state electrochemical systems, such as Ni-MH, Zn-air fuel cells,


G.M. Wu, S.J. Lin and C.C. Yang446

Al-air fuel cells, Zn-MnO2 and Al-MnO2 cells, were assembled with anodes, cathodes andalkaline solid polymer electrolyte membranes. The electrochemical cells showed excellent cellpower density and high electrode utilization. Therefore, these PVA-based solid polymerelectrolyte membranes have great advantages in the applications for all-solid-state alkalinefuel cells. Some other potential applications include small electrochemical devices, such assupercapacitors and 3C electronic products.

Introduction

In 1973, Wright et al. [1-3] discovered a class of solid polymer electrolytes (SPE) composedof polyethylene oxide (PEO) and alkali metal ions that exhibited conductivity at roomtemperature. 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 electrolytecharacteristics in lithium polymer batteries for 3C electronic products, and even for electricvehicles [5]. For the use of SPE in lithium battery, the polymer electrolytes usually containlithium salts such as LiClO4, LiBF4, LiPF6, LiCF3SO3 [6,7] and a non-aqueous characteristicis necessary. During the charge and discharge processes, the charge-carrier ion of Li-polymerelectrolyte is transported by lithium. However, the alkaline SPE system is totally differentfrom the lithium polymer electrolytes. It is an aqueous system and contains water. Thedissolved hydroxide (OH-) ion is transported from cathode to anode in SPE during thedischarge process and the electric current is thus generated. Fig. 1 shows the schematicillustration of alkaline H2/O2 fuel cell and Table 1 summarizes the characteristics for differentfuel cell systems. The alkaline fuel cell and Zn/air battery can be operated at ambienttemperature with stable electrochemical properties and lower cost.

Hydrogen inlet

Water outlet

Ve- e-

Cathode Anode

Oxygen inlet

Hydroxyl OH- ions

Electrolyte

e-

e-

Figure 1. Schematic illustration of alkaline H2/O2 fuel cell.

Developments of Alkaline Solid Polymer Electrolyte Membranes… 447

Table 1. The characteristic properties for different fuel cell systems.

Fuel cell typeOperating

temperature(oC)

Mobileion Characteristics

Alkaline fuel cell (AFC) R.T.-150 OH- Low cost, low operating temperature andrelatively low power applications ~50 kW.

Zn/air, Al/airMetal/air semi-fuel cell R.T.-100 OH-

Low cost, low operating temperature,mechanically rechargeable and low powerapplications ~10 kW.

Proton exchange membranefuel cell (PEMFC) 80-100 H+ Relatively low operating temperature, higher

cost.

Phosphoric acid fuel cell(PAFC) 200-300 H+ Relatively high operating temperature,

corrosion might be accompanied.

Solid oxide fuel cell(SOFC) 500-1000 O2- High operating temperature and wide power

applications ~100 MW.

Molten carbonate fuel cell(MCFC) 700-1000 CO3

2- High operating temperature and wide powerapplications ~100 MW.

The traditional aqueous alkaline batteries have some disadvantages that inhibit itsadvanced applications: (1) leakage and evaporation of liquid electrolyte, (2) corrosion ofelectrode, (3) instability in chemical and electrochemical characteristics, and (4) large sizeand heavy weight that reduce the energy and power density of battery. On the other hand, thealkaline SPEs have low cost, thin thickness, high flexibility, good film-forming characteristic,reasonable mechanical strength, no electrolyte leakage problem, and the high ionicconductivity at room temperature has been developed. For the past decade, compared to thelithium battery SPE, there are only few literatures to discuss these novel alkaline SPEmaterials. Fauvarque et al. [8-13] developed water-containing alkaline solid polymerelectrolytes based on PEO and KOH by a solvent casting method. The ionic conductivity wasup to 10-3 S cm-1. They demonstrated good cyclability of 60 cycles without short circuit andthe 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 membraneproperties 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)-KOHwith interpenetrating network (IPN) [15] showed high ionic conductivity and veryencouraging electrochemical performance. However, further developments are still requiredto improve the ionic conductivity, alkaline sustainability, mechanical properties, and waterholding stability. This chapter describes the recent advances in alkaline SPEs primarily basedon PVA polymers. The developments in PEO-based alkaline SPE and gel-type alkaline SPEare also briefly reviewed. The applications in electrochemical cells, especially metal-air fuelcells, are discussed.

Although Hassan et al. [16] continued to study PEO/KOH SPE, the ionic conductivityreached only about 3.1×10-5 S cm-1. The effect of KOH concentration on the ionicconductivity of PEO polymer electrolyte could be attributed to the decreased crystallinity of


PEO polymer with the increasing KOH concentration. This also increased the concentrationof conductive OH- ions in the polymer matrix. The optimum KOH concentration wascontrolled at 40-60 wt.% in PEO matrix. The amorphous phase in polymer matrix is thereforefavorable for ion transportation and movement.

For the first time, a totally solid-state electric double layer capacitor (EDLC) wasfabricated using PEO-KOH-H2O as the SPE and the polymer electrolyte could replace largeamount of liquid KOH electrolyte [17,18]. The ideal rectangular shape of cyclic voltammetryresult for this solid-state EDLC was obtained, and the real value of specific capacitance was90 F g-1. It was only slightly lower than that of liquid electrolyte supercapacitor, and it mightbe related to the electrode material and structure.

Alkaline PEO-based SPE may be limited by the high crystallinity and poor thermalstability. Vassal et al. [11] intended to use amorphous copolymer, poly(epichlorohydrin-co-ethylene oxide) P(ECH-co-EO), and presented good electrochemical properties withoutPEO’s disadvantages. The ionic conductivity value at room temperature was around 10-3 Scm-1. The all solid-state alkaline Ni/MH and Zn/air electrochemical cells also exhibited stablebehavior.

Inasmuch as the use of epichlorohydrin concept, Agel et al. [13] developed a new andcheap type of anion exchange membranes (AEM) by preparing the polyepichlorohydrin(PECH) graft quaternary amines (DABCO, TEA) for use in alkaline cells. It’s a quasi-gasimpervious 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 polymerside chains. For the first time, the alkaline SPE employed in alkaline fuel cell, the test resultsexhibited 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 theAEM MORGANE®-ADP membranes (Solvay, S.A.), a cross-linked fluorinated polymer withthe exchange group of quaternary ammonium, in direct methanol fuel cell (DMFC) and theyinvestigated the properties for preparing membrane electrode assembly. It was described thatthe DMFC with anionic exchange membrane in alkaline media showed better performancethan in acid media. This was attributed to that methanol oxidation catalyst has been lessstructure 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, whencompared to Nafion® membrane, it still showed higher resistance and needs to solve the mainhindrance 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 theelectrodes. But the poor mechanical property, high evaporation and easy leakage stalled thewide use in battery systems. Feuillade [20] studied PAN gel polymer electrolyte with physicalcrosslinking. Tsuchida et al. [21] prepared PVDF gel polymer electrolyte and found higherionic conductivity at 10-3 S cm-1 at ambient temperature. More recently, Girish [22] reportedPAN gelled polymer electrolyte with propylene (PC), ethylene carbonate (EC) and zinctrifluoromethane sulfonate (ZnTfs). The ionic conductivity was about 2×10-3 S cm-1 and thegelled polymer electrolyte exhibited good charge/discharge cyclibility for Zn|GPE|γ-MnO2

cell. The successful preparation of gelled polymer electrolyte required a cross-linkingprocedure, physical or chemical [23]. However, physical crosslinking gel may encounterserious problems at high temperature environment and under long-term storage, likedissolution and phase separation. The gelled polymer electrolyte can be prepared from low


molecular weight poly(vinyl alcohol), PVA and poly(acrylic acid), PAA. There exists aninteresting preparation method by the freezing/thawing techniques on PVA/PAA polymerblends [24,25]. They can absorb water more than 5 times of the polymer weight, but untilnow there have been no literature data for alkaline SPE preparation. For the chemicalcrosslinking preparation, we have previously introduced a free radical polymerization methodfor PAA. It requires a crosslinking agent with bi-functional or tri-functional groups to reactwith PAA and to form a stable structure of gelled polymer. Iwakura et al. [26] usedcommercial crosslinked PAA (Aldrich, #43532-5) to absorb KOH aqueous solution andapplied it on Ni-MH battery.

For the experimental results, the ionic conductivity value of gel polymer electrolytes hasbeen as high as that of the KOH aqueous solution (about 0.6 S cm-1). This is the typicalcharacteristic for gelled polymer electrolyte because the crosslinked structure can absorb andhold a large amount of KOH aqueous solution. The higher ionic conductivity is good for thebattery 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 withcrosslinked PAA/KOH. The hydrogel polymer electrolytes have been used on the electricdouble layer capacitor. The solid-state capacitor remained high discharge capacitance of 80 Fg-1 after 20,000 charge/discharge cycle. For low temperature tests, Zhang et al. [29] used PVASPE as gelled polymer electrolyte (σ=0.097 S cm-1) in the NiO/activated hybridsupercapacitor. The system showed higher specific capacitance of 65 F g-1 at 40 oC and theyeven obtained 30 F g-1 at low temperature of -20 oC.

2. Polyvinyl Alcohol (PVA)-Based Alkaline SPE

Polyvinyl alcohol, PVA or PVOH, is a nontoxic and highly hydrophilic polymeric materialwith semi-crystallinity due to the high density of hydroxy (-OH) groups [30,31]. It contains arandom distribution of syndiotactic and isotactic units [32], and has a high meltingtemperature of Tm=230 oC as a result of the high level of hydrogen bonding on every secondcarbon atom in the crystal structure. This hydrogen bonding is crucial in affecting the watersolubility for PVA polymer and it is also the major advantage for use in alkaline SPE. PVAposes some general characteristics, such as good strength, high water absorbility, low foulingpotential, long-term temperature and pH stability. Earlier work was prepared by mixingdissolved PVA solution with KOH to form the PVA-KOH-H2O alkaline SPE [14]. The ionicconductivity value reached the level of 10-3 S cm-1 with the composition of 42.2 wt.% PVA,29.9 wt.% KOH and 27.9 wt.% H2O. It was also shown broad electrochemical potentialstability windows at the metal-SPE interface. We studied PVA/KOH SPE and found that themolecular weight of PVA has been a major issue in preparing alkaline SPE. The highmolecular weight PVA provided low ionic conductivity due to the increase in polymercrystallinity. The ionic conductivity is examined by AC impedance analysis. Analysis of theAC spectra yields information about the properties of the polymer electrolyte, such as bulkresistance (Rb). The value is then converted to the ionic conductivity σ by the equation σ = l /(Rb×A), where l is the SPE film thickness (cm), A is the area of the blocking electrode (cm2),and Rb is the bulk resistance (ohm). For the demand in both ionic conductivity andmechanical strength, the optimal range of molecular weight of PVA polymer is at 75,000-


80,000. Fig. 2 displays typical AC impedance spectra for the alkaline PVA/KOH SPE atdifferent temperatures ranging from -20 to 90 oC. The ionic conductivity is highly dependenton the temperature and is followed by the Arrhenius equation: σ = σo × exp(-Ea/RT).

Zre''/ohm

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

-Zim

''/ohm

0

1

2

3

4

5

6

7

8-20oC-10oC 0oC10oC20oC30oC40oC50oC60oC70oC80oC90oC

-20oC-10OC

0oC10oC

20oC30oC

40oC

50oC60oC

70oC

80oC

90oC

Figure 2. AC impedance spectra for alkaline PVA/KOH SPE at -20 to 90 oC.

2.1. Preparation of Alkaline Polymer Battery

The best way to verify the applicability of SPE is to assemble it in A real practical batterysystem. The alkaline zinc manganese dioxide cell (Zn/MnO2) has many advantages of highpower density, longer shelf life, and better dimensional stability. It can be used undercontinuous or intermittent demand, and has been advanced to the portable battery market. Thedisadvantages that may limit its use in some devices are the slopping discharge curve andlower energy density. The Zn/air battery uses oxygen directly from the atmosphere to produceelectrochemical energy, and is thus classified as semi-fuel cell. Much research has beenfocused on this system due to its high energy density and flat discharge voltage. Aluminumelectrode provides another potential battery material because of its high theoretical Ahcapacity, high voltage and high specific energy. In this chapter, we assembled alkalinealuminum 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 cellstructure with alkaline SPE membrane.

The preparation procedure is briefly described for carbon-based air cathode and Zn gelledanode. 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

Zre’ / ohm


about 30 min at 300 oC. The active layer of the air electrode was then sprayed with a mixtureof 15 wt.% PTFE binder, 10–25 wt.% of KMnO4+MnO2 (2:1), 60 wt.% Vulcan XC-72R, andappropriate amount of isopropyl alcohol onto the gas-diffusion layer. The air electrode withboth the gas diffusion layer and the active layer was finally sintered for about 30 min at 360oC for binding. The thickness of the air cathode electrode was about 0.4-0.6 mm. On the otherhand, Zn anode gel was prepared according to the following formulation: 65 wt.% Znpowders, 0.5-1 wt.% Carbopol 940 gelling agent, 32 wt.% KOH, and some metal additivesused as hydrogen inhibitor.

Total reaction of zinc air battery: Zn + 1/2O2 ?ZnO EO = 1.65 V

Zinc gelled anode: Zn + 4OH¯ ?Zn(OH)2-

4 + 2e¯ Zn(OH)2-

4 ?ZnO + 2OH¯ + H2O EO = 1.25 V

Air cathode:O2+ 2H2O + 4e¯ ?4OH¯

EO = 0.40 V

PVA-based alkaline solidpolymer electrolyte

membrane

OH-

Air hole O2

Figure 3. The scheme of a solid state Zn/air cell with alkaline PVA SPE membrane.

2.2. PVA/KOH/H2O SPE

The prepared PVA/KOH/H2O SPE was employed for both Ni/MH and Zn/air batteries. Fig. 4shows typical charge and discharge curves of all solid-state Ni/MH battery. The resultsexhibited 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 successfullyassembled 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-termdischarge, the water content might lose quickly and still cause deformation of SPEmembrane. Therefore, the composite PVA/KOH alkaline SPE with glass-fiber-mat (GF) wasdeveloped to prevent the film from deformation and could enhance the mechanical properties[34]. It was further evidenced that the reinforcing GF improved membrane dimensionalstability without significantly affecting its ionic conductivity value. We compared thedischarge curves of solid state Zn/air cells with PVA/KOH/GF SPE and commercial PP/PEand cellulose separators at C/10. The results showed that at the high discharge rate, the Zn/aircell with PVA/KOH/GF SPE could reach higher Zn utilization of 86% and higher plateauvoltage than those of the other commercial separators.


time (hrs)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

E (m

V)

700

800

900

1000

1100

1200

1300

1400

1500

charge curve

discharge curve

time (hrs)

0 5 10 15 20 25 30 35 40 45 50 55

E (m

V)

300

400

500

600

700

800

900

1000

1100

1200

1300

1400

1500

1600

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 dischargecurves.

The PEO-PVA-H2O alkaline SPE was further developed and applied in the Zn/air andNi/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, after40 cycles of charge/discharge tests, the Ni(OH)2 active material still had an average capacitydensity of 250 mAh g-1. However, the PVA-PEO blend could form a porous and brittlestructure membrane. This is not favorable for Zn/air battery application, because the zincdendrite would penetrate through the air cathode and cause short-circuit during the discharge.


The composite model of PEO-PVA-H2O SPE with GF could be employed to improve thestability [36]. The zinc utilization for Zn/air cell was achieved at 84% when discharged at 150mA.

2.3. PVA/PECH SPE

In this section, PVA was blended with polyepichlorohydrin (PECH) in DMSO solution toprepare the PVA/PECH blend polymer membrane. The blend membrane was immersed in 6M 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 wasfound to be 1:0.2. This polymer blend formed a uniform and homogeneous film. High PECHcontent, such as PVA:PECH (1:1), resulted in phase separation morphology. The solid-stateZn/air batteries with PVA/PECH blend polymer electrolytes have been assembled and the testresults are listed in Table 2.

Table 2. The results of solid-state Zn-air cells with PVA-PECH solid polymerelectrolytes at C/10 discharge rate.

Zn-air cellItem PVA-PECH

(1:0) SPEPVA-PECH(1:0.2) SPE

PVA-PECH(1:0.5) SPE

Theo. capacity (mAh) 1,476 1,476 1,476Discharge current (mA) 150 150 150Discharge time (hr) 8.9 8.6 7.7Real capacity (mAh) 1338 1290 1160Zn utilization (%) 89 86 77

If the PECH content was further increased to the blend ratio of 1:0.5 to 1:1, phaseseparation was observed and the film exhibited a sticky surface. These issues posed problemsfor 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 theblend ratio of 1:1 sample exhibited the lowest KOH content and thus low ionic conductivity.

2.4. PVA/TEAC SPE

The anionic exchange concept has been employed on PVA polymer electrolyte to improve theionic conductivity of alkaline PVA-based SPE. The solvent casting method started byblending PVA and quaternary amine, such as tetraethyl ammonium chloride (TEAC) indimethyl sulfoxide (DMSO) to form PVA/TEAC blend polymer membrane [38]. Afterimmersed 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 greatlyreduced. It was suggested that the addition of TEAC into the PVA polymer matrix increasedthe domain of amorphous region and then the hydrophilicity. The XRD spectra are evidencedin Fig. 5.


2θ (degree)

10 20 30 40 50 60 70 80100

200

300

400

500

600

700

800

900 (a) PVA film(b) PVA/TEAC=1:0.2(c) PVA/TEAC=1:1(d) PVA/TEAC=1:2

inte

nsity

(arb

itrar

y un

it)

(a)

(b)

(c)

(d)

Figure 5. XRD spectra for PVA/TEAC with different blend ratios.

Table 3. The conductivity values (S cm-1) of alkaline PVA/PECH SPE at differenttemperatures.

Blend ratioT (oC) PVA:PECH

=1:0.2PVA:PECH

=1:0.5PVA:PECH

=1:130 0.0074 0.0042 0.001240 0.0081 0.0049 0.001850 0.0083 0.0053 0.002560 0.0091 0.0058 0.002970 0.0096 0.0066 0.0032

Table 4. The conductivity values (S cm-1) of alkaline PVA/TEAC SPE at differenttemperatures.

Blend ratiosT (oC) PVA/TEAC

= 10.2PVA/TEAC

=11PVA/TEAC

=1230 0.0106 0.0219 0.045940 0.0138 0.0223 0.046250 0.0164 0.0233 0.046460 0.0201 0.0249 0.048070 0.0224 0.0259 0.0497

These SPE films also exhibited uniform and homogeneous surface. No phase separationwas observed for all the blend PVA/TEAC polymer films, and all SPE films appearedtranslucent. As a result, the conductivity value of the alkaline PVA/TEAC SPE of PVA:TEAC = 1:2 (σ = 0.0459 S cm-1) was close to that of alkaline PVA polymer electrolyte at

( CH2-CH )n

OH


ambient temperature. The ionic conductivity of blend PVA/PECH and PVA/TEAC withdifferent blend ratios at different temperatures are displayed in Tables 3 and 4, respectively.

2.5. PVA/PAA SPE

In order to obtain higher ionic conductivity and suitable mechanical strength for alkaline PVASPE, we need to further enhance its water absorbility and hydrophilicity. The preparation ideawas to blend two highly hydrophilic polymers, polyvinyl alcohol (PVA) and polyacrylic acid(PAA) to form a cross-linked (interpenetrating) SPE membranes. The high hydrophilicity ofPAA 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 popularfor polymer membrane applications. However, its high water absorbing ability and strongwater 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 theapplicability of PAA for membranes: including grafting, crosslinking and blending. Habert[42] reported grafting PAA onto polysulfone membrane for dialysis and desalination viareverse osmosis. Ito et al. [43] grafted PAA onto the polycarbonate membranes to reduce thepore size. For the preparation of crosslinked PAA, the chain mobility and relaxation modewere studied [44]. Only few studies focused on the behavior and structure of PVA/PAA blendpolymers [45-48]. We synthesized cross-linked PAA and blended it with PVA to form theIPN structure of PVA/PAA membrane.

For the preparation procedure, the acrylic acid monomer with cross-linker was firstlyblended with PVA polymer. A free radical polymerization was carried out. The solid-statePVA/PAA polymer membranes were obtained by a solution casting method. The appropriateweight ratios of acrylic acid monomer (PVA:PAA varied from 10:3 to 10:7.5 PAA) and 1wt.% (vs. AA monomer) cross-linking agent, such as triallylamine, N,N-dimethyl acrylamide,glutaraldehyde or N,N-methylene bisacryalmide (MBAC) [44,49], were directly dissolvedand mixed in distilled water with agitation for 12 h at 60 oC. The AA monomer solution wasadded with 75 mol.% KOH (vs. AA monomer). The resulting mixture was then blended withPVA polymer aqueous solution until completely homogeneous. The appropriate weightpercent (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 undercontinuous stirring condition at 90 oC for 2 h. The resulting homogeneous polymer solutionwas then poured onto a PTFE or glass plate. The excess water solvent was evaporated slowlyat 80 oC in a vacuum oven. The resulting PVA/PAA blend membranes showed homogeneousand semi-transparent structure.

The absorption of KOH electrolyte by the SPE membrane is very crucial, because thehigher KOH electrolyte content is beneficial for higher ionic conductivity. It would benecessary to have higher concentration of hydroxyl ions (OH-) in polymer matrix. After PVAwas blended with PAA, the typical (101) crystalline peak of PVA at 2θ=19o was significantlyreduced [50,51]. The XRD spectra for the PVA/PAA membranes were shown in Fig. 6. Thisis very helpful to enhance the ionic transport property since there is more amorphous domainavailable for ion transport. It also improves the absorbility of KOH solution in PVA/PAAmembrane due to the higher hydrophilicity property of PAA.


2θ / degree

10 15 20 25 30 35 40 45 50 55 600

200

400

600

800

1000(a) Pure PVA film(b) PVA : PAA = 10 : 3(c) PVA : PAA = 10 : 5(d) PVA : PAA = 10 : 7.5

(a)

(b)

(c)

(d)

inte

nsity

(arb

itrar

y un

it)

Figure 6. XRD spectra for the PVA/PAA polymer membranes.

The PVA/PAA blend membranes exhibited novel absorption behavior that is sensitive tothe aqueous media and the composition of blend polymer. Table 5 lists the absorptionpercentage data of PVA/PAA polymer membranes under various solutions. It has beenevidenced 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 isattributed 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 roomtemperature.

Blend ratiosMedium PVA:PAA

(10:0)PVA:PAA

(10:3)PVA:PAA

(10:5)PVA:PAA

(10:7.5)6 M CH3OH 37 103 127 2062 M H2SO4 --- 161 164 2247 M KOH 49 207 255 413D.I. water 41 68 98 195

The higher absorption of KOH directly influenced the ionic conductivity of thePVA/PAA SPE. Table 6 summarized the ionic conductivity data. Very good ambienttemperature ionic conductivity was achieved. The results also demonstrated that the IPNstructure had good electrolyte holding capability even at higher temperature. Water channelsmay be provided for hydroxyl ion (OH-) to move from the cathode side to the anode sideduring discharge [39, 49, 52].


Table 6. The ionic conductivity (S cm-1) of PVA/PAA SPE.

Blend ratiosT (oC) PVA : PAA

(10:0)PVA : PAA

(10:3)PVA : PAA

(10:5)PVA : PAA

(10:7.5)25 0.072 0.142 0.256 0.30140 0.096 0.151 0.289 0.34150 0.108 0.163 0.300 0.38160 0.121 0.178 0.321 0.42270 0.132 0.190 0.352 0.45680 0.142 0.201 0.379 0.48890 - 0.211 0.410 0.510

Cyclic voltammetry was used to evaluate the electrochemical stability of the alkalinePVA/PAA SPE. The sweep potential was firstly scanned in the positive-going direction andthen reversed. Fig. 7 shows the cyclic voltammograms for the Zn| PVA/PAA| Zn symmetriccell with the PVA/PAA SPE of 10:5 for cyclic sweeping in the –0.50 to + 0.50 V range. Thecathodic 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 densityincrease apparently with increasing cycles. Fig. 8 shows the results with scanning rate from 1to 50 mV s-1. The ions mass transport along the water channels in the polymer matrix isunobstructed, facile and quite reversible [53]. On the other hand, the study on Al/air cell wasnot as impressive. The magnitude of the peak reduction and peak oxidation current densitiesdecreased slightly with increasing sweeping cycle. A deactivation layer (oxide) might beformed on Al electrode while the sweeping cycle was increased.

E (V)

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5

i (m

A c

m-2

)

-80

-60

-40

-20

0

20

40

60

80

cycle 12

cycle 9cycle 6cycle 3cycle 2

cycle 1

Figure 7. The cyclic voltammograms for Zn| PVA/PAA| Zn symmetric cell.


E (V)

-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

i (m

A c

m-2

)

-150

-100

-50

0

50

100

150scan rate : 1 mV/sscan rate : 5 mV/sscan rate : 10 mV/sscan rate : 20 mV/sscan rate : 50 mV/s

1 mV/s5 mV/s

10 mV/s

20 mV/s

50 mV/s

Figure 8. The cyclic voltammograms for Zn| PVA/PAA| Zn symmetric cell with increasing scanningrate.

Zre / ohm

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

-Zim

/ oh

m

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

(a) PVA : PAA = 10 : 3(b) PVA : PAA = 10 : 5(c) PVA : PAA = 10 : 7.5

a)

Zre / ohm

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

-Zim

/ oh

m

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

(a) PVA : PAA = 10 : 3(b) PVA : PAA = 10 : 5(c) PVA : PAA = 10 : 7.5

b)

Figure 9. AC impedance spectra of Zn-air cells (a) before and (b) after C/10 discharge.


It was found that the bulk resistance (Rb) of Zn/air cells with the different blend ratios ofPVA/PAA polymer electrolyte membrane was in the range of 0.21-0.26 ohm before the C/10discharge, as shown in Fig. 9. However, the bulk resistance increased slightly to 0.26-0.28ohm after the discharge.

Fig. 10 shows the electrochemical performance of solid-state Zn/air cell as a function ofdischarge current. The results also includes the corresponding power density curves. The peakpower density of Zn/air cell (over 50 mW cm-2) is much higher than that of gelled polymerZn/air cell (only 6-7 mW cm-2) [54,55].

i (mA cm-2)

0 10 20 30 40 50 60 70 80 90 100

Cel

l vol

tage

(V)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

Pow

er d

ensi

ty (m

W c

m-2

)

0

10

20

30

40

50

60

(a) PVA : PAA = 10 : 3

(b) PVA : PAA = 10 : 5

(a) PVA : PAA = 10 : 3

(b) PVA : PAA = 10 : 5

(c) PVA : PAA = 10 : 7.5

(c) PVA : PAA = 10 : 7.5

Figure 10. The cell potential of Zn/air cells vs. discharge current density using the various PVA/PAASPE.

2.6. Ionic Transport Number Study

The ionic transport numbers for alkaline PVA-based SPE are very important because the highanionic transport number of SPE can limit the carbonation problems [56]. The anionictransport number (t-) was measured by the dynamic Hittorf’s method [57]. A test cell, asshown in Fig. 11, with two Pt electrodes was made for electrolysis and the electrolysis currentwas imposed by a power supply. All PVA-based polymer electrolyte membranes were locatedand fixed at two separated compartments with the same 1 M KOH solution. The reactionoccurred at the Pt-cathode, producing O2, H2 and OH- while consuming water and OH-. Thebalance of OH- ions in each compartment led to OH- transport number after a fixed amount ofcharge was passed through the polymer membranes.

After the electrolysis, the concentration of OH- ion in both compartments was determinedby titration using 1 M HCl stock solution. The cationic and anionic transport numbers couldbe calculated by the following two equations:


t+ = nF / It (C0V0 − C+V+)

t− = nF / It (C−V− − C0V0)

where n is the charge of OH− ion, F is Faraday’s number, I is the applied current duringelectrolysis, t is the electrolysis time, C0, C+, C− are the OH− ion concentration at the initialand at the final of the electrolysis in the anodic and cathodic compartments, V0 is the initialvolume in each compartment, and V+ and V− are the final volume in the anodic and cathodiccompartments, respectively.

Pt-anodeO2

4OH-?O2+4e-+2H2O

Pt-cathodeH2

2H2O+2e-?H2+2OH-

SPE film

Figure 11. The experimental apparatus for dynamic transport number measurements.

The anionic transport numbers of alkaline PVA-based SPE at 25 oC measured at 20 mAcm-2 by Hittorf’s method are summarized in Table 7 in 1 M KOH, 1 M NaOH and 1 M LiOHsolutions. It has been noted that the higher the applied current density, the higher is the valuefor anionic transport number. High alkali electrolyte concentration would also cause thesystem to be more complex and can influence the ion transport ability. Fig. 12 shows that thevalue of t- decreases significantly while the alkali electrolyte concentration goes beyond 1 M.

Table 7. The anionic transport number data for different alkalinePVA-based SPE at 25 oC.

Alkali solutionType of membraneLiOH NaOH KOH

PVA 0.82 0.85 0.95PVA/Glass-fiber 0.82 0.84 0.95PVA/PECH (1:1) 0.73 0.80 0.86PVA/TEAC (1:2) 0.93 0.95 0.99PVA/PAA (10:5) 0.92 0.96 0.99PVA/PAA/PP/PE (10:5) 0.88 0.95 0.98PP/PE (after sulfonation) 0.80 0.84 0.89PP/PE separator 0.73 0.77 0.79


Concentration (mol L-1)

0 1 2 3 4 5 6

t-

0.70

0.75

0.80

0.85

0.90

0.95

1.00

LiOH(aq) KOH(aq) NaOH(aq)

Figure 12. The alkali metal salt effect on ionic transport number for alkaline PVA solid polymerelectrolyte.

Experimental results indicate that the ionic transport property for the PVA-based SPE ishighly dependent on both the alkali salt and the solution concentration. All the values ofanionic transport numbers for different PVA-based SPEs with KOH solution are much higherthan those with alkaline NaOH and LiOH solutions. This trend is consistent with literatureresults [58] that ion movement in polymer is related to polymer segmental motion, and theorder 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 ionicconductivity and high anionic transport number. These membranes consist of a polymermatrix swollen with water and KOH electrolyte solution, and this may directly influence theSPE on cell performance.

2.7. NON-Woven PP/PE Separator

Non-woven separators play important role in the alkaline battery system [63]. It slightlyabsorbs 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 separatorcannot completely prevent Zn dendrite growth from penetrating through the other side, andthis may cause the battery system in danger. The higher KOH electrolyte content in theseseparators may easily cause leakage of electrolyte or dry-out of electrolyte. This would limitthe shelf-life of battery. Although multi-layers of non-woven cloth separators may slightlysolve some problems, the thicker multi-layers could still result in less electrode activematerials which decreases electrochemical performance.

In the following section, we present a simple method to prepare high ionic conductingmembrane separators with lower cost non-woven polypropylene/polyethylene (PP/PE)


membrane through sulfonation [64-68]. The non-woven PP/PE separator membrane was insheet form and was composed of numerous filaments of 10-20 μm in diameter. The filamenthas 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. FTIRspectra are shown in Fig. 13 for the PP/PE membrane samples before and after sulfonationtreatment. Several additional transmittance bands appeared for the sulfonated membranes.The peaks at around wavenumber of 1150–1200 cm-1 were from SO3- asymmetric stretchingmodes [69]. The characteristic absorption bands at 550–585 cm-1 can be assigned to thedeformation of S–O bonds [70]. These band regions are in accordance with those foralkanesulfonic acid, C–SO3H. These peaks indicated that the sulfonic acid groups had beensuccessfully incorporated to the surface of polymer membrane. No significant change wasobserved for the peaks at higher wavenumbers, such as 2800–3000 cm−1, which were mainlyfrom C–C and C–H bonds. In addition, the sulfonation reaction might happen on both thesurface of polyethylene on the shell and the exposed polypropylene, originally in the core[71,72].

Wavenumber (cm-1)

400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600

Tran

smitt

ance

(arb

itrar

y un

it)

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0(a) Sulfonation time 0 h(b) Sulfonation time 3 h(c) Sulfonation time 9 h(d) Sulfonation time 72 h

550~585 cm-11150~1200 cm-1

(d)

(b)

(c)

(a)

Figure 13. FTIR spectra of the non-woven PP/PE membrane samples.

The SEM micrograph of the highly sulfonated PP/PE membrane is shown in Fig. 14. Thenon-woven membrane sheet had numerous 10–20 μm filaments that were bound together.The uniform porosity provided adequate channels for the transport of ions. After thesulfonation, the diameter of the filaments was slightly reduced. The surface polyethylenelayer was somewhat etched away. The longer the sulfonation time, the deeper the sulfonationreaction could take place. The degree of sulfonation was thus enhanced by more sulfonic acidgroups grafted on the surface.


300 μm

Figure 14. SEM micrographs of the highly sulfonated PP/PE membrane.

Fig. 15 displays the KOH electrolyte absorption curves for the non-woven PP/PEmembranes after immersion in 32 wt.% KOH solution. These sulfonated non-wovenmembranes can reach the equilibrium absorption within 1 h. The sulfonation treatment helpedincreasing the percent of absorption from 100% to 185%. The AC impedance spectra for thenon-woven PP/PE membrane at 25oC with different sulfonation times are shown in Fig. 16.

Immersion time (h)

0 5 10 15 20 25 30 35 40

Abso

rptio

n ra

tio (%

)

0

20

40

60

80

100

120

140

160

180

200

220

(a) sulfonation time 0 h(b) sulfonation time 3 h(c) sulfonation time 72 h(d) sulfonation time 128 h

(a)

(b)

(c)

(d)

Figure 15. The absorption curves of non-woven membranes using 32 wt.% KOH solution.


Z're / ohm

0 3 6 9 12 15 18 21 24 27 30

-Z'' im

/ oh

m

0

2

4

6

8

10

12

14

16

18

20

(a) sulfonation time 0 h(b) sulfonation time 3 h(c) sulfonation time 9 h(d) sulfonation time 18 h(e) sulfonation time 72 h(f) sulfonation time 128 h

Figure 16. AC impedance spectra of non-woven PP/PE membrane separators.

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 wasevidenced that the ionic conductivity for the sulfonated PP/PE separator sample increasedwith sulfonation time. As a result, the ionic conductivity of the highly sulfonated PP/PEseparator was as high as 0.0175 S cm-1. It was approximately two times that of the non-sulfonated sample (σ = 0.0088 S cm-1). The sulfonated non-woven PP/PE membranesexhibited high ionic conductivity in the alkaline systems.

2.8. Composite PVA/PAA/PP/PE SPE

Although the alkaline PVA/PAA SPE has high ionic conductivity, it also exhibits lowermechanical strength and large deformation after absorbing KOH electrolyte. On the otherhand, the sulfonated PP/PE separator membrane shows high hydrophilicity with moderateionic conductivity. Therefore, we tried to prepare a new class of composite polymerelectrolyte membranes based on sulfonated non-woven PP/PE sheets and PVA/PAApolymeric blends. This composite polymer membrane would offer high mechanical strengthand good ionic conductivity. The composite membrane is composed of PP/PE separatorsheets as reinforcements that are coated with homogeneous polymer PVA/PAA blends. Thecomposite PVA/PAA polymer membranes were also immersed in 32 wt.% KOH solution tobecome freestanding composite solid polymer electrolyte membranes [73].

The SEM micrographs for the PVA/PAA (10:5) composite polymer membranes withhighly sulfonated PP/PE separator are shown in Fig. 17. It has been found no phaseseparation. The improvement in hydrophilicity of PP/PE separator produced no visible voidson the surface and the cross-section. The binding between PP/PE and PVA/PAA polymer hasbeen very well.


(a) (b)

Figure 17. SEM micrographs for the PVA/PAA composite polymer membrane reinforced withsulfonated PP/PE separator (a) top surface and (b) cross-section.

E (V)

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5

i (m

A cm

-2)

-150

-100

-50

0

50

100

150cycle 1cycle 10cycle 20cycle 30cycle 40cycle 50cycle 60cycle 70cycle 80cycle 90cycle 100

cycle 1

cycle 10

cycle 100

increasing sweeping cycle

E (V)

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5

i (m

A cm

-2)

-250

-200

-150

-100

-50

0

50

100

150

200(a) s-PP/PE / PVA:PAA=10:3(b) s-PP/PE / PVA:PAA=10:5(c) s-PP/PE / PVA:PAA=10:7.5

(a)

(c)

(b)

(a)

(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.


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 thePVA/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 mVs-1. It can be clearly evidenced that even after 100 cycles, the cathodic and anodic peaks arestill symmetrical between Eca= -0.12V and Ean= 0.12V. This suggests high electrochemicalstability at zinc electrode. The entire sweeping curves are stable and smooth, and thereduction current density and the oxidation current density apparently increase withincreasing sweeping cycle. In addition, the CV results at the 100th sweeping cycle suggesthigher reduction and oxidation current densities for the higher PAA content sample. Thesymmetrical cathodic and anodic peaks would be obtained.

The solid-state Zn/air cell using only sulfonated PP/PE separator showed dischargecapacity of 728 mAh and lower Zn anode utilization. This indicates that during the dischargeprocess, the expansion of zinc anode may produce Zn dendrite that would easily penetratethrough 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 veryhigh discharge capacity of 1507 mAh. The zinc utilization is alos as high as 96%. Thisindicates that the composite PVA/PAA/PP/PE SPE has uniform morphology that caneffectively impede Zn dendritic penetration and thus extend cell life. However, it is necessaryto keep in mind that the Zn/air or Al/air cell is a semi-open system. The cell’s electrolyte orwater can easily evaporate into the ambient during long-term operation. Nevertheless, it willdirectly 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 probablycaused by the unbalanced tensile stress inside the membrane, while KOH electrolyteevaporates and loses during high-rate or long-term discharge operation.

i (mA cm-2)

0 20 40 60 80 100 120 140 160

E (V

)

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0P

ower

den

sity

(mW

cm

-2)

-20

0

20

40

60

80

100

(b) s-PP/PE / PVA:PAA=10:5

(a) s-PP/PE / PVA:PAA=10:3

(c) s-PP/PE / PVA:PAA=10:7.5



Figure 19. The electrochemical performance of Zn/air cells using composite PVA/PAA/PP/PE SPE.


Figs. 19 and 20 show the electrochemical performance of the solid-state Zn/air cells andAl/air cells, respectively. The results in cell power density curves are also displayed. It wasfound that the solid-state Zn/air cell with the composite solid polymer electrolyte membranewith composition ratio of PVA/PAA=10:5 had the highest power density of 90 mW cm-2. It isabout two times higher than that of the Zn/air cell assembled from PVA/PAA solid polymerelectrolyte (only about 40-50 mW cm-2). The value is also much higher than that of gelledZn/air cell (only 6-7 mW cm-2). In addition, the results of Al/air cell exhibited tremendousimprovement from 1-2 mW cm-2 to 70 mW cm-2. It can be attributed to the good bindingbetween sulfonated PP/PE and PVA/PAA polymer blends. Thus, the excellent dimensionalstability of composite solid polymer electrolyte membrane can sustain higher dischargecurrent density in battery applications. In addition, Table 8 summarizes the characteristicproperties of different alkaline PVA-based solid polymer electrolytes and commercial non-woven PP/PE separator.

i (mA cm-2)

0 10 20 30 40 50 60 70 80 90 100 110 120

E (V

)

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

2.25

2.50

2.75

3.00

Pow

er d

ensi

ty (m

W c

m-2

)

-40

-20

0

20

40

60

80


(c) s-PP/PE / PVA:PAA=10:7.5(a) s-PP/PE / PVA:PAA=10:3




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 dischargecurrent 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, ifthe cell is firmly sealed. It has been found that the power density of the Zn/MnO2 andAl/MnO2 cells is increased when the PVA/PAA blend ratio is increased. This was due to thehigh ionic conductivity of the alkaline composite PVA/PAA/PP/PE solid polymer electrolytemembranes. For Al/MnO2 cell, it was found that the cell had the highest power density of 110mW cm-2 at 120 mA cm-2. This particular cell was assembled with PVA/PAA (10:7.5)


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 specificenergy density. These solid-state alkaline cells with alkaline composite PVA/PAA/PP/PEsolid polymer electrolyte membranes indeed exhibit excellent electrochemical performance.

i (mA cm-2)

0 10 20 30 40 50 60 70 80 90 100 110 120

E (V

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

Pow

er d

ensi

ty (m

W c

m-2

)

-40

-20

0

20

40

60

80

100

(a) s-PP/PE/PVA:PAA=10:3 (b) s-PP/PE/PVA:PAA=10:5

(c) s-PP/PE/PVA:PAA=10:7.5

(c) s-PP/PE/PVA:PAA=10:7.5

(b) s-PP/PE/PVA:PAA=10:5

(a) s-PP/PE/PVA:PAA=10:3

Figure 21. The electrochemical performance of solid-state Zn/MnO2 cells using compositePVA/PAA/PP/PE SPE.

i (mA cm-2)

0 20 40 60 80 100 120 140 160 180 200 220

E (V

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

2.8

3.0

Pow

er d

ensi

ty (m

W/c

m2 )

0

10

20

30

40

50

60

70

80

90

100

110

120


(b) s-PP/PE / PVA:PAA=10 : 5

(a) s-PP/PE / PVA:PAA=10 : 3

(a) s-PP/PE / PVA:PAA=10 : 3

(b) s-PP/PE / PVA:PAA=10 : 5

(c) s-PP/PE / PVA:PAA=10 : 7.5

Pow

er d

ensi

ty (m

W c

m-2

)

Figure 22. The electrochemical performance of solid-state Al/MnO2 cells using compositePVA/PAA/PP/PE SPE.


Table 8. The characteristic properties of different alkaline PVA-based solid polymerelectrolytes and commercial non-woven PP/PE separator.

Property

SPE type σ(S cm-1)

t-

(1M KOH)

Mechanicalstrength(MPa)

UtilizationZn/airat C/10

(%)

Max.Zn/air cell power

density(mW cm-2)

PVA(PVA:KOH=30:40) 0.047 0.95 3.8 82 30

PVA/Glass-fiber 0.047 0.92 9.5 84 30

PVA/PEO (2:8) 7.4×10-6 - 1.3 63 8

PVA/PEO/Glass-fiber (2:8) 7×10-6 - 8.8 68 6

PVA/PECH (1:0.5) 4.6×10-3 0.86 58.9 77 10

PVA/TEAC (1:1) 0.022 0.99 0.14 85 80

PVA/PAA (10:5) 0.256 0.99 2.3 85 43

PVA/PAA/PP/PE (10:5) 0.165 0.98 11.9 96 96

PP/PE (sulfonated) 0.016 0.89 4.4 46 25

PP/PE separator 0.009 0.79 5.5 78 15

3. Conclusion

The preparation and the characteristic properties of different alkaline solid polymerelectrolytes have been presented and discussed. The PEO-based SPE was firstly proposed andused for the alkaline SPE. Although the ionic conductivity can be improved by increasingKOH content, he brittle characteristics and porous structure may limit its practicability inelectrochemical applications. The alkaline PVA-based polymer SPE has been successfullyintroduced and developed since 2000. When carefully prepared, the ionic conductivity can beas high as 0.301 S cm-1 for the PVA/PAA solid polymer electrolyte. The alkaline compositesolid polymer electrolyte membrane (CSPEM) based on sulfonated PP/PE separator andPVA/PAA may solve the problems in deformation, shrinkage or expansion. It still shows highionic conductivity and good mechanical strength. For Zn/air battery application, the powerdensity was as high as 90 mW cm-2, and thus the PVA-based composite SPE has highpotential for use in alkaline battery system. The gelled polymer electrolyte (GPE) has thehighest ionic conductivity of 0.60 S cm-1 at room temperature, which is close to KOHaqueous 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 majorissues for GPE lie in low mechanical strength, low thermal stability and high aqueousevaporation and leakage rates. The electrochemical cell may require firm sealing or packingfor applications.


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.


[34] Yang, C. C.; Lin, S. J. Mater. Lett. 2002, 57, 873.[35] Yang, C. C. J. Power Sources 2002, 109, 22.[36] Yang, C. C.; Lin, S. J. J. Power Sources 2002, 112, 497.[37] Yang, C. C.; Lin, S. J.; Hsu, S. T. J. Power Sources 2003, 122, 210.[38] Yang, C. C.; Wu, G. M.; Lin, S. J. J. Appl. Electrochem. 2006, 36, 655.[39] Choudhury, N. A.; Shukla, A. K.; Sampath, S.; Pitchuman, S. J. Electrochem. Society

2006, 153, A614.[40] Degouw, A. M.; Prins, J.; Dingermans, J. European Patent 1983, EP 68,530.[41] Lindsjo, M. C.; Ekman K. B.; Nasman, J. H. Biomaterials 1996, 17, 913.[42] Habert, A. C.; Burns, C. M.; Huang, R. Y. M. J. Appl. Polym. Sci. 1979, 24, 801.[43] Ito, Y.; Kotera, S.; Inaba, M.; Kono, K.; Imanishi, Y. Polymer 1990, 31, 2157.[44] Mayoux, C.; Dandurand, J.; Lacabanne, C. Thermochim. Acta 2004, 421, 43.[45] Daniliuc, L.; David, C. Polymer 1996, 37, 5219.[46] Gudeman, L. F.; Peppas, N. A. J. Membr. Sci. 1995, 107, 239.[47] Marra, S. P.; Ramesh, K. T.; Douglas, A. S. Mater. Sci. Eng. C 2001, 14, 25.[48] Kokufuta, E.; Yamauchi, T.; Osada, Y. Polym. Gels Networks 1995, 3, 397.[49] Zhu, X.; Yang, H.; Cao, Y.; Ai, X. Electrochim. Acta 2004, 49, 2533.[50] Assendert, H. E.; Windle, A. H. Polymer 1997, 39, 4295.[51] Assendert, H. E.; Windle, A. H. Polymer 1997, 39, 4303.[52] Ferrando, W. A. J. Power Sources 2004, 130, 309.[53] Wu, G. M.; Lin, S. J.; Yang, C. C. J. Membr. Sci. 2006, 280, 802.[54] Othman, R.; Basirun, W. J.; Yahaya, A. H.; Arof, A. K. J. Power Sources 2001, 103, 34.[55] Othman, R.; Yahaya, A.H.; Arof, A.K. J. Appl. Electrochem. 2002, 32, 1347.[56] Agel, E.; Bouet, J.; Fauvarque, J. F. J. Power Sources 2001, 101, 267.[57] Yang, C. C.; Lin, S. J.; Wu, G. M. Mater. Chem. Phys. 2005, 92, 251.[58] Okamoto, Y.; Yeh, T. F.; Lee, H. S.; Skotheim, T. A. J. Polym. Sc.: Part A Polym.

Chem. 1993, 31, 2573.[59] Zhang, G. Q.; Zhang, X. G. Solid State Ionics 2003, 160, 155.[60] Mohamad, A. A.; Mohamed, N. S.; Alias, Y.; Arof, A. K. J. Alloys Compounds 2002,

337, 208.[61] Mohamad, A. A.; Mohamed, N. S.; Yahya, M .Z. A.; Othman, R.; Ramesh, S.; Alias,

Y.; Arof, A. K. Solid State Ionics 2003, 156, 171.[62] Yuan, A.; Zhao, J. Electrochim. Acta 2006, 51, 2454.[63] Smitha, B.; Sridhar, S.; Khan, A.A. J. Membr. Sci. 2005, 259, 10.[64] Tricoli, V.; Carretta, N. Electrochem. Commun. 2002, 4, 272.[65] Bergbreiter, D. E.; Kabza, K. J. Am. Chem. Soc. 1991, 113, 1447.[66] Sasikumar, G.; Ihm, J. W.; Ryu, H. Electrochim. Acta 2004, 50, 598.[67] Kaneko, M.; Sato, H. Macromol. Chem. Phys. 2005, 206, 456.[68] Wu, G. M.; Lin, S. J.; Yang, C. C. J. Membr. Sci. 2006, 284, 120.[69] Sowa, M. G.; Fischer, D.; Eysel, H. H.; Mantsch, H. H. J. Mol. Struct. 1996, 379, 77.[70] Tada, H.; Ito, S. Langmuir 1997, 13, 3982.[71] Ihata, J. J. Polym. Sci. Part A: Polym. Chem. 1988, 26, 167.[72] Fonseca, C.; Perena, J. M.; Fatou, J.G.; Bello, A. J. Mater. Sci. 1985, 20, 3283.[73] Wu, G. M.; Lin, S. J. Proceeding SAMPE Conf. Long Beach, CA 2006.

INDEX

A

access, 249, 285, 401, 414 accounting, 165, 297, 301, 358 accuracy, 12, 13, 66, 303, 320 acetic acid, 147 acetone, 416, 427 achievement, 5 acid, viii, ix, x, 10, 11, 135, 136, 137, 138, 139, 140,

142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 156, 159, 163, 173, 220, 221, 223, 224, 225, 226, 231, 233, 235, 236, 237, 247, 248, 255, 256, 264, 266, 269, 285, 361, 382, 383, 385, 388, 391, 392, 394, 395, 396, 399, 401, 403, 404, 408, 409, 411, 414, 415, 416, 419, 420, 425, 428, 440, 447, 448, 449, 455, 462

acidity, 137, 151, 156, 384 acrylic acid, 449, 455 activation, x, 14, 73, 80, 93, 95, 96, 105, 107, 109,

110, 111, 113, 114, 115, 203, 264, 265, 274, 289, 290, 291, 292, 295, 299, 300, 302, 303, 304, 311, 314, 317, 318, 337, 348, 360, 376, 385, 453

activation energy, 114, 314, 337, 453 active site, 256, 401, 405, 415 adaptation, 79 additives, 401, 451 adhesion, 227, 252 adjustment, 198, 321 ADP, 448 adsorption, 4, 96, 208, 254, 255, 256, 263, 264, 265,

268, 269, 387, 413, 416, 419, 427 aerospace, 175 agent, 251, 382, 383, 390, 391, 421, 449, 451, 455 aggregation, 383, 420, 421, 432 aggregation process, 383 agriculture, 455 AIBN, 455 alcohol, xi, 233, 251, 412, 445, 447, 449, 451, 455

alcohols, 212 algorithm, x, 162, 169, 170, 180, 186, 201, 205, 274,

300, 301, 302, 303, 304, 318, 320, 361, 376 alkaline media, 254, 256, 263, 264, 269, 448 alkane, 159 alloys, 220, 248, 255, 263, 264, 266, 267, 269, 382,

386, 387, 390, 415 alternative(s), viii, 10, 121, 135, 136, 137, 148, 157,

223, 224, 233, 236, 241, 247, 249, 279, 297, 348, 412, 437

alternative hypothesis, 121 aluminum, 220, 286, 450 ambient air, 237, 326, 368 amines, 448 ammonia, 175 ammonium, 159, 397, 398, 448, 453, 455 anion, 448 anisotropy, 221 ANOVA, 76, 79 aqueous solutions, 264, 417, 418 aromatics, 157 Arrhenius equation, 450 asbestos, 248 aspect ratio, 221, 232, 383, 406, 427, 431 assessment, 322, 348 assignment, 175, 207, 268 assumptions, 94, 128, 169, 292, 305, 355 atmospheric pressure, 154, 164, 165 atoms, 147, 219, 256, 259, 269, 383, 385, 394, 397,

399, 423, 430, 437 attention, 10, 11, 73, 77, 80, 137, 291, 294, 381, 412 automobiles, 10, 214 autonomy, 212, 213 availability, 66, 72, 75, 192, 314

B

bandwidth, 176

Index 474

barriers, 11 basicity, 384 batteries, vii, 9, 11, 212, 446, 447, 451, 452, 453 behavior, ix, 73, 80, 92, 104, 105, 132, 133, 159,

162, 169, 170, 176, 180, 192, 200, 247, 249, 257, 264, 269, 295, 313, 320, 348, 355, 368, 372, 376, 409, 448, 455, 456

bending, 290, 348 beneficial effect, 432 benzene, 144 benzoyl peroxide, 455 binding, 423, 435, 451, 464, 467 binding energies, 423 binding energy, 435 biocompatibility, 437 biomass, 162, 173, 174, 175, 201, 206, 208 biomolecules, 437 Blacks, 415, 419 bleeding, 387 blends, 137, 150, 157, 449, 464, 467 blocks, 148, 217, 387 Boltzmann constant, 97, 114 bonding, 142, 147, 148, 216, 222, 227, 437, 449, 455 bonds, 144, 147, 462 boundary surface, 316 bounds, 162, 175 breathing, 228, 229, 230, 231 building blocks, 217 bulk materials, 269 burn, 100 burning, 10 burnout, 354, 367

C

Canada, 159 candidates, 157, 162, 388 capillary, 250, 284, 300, 355, 358, 366, 367, 376 carbon, x, xi, 4, 12, 147, 163, 168, 174, 218, 228,

232, 233, 237, 240, 241, 250, 251, 252, 253, 254, 255, 265, 266, 267, 269, 281, 285, 286, 297, 381, 383, 384, 385, 387, 389, 391, 392, 393, 394, 395, 396, 397, 398, 399, 400, 401, 402, 403, 404, 405, 406, 407, 408, 409, 410, 411, 412, 413, 414, 415, 416, 417, 418, 419, 420, 421, 422, 424, 425, 426, 427, 428, 429, 430, 431, 432, 433, 434, 435, 436, 437, 439, 440, 445, 449, 450

carbon dioxide, 163, 168, 174, 297, 392, 403 carbon materials, x, 411, 412, 414, 415, 416, 418,

426, 427, 428, 429, 430, 432, 433, 434, 437, 440 carbon monoxide, 174 carbon nanotubes, x, 218, 407, 408, 409, 411, 415,

440

carbonization, 383, 414, 415 carrier, 147, 248, 446 case study, 93 casting, 137, 147, 148, 150, 217, 224, 251, 447, 453,

455 catalysts, x, xi, 10, 11, 12, 163, 227, 250, 254, 255,

264, 265, 267, 269, 270, 381, 387, 391, 393, 407, 408, 409, 410, 411, 414, 415, 417, 420, 423, 424, 427, 428, 429, 430, 431, 432, 433, 435, 436, 438, 439, 440, 448

catalytic activity, 10, 254, 266, 285, 289, 385, 412, 415

categorization, 293 cation, 144 CE, 160 cell assembly, 225, 266 cellulose, 451 ceramic, 248, 406 chain mobility, 455 chain scission, 144 channels, x, 12, 13, 16, 17, 137, 149, 153, 154, 156,

214, 222, 225, 226, 227, 232, 233, 240, 273, 282, 283, 284, 286, 297, 298, 299, 300, 301, 303, 304, 305, 316, 322, 323, 335, 338, 355, 356, 361, 364, 366, 376, 456, 457, 462

charge density, 151 chemical composition, 72, 413 chemical energy, vii, 9, 165, 278, 279, 302 chemical properties, viii, 135, 136, 405, 413 chemical reactions, 217 chemical structures, 388 chromium, 217, 220 circulation, 240, 298 classification, 280 cleaning, 175, 216 cluster bomb, 437 clustering, 143 clusters, 151, 412, 417, 418, 423, 440 C-N, 434, 436, 440 CO2, 11, 12, 13, 14, 15, 18, 76, 163, 174, 205, 208,

227, 231, 249, 255, 266, 275, 279, 311, 315, 385, 392, 402, 403

coal, ix, 10, 161, 174 coatings, 406 cobalt, 10, 254 codes, 297 colloidal particles, 385 combustion, vii, 4, 161, 165, 174, 203, 204, 206,

278, 281, 382, 385 combustion chamber, 165 communication, 212 compatibility, 233 compensation, 193

Index 475

complement, 179 complexity, 113, 164, 216, 223, 247, 249, 269, 294,

361, 413 complications, 249, 269 components, vii, 3, 4, 9, 10, 11, 18, 19, 20, 21, 29,

30, 31, 34, 44, 53, 54, 65, 98, 149, 150, 165, 179, 180, 183, 192, 194, 200, 211, 213, 219, 222, 230, 279, 283, 288, 305, 313, 358, 376, 402

composites, xi, 158, 388, 445 composition, 72, 94, 95, 102, 103, 137, 162, 164,

171, 173, 174, 201, 236, 249, 255, 269, 306, 307, 318, 413, 423, 434, 449, 452, 456, 465, 467

compounds, 157, 174, 175, 220, 413, 415 computation, 12, 179, 207 computational fluid dynamics, 300, 376 computational grid, 301 computers, 11, 211 computing, 162, 179, 301, 304, 317 concentration, vii, 4, 9, 12, 14, 18, 21, 22, 23, 25, 26,

27, 31, 32, 35, 36, 37, 48, 51, 53, 54, 56, 57, 67, 95, 111, 139, 144, 145, 147, 150, 152, 164, 203, 221, 258, 259, 260, 262, 265, 268, 274, 275, 277, 284, 285, 289, 290, 291, 292, 294, 295, 297, 298, 300, 301, 302, 303, 314, 319, 320, 322, 325, 326, 338, 348, 359, 360, 368, 376, 385, 417, 418, 420, 440, 447, 448, 455, 459, 460, 461

conception, 213, 214, 216, 245 condensation, 141, 175, 220, 275, 300, 329, 355,

357, 360, 363, 364, 366, 368, 369, 376 conditioning, 75, 170, 176, 227 conduction, 145, 284, 286, 299, 314 conductivity, viii, x, xi, 58, 135, 136, 137, 142, 144,

145, 147, 148, 149, 150, 151, 152, 153, 154, 156, 159, 180, 232, 233, 254, 275, 277, 285, 286, 287, 291, 292, 297, 299, 301, 305, 308, 309, 310, 312, 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, 456, 457, 461, 464, 467, 469

conductor, 145, 226, 305 confidence, 108, 109 configuration, 148, 163, 266, 268 Congress, 132 connectivity, 225 conservation, 165, 355, 356 constant rate, 186, 198 constraints, 310 construction, 291 consumers, 222 consumption, vii, viii, 9, 10, 20, 30, 53, 67, 71, 72,

73, 76, 77, 78, 79, 80, 81, 82, 83, 84, 86, 87, 88, 89, 90, 91, 92, 93, 94, 102, 103, 104, 110, 113,

133, 165, 204, 291, 294, 297, 298, 314, 325, 326, 338, 363, 368, 403

contaminants, 222 contamination, 175 continuity, 297, 298, 305, 309, 355, 359 control, 5, 18, 37, 72, 76, 78, 79, 80, 88, 113, 133,

139, 153, 154, 162, 163, 166, 167, 169, 170, 171, 172, 173, 175, 176, 179, 180, 183, 184, 185, 192, 198, 200, 201, 202, 204, 205, 206, 207, 208, 241, 250, 262, 265, 266, 301, 313, 316, 359, 360, 394, 400, 408, 409, 426

convergence, 300, 317, 318, 320 conversion, vii, 10, 75, 165, 174, 203, 278, 279, 302 cooling, 12, 83, 88, 89, 90, 216, 286, 300, 301 copolymers, 157, 159 copper, 12, 232 correlation, 91, 92, 96, 109, 110, 113, 308, 309, 357,

358 correlations, 91, 292 corrosion, x, 175, 230, 232, 233, 254, 286, 381, 383,

392, 402, 403, 404, 405, 406, 409, 411, 412, 415, 447

costs, 136, 361, 413 coupling, 93, 167, 168, 376 coverage, 256, 257, 261, 263, 265, 266, 396 cristallinity, 139, 144, 153 crystal structure, 449 crystalline, 406, 413, 417, 419, 420, 422, 428, 453,

455 crystallinity, 447, 448, 449 current limit, viii, 9, 15, 17, 54 CVD, 217, 218, 219, 229 cycles, 162, 163, 200, 207, 208, 209, 287, 294, 406,

447, 451, 452, 457, 466 cycling, 5

D

damping, 176, 177 danger, 461 data collection, 79 DBP, 416, 427 decay, 4, 5, 267, 401, 402, 405, 406, 409 decomposition, 143, 144, 147, 383, 397, 398, 406 decomposition temperature, 398 deconvolution, 417 defects, 4, 383, 404 definition, 176, 214 deformation, 222, 347, 348, 451, 462, 464, 469 degradation, 75, 113, 137, 138, 142, 144, 157, 287,

409, 419 dehydration, 145, 149, 294, 299 delivery, 165, 222, 224, 236

Index 476

demand, 170, 176, 186, 194, 195, 197, 198, 199, 200, 201, 202, 203, 302, 449, 450

Denmark, 133, 206 density values, viii, 71, 73 Department of Energy, 206 dependent variable, viii, 71, 72, 73, 76, 77, 78, 79,

80, 81, 113, 115, 116 depolarization, 11 deposition, 153, 214, 216, 217, 218, 219, 220, 222,

227, 240, 253, 254, 285, 348, 383, 384, 385, 388, 389, 390, 393, 402, 406, 412, 413, 415, 420, 421, 422, 427, 429, 434, 437

deposition rate, 218 deposits, 175 designers, 13, 291, 295, 297, 303 desorption, 4, 96 deviation, 81, 83, 85, 86, 169, 171, 172, 179, 202,

204, 313 devolatilization, 174 dialysis, 455 dielectrics, 218 differential equations, 165, 193, 298 diffraction, 416, 417, 418, 419, 421, 428, 429, 438 diffusion, ix, 12, 13, 14, 93, 95, 96, 97, 98, 99, 100,

101, 102, 107, 111, 113, 115, 216, 232, 236, 247, 249, 250, 251, 258, 259, 260, 262, 263, 265, 266, 267, 268, 269, 275, 277, 281, 282, 283, 284, 285, 286, 287, 290, 294, 295, 297, 298, 299, 300, 301, 303, 304, 305, 306, 309, 310, 312, 314, 315, 316, 318, 320, 326, 329, 335, 337, 338, 342, 347, 348, 354, 355, 357, 358, 359, 360, 363, 364, 366, 367, 368, 370, 371, 372, 376, 387, 408, 409, 450, 451

diffusion region, 297, 301 diffusivities, 107, 310 diffusivity, 96, 100, 114, 312, 315, 326, 338, 359,

361, 368 dilute gas, 98 dimensionality, 318 dimethylformamide, 139 dimethylsulfoxide, 139 direct measure, 103 disaster, ix, 273, 280 disclosure, 81 discontinuity, 183, 400 discretization, 320 dispersion, x, 4, 152, 233, 254, 381, 385, 388, 404,

406, 407, 414, 440 displacement, 347, 348, 351, 353 dissociation, 136, 145, 259 dissolved oxygen, 295 distilled water, 251, 416, 427, 455 distribution, viii, ix, x, 4, 19, 20, 29, 71, 72, 73, 74,

77, 79, 80, 81, 82, 84, 87, 88, 89, 90, 92, 102, 103,

104, 107, 108, 109, 110, 111, 113, 114, 151, 161, 162, 163, 171, 179, 180, 186, 188, 192, 193, 194, 195, 196, 198, 200, 201, 202, 205, 206, 224, 231, 250, 274, 295, 298, 300, 302, 303, 304, 310, 311, 318, 325, 326, 327, 329, 331, 333, 334, 335, 337, 338, 339, 341, 342, 344, 345, 347, 348, 350, 355, 360, 367, 368, 369, 370, 371, 372, 373, 374, 375, 376, 408, 412, 414, 418, 430, 449

divergence, 305 DMF, 139, 146, 147, 148, 150 doping, viii, 135, 136, 149, 216 dry ice, 251 drying, 174, 249, 321, 331, 354, 367, 372, 398 durability, 3, 5, 247, 347, 405, 406, 407, 409 duration, 399 dynamical properties, 169 dynamical systems, 194

E

education, 274 election, 201 electric conductivity, 312 electric current, 446 electric field, 314 electrical conductivity, x, 254, 286, 411, 415, 431 electrical power, 193, 194, 202, 203 electrical resistance, 240 electricity, vii, 9, 10, 18, 148, 162, 163, 168, 173,

192, 208, 219, 278 electrocatalysis, 249, 408 electrocatalyst, 227, 249, 250, 254, 255, 409, 412,

413 electrochemical impedance, 267 electrochemical reaction, vii, 9, 13, 78, 96, 156, 250,

254, 283, 286, 288, 289, 291, 295, 301, 304, 314, 323, 325, 331, 337, 355, 357, 364, 376, 382, 425, 432

electrochemistry, 154, 156, 303, 407, 437 electrode/electrolyte interfaces, 13, 15 electrodeposition, 217, 253 electrodes, vii, ix, x, 4, 5, 10, 12, 17, 21, 53, 96, 100,

106, 156, 163, 173, 213, 221, 222, 224, 225, 226, 227, 228, 230, 232, 237, 247, 248, 249, 250, 251, 252, 253, 254, 263, 264, 265, 266, 267, 268, 269, 285, 291, 297, 301, 302, 304, 305, 310, 314, 335, 337, 338, 355, 382, 385, 386, 387, 389, 391, 393, 394, 395, 396, 398, 399, 404, 406, 407, 408, 409, 410, 411, 414, 415, 417, 425, 426, 427, 428, 433, 440, 448, 459

electrolysis, 237, 459, 460 electrolyte, vii, viii, ix, xi, 4, 5, 10, 11, 13, 14, 15,

41, 58, 78, 106, 114, 135, 136, 142, 144, 149, 152,

Index 477

159, 163, 173, 180, 233, 236, 247, 248, 249, 250, 255, 264, 265, 266, 268, 269, 280, 285, 291, 301, 314, 361, 367, 385, 386, 387, 400, 407, 408, 409, 412, 427, 433, 445, 446, 447, 448, 449, 451, 453, 454, 455, 456, 459, 460, 461, 463, 464, 466, 467, 468, 469

electron, 10, 139, 154, 155, 220, 239, 286, 289, 290, 291, 314, 337, 388, 399, 401

electron microscopy, 388 electronic structure, 383 electrons, 4, 106, 154, 155, 156, 163, 174, 203, 225,

276, 281, 283, 284, 285, 289, 337, 382, 383, 385, 387, 388, 426

electroplating, 219, 220 embossing, 222 emergence, 212, 214 emission, 163 encouragement, 274 endothermic, 84, 88, 90, 282 energy, vii, viii, ix, 9, 10, 11, 12, 13, 16, 17, 18, 20,

21, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 39, 40, 41, 42, 43, 44, 46, 47, 48, 50, 51, 53, 55, 56, 57, 59, 60, 62, 63, 64, 65, 66, 67, 75, 76, 97, 114, 135, 154, 156, 161, 162, 165, 173, 180, 195, 196, 212, 213, 218, 221, 223, 231, 247, 273, 275, 278, 279, 281, 288, 289, 298, 302, 306, 313, 314, 331, 337, 355, 356, 359, 381, 412, 426, 435, 447, 450, 453, 468

energy consumption, vii, 9, 20, 53 energy density, vii, 9, 11, 16, 17, 18, 21, 24, 25, 26,

27, 28, 29, 30, 31, 32, 33, 34, 35, 39, 41, 42, 43, 44, 46, 47, 50, 51, 53, 56, 57, 59, 60, 62, 63, 64, 65, 67, 223, 450, 468

energy efficiency, vii, viii, 9, 10, 11, 13, 17, 21, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 39, 40, 41, 42, 43, 44, 46, 47, 50, 51, 53, 55, 56, 57, 59, 60, 62, 63, 65, 67

England, 205, 407, 410 entropy, 276, 288, 311, 331 environment, ix, 72, 135, 222, 254, 264, 265, 269,

270, 273, 287, 291, 302, 303, 331, 382, 391, 393, 448

environmental impact, ix, 163, 273, 302 epitaxial growth, 219 equilibrium, 83, 87, 89, 159, 193, 195, 197, 203,

257, 259, 260, 265, 289, 290, 291, 310, 313, 314, 355, 356, 359, 416, 463

equipment, 16, 18, 34, 175, 302 ester, 420, 437 estimating, 126, 206 etching, 214, 216, 217, 220, 221, 222, 227, 230, 233,

237, 239, 240 ethanol, 223, 224, 227, 235, 237, 251

ethylene, 415, 416, 428, 448 ethylene glycol, 415, 416, 428 Europe, 214 evacuation, 437 evaporation, 219, 220, 275, 277, 300, 303, 357, 359,

360, 362, 363, 364, 376, 447, 448, 467, 469 evolution, 225, 258, 262, 263, 294, 296 experimental condition, 103, 249, 258, 268 exposure, 217 expulsion, 389 extraction, 416, 427 extrapolation, 262, 269

F

fabrication, ix, xi, 4, 158, 211, 212, 213, 214, 216, 225, 226, 227, 228, 229, 231, 236, 237, 240, 250, 252, 269, 273, 303, 397, 445

failure, 302, 331, 372, 401, 405 feature selection, 201 feedback, 169 fibers, 4, 383, 407 filament, 462 fillers, 149 film, 151, 216, 217, 219, 220, 224, 225, 226, 229,

232, 241, 249, 251, 252, 265, 269, 406, 408, 410, 417, 427, 428, 437, 447, 449, 451, 453, 454, 455, 456, 460

film thickness, 449 films, 214, 216, 217, 218, 219, 226, 229, 454 filters, 76, 214 filtration, 252, 393 financial performance, viii, 161 flexibility, 137, 148, 162, 447 floating, 266 flocculation, 252 flood, 354, 367 flooding, 236, 286, 295, 299, 300, 301, 321, 392,

396, 397, 401, 407 flow field, x, 230, 232, 233, 273, 274, 286, 287, 298,

299, 300, 301, 303, 305, 317, 318, 322 fluid, ix, 83, 88, 89, 104, 273, 298, 300, 301, 302,

303, 316, 376 fluidized bed, 174 fluorine, 437, 441 focusing, 137 foils, 230, 231, 233, 241 forgetting, 171, 202 formaldehyde, 383, 388, 390, 393, 428 fossil, 10, 76, 205, 281 fossil fuels, 10, 281 fouling, 449 France, 133, 211, 214, 377, 378, 441, 443

Index 478

free energy, 154, 275, 288, 289 freedom, 120, 121, 122, 126, 128, 129, 130 freezing, 449 FTIR, 440, 462 FT-IR, 420, 434, 437, 438 FT-IR, 440 FT-IR, 441 fuel efficiency, 148 functionalization, x, 150, 381, 405, 407

G

gallium, 219 gas diffusion, 12, 13, 249, 250, 251, 266, 267, 268,

277, 282, 283, 285, 286, 287, 294, 295, 297, 298, 299, 300, 301, 303, 304, 305, 309, 310, 314, 316, 318, 335, 342, 348, 354, 355, 357, 359, 363, 364, 366, 367, 368, 371, 372, 376, 408, 409, 450, 451

gas phase, 174, 255, 276, 277, 295, 310, 329, 355, 356, 357, 358, 359, 360, 363, 366, 376

gases, 95, 98, 134, 165, 174, 200, 203, 204, 212, 218, 219, 221, 237, 250, 276, 284, 285, 286, 289, 297, 298, 299, 300, 301, 304, 305, 355, 376, 392, 416, 437

gasification, 162, 173, 174, 175, 201, 208 gasoline, 136 gel, 12, 150, 154, 251, 447, 448, 449, 451 generation, viii, ix, 19, 20, 75, 153, 161, 162, 173,

188, 192, 196, 201, 205, 206, 207, 227, 248, 276, 278, 297, 301, 302, 303, 311, 331, 335, 347, 372, 382, 405, 407

Georgia, 205, 206 germanium, 218 Germany, 75, 206, 207, 243 glass, xi, 142, 214, 217, 222, 225, 227, 228, 232,

241, 445, 451, 455 glass transition, 142 glass transition temperature, 142 global markets, 11 glycol, 415, 416, 428 goals, 5, 413 gold, 220, 221, 225, 226, 227, 230, 231, 232, 241,

265 graph, 87, 103, 431 graphene sheet, 383, 403 graphite, x, 12, 20, 233, 286, 287, 305, 310, 313,

411, 412, 414, 415, 426, 440 gravimetric analysis, 395, 396 groups, x, 136, 137, 138, 139, 142, 143, 144, 145,

147, 148, 149, 150, 151, 152, 153, 156, 157, 217, 236, 237, 285, 381, 382, 383, 384, 385, 391, 392, 393, 394, 395, 396, 397, 398, 399, 401, 403, 404, 405, 406, 407, 409, 414, 415, 416, 420, 422, 424,

429, 434, 436, 437, 439, 440, 441, 449, 455, 456, 462

growth, 211, 219, 402, 404, 461 growth rate, 219 guidance, 5 guidelines, 348 gyroscope, 214

H

Hawaii, 133 H-bonding, 147, 437 HE, 139 heat, vii, ix, x, 4, 9, 10, 11, 18, 27, 41, 75, 76, 80, 88,

154, 155, 162, 163, 168, 173, 174, 203, 213, 216, 250, 254, 273, 275, 276, 281, 282, 286, 299, 300, 301, 302, 303, 304, 309, 310, 311, 316, 331, 332, 335, 347, 359, 360, 364, 372

heat capacity, 275, 309 heat release, 301 heat removal, 286 heat transfer, 88, 216, 276, 299, 300, 303, 304, 310,

359 heating, 115, 175, 203, 204, 220, 275, 301, 406 height, 318, 418, 419 high power density, 230, 232, 331 hip, 216 homogeneity, 81, 90, 92 Hong Kong, 225 hot pressing, 252 humidity, 5, 11, 16, 18, 144, 145, 153, 277, 294, 297,

298, 299, 301, 302, 303, 313, 319, 355, 359, 360, 363

hybrid, 11, 150, 151, 153, 159, 161, 162, 163, 168, 170, 176, 201, 206, 207, 208, 252, 449

hydrazine, 390 hydrides, 212, 223 hydrocarbons, 11, 255, 278, 279 hydrofluoric acid, 220 hydrogen, vii, ix, 4, 10, 96, 103, 114, 115, 132, 135,

139, 142, 144, 147, 148, 154, 156, 157, 163, 174, 178, 182, 202, 203, 223, 236, 247, 249, 250, 255, 256, 258, 259, 260, 261, 262, 263, 264, 265, 266, 267, 268, 274, 275, 276, 281, 283, 285, 292, 295, 297, 299, 306, 307, 311, 314, 315, 318, 322, 326, 338, 360, 366, 368, 369, 370, 382, 383, 387, 391, 394, 397, 402, 405, 406, 408, 413, 449, 451, 455

hydrogen atoms, 147, 256, 394, 397 hydrogen bonds, 144, 147 hydrogen gas, 281 hydrogen peroxide, 236, 405, 406 hydrolysis, 153, 393 hydrophilicity, 137, 142, 148, 151, 453, 455, 464

Index 479

hydrophobicity, 4, 5, 137, 148, 399, 427 hydroquinone, 139, 140 hydroxide, 220, 446 hydroxyl, 137, 391, 401, 420, 422, 437, 455, 456 hydroxyl groups, 401 hypothesis, 93, 96, 103, 104, 120, 121 hysteresis, 184

I

identification, 162, 180, 186, 208 illumination, 217 images, 384, 425 immersion, 463 implementation, 162, 355, 376 impregnation, 412, 413, 427 impurities, 249, 255, 258, 414, 418, 422 in situ, 153, 383, 409 independent variable, 72, 77, 78, 113, 116, 118 indication, 4, 113 indigenous, 162 indium, 406, 410 induction, 219 industry, viii, ix, 161, 206, 218, 229, 273, 280, 455 inertia, 165, 166, 194, 195, 202 infrastructure, 247 inhibition, 268 inhibitor, 451 input, 20, 80, 162, 166, 168, 171, 180, 181, 182, 184,

185, 186, 193, 194, 197, 198, 203, 237, 294, 295, 434

insight, ix, x, 179, 273, 274, 298, 303 instability, 193, 194, 200, 447 insulation, 78, 226, 237, 285 insulators, 220 integration, vii, 9, 11, 162, 168, 201, 207, 222, 224,

236 intensity, 139, 147, 416, 420, 423, 428, 429, 434,

435, 437, 440, 453, 454, 456 interaction, ix, x, 72, 77, 113, 122, 123, 124, 125,

126, 127, 128, 129, 142, 144, 145, 147, 151, 152, 153, 161, 252, 273, 298, 303, 381, 399, 402, 404, 423

interaction effect, 72, 113, 124, 125, 126 interaction effects, 72, 113 interactions, 72, 127, 128, 129, 147, 150, 158, 179,

291, 383, 400 interface, viii, 13, 114, 137, 153, 161, 200, 236, 250,

266, 291, 297, 311, 314, 316, 355, 359, 364, 366, 382, 386, 398, 449

interference, 269 interpretation, 125, 249 interval, 80

intervention, 194 ion bombardment, 384 ion transport, 448, 455, 460, 470 ionization, 256 ions, 13, 152, 163, 173, 219, 220, 221, 280, 388,

446, 448, 455, 457, 459, 462 IR, 416, 420, 434, 437, 438, 440, 441 IR spectra, 420, 434, 438 IR spectroscopy, 420 iron, 254 irradiation, 415 isolation, 222 Italy, viii, 71, 72, 73, 74, 132, 133, 135, 244, 245

J

Japan, 214, 225, 243, 427

K

K+, 461 kinetic parameters, 263, 265, 267 kinetics, ix, x, 10, 38, 164, 223, 236, 247, 264, 265,

266, 269, 270, 273, 289, 294, 295, 298, 300, 301, 303, 314, 376, 402, 427

KOH, xi, 220, 226, 228, 237, 248, 264, 281, 416, 445, 447, 448, 449, 450, 451, 453, 455, 456, 459, 460, 461, 463, 464, 466, 469

Korea, 242, 244, 411, 416, 427, 441, 443

L

laminar, 236, 241, 305, 322 land, 286, 298, 299, 300, 301, 302, 304, 316, 325,

326, 329, 335, 337, 338, 347, 366, 368 laser ablation, 230, 231 leakage, 78, 101, 447, 448, 461, 469 leaks, 223 LED, 230 lending, viii, 135 lice, 351, 353 life cycle, 287 lifetime, 4, 211, 212, 254, 280, 302, 347, 401 likelihood, 5 limitation, 31, 295, 298 linear model, 180, 185 linear systems, 201 linkage, 384, 392, 397, 398, 437 liquid phase, 276, 355, 356, 357, 358, 359, 360, 364,

366 liquids, 134, 301 lithium, 75, 173, 248, 446, 447

Index 480

lithography, 216, 217, 221, 229, 232 location, 93, 165, 176, 200, 288, 395 low temperatures, ix, 273

M

management, 163, 213, 225, 227, 241, 250, 269, 292, 295, 297, 331, 348, 354, 355, 407

manganese, 450 manifolds, 228 manufacturing, 6, 214, 216, 225, 238, 252, 253, 287 market, 211, 212, 229, 450 markets, 11 masking, 217, 220, 226, 237 mass loss, 403 material surface, 437 matrix, 118, 119, 139, 144, 151, 152, 153, 163, 171,

202, 208, 248, 285, 310, 448, 453, 455, 457, 461 Maxwell equations, 295, 358 measurement, 16, 72, 103, 170, 197, 249, 389, 416 measures, 127 mechanical energy, 195 mechanical properties, 137, 139, 144, 150, 151, 156,

447, 451 mechanical stress, 5, 302 mechanical structure, 214 media, 174, 236, 241, 254, 256, 263, 264, 268, 269,

309, 310, 448, 456 melt, 173 melting, 219, 449 membranes, viii, xi, 10, 11, 135, 136, 137, 142, 144,

145, 147, 148, 149, 150, 151, 152, 153, 154, 155, 157, 158, 159, 160, 214, 222, 223, 226, 236, 237, 238, 239, 241, 347, 408, 445, 446, 448, 453, 455, 456, 459, 461, 462, 463, 464, 466, 467, 468

memory, 216 metal oxide, 212, 287 metal oxides, 212 metal salts, 219 metals, 213, 220, 223, 254, 255, 286, 382, 386, 414 methane, 83, 89, 163, 174, 178 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, 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, 152, 153, 154, 157, 158, 159, 223, 225, 227, 228, 233, 235, 236, 280, 297, 382, 383, 407, 408, 412, 413, 414, 415, 425, 426, 427, 430, 436, 439, 440, 448

methyl groups, 147 microelectronics, ix, 211, 213, 214, 229, 230, 437

microfabrication, ix, 211, 213, 214, 216, 217, 221, 224, 225, 231, 236, 237, 240

micrometer, 236 microscope, 3, 239 microstructure, 148, 154, 407, 419, 437 microstructures, 96, 137, 138, 148, 156 migration, 144 military, vii miniaturization, ix, 211, 213, 216, 236, 240 mixing, 251, 382, 386, 392, 417, 427, 428, 430, 431,

440, 449 mobility, 147, 151, 395, 400, 455 modeling, viii, x, 13, 71, 72, 76, 134, 162, 164, 165,

168, 173, 193, 201, 206, 207, 273, 291, 292, 294, 295, 296, 298, 299, 318, 362, 376

models, ix, x, 66, 71, 72, 76, 77, 78, 79, 81, 96, 101, 107, 116, 122, 132, 133, 162, 165, 169, 179, 180, 185, 186, 192, 195, 201, 206, 207, 208, 209, 269, 273, 274, 291, 292, 294, 295, 297, 298, 300, 301, 302, 303, 321, 371, 376

modules, 180, 402 modulus, 277, 319, 320 moisture, 174, 222, 232, 313, 393 moisture content, 313 mold, 222 mole, 203, 277, 292, 306, 319, 329 molecular mass, 97, 202 molecular weight, 16, 97, 114, 137, 308, 449 molecules, 97, 98, 114, 148, 150, 218, 221, 236, 237,

361, 393 molybdenum, 254, 255 momentum, 165, 305, 310, 355, 356, 357, 361 monograph, 383 monomer, 138, 140, 141, 144, 146, 388, 455 monomers, 136, 138 morphology, 3, 392, 413, 415, 426, 427, 433, 453,

466 motion, 193, 314, 461 motivation, 10, 249 movement, 448, 461 multidimensional, ix, 273 multiplication, 119

N

Na+, 461 nanocomposites, 414 nanofibers, x, 383, 411, 414, 415, 426, 440 nanomaterials, 415 nanometer, 214, 432 nanometer scale, 214, 432 nanometers, 383

Index 481

nanoparticles, x, 388, 389, 407, 408, 411, 414, 415, 421, 422, 425, 427, 428, 429, 430, 431, 432, 440

nanostructured materials, 407 nanotube, 410 nanotubes, x, 218, 383, 404, 407, 408, 409, 411, 415,

440 natural gas, 10, 72, 74, 75, 76, 132, 163 natural resources, 10, 278 negativity, 399 neglect, 295 Netherlands, 75, 417, 428 network, 150, 185, 192, 193, 202, 206, 222, 294, 447 neural network, 294 nickel, 10, 163, 220, 254, 437 nitric acid, 385, 408 nitrogen, 48, 159, 175, 307, 318, 371, 384, 393, 403,

435, 436, 437 nitrogen oxides, 175 NMR, 139, 140 noble metals, 223, 254, 255 nodes, ix, 247, 249 noise, 171, 176, 279 normal distribution, 171 nucleation, 219, 384 null hypothesis, 121

O

observations, 116, 117, 118, 119, 120, 122, 123, 125, 127, 130

oil, 76, 381 operator, 76, 78, 81, 186 optimization, vii, ix, 9, 76, 132, 133, 175, 273, 294,

303, 413 ordinary differential equations, 165 organic compounds, 413 organic solvent, 144, 153, 388 organic solvents, 144, 153, 388 orientation, 219 osmosis, 455 output, viii, ix, 9, 12, 20, 21, 29, 30, 31, 33, 34, 41,

44, 46, 47, 51, 54, 55, 56, 57, 64, 67, 80, 154, 155, 162, 166, 167, 169, 170, 171, 172, 180, 182, 183, 184, 185, 187, 193, 194, 195, 197, 198, 200, 202, 203, 204, 227, 273, 280, 294, 314, 315

oxidation, ix, 10, 11, 13, 21, 27, 37, 40, 42, 77, 150, 163, 174, 213, 216, 217, 219, 247, 249, 255, 257, 258, 259, 262, 265, 266, 267, 283, 382, 384, 385, 387, 391, 392, 402, 403, 404, 405, 408, 410, 414, 415, 423, 426, 427, 430, 431, 436, 437, 439, 440, 448, 457, 466

oxides, 153, 156, 157, 175, 212 oxyfluorination, x, 411, 437, 438, 439, 441

oxygen, vii, viii, 9, 10, 12, 13, 14, 31, 37, 41, 42, 48, 51, 58, 66, 67, 93, 114, 115, 147, 148, 154, 156, 163, 168, 174, 202, 203, 213, 233, 236, 258, 275, 276, 281, 283, 285, 291, 292, 295, 299, 300, 301, 306, 307, 311, 314, 315, 318, 321, 322, 323, 325, 326, 338, 348, 354, 360, 364, 366, 367, 368, 369, 371, 372, 382, 385, 391, 399, 403, 408, 415, 418, 420, 422, 424, 425, 430, 437, 441, 450

oxygen consumption, 326 oxygen consumption rate, 326 ozone, 237

P

PAA, xi, 445, 447, 449, 455, 456, 457, 458, 459, 460, 464, 465, 466, 467, 468, 469

packaging, 216, 230 palladium, 226, 227, 254 PAN, 448 parameter, 73, 77, 80, 95, 96, 106, 107, 116, 137,

144, 145, 148, 164, 169, 175, 177, 178, 195, 201, 203, 277, 294, 319, 321, 362, 363, 367, 413, 419, 423

parameter estimation, 80, 96, 107 parents, 274 partial differential equations, 165 particles, x, 4, 5, 173, 228, 250, 252, 254, 269, 285,

288, 381, 382, 383, 384, 385, 386, 387, 388, 389, 390, 391, 392, 396, 397, 398, 400, 402, 403, 404, 406, 408, 411, 414, 415, 418, 425, 426, 427, 428, 432, 440

passivation, 221 passive, 226, 227, 231, 235, 237 patents, 250 pathways, 145, 286, 403 patterning, 221, 222, 230, 232 pedigree, viii, 71, 73, 81, 112, 113, 114 PEEK, viii, 135, 136, 137, 138, 139, 140, 141, 142,

143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159

periodicity, 316 permeability, xi, 136, 146, 148, 149, 152, 154, 158,

275, 299, 300, 319, 320, 357, 363, 367, 368, 371, 372, 445

permeation, 148, 157, 159 permit, 144, 219, 268 peroxide, 236, 405, 406, 407, 455 pH, 181, 182, 236, 251, 257, 260, 261, 416, 419,

421, 428, 449 phase boundaries, 385, 386, 387, 402 phenol, 144, 391 photoelectron spectroscopy, 416, 440

Index 482

photolithography, 216, 217, 218, 222, 227, 232, 233, 236, 237, 240

photoluminescence, 221 physical mechanisms, 355, 376 physical properties, x, 217, 250, 269, 381 physicochemical properties, 271, 407 physico-chemical properties, 413 physics, 295, 300, 302, 303, 357, 376 pitch, 414 planning, 78, 192 plants, viii, ix, 161, 162, 163, 169, 206, 209, 248,

278 plasma, x, 218, 220, 221, 383, 384, 391, 399, 407,

409, 411, 434, 435, 436, 437, 440 platinum, x, xi, 10, 41, 135, 163, 213, 220, 221, 225,

248, 249, 253, 254, 255, 263, 264, 266, 267, 269, 285, 295, 400, 407, 408, 409, 411, 412, 415, 417, 418, 421, 422, 429, 430, 440

PMMA, 231, 232, 233, 235 Poincaré, 211 poison, 405 polar groups, 151 polarity, 437 polarization, viii, 5, 71, 73, 80, 88, 93, 96, 102, 104,

105, 114, 115, 133, 134, 154, 164, 203, 255, 260, 282, 290, 291, 295, 320, 321, 322, 348, 415

pollution, 76, 279, 382 polycarbonate, 455 polycondensation, 146 polydimethylsiloxane, 231 polyether, 158 polyetheretherketone, viii, 135, 136, 159 polyetheretherketone (PEEK), viii, 135, 136 polyimide, 233 polymer, viii, ix, xi, 10, 14, 15, 58, 135, 136, 137,

138, 139, 140, 142, 143, 144, 145, 148, 149, 150, 151, 152, 153, 154, 157, 158, 159, 212, 214, 220, 222, 225, 230, 232, 233, 241, 247, 248, 264, 266, 268, 282, 283, 284, 285, 298, 312, 348, 355, 361, 367, 382, 383, 388, 389, 390, 391, 400, 401, 407, 408, 409, 412, 426, 428, 433, 445, 446, 447, 448, 449, 451, 453, 454, 455, 456, 457, 459, 461, 462, 464, 465, 466, 467, 468, 469

polymer blends, 467 polymer chains, 400 polymer electrolytes, 446, 448, 449, 453, 467 polymer film, 151, 454 polymer films, 454 polymer matrix, 151, 153, 448, 453, 455, 457 polymer solubility, 139 polymer structure, 137 polymeric blends, 464 polymeric chains, 145

polymeric membranes, 136 polymerization, 152, 157, 388, 400, 409, 449, 455 polymerization time, 152 polymers, viii, xi, 135, 136, 138, 139, 142, 144, 145,

146, 148, 149, 153, 156, 157, 158, 213, 217, 231, 233, 236, 241, 387, 388, 390, 400, 408, 445, 447, 455, 461

polypropylene, 449, 461, 462 polystyrene, 10 polyvinyl alcohol, xi, 445, 447, 455 poor, 17, 20, 21, 31, 40, 44, 122, 144, 149, 150, 195,

197, 223, 225, 226, 233, 286, 448 porosity, 4, 115, 152, 250, 251, 286, 292, 294, 299,

319, 383, 413, 426, 462 porous materials, 286 porous media, 241, 309, 310 porphyrins, 10 potassium, 173, 220, 248, 455 potassium persulfate, 455 power, vii, viii, ix, xi, 9, 10, 11, 12, 13, 19, 20, 21,

22, 26, 27, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 41, 42, 44, 46, 47, 48, 51, 53, 54, 56, 57, 58, 60, 61, 63, 65, 66, 67, 74, 75, 76, 113, 155, 157, 161, 162, 163, 165, 166, 167, 168, 169, 170, 171, 172, 173, 176, 179, 180, 181, 183, 184, 185, 186, 188, 192, 193, 194, 195, 196, 197, 198, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 211, 216, 221, 222, 224, 225, 227, 228, 229, 230, 231, 232, 233, 240, 241, 248, 273, 276, 278, 279, 280, 281, 282, 285, 294, 300, 302, 315, 320, 331, 348, 354, 376, 382, 394, 395, 396, 397, 399, 412, 414, 415, 445, 446, 447, 448, 450, 459, 467, 469

power generation, 20, 162, 192, 206, 207, 227, 248, 302, 382

prediction, x, 132, 186, 202, 205, 274, 302, 303, 304, 318, 376, 385

prediction models, 302 president, 274 pressure, vii, viii, 4, 9, 20, 31, 48, 58, 67, 93, 94, 96,

97, 98, 99, 114, 115, 122, 154, 164, 165, 168, 169, 175, 181, 182, 203, 204, 208, 214, 221, 227, 250, 268, 276, 289, 291, 294, 295, 297, 299, 300, 303, 305, 306, 311, 313, 315, 316, 317, 319, 355, 356, 358, 360, 361, 363, 364, 366, 367, 376, 383, 416, 434, 437

priming, 227 probability, 127, 156 production, 27, 76, 137, 174, 214, 285, 291, 294,

301, 311, 331, 338, 354, 360, 364, 413 production costs, 413 program, ix, 165, 273, 274, 417 proportionality, 266 propylene, 448

Index 483

protons, 13, 14, 137, 139, 145, 148, 155, 156, 260, 281, 283, 284, 285, 314, 338, 355, 382, 385, 387, 388, 389, 391, 399

prototype, 72, 73, 230 PTFE, 232, 233, 235, 250, 251, 252, 253, 265, 267,

382, 386, 389, 398, 399, 450, 451, 455 pulse, 12, 30, 183, 186, 196 pumps, vii, 9, 18, 227 pure water, 358 PVA, xi, 445, 446, 447, 449, 450, 451, 452, 453,

454, 455, 456, 457, 458, 459, 460, 461, 464, 465, 466, 467, 468, 469

pyrolysis, 144, 174

Q

quality improvement, 206 quartz, 214 quasi-equilibrium, 259 quaternary ammonium, 448 quinone, 391

R

radial distribution, 192 radiation, 217, 227, 416 radical polymerization, 449, 455 radio, 220, 221, 399, 434 random errors, 118 range, 4, 20, 21, 38, 51, 60, 63, 65, 72, 79, 104, 105,

111, 134, 137, 145, 151, 167, 168, 170, 173, 180, 181, 187, 194, 222, 241, 248, 254, 259, 278, 285, 294, 301, 382, 383, 422, 428, 431, 432, 449, 457, 459, 464, 466

reactant, 4, 5, 10, 13, 76, 95, 96, 197, 203, 250, 280, 285, 286, 289, 295, 297, 298, 299, 300, 304, 313, 314, 337, 363, 368, 371, 372, 376, 385, 386, 387, 401, 414

reactants, vii, ix, 4, 13, 15, 94, 115, 154, 156, 165, 180, 182, 222, 224, 249, 250, 254, 273, 283, 285, 288, 291, 295, 303, 306, 311, 314, 322, 348, 363, 370, 376, 383, 385

reaction mechanism, 263, 289 reaction order, 96 reaction rate, 250, 262, 289, 291, 301 reaction time, 144, 437, 440 reaction zone, 249 reactivity, 249, 257, 267, 268, 437 reality, 5, 155, 385 recovery, 168, 205, 208 recycling, 18

reduction, 10, 13, 37, 41, 42, 80, 84, 87, 122, 144, 151, 174, 195, 201, 213, 241, 258, 301, 354, 357, 382, 383, 384, 388, 389, 390, 391, 393, 394, 396, 399, 403, 405, 407, 408, 412, 413, 415, 416, 418, 421, 422, 425, 427, 428, 429, 440, 457, 466

reflection, 419 reforms, 163 regional, 193 regression, 71, 72, 76, 77, 78, 79, 81, 109, 111, 112,

113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 126, 127, 129, 132, 133

regression analysis, 114 regulation, 89, 192, 193, 206 regulators, 190, 193 reinforcement, 5, 401 relationship, xi, 11, 40, 44, 76, 116, 120, 156, 183,

249, 292, 445 relationships, 18, 19, 182, 183, 263, 265, 292, 294 relaxation, 455 relevance, 11, 226, 236 reliability, vii, viii, 12, 13, 161, 302 renewable energy, 381 reparation, 440 replacement, 76, 241, 281 replication, 221, 222 reserves, 381 residuals, 119, 120, 127 resistance, x, 4, 5, 15, 16, 17, 20, 22, 34, 37, 41, 53,

65, 67, 81, 96, 106, 107, 142, 150, 153, 155, 225, 233, 240, 254, 255, 267, 268, 286, 287, 292, 299, 301, 312, 337, 338, 372, 383, 390, 392, 396, 404, 405, 411, 412, 415, 432, 433, 440, 448, 449, 459, 464

resolution, 221, 320, 384 resources, ix, 10, 161, 273, 278, 304 response time, 180, 198, 203 retardation, 401 retention, 145, 151 RIE, 221, 225, 226, 227, 229, 230, 237 rings, 139, 228, 406 risk, 213 robustness, 172, 176, 178 rolling, 251, 252, 285 Romania, 71 room temperature, xi, 145, 147, 151, 225, 232, 235,

267, 412, 417, 437, 445, 447, 448, 469 roughness, 437 Royal Society, 389, 394 rural areas, 192 ruthenium, 27, 40, 41, 254, 255

Index 484

S

safety, viii, 72, 76, 132, 161, 193, 212, 213, 216, 222 salt, 151, 159, 237, 394, 395, 397, 398, 461 salts, 157, 173, 219, 409, 446 sample, 76, 116, 140, 142, 143, 149, 218, 416, 418,

419, 430, 453, 456, 464, 466 sampling, 186 Samsung, 229 saturation, x, 274, 276, 300, 301, 306, 329, 358, 360,

363, 364, 366, 367, 376 scaling, 253, 300 search, 213, 412 searching, 212 second generation, 173 seed, 219 segregation, 3, 152 selectivity, 149, 220 SEM micrographs, 431, 432, 463, 464, 465 semiconductor, 218, 219 semiconductors, 220 sensing, 407 sensitivity, viii, ix, 71, 72, 73, 77, 80, 81, 82, 83, 84,

85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 99, 100, 101, 102, 103, 104, 113, 133, 134, 194, 273, 298, 303

sensors, vii, 9, 11, 18, 214 separation, 137, 138, 148, 149, 150, 151, 156, 157,

222, 227, 285, 448, 453, 454, 464 series, vii, xi, 9, 11, 16, 17, 21, 66, 72, 74, 123, 137,

203, 224, 445, 447 service quality, viii, 161 severity, 193 shape, 87, 102, 231, 232, 265, 347, 348, 351, 353,

359, 432, 448 sharing, 201, 305 shear, 251, 358 side effects, 27 sign, 126, 128 signals, 139, 167, 214 signs, 126, 127 silane, ix, 211, 214, 236, 237, 241, 392, 393, 399,

409 silicon, ix, 12, 211, 213, 214, 217, 218, 219, 220,

221, 222, 224, 226, 227, 228, 229, 230, 233, 235, 236, 237, 238, 239, 241, 243, 248

silver, 220, 232 simulation, vii, ix, 9, 11, 12, 13, 22, 24, 27, 28, 29,

31, 32, 38, 39, 42, 43, 45, 46, 48, 50, 52, 53, 54, 55, 58, 60, 61, 62, 63, 64, 96, 163, 165, 169, 170, 171, 173, 176, 179, 180, 193, 200, 205, 206, 207, 208, 269, 273, 303, 304, 320, 322, 379

Singapore, 132

sintering, 252, 384 SiO2, 218, 225, 237 sites, x, 145, 150, 151, 250, 254, 256, 267, 381, 383,

384, 385, 387, 393, 405, 414, 415 skin, 406 sodium, 151, 157, 248, 251, 388, 418 soft lithography, 221 software, 193, 207, 294 SOI, 219 solid matrix, 285, 310 solid oxide fuel cells, 133, 382 solid phase, 151 solid state, xi, 180, 445, 450, 451 solubility, 136, 139, 144, 150, 151, 449 solvents, 136, 137, 144, 147, 148, 150, 151, 153,

251, 388, 413 sorption, 152 Spain, 161, 247, 441 species, ix, x, 96, 97, 114, 150, 174, 203, 257, 258,

273, 274, 289, 291, 292, 295, 297, 298, 300, 303, 304, 305, 306, 307, 308, 309, 316, 322, 324, 355, 356, 358, 363, 376, 412, 423, 424, 426, 430

specific heat, 203, 309 specific surface, 310, 416, 419, 427, 428, 429, 432,

440 spectroscopy, 267, 390, 391, 416, 420, 428 spectrum, 140, 424 speed, 78, 165, 180, 193, 194, 195, 202, 204 spin, 217 sputtering, 219, 220, 226, 227, 232, 233, 253 stability, viii, 3, 5, 10, 136, 137, 139, 149, 150, 151,

153, 161, 162, 176, 177, 178, 193, 194, 195, 198, 200, 201, 202, 205, 207, 222, 248, 285, 382, 384, 390, 404, 406, 407, 409, 410, 412, 447, 448, 449, 450, 451, 453, 457, 466, 467, 469

stabilization, 175, 209, 229, 426 stages, 123, 175, 192, 248 standard deviation, 81, 83, 85, 86, 171, 172, 202 starvation, 17, 21, 185 statistical analysis, viii, 71, 72, 113, 127 steel, 230, 233, 235, 286 stock, 459 stoichiometry, 86, 87, 96, 99, 113 storage, 10, 195, 206, 232, 281, 390, 448 strain, 313 strategies, 144, 208 strength, ix, x, 145, 151, 252, 273, 298, 303, 318,

381, 447, 449, 453, 455, 464, 469 stress, 5, 313, 347, 350, 466 stretching, 462 substitution, ix, 140, 141, 153, 161 substitution reaction, 140, 141 substrates, 213, 217, 219, 221, 223, 225, 227, 232

Index 485

subtraction, 126 sulfate, 397, 398, 409 sulfur, 279, 384, 385, 397, 401, 418, 422, 430 sulfuric acid, 10, 226, 233, 235, 237, 425, 462 summer, 75 Sun, 158, 244, 442 supply, ix, 11, 181, 193, 195, 198, 200, 201, 206,

211, 212, 213, 216, 223, 224, 225, 233, 298, 299, 314, 322, 459

surface area, 4, 12, 222, 249, 254, 255, 263, 264, 267, 269, 285, 291, 310, 359, 360, 382, 383, 385, 386, 391, 402, 404, 406, 412, 413, 414, 416, 419, 421, 427, 428, 429, 430, 431, 432, 440

surface chemistry, 413, 437 surface diffusion, 96 surface layer, 383 surface modification, 407, 414, 437 surface properties, 412, 441 surface treatment, 411 sustainability, 447 swelling, 144, 151, 152, 153, 156, 313, 348 switching, 184 Switzerland, 132, 133, 134 symbols, 124 symmetry, 89, 156, 264, 316, 361 symptom, 4 synergistic effect, 255 synthesis, 138, 140, 146, 151, 159, 169, 174, 412,

413, 415, 416, 427, 428 system analysis, 186 systems, ix, xi, 3, 12, 13, 20, 31, 66, 67, 73, 135,

136, 149, 161, 162, 163, 166, 167, 173, 179, 185, 186, 187, 188, 192, 193, 194, 195, 197, 201, 205, 206, 207, 208, 209, 212, 213, 214, 218, 222, 254, 273, 278, 279, 280, 281, 292, 294, 302, 303, 347, 407, 413, 445, 446, 447, 448, 464

T

Taiwan, 445 tar, 174, 175 technical carbon, 265 technology, ix, xi, 10, 135, 163, 173, 192, 201, 205,

208, 212, 214, 216, 219, 221, 225, 227, 229, 230, 232, 233, 236, 243, 247, 273, 302, 407, 408, 410, 437, 445

telecommunications, 11, 214 TEM, 384, 388, 389, 425 temperature dependence, 172, 179 temperature gradient, 298, 348 tensile stress, 466 tension, 358 terminals, 196

test procedure, 121 test statistic, 121, 122 TGA, 143, 144 theory, 163, 164, 175, 205, 261, 292, 306, 313 thermal decomposition, 383, 397 thermal energy, 75, 161, 162 thermal expansion, 254 thermal oxidation, 217 thermal stability, 136, 137, 139, 150, 285, 384, 406,

469 thermal treatment, 384, 398, 437 thermodynamic equilibrium, 356, 359 thermodynamic function, 288 thermodynamics, 10, 80, 159, 385 thin films, 216, 217, 229 threat, 10 three-dimensional model, 298, 299, 300, 322 threshold, 184 time, vii, 5, 6, 12, 29, 44, 72, 80, 122, 127, 137, 144,

151, 152, 153, 156, 162, 165, 169, 170, 171, 175, 176, 177, 179, 180, 181, 183, 185, 186, 194, 195, 196, 198, 201, 203, 204, 205, 207, 208, 209, 218, 221, 278, 281, 291, 295, 297, 303, 320, 348, 361, 388, 391, 395, 404, 414, 416, 419, 434, 436, 437, 440, 448, 452, 453, 460, 462, 463, 464

time frame, 194 tin, 254, 255, 406, 410 tin oxide, 406, 410 titanium, 218, 220, 231, 233 toluene, 251 TPA, 151, 156 tracking, 169, 195 transformation, 147 transition, 10, 75, 142, 269, 386 transition metal, 10, 386 transition temperature, 142 transmission, viii, 161, 163, 193, 196, 204, 205, 388 transmission electron microscopy, 388 transport, ix, x, xi, 4, 5, 11, 15, 134, 137, 145, 153,

156, 157, 159, 236, 248, 249, 250, 258, 264, 268, 269, 273, 274, 283, 284, 286, 291, 292, 294, 295, 297, 298, 299, 300, 302, 303, 304, 305, 309, 310, 312, 314, 318, 320, 321, 322, 325, 331, 337, 338, 344, 348, 354, 355, 357, 358, 361, 363, 367, 370, 372, 376, 382, 387, 395, 396, 399, 401, 408, 413, 414, 445, 455, 457, 459, 460, 461, 462, 470

transport processes, 258, 284, 292, 300, 302, 304, 318

transportation, ix, 10, 273, 280, 281, 448 trees, 173 trend, 89, 92, 109, 269, 461 trial, 310 tungsten, 218, 220, 254

Index 486

U

UK, 270, 271, 273, 274 uncertainty, 162, 172, 175, 178 underlying mechanisms, 368 uniform, 5, 72, 87, 89, 104, 173, 217, 251, 301, 302,

303, 313, 335, 338, 348, 415, 453, 454, 462, 466 universal gas constant, 115, 203 users, 20 UV, 217, 232, 237 UV light, 217

V

vacuum, 220, 222, 253, 416, 427, 455 validation, 133, 206, 318 values, viii, 27, 42, 48, 58, 60, 62, 71, 73, 79, 80, 81,

83, 84, 86, 87, 88, 90, 91, 92, 98, 101, 102, 103, 104, 107, 108, 109, 110, 111, 112, 113, 122, 126, 127, 139, 142, 144, 145, 147, 150, 151, 152, 154, 155, 171, 257, 258, 260, 267, 298, 299, 306, 316, 318, 319, 321, 326, 337, 338, 347, 362, 416, 418, 419, 420, 430, 432, 440, 454, 461, 464

vapor, 104, 159, 219, 220, 250, 283, 305, 306, 307, 318, 323, 324, 329, 354, 355, 359, 361, 363, 364, 366, 368, 369, 371, 372, 376, 382, 383, 406

variability, 122, 128, 129 variable, 76, 77, 78, 79, 80, 81, 99, 115, 116, 120,

122, 133, 166, 167, 208, 292, 295, 316, 318, 348, 361, 388, 399

variables, viii, 71, 72, 73, 76, 77, 78, 80, 113, 115, 116, 118, 120, 121, 122, 125, 127, 128, 137, 150, 167, 173, 184, 195, 202, 294, 316

variance, 79, 80, 120, 121, 122, 125, 128, 131, 132, 169, 171, 176, 200, 202

variation, 3, 5, 51, 60, 79, 80, 83, 86, 87, 89, 90, 93, 94, 99, 122, 131, 155, 162, 169, 170, 171, 176, 200, 202, 299, 300, 311, 322, 337, 338, 358, 361, 367

vector, 118, 119, 183, 184, 195, 204, 276 vehicles, 211, 279, 281, 446 velocity, x, 145, 274, 276, 298, 305, 310, 316, 322,

323, 324, 358, 361, 363, 364, 365, 366

viscosity, 251, 308, 357, 366 visualization, 207 volumetric changes, 236

W

water absorption, 142 water evaporation, 467 water vapor, 104, 159, 305, 306, 307, 318, 323, 324,

329, 355, 361, 363, 364, 366, 368, 369, 371, 376 wear, 5 weight loss, 142, 143, 144 weight ratio, 440, 455 welding, 10 wettability, 136, 391 wetting, 355, 372, 376 wind, 10 windows, 449 winter, 75 wood, 173, 209 writing, 186

X

XPS, 414, 416, 422, 423, 425, 434, 435, 440 X-ray diffraction, 416, 417, 421, 428, 429 X-ray diffraction (XRD), 416 X-ray photoelectron spectroscopy (XPS), 416, 440 XRD, 416, 418, 419, 420, 425, 428, 438, 439, 441,

453, 454, 455, 456

Y

yield, 122, 175 yttria-stabilized zirconia, 248

Z

zinc, 448, 450, 451, 452, 453, 466 zirconia, 153, 156, 157, 248 zirconium, 153, 154 ZnO, 451

1600216692 fuel cell

Documents