Transport Phenomena in Fuel Cells

TRANSPORT PHENOMENA
IN FUEL CELLS
WIT Press publishes leading books in Science and Technology.

Visit our website for the current list of titles.
www.witpress.com
WITeLibrary
Home of the Transactions of the Wessex Institute, the WIT electronic-library provides the
international scientific community with immediate and permanent access to individual
papers presented at WIT conferences. Visit the WIT eLibrary at www.witpress.com
International Series on Developments in Heat Transfer
Objectives
The Developments in Heat Transfer book Series publishes state-of-the-art

books and provides valuable contributions to the literature in the field of heat
transfer. The overall aim of the Series is to bring to the attention of the
international community recent advances in heat transfer by authors in
academic research and the engineering industry.
Research and development in heat transfer is of significant importance
to many branches of technology, not least in energy technology. Developments
include new, efficient heat exchangers, novel heat transfer equipment as well
as the introduction of systems of heat exchangers in industrial processes.
Application areas include heat recovery in the chemical and process industries,
and buildings and dwelling houses where heat transfer plays a major role.
Heat exchange combined with heat storage is also a methodology for
improving the energy efficiency in industry, while cooling in gas turbine
systems and combustion engines is another important area of heat transfer
research.
To progress developments within the field both basic and applied
research is needed. Advances in numerical solution methods of partial
differential equations, high-speed, efficient and cheap computers, advanced
experimental methods using LDV (laser-doppler-velocimetry), PIV (particle-
image-velocimetry) and image processing of thermal pictures of liquid
crystals, have all led to dramatic advances during recent years in the solution
and investigation of complex problems within the field.
The aims of the Series are achieved by contributions to the volumes
from invited authors only. This is backed by an internationally recognised
Editorial Board for the Series who represent much of the active research
worldwide. Volumes planned for the series include the following topics:
Compact Heat Exchangers, Engineering Heat Transfer Phenomena, Fins
and Fin Systems, Condensation, Materials Processing, Gas Turbine Cooling,
Electronics Cooling, Combustion-Related Heat Transfer, Heat Transfer in
Gas-Solid Flows, Thermal Radiation, the Boundary Element Method in Heat
Transfer, Phase Change Problems, Heat Transfer in Micro-Devices, Plate-
and-Frame Heat Exchangers, Turbulent Convective Heat Transfer in Ducts,
Enhancement of Heat Transfer and other selected topics.
Series Editor
B. Sundén
Lund Institute of Technology
Box 118
22100 Lund
Sweden
Associate Editors
C.I. Adderley S. del Guidice

Rolls Royce and Associates Limited University of Udine
UK Italy
E. Blums M. Faghri
Latvian Academy of Sciences The University of Rhode Island
Latvia USA
C.A. Brebbia P.J. Heggs

Wessex Institute of Technology UMIST
UK UK
G. Comini C. Herman
University of Udine John Hopkins University
Italy USA
R.M. Cotta D.B. Ingham

COPPE/UFRJ, University of Leeds
Brazil UK
L. De Biase Y. Jaluria
University of Milan Rutgers University
Italy USA
G. De Mey S. Kotake
University of Ghent University of Tokyo
Belgium Japan
G. de Vahl Davies P.S. Larsen

University of New South Wales Technical University of Denmark
Australia Denmark
D.B. Murray A.C.M. Sousa
Trinity College Dublin University of New Brunswick
Ireland Canada
A.J. Nowak D.B. Spalding

Silesian University of Technology CHAM
Poland UK
K. Onishi J. Szmyd
Ibaraki University University of Mining and Metallurgy
Japan Poland
P.H. Oosthuizen E. Van den Bulck

Queen’s University Kingston Katholieke Universiteit Leuven
Canada Belgium
W. Roetzel S. Yanniotis
Universtaet der Bundeswehr Agricultural University of Athens
Germany Greece
B. Sarler
University of Ljubljana
Slovenia
TRANSPORT PHENOMENA
IN FUEL CELLS
Editors
B. Sundén
Lund Institute of Technology, Sweden.
M. Faghri
University of Rhode Island, USA.
TRANSPORT PHENOMENA
IN FUEL CELLS
Series: Developments in Heat Transfer, Vol. 19
Editors: B. Sundén and M. Faghri
Published by
WIT Press
Ashurst Lodge, Ashurst, Southampton, SO40 7AA, UK
Tel: 44 (0) 238 029 3223; Fax: 44 (0) 238 029 2853
E-Mail: witpress@witpress.com
http://www.witpress.com
For USA, Canada and Mexico
WIT Press
25 Bridge Street, Billerica, MA 01821, USA
Tel: 978 667 5841; Fax: 978 667 7582
E-Mail: infousa@witpress.com
http://www.witpress.com
British Library Cataloguing-in-Publication Data
A Catalogue record for this book is available

from the British Library
ISBN: 1-85312-840-6
ISSN: 1369-7331
Library of Congress Catalog Card Number: 2004116358
No responsibility is assumed by the Publisher, the Editors and Authors for any injury
and/or damage to persons or property as a matter of products liability, negligence or
otherwise, or from any use or operation of any methods, products, instructions or ideas
contained in the material herein.
© WIT Press 2005.
Printed in Great Britain by Athenaeum Press Ltd.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the Publisher.
Contents
Preface xv
Chapter 1:
Multiple transport processes in solid oxide fuel cells 1
P.-W. Li, L. Schaefer & M.K. Chyu
1 Introduction......................................................................................... 1
2 Thermodynamic and electrochemical fundamentals for
solid oxide fuel cells ........................................................................... 3
2.1 Operation with hydrogen fuel ..................................................... 4
2.2 Operation with methane through internal reforming and
shift reactions.............................................................................. 8
3 Electrical potential losses .................................................................... 11
3.1 Activation polarization................................................................ 12
3.2 Ohmic loss ................................................................................. 14
3.3 Mass transport and concentration polarization ........................... 18
4 Computer modeling of a tubular SOFC............................................... 19
4.1 Outline of a computation domain................................................ 21
4.2 Governing equations and boundary conditions ........................... 22
4.3 Numerical computation............................................................... 26
4.4 Typical results from numerical computation for tubular SOFCs.. 27
4.4.1 The SOFC terminal voltage ............................................. 27
4.4.2 Cell temperature distribution ........................................... 30
4.4.3 Flow, temperature and concentration fields ..................... 31
5 Concluding remarks ............................................................................ 34
Chapter 2:
Numerical models for planar solid oxide fuel cells 43
S.B. Beale
1 Introduction......................................................................................... 44
1.1 History and types of solid oxide fuel cell ................................... 44
1.2 Survey of modeling techniques ................................................... 45
1.3 Thermodynamics of solid oxide fuel cells................................... 46
1.4 Cell voltage and current .............................................................. 49
1.5 Activation losses ......................................................................... 52
1.6 Diffusion losses........................................................................... 54
1.7 Basic computational algorithm................................................... 57
2 Computer schemes .............................................................................. 58
2.1 General scalar equation............................................................... 59
2.2 Continuity ................................................................................... 60
2.3 Momentum ................................................................................. 60
2.4 Heat transfer................................................................................ 62
2.5 Mass transfer............................................................................... 64
2.6 Numerical integration schemes ................................................... 64
2.7 Iterative procedure...................................................................... 65
2.8 Additional chemistry and electrochemistry................................. 66
2.9 Porous media flow ...................................................................... 67
2.10 Current and voltage distribution.................................................. 68
2.11 Advanced diffusion models ........................................................ 70
2.12 Thermal radiation........................................................................ 71
3 Stack models ....................................................................................... 73
4 Closure................................................................................................ 75
Chapter 3:
Electrochemical and thermo-fluid modeling of a tubular solid oxide
fuel cell with accompanying indirect internal fuel reforming 83
K. Suzuki, H. Iwai & T. Nishino
1 Introduction......................................................................................... 84
2 General remarks on the mechanism of IIR-T-SOFC ........................... 86
2.1 Tubular cell................................................................................. 86
2.2 Internal reforming process .......................................................... 88
2.3 Electrochemical process.............................................................. 88
2.4 Purpose and key points of the analysis........................................ 90
3 Numerical modeling............................................................................ 91
3.1 Computational domain and general assumptions for
heat and mass transfer ................................................................ 91
3.2 Model for electrochemical reactions ........................................... 93
3.3 Model for internal fuel reforming ............................................... 95
3.4 Governing equations of velocity, temperature and
concentration fields and boundary conditions ............................. 95
3.5 Discretization scheme................................................................. 98
3.6 Equations for electric potential and electric circuit .................... 101
3.7 Mass production or consumption rate of each chemical
species through electrochemical and reforming reactions............. 102
3.8 Model for thermodynamic heat generation rates ......................... 103
3.9 Ohmic heat generation ................................................................ 104
3.10 Radiation model.......................................................................... 104
3.11 Overall picture of the model ....................................................... 107
4 Results and discussion......................................................................... 108
4.1 Results for a cathode-supported tubular SOFC
without accompanying indirect internal reforming
[16,43,44].................................................................................. 108
4.2 Results for the Base case with accompanying indirect
internal reforming ....................................................................... 113
4.2.1 Thermal and concentration fields..................................... 115
4.2.2 Electric potential and current fields ................................. 116
4.2.3 Power generation characteristics...................................... 119
4.3 Strategies for the ideal thermal field ........................................... 121
4.3.1 Effect of gas inlet temperature ......................................... 121
4.3.2 Effect of air flow rate....................................................... 123
4.3.3 Effect of density distribution of catalyst ......................... 123
5 Conclusions......................................................................................... 125
Chapter 4:
On heat and mass transfer phenomena in PEMFC and SOFC and
modeling approaches 133
J. Yuan, M. Faghri & B. Sundén
1 Introduction......................................................................................... 133
2 Fuel cell modeling development......................................................... 135
2.1 Basics of SOFCs and PEMFCs................................................... 135
2.2 Modeling development............................................................... 137
2.2.1 Modeling approaches....................................................... 137
2.2.2 Various existing models ................................................... 138
3 Main processes in SOFCs and PEMFCs ............................................. 141
3.1 Gas transport............................................................................... 142
3.2 Electrochemical reactions ........................................................... 142
3.3 Heat transfer ............................................................................... 143
3.4 Various transport processes in the electrodes
(porous layers) ............................................................................ 143
3.5 Other processes appearing in fuel cell components..................... 144
4 Processes and issues in SOFC and PEMFC ........................................ 144
4.1 Water management in PEMFCs .................................................. 144
4.2 Fuel reforming issues in SOFC................................................... 145
5 Modeling methodologies for transport processes in
SOFC and PEMFC.............................................................................. 146
5.1 General considerations................................................................ 146
5.2 Assumptions................................................................................ 147
5.3 Governing equations ................................................................... 147
5.4 Boundary and interfacial conditions............................................ 149
5.5 Additional equations ................................................................... 150
5.6 Solution methodology................................................................. 151
6 Results and discussions ....................................................................... 152
6.1 Mass transfer effects on the gas flow and heat transfer ............... 152
6.2 Porous layer effects on the transport processes ........................... 155
6.2.1 Transport processes in PEMFCs ...................................... 155
6.2.2 Transport processes in SOFCs ......................................... 160
6.3 Two-phase flow and its effects on the cell performance ............. 163
7 Conclusions......................................................................................... 168
Chapter 5:
Two-phase transport in porous gas diffusion electrodes 175
S. Litster & N. Djilali
1 Introduction......................................................................................... 175
1.1 PEM fuel cells ............................................................................ 175
1.2 Porous media .............................................................................. 177
1.3 Porous media in PEMFC electrodes ........................................... 178
2 Single-phase transport......................................................................... 181
2.1 Transport of a single-phase with a single component ................. 181
2.1.1 Permeability .................................................................... 181
2.2 Transport of a single-phase with two components ...................... 182
2.2.1 Effective diffusivity in porous media............................... 183
2.2.2 Determination of the binary diffusion coefficient .............. 185
2.2.3 Transport of a single-phase with more than two
components...................................................................... 185
2.3 Knudsen diffusion....................................................................... 186
2.4 Determination of Knudsen diffusivity........................................ 187
2.5 Knudsen transition regime .......................................................... 187
2.5.1 Comparison of the diffusivities ........................................ 187
3 Two-phase systems.............................................................................. 188
3.1 Two-phase regimes ..................................................................... 189
3.2 Hydrodynamics and capillarity in two-phase systems................. 191
3.2.1 Capillary pressure curves ................................................. 195
3.3 Relative permeability.................................................................. 197
4 Multiphase flow models...................................................................... 200
4.1 Multi-fluid model ....................................................................... 200
4.1.1 Phase change.................................................................... 201
4.1.2 Application ...................................................................... 203
4.2 Mixture model ............................................................................ 205
4.3 Moisture diffusion model............................................................ 207
4.4 Porosity correction model ........................................................... 208
4.5 Evaluation of the multiphase models in the literature ................ 208
5 Outstanding issues and conclusions .................................................... 208
Chapter 6:
Numerical simulation of proton exchange membrane fuel cell 215
T.C. Jen, T.Z. Yan & Q.H. Chen
1 Introduction......................................................................................... 216
2 One-dimensional (1-D) model ............................................................ 216
2.1 General 1-D model ..................................................................... 217
2.1.1 Model description ........................................................... 217
2.1.2 Model assumptions .......................................................... 218
2.1.3 Governing equations........................................................ 218
2.1.4 Catalyst layers ................................................................. 219
2.1.5 Membrane........................................................................ 220
2.1.6 Boundary conditions........................................................ 221
2.1.7 Results and discussion ..................................................... 223
2.1.8 Summary.......................................................................... 223
2.2 General 2-D model ..................................................................... 224
2.2.1 Model description and assumptions................................. 227
2.2.2 Mathematical model ........................................................ 227
2.2.4 Numerical procedures ..................................................... 230
2.2.6 Concluding remarks......................................................... 232
2.3 Three-dimensional (3-D) model.................................................. 233
2.3.1 Model development ......................................................... 233
2.3.2 Mathematical model ........................................................ 234
2.3.4 Discretization strategies................................................... 238
2.3.5 Solution algorithms ......................................................... 238
2.4 Summary and conclusion ........................................................... 244
Chapter 7:
Mathematical modeling of fuel cells: from analysis to numerics 247
M. Vynnycky & E. Birgersson
1 Introduction ........................................................................................ 247
2 PEFC................................................................................................... 250
2.1 Mathematical formulation for flow in the cathode ...................... 251
2.1.1 Channel............................................................................ 251
2.1.2 Porous backing ................................................................ 253
2.2 Nondimensionalization ............................................................... 256
2.3 Parameters .................................................................................. 258
2.4 Narrow-gap approximation......................................................... 258
2.5 Further simplifications and observations .................................... 261
2.6 Numerics and results .................................................................. 264
2.6.1 Effect of A and Q ............................................................ 264
2.6.2 'Polarization surfaces' ...................................................... 268
3 DMFC ................................................................................................. 270
3.1 Mathematical formulation for flow in the anode......................... 271
3.1.1 Channel............................................................................ 271
3.1.2 Porous backing ................................................................ 271
3.2 Nondimensionalization ............................................................... 272
3.3 Parameters .................................................................................. 273
3.4 Narrow-gap approximation ......................................................... 273
3.5 Further simplifications and observations..................................... 273
3.6 Numerics and results................................................................... 277
4 Conclusions......................................................................................... 278
Chapter 8:
Modeling of PEM fuel cell stacks with hydraulic network approach 283
J.J. Baschuk & X. Li
1 Introduction......................................................................................... 284
2 Model formulation .............................................................................. 285
2.1 Stack flow model ........................................................................ 288
2.2 Manifold pressure loss ................................................................ 290
2.3 Cell pressure loss ........................................................................ 292
2.4 Mass consumed in the catalyst layers.......................................... 293
2.5 Boundary conditions ................................................................... 294
3 Numerical procedure .......................................................................... 294
3.1 Outer iteration............................................................................. 295
3.2 Inner iteration ............................................................................. 297
3.3 Numerical procedure summary .................................................. 298
4 Results and discussion......................................................................... 298
5 Conclusions......................................................................................... 309
Chapter 9:
Two-phase microfluidics, heat and mass transport in direct
methanol fuel cells 317
G. Lu & C.-Y. Wang
1 Introduction......................................................................................... 317
2 Fundamentals of DMFC...................................................................... 320
2.1 Cell components and polarization curve .................................... 320
2.2 Thermodynamics......................................................................... 322
2.3 Methanol oxidation and oxygen reduction kinetics..................... 324
3 Two-phase flow phenomena................................................................ 325
3.1 Bubble dynamics in anode ......................................................... 325
3.1.1 Flow visualization................................................................ 325
3.1.2 Bubble diameter and drift velocity....................................... 328
3.1.3 Pressure drop ....................................................................... 329
3.2 Liquid water transport in cathode............................................... 329
3.2.1 Flooding in the cathode........................................................ 329
3.2.2 Flooding visualization.......................................................... 331
4 Mass transport phenomena ................................................................. 333
4.1 Methanol crossover ........................................................................ 333
4.2 Water management in portable DMFC systems.............................. 335
5 Heat transport.......................................................................................... 337
6 Mathematical modeling and experimental diagnostics ............................ 339
6.1 Mathematical modeling .................................................................. 340
6.2 Experimental diagnostics ................................................................ 341
6.3 Model validation............................................................................. 343
7 Application: micro DMFC ...................................................................... 344
8 Summary and outlook ............................................................................. 350
This page intentionally left blank
Preface
Fuel cells are expected to play a significant role in the next generation of energy
systems and road vehicles for transportation. However, substantial progress is
required in reducing manufacturing costs and improving the performance. Solid
Oxide fuel cells (SOFC), Proton Exchange Membrane fuel cells (PEMFC) and
Direct Methanol fuel cells (DMFC) are of current interest.
Many of the associated heat and mass transport processes are not well
understood and include multidimensional flow and heat transfer in multi phase
flows, multicomponent transport of gaseous species in porous media and
electrochemical reactions including heat generation. Depending on the fuel being
used, modifications in the design of the next generation of fuel cells are needed.
Therefore, additional transport processes even at micro scale level need to be
investigated. Other important considerations are air management system for
oxygen supply, water management and recovery or rejection of heat of exhaust
products.
Currently an extensive amount of research and development activities are
carried out for fuel cells worldwide. The dissemination of results is through
various generic and specialized journals and conference proceedings. There is no
comprehensive book available to address the analysis of transport phenomena in
fuel cells.
This book aims to contribute to the understanding of the transport processes
in SOFC, PEMFC and DMFC fuel cells. The nine chapters cover a wide range of
topics and are invited contributions from some prominent scientists in the field.
The first chapter presents the thermodynamic and electrochemical
fundamentals of SOFC. Efficiency, energy distribution, chemical equilibrium,
losses of electrical potential, ohmic losses and losses due to mass transfer
resistance are discussed. Modeling and numerical simulations of the coupled
transport processes which determine the local and overall electromotive force in
a SOFC are provided. Chapter 2 discusses various numerical techniques to model
single-cells and stacks of planar SOFC. In chapter 3 a numerical model for
analysis of a tubular SOFC including indirect internal reforming is presented.
Fundamental results and strategies to reduce the maximum temperature and
temperature gradient of the cell are discussed. Chapter 4 provides numerical
analysis of heat, mass transfer (species flow), two-phase transport and effects on
the performance in SOFC and PEMFC. In chapter 5 information on transport
phenomena in the electrodes of PEMFC is provided. The physical characteristics
of such electrodes are also discussed. The focus is on two-phase flow in porous
media, with a discussion of driving forces and various flow regimes. Also, the
mathematical models are summarized. Chapter 6 presents mathematical models
and numerical simulations of PEMFC to evaluate effects of various designs and
operating parameters on the fuel cell performance. The three-dimensional model
can be used for optimisation of design and operation and serve as a building
block for modeling and understanding of PEMFC stacks and systems. In chapter
7 scaling analysis, nondimensionalization and asymptotic techniques are used to
identify the governing parameters in order to obtain a simplified model.
Illustrations are provided for PEMFC and DMFC. In chapter 8 a mathematical
model for a PEMFC stack is formulated. Distributions of pressure, fuel and
oxidant mass flow rates in the stack are determined by a hydraulic network
analysis. Finally chapter 9 provides an overview of the latest developments in the
DMFC technology. Experimental and modeling works to elucidate critical
transport phenomena, including two-phase mierofluidics, heat and mass transport
are presented.
All of the chapters follow a unified outline and presentation to aid
accessibility and the book provides invaluable information for both graduate
research and R & D engineers at industry and consultancy.
We are grateful to the authors and reviewers for their excellent contributions.
We also appreciate the cooperation and patience provided by the staff of WIT
Press and for their encouragement and assistance in producing this volume. The
editors would also like to thank the Wenner-Gren Center Foundation in Sweden
for financial support.
B. Sundén and M. Faghri

2005
CHAPTER 1
Multiple transport processes in solid

oxide fuel cells
P.-W. Li, L. Schaefer & M.K.. Chyu
Department of Mechanical Engineering, University of Pittsburgh, USA.
Abstract
In this topic, three important issues are discussed which concern the theoretical
fundamentals and practical operation of a solid oxide fuel cell. The thermodynamic
and electrochemical fundamentals of a fuel cell are reviewed in the Section 2. These
fundamentals concern the ideal efficiency and energy distribution of a fuel cell’s
conversion of chemical energy directly into electrical energy through the oxidation
of a fuel. Issues of the chemical equilibrium for a solid oxide fuel cell with internal
reforming and shift reactions (in case of methane or natural gas being used as the
fuel), are also discussed in detail in this section. The losses of electrical potential
in the practical operation of a fuel cell are elucidated in the third section, which
includes a discussion about activation polarization, Ohmic loss, and the losses due
to mass transport resistance. In the fourth section, the coupled processes of flow,
heat/mass transfer, chemical reaction, and electrochemistry, which influence the
performance of a fuel cell, are analyzed, and modeling and numerical computation
for the fields of flow, temperature, and species concentration, which collectively
determine the local and overall electromotive force in a solid oxide fuel cell, are
examined in detail.
1 Introduction
A fuel cell is a device that converts the chemical energy of a fuel oxidation reaction
directly into electricity. It is substantially different from a conventional thermal
power plant, where the fuel is oxidized in a combustion process and a thermal-
mechanical-electrical energy conversion process is employed. Therefore, unlike
heat engines that are subjected to the Carnot cycle efficiency limitation, fuel cells can
have energy conversion efficiencies generally higher than that of heat engines [1].
2 Transport Phenomena in Fuel Cells
Figure 1: Principle of operation of a SOFC.
Ideally, the Gibbs free energy change of fuel oxidation is directly converted into
electricity [1, 2] in a fuel cell.
As is common in many kinds of fuel cells, the core component of a solid oxide
fuel cell (SOFC) is a thin gas-tight ion conducting electrolyte layer sandwiched by
a porous anode and cathode, as shown in Fig. 1. For a SOFC, this electrolyte is a
solid oxide material that only allows the passage of charge-carrying oxide ions. To
produce useful electrical work, free electrons released in the oxidation of a fuel at
the anode must travel to the cathode through an external load/circuit. Therefore,
the electrolyte must conduct ions while preventing electrons released at the anode
from returning back to the cathode by the same route. The oxide ions are driven
across the electrolyte by the chemical potential difference on the two sides of the
electrolyte, which is due to the oxidation of fuel at the anode. This difference in the
chemical potential is proportional to the electromotive force across the electrolyte,
which, therefore, sets up a terminal voltage across the external load/circuit.
The solid oxide electrolyte has sufficient ion conductivity only at high tempera-
tures (from 600–1000 ◦ C). The high operating temperature of a SOFC also ensures
rapid fuel-side reaction kinetics without requiring an expensive catalyst. In addi-
tion, the high temperature exhaust from a SOFC can be directed to a gas turbine
(GT); thus, using a SOFC-GT hybrid system, one can achieve an efficiency of at
least 66.3% based on the lower heating value (LHV, which means that the electro-
chemical product, water, is in a gaseous state) of the SOFC [3–6]. Since it operates
via transport of oxide ions rather than that of fuel-derived ions, in principle, a SOFC
can be used to oxidize a number of gaseous fuels. In particular, a SOFC can consume
CO as well as hydrogen as its fuel, and therefore can be fueled with reformer gas
containing a mix of CO and H2 [7, 8]. Recently, ammonia has also been reported
as a fuel for SOFCs [9].
Since a SOFC operates under high temperatures, its energy conversion efficiency
and component safety are both of concern to industry. In the following sections,
the issues to be discussed will include: (1) the thermodynamic and electrochemi-
cal fundamentals of the energy conversion and species variation, (2) the potential
losses in practical operation, (3) the influence of fluid flow and heat and mass
transfer on operational efficiency and safety, and (4) the creation of a numerical
model to simulate the performance and the fields of flow, temperature, and species
concentration.
2 Thermodynamic and electrochemical fundamentals for

solid oxide fuel cells
To study the energy conversion efficiency and distribution of the conversion pro-
cesses in a fuel cell, one must understand the basic principles. The chemical potential
and, thereof, electromotive force across the electrolyte involve the interrelation of
thermodynamics, electrochemistry, ion/electron conduction, and heat/mass trans-
fer. In this section, the fundamentals of thermodynamics and the electrochemistry
for a solid oxide fuel cell system are reviewed.
The isothermal oxidation of a fuel A with oxidant B can be expressed by the
following equation:
aA + bB + · · · → xX + yY + · · ·. (1)
The systematic changes of enthalpy, Gibbs free energy, and entropy production in
the reaction are related by
H = T S + G. (2)
In a solid oxide fuel cell, the operating temperature is from 600 ◦ C to 1000 ◦ C and
the pressure of gases is relatively not high. Thus, the gas species of reactants and
products can be treated as ideal gases, which allows the chemical enthalpy change
to be expressed as:
H = (xhX + yhY + · · ·) − (ahA + bhB + · · ·), (3)
where the h is the specific enthalpy. When a gas is pure, ideal, and at 1 atm, it is said
to be in its standard state. The standard state is designated by writing a superscript 0
after the symbol of interest [10]. The Gibbs free energy which pertains to one
mole of a chemical species is called the chemical potential. For an ideal gas at
temperature of T and pressure of p, the chemical potential is expressed as:
p
g = g 0 + RT ln , (4)
p0
where R is the gas constant and p0 is the standard pressure of 1 atm. One may omit the
p0 in the denominator of the logarithm in eqn (4), but in such a case, the pressure
p must be measured in atm. The systematic change of the Gibbs free energy in
eqn (1) can be expressed in terms of the standard state Gibbs free energy and the
partial pressures of the reactants and products:
G = (xgX + ygY + · · ·) − (agA + bgB + · · ·)

(pX /p0 )x (pY /p0 )y · · ·
= [xgX0 + ygY0 + · · ·] − [agA0 + bgB0 + · · ·] + RT ln
(pA /p0 )a (pB /p0 )b · · ·
(pX /p0 )x (pY /p0 )y · · ·
= G 0 + RT ln , (5)
(pA /p0 )a (pB /p0 )b · · ·
where
G 0 = (xgX0 + ygY0 + · · ·) − (agA0 + bgB0 + · · ·), (6)
which is the Gibbs free energy change of the standard reaction at temperature T (i.e.,
with each reactant supplied and each product removed at the standard atmospheric
pressure, p0 = 1 atm).
The theoretical electromotive force (EMF) induced from the chemical potential
(G) is the Nernst potential:
−G −G0 RT (pA /p0 )a (pB /p0 )b · · ·

E= = + ln , (7)
ne F ne F ne F (pX /p0 )x (pY /p0 )y · · ·
where F(=96486.7 C/mol) is Faraday’s constant. The first part of the right-
hand side of the standard reaction is also called the ideal potential, which is
denoted by:
−G0
E0 = , (8)
ne F
where ne is the number of electrons derived from a molecules of the fuel, when
the fuel is oxidized in the reaction of eqn (1). While the Gibbs free energy change,
−G, converts to electrical power, the entropy production, −T S, is the thermal
energy that is released in the electrochemical oxidation of the fuel. Both the h and
g 0 are solely functions of temperature for ideal gases, which are given in Tables 1(a)
and 1(b) for the gas species involved in the reactions of a SOFC.
While the electromotive force in a fuel cell is determinable from the chemical
potentials as discussed above, the current to be withdrawn from a fuel cell, denoted
by I , is directly related to the molar consumption rate of fuel and oxidant through
the following expressions:
I I
mfuel = ; mO2 = , (9)
nfuel
e F nO 2
e F
where nO 2
e is for oxygen, and is the number of electrons per b molecules of oxygen
obtained in the electrochemical reaction (in eqn (1)), and nfuel
e is the number of
electrons derived per a molecules of the fuel.
2.1 Operation with hydrogen fuel
If a SOFC operates on hydrogen gas, the oxidation of hydrogen is the only electro-
chemical reaction in the fuel cell, which may be expressed by the following chemical
equation:
1
H2 + O2 = H2 O (gas). (10)
2
Table 1(a): Enthalpy and standard state Gibbs free energy of species.
Formula CO CO2 H2
weight 28.01 44.01 2.016
T h g◦ h g◦ h g◦
(K) (kJ/k mol) (kJ/k mol) (kJ/k mol) (kJ/k mol) (kJ/k mol) (kJ/k mol)
298.15 −110530 −169474 −393510 −457254 0 −38968

300 −110476 −169816 −393441 −457641 53 −39217
320 −109892 −173796 −392687 −461967 630 −41866
340 −109309 −177819 −391916 −466308 1209 −44521
360 −108725 −181877 −391128 −470688 1791 −47241
380 −108141 −185927 −390326 −475142 2373 −49953
400 −107555 −190035 −389507 −479627 2959 −52721
420 −106967 −194201 −388675 −484141 3544 −55508
440 −106377 −198337 −387827 −488719 4131 −58305
460 −105887 −202671 −386966 −493318 4715 −61157
480 −105195 −206763 −386094 −497982 5298 −64062
500 −104599 −210999 −385205 −502655 5882 −66968
550 −103102 −221737 −382938 −514498 6760 −74970
600 −101588 −232568 −380603 −526583 8811 −81849
650 −100053 −243573 −378207 −538822 10278 −89432
700 −98507 −254677 −375756 −551316 11749 −97171
750 −96938 −265838 −373250 −564800 13223 −104977
800 −95353 −277193 −370704 −576704 14702 −112898
850 −93749 −288569 −368112 −589622 16186 −121004
900 −92129 −300119 −365480 −602720 17676 −129114
950 −90499 −311659 −362821 −615996 19175 −137290
1000 −88840 −323340 −360113 −629413 20680 −145520
1100 −85495 −346965 −354626 −656576 23719 −162291
1200 −82100 −370940 −349037 −684317 26797 −179363
1300 −78662 −395082 −343362 −712432 29918 −196672
1400 −75187 −419587 −337614 −741094 33082 −214158
1500 −71680 −444280 −331805 −770105 36290 −231910
1600 −68145 −469265 −325941 −799541 39541 −249899
1700 −64585 −494515 −320030 −829350 42835 −268095
1800 −61004 −519824 −314079 −859479 46169 −286471
1900 −57404 −545324 −308091 −889871 49541 −305189
The Nernst potential from this electrochemical reaction will be:

−G(H RT
0
2 +1/2O 2 =H2 O)
E(H2 +1/2O2 =H2 O) = + ln( pH2 /p0 )anode
2F 2F

+ ln( pO2 /p0 )0.5
cathode − ln( pH2 O /p )anode .
0
(11)
The ideal chemical potentials at the temperature T (K) can be calculated from
the data given by handbooks [11]. As a convenient reference, Table 1(c) gives the
Table 1(b): Enthalpy and standard state Gibbs free energy of species.
Formula O2 H2 O (Gas) CH4

weight 31.999 18.015 16.043
T h g◦ h g◦ h g◦
(K) (kJ/k mol) (kJ/k mol) (kJ/k mol) (kJ/k mol) (kJ/k mol) (kJ/k mol)
298.15 0 −61151 −241814 −298105 – –

300 54 −61536 −241752 −298452 −74448 −130398
320 643 −65661 −241079 −302263 −73718 −134166
340 1234 −69826 −240404 −306126 −72974 −137948
360 1828 −74024 −239726 −309998 −72213 −141801
380 2425 −78325 −239045 −313943 −71432 −145684
400 3025 −82495 −238362 −317882 −70631 −149591
420 3629 −86797 −237675 −321885 −69808 −153598
440 4236 −91112 −236985 −325865 −68962 −157578
460 4847 −95433 −236291 −329947 −68094 −161658
480 5463 −99849 −235592 −334040 −67202 −165698
500 6084 −104266 −234889 −338139 −66287 −169837
550 7653 −115382 −233115 −348890 −63892 −180327
600 9244 −126656 −231313 −359173 −61356 −191016
650 10859 −138056 −229493 −369828 −58671 −201899
700 12499 −149551 −227622 −380712 −55853 −213073
750 14158 −161117 −225732 −391707 −52897 −224347
800 15835 −172885 −223812 −402852 −49818 −235898
850 17531 −184684 −221860 −414130 −46613 −247638
900 19241 −196669 −219876 −425526 −43296 −259566
950 20965 −208745 −217860 −436930 −39866 −271666
1000 22703 −220897 −215814 −448514 −36336 −283936
1100 26212 −245378 −211623 −471993 −28981 −309041
1200 29761 −270239 −207308 −495908 −21274 −334834
1300 33344 −295426 −202872 −520072 −13254 −361264
1400 36957 −320883 −198321 −544681 −4956 −388416
1500 40599 −346551 −193663 −569563 3587 −416113
1600 44266 −372374 −188906 −594826 12347 −444293
1700 47958 −398632 −184056 −620276 21295 −473235
1800 51673 −424967 −179121 −646221 30406 −502574
1900 55413 −451507 −174108 −672288 39658 −532432
ideal chemical potentials and enthalpies for the gas species that are typically utili-
zed in a SOFC. Recognizing the electrochemical equilibrium in the anode gas
mixture:

G = −G(H 0
2 +1/2O2 =H2 O)
+ RT ln( pH2 /p0 )anode + ln( pO2 /p0 )0.5
anode

− ln( pH2 O /p0 )anode = 0. (12)
Table 1(c): Change of enthalpy and standard state Gibbs free energy of reactions.
Reaction H2 + 1/2O2 = H2 O (gas) CH4 + H2 O = 3H2 + CO CO + H2 O = H2 + CO2
T H G 0 E0 H G 0 H G 0
(K) (kJ/mol) (kJ/mol) (V) (kJ/mol) (kJ/mol) (kJ/mol) (kJ/mol)
298.15 −241.814 −228.561 1.184 – – – –

300 −241.832 −228.467 1.184 205.883 141.383 −41.160 −28.590
320 −242.031 −227.567 1.179 206.795 137.035 −41.086 −27.774
340 −242.230 −226.692 1.175 207.696 132.692 −40.994 −26.884
360 −242.431 −225.745 1.170 208.587 128.199 −40.886 −26.054
380 −242.631 −224.828 1.165 209.455 123.841 −40.767 −25.225
400 −242.834 −223.914 1.160 210.315 119.275 −40.631 −24.431
420 −243.034 −222.979 1.155 211.148 114.758 −40.489 −23.563
440 −243.234 −222.004 1.150 211.963 110.191 −40.334 −22.822
460 −243.430 −221.074 1.146 212.643 105.463 −40.073 −21.857
480 −243.622 −220.054 1.140 213.493 100.789 −40.009 −21.241
500 −243.813 −219.038 1.135 214.223 96.073 −39.835 −20.485
550 −243.702 −216.229 1.121 214.185 82.570 −39.961 −18.841
600 −244.746 −213.996 1.109 217.514 72.074 −38.891 −16.691
650 −245.201 −211.368 1.095 218.945 59.858 −38.383 −14.853
700 −245.621 −208.766 1.082 220.215 47.595 −37.878 −13.098
750 −246.034 −206.172 1.068 221.360 35.285 −37.357 −12.232
800 −246.432 −203.512 1.055 222.383 22.863 −36.837 −9.557
850 −246.812 −200.784 1.040 223.282 10.187 −36.317 −7.927
900 −247.173 −198.078 1.026 224.071 −2.369 −35.799 −6.189
950 −247.518 −195.268 1.012 224.752 −14.934 −35.287 −4.697
1000 −247.846 −192.546 0.998 225.350 −27.450 −34.779 −3.079
1100 −248.448 −187.013 0.969 226.266 −52.804 −33.789 0.091
1200 −248.986 −181.426 0.940 226.873 −78.287 −32.832 3.168
1300 −249.462 −175.687 0.910 227.218 −103.762 −31.910 6.050
1400 −249.882 −170.082 0.881 227.336 −128.964 −31.024 9.016
1500 −250.253 −164.378 0.852 227.266 −154.334 −30.172 11.828
1600 −250.580 −158.740 0.823 227.037 −179.843 −29.349 14.651
1700 −250.870 −152.865 0.792 226.681 −205.289 −28.554 17.346
1800 −251.127 −147.267 0.763 226.218 −230.442 −27.785 20.095
1900 −251.356 −141.346 0.732 225.669 −256.171 −27.038 22.552
Substituting eqn (12) into eqn (11), the Nernst potential in another form for the
electrochemical reaction of eqn (10) is obtained:
RT
E(H2 +1/2O2 =H2 O) = cathode − ln( pO2 /p )anode .
ln( pO2 /p0 )0.5 0 0.5
(13)
2F
Since the oxygen partial pressure at the anode is very low (on the order of 10−22
bar) due to the anode reaction [2], it does not cause an appreciable effect on the
partial pressures of the other major species in the anode flow. Therefore, when
calculating the partial pressures of hydrogen and water vapor for determining the
Nernst potential of the electrochemical reaction of eqn (10), the oxygen partial pres-
sure in anode flow stream is ignorable. Hereafter, for an electrochemical reaction
as in eqn (1), the common practice in determining the Nernst potential will be to
use eqn (7), in which the partial pressure of oxygen on the cathode side and those
of the fuel and product species on the anode side are used.
The molar consumption rate of hydrogen and oxygen in the electrochemical
reaction of eqn (10) can be easily derived from eqn (9) as:
I I
mH2 = ; mO2 = . (14)
2F 4F
2.2 Operation with methane through internal reforming and shift reactions
As previously mentioned, it is necessary to have a high operating temperature in a

solid oxide fuel cell in order to maintain sufficient ionic conductivity for the solid
oxide electrolyte [2, 4]. This provides a favorable environment for the reforming of
hydrocarbon fuels like methane. In fact, since a solid oxide fuel cell operates based
on the transport of oxide ions through the electrolyte layer from the cathode side
to the anode side, the reforming products of hydrogen and carbon monoxide in the
fuel channel can both serve as fuels. Given this advantage, solid oxide fuel cells can
directly utilize hydrocarbon fuels or, at least, methane as a pre-reformed or partly
reformed gas with components of CH4 , CO, CO2 , H2 and H2 O. Therefore, the fuel
reforming and shift reactions will occur in the fuel channel in a solid oxide fuel
cell. The anode is, in fact, a good material to serve as the catalyst for such chemical
reactions, since the high temperature in a SOFC means that no noble metals are
needed for a catalyst [12].
If there are five gas species, CH4 , CO, CO2 , H2 , and H2 O, in the fuel channel,
the solid oxide fuel cell will operate with internal reforming and shift reactions.
Therefore, the electrochemical reaction and the coexisting chemical reactions of
reforming and shift need to be considered for determining the species’mole fractions
(which are crucial to the electromotive forces in the fuel cell).
Reforming : CH 4 + H2 O ↔ CO + 3H2 . (15)
Shift : CO + H2 O ↔ CO2 + H2 . (16)

Since the high operating temperature of a SOFC ensures rapid fuel reaction kinetics,
it is a common practice to assume that the reforming and shift reactions are in
chemical equilibrium [4] when determining the mole fractions of the species, which
makes the computation significantly convenient. From the concept of chemical
equilibrium, the reactants and products must satisfy the condition of G = 0.
Therefore, the mole fractions or partial pressures of the five gas species in the fuel
stream are related through the following two simultaneous equations [13]:
p 3
H2 pCO 0
p0 p0 Greforming
KPR = p p = exp − , (17)
CH4 H2 O RT
p0 p0
p p
CO2 H2
0
Gshift
p0 p0
KPS = p = exp − . (18)
pCO H2 O RT
p0 p0
The dominant electrochemical reaction has been reported to be the oxidation

of H2 [12], which is primarily responsible for the electromotive force. However,
at the same time, the electrochemical oxidation of the CO is also possible, and
likely occurs to some extent in the solid oxide fuel cell. It has been reported that
fuel cells operated by using mixtures of CO and CO2 have shown that the electro-
chemical oxidation of CO is an order of magnitude slower than that of hydrogen
[14]. Nevertheless, there is no necessity to distinguish whether the electrochemical
oxidation process involves H2 or CO in order to formulate the electromotive force.
The following discussion will clarify this point.
When the shift reaction of eqn (16) in the anodic gas is in chemical equilibrium,
there is

pCO2 pH 2
G = gCO2 + RT ln
0
+ gH2 + RT ln
0
p0 p0

pCO pH2 O
− gCO0
+ RT ln + g 0
H2 O + RT ln = 0. (19)
p0 p0
Rearranging this equation gives:

pCO2 pCO
0
gCO + RT ln − g 0
CO + RT ln
2
p0 p0

pH2 O pH2
= gH2 O + RT ln
0
− gH2 + RT ln
0
. (20)
p0 p0
1/2
Subtracting a term of [(1/2)gO0 2 +RT ln( pO2 /p0 )cathode ] from both sides of eqn (20),
results in:

1 0 pCO2 pCO pO2 1/2
gCO2 − gCO − gO2 + RT ln
0 0
− RT ln − RT ln
2 p0 p0 p0 cathode

1 pH 2 O
= gH0 2 O − gH0 2 − gO0 2 + RT ln
2 p0

pH2 pO2 1/2
− RT ln − RT ln . (21)
p0 p0 cathode
It is easy to see that the left-hand side of eqn (17) is the Gibbs free energy change
of the electrochemical oxidation of CO, and the right-hand side is that for H2 .
Dividing by (2F) on both sides, eqn (17) is further reduced to:
E(H2 +1/2O2 =H2 O) = E(CO+1/2O2 =CO2 ) , (22)

where E(H2 +1/2O2 =H2 O) is given in eqn (11), while the Nernst potential for the
electrochemical oxidation of CO is
−G(CO+1/2O RT
0
2 =CO2 )
E(CO+1/2O2 =CO2 ) = +
ln (pCO /p0 )anode
2F 2F

+ ln (pO2 /p0 )0.5 0
cathode − ln (pCO2 /p )anode . (23)
It is preferable that the EMF of an internal reforming SOFC be calculated from the
electrochemical oxidation of H2 ; however, the species’consumption and production
are the results collectively determined from the reactions of eqns (10), (15) and (16).
The above discussion clearly indicates that the electrochemical reaction can be
assumed to be driven by the hydrogen, and the electrochemical fuel value of CO
is readily exchanged for hydrogen by the shift reaction under chemical equilib-
rium. Therefore, only H2 is considered as the electrochemical fuel in the following
analysis, and CO only takes part in the shift reaction.
For convenience, the mole flow rates of CH4 , CO and H2 are denoted by their
formulae. Assuming that, x̄, ȳ, and z̄ are the mole flow rates, respectively, for the
CH4 , CO, and H2 that are consumed in the three reactions given by eqns (15), (16)
and (10) in the fuel channel, the coupled variations of the five species between the
inlet and the outlet of an interested section of fuel channel are in the following
forms [8, 15]:
CH4 out = CH4 in − x̄, (24)

COout = COin + x̄ − ȳ, (25)
CO2 out
= CO2 + ȳ, in
(26)
H2 out = H2 in + 3x̄ + ȳ − z̄, (27)
H2 Oout = H2 Oin − x̄ − ȳ + z̄. (28)
The overall mole flow rate of fuel, denoted by Mf , will vary from the inlet to the
outlet of the section of interest in the fuel channel in the form of
Mfout = Mfin + 2x̄. (29)
Meanwhile, the partial pressures of the species, proportional to the mole fractions,
must satisfy eqns (15) and (16) at the outlet of the section, which thus gives:
3
COin +x̄−ȳ H2 in +3x̄+ȳ−z̄ p2
Mfin +2x̄ Mfin +2x̄ p0
KPR = , (30)
CH4 in −x̄ H2 Oin −x̄−ȳ+z̄
Mfin +2x̄ Mfin +2x̄

H2 in +3x̄+ȳ−z̄ CO2 in +ȳ
KPS = , (31)
COin +x̄−ȳ H2 Oin −x̄−ȳ+z̄
where the p is the overall pressure of the fuel flow in the section of interest.
Since, as discussed in the preceding section, the oxidation of H2 is responsible
for the electrochemical reaction, the consumption of hydrogen is directly related to
the charge transfer rate, or current, I , across the electrolyte layer:
z̄ = I /(2F). (32)
From the electrochemical reaction, the molar consumption of oxygen on the

cathode side can be calculated by using eqn (14). By finding a simultaneous solution
for eqns (30)–(32), the species variations, x̄, ȳ and z̄, can be determined. Finally,
with the reacted mole numbers of CH4 and CO determined, the heat absorbed in
the reforming reaction and released from the shift reaction can be obtained:
QReforming = H Reforming · x̄, (33)

Q Shift = H Shift · ȳ. (34)
Nevertheless, prior to finding a solution for eqns (30) and (31), the electric current
of the fuel cell in eqn (32) must be known. This demonstrates that the processes in a
SOFC feature a strong coupling of the species molar variation and the electromotive
force, as well as interdependency of the ion conduction and current flow. The ion
transfer rate or current conduction in a SOFC will be discussed in Section 3.
3 Electrical potential losses
The ideal efficiency is never attained in practical operation for any fuel cell. In fact,
there are three potential drops in a fuel cell that cause the actual output potential to
be lower than the ideal electromotive forces of the electrochemical reaction. The
nature of the fuel cell performance in response to loading condition can be realized
by its polarization curve, typically shown as in Fig. 2.
With an increase in current density, the cell potential experiences three kinds
of potential losses due to different dominant resistances. The potential drop due
to the activation resistance, which is the activation polarization, is associated with
the electrochemical reactions in the system. Another potential drop comes from
the ohmic resistance in the fuel cell components, when the ions and electrons are
conducted in the electrolyte and electrodes, respectively. The third drop, which can
be sharp at high current densities, is attributable to the mass transport resistance,
or concentration polarization, in the flow of the fuel and oxidant. It is known from
observing the Nernst equation that the electromotive force of a fuel cell is a function
of the temperature and the gas species’ partial pressures at the electrolyte/electrode
Figure 2: Over-potential in the operation of a fuel cell.
interfaces, which are directly proportional to their mole fractions. It is important

to note that, in the fuel stream, fuel must be transported or diffused from the core
region of the stream to the anode surface, and, also, the product of the electrochem-
ical reaction must conversely be transported or diffused from the reaction site to
the core region of the fuel flow. On the cathode side, oxygen must be transported
and diffused from the core region of airflow to the cathode surface. Along with
the fuel and air streams, the consumption of reactants or development of prod-
ucts will make the mole fractions of reactants decrease and those of the product
increase. Due to these resistances in the mass transport process, the feeding of reac-
tants and removing of products to/from the reaction site can only proceed under a
large concentration gradient between the bulk flow and the electrode surface when
the current density is high, which therefore induces a sharp drop in the fuel cell
potential.
As a consequence of all the above-mentioned potential drops, extra thermal
energy will be released together with the heat (−T S) from systematic entropy
production. The heat transfer issues in a solid oxide fuel cell will be considered
later in Section 4.
3.1 Activation polarization
The activation polarization is the electronic barrier that must be overcome prior to
current and ion flow in the fuel cell. Chemical reactions, including electrochemical
reactions, also involve energy barriers, which must be overcome by the reacting
species. The activation polarization may also be viewed as the extra potential nec-
essary to overcome the energy barrier of the rate-determining step of the reaction
to a value such that the electrode reaction proceeds at a desired rate [16–18].
The Butler-Volmer equation is a well-known expression for the activation polar-

ization, ηAct :

ne FηAct ne FηAct
i = i0 exp β − exp −(1 − β) , (35)
RT RT
where β, which is usually 0.5 for the fuel cell application [16], is the transfer
coefficient; i is the actual current density in the fuel cell; and i0 is the exchange
current density. The transfer coefficient is considered to be the fraction of the
change in polarization that leads to a change in the reaction-rate constant. The
exchange current density, i0 , is the forward and reverse electrode reaction rate
at the equilibrium potential. A high exchange current density means that a high
electrochemical reaction rate and good fuel cell performance can be expected. The
ne in eqn (35) is the number of electrons transferred per reaction, which is 2 for
the reaction of eqn (10). Substituting the value of β = 0.5 into eqn (35), one can
obtain a new expression as follows:

ne FηAct
i = 2i0 sinh (36)
2RT
from which the activation polarization can be expressed as:

 
2
2RT i 2RT i i
ηAct = sinh−1 or ηAct = ln  + + 1.
ne F 2i0 ne F 2i0 2i0
(37)
For a high activation polarization, eqn (37) can be approximated as the simple and
well-known Tafel equation [16]:

2RT i
ηAct = ln . (38)
ne F i0
On the other hand, if the activation polarization is small, eqn (37) can be approxi-
mated as the linear current-potential expression [16]:
2RT i
ηAct = . (39)
ne F i0
Nevertheless, eqn (37) is recommended for its integrity and accuracy in calculating
the activation polarization.
The value of the exchange current density (i0 ) is different for the anode and
cathode, and is also dependent on the electrochemical reaction temperature, the
partial pressures of the gases [18, 19], and the electrode materials. The determina-
tion for i0 shows diversity in different literature [16–22]. There are formulations
available in the literature [18–20], but some parameters used in the formulation
are not well documented. On the other hand, empirical estimation of i0 is also
made in literature, like the work of Chan et al. [16]. An i0 of 5300 A/m2 for the
anode and 2000 A/m2 for the cathode for a SOFC were used; however there was
no comment on the selection methodology for setting these values in the paper
[16]. Keegan et al. [17] also adjusted the i0 so as to obtain a simulation result to
satisfy their experimental data; no report, however, is given about the adjusted i0
values in their paper. The present authors used a slightly higher value of 6300 A/m2
and 3000 A/m2 [23, 24], respectively, for the i0 of the anode and cathode, which
resulted in very good agreement between the numerical simulated cell terminal
voltage and experimental results from different researchers [25–28]. Nevertheless,
i0 varies according to the temperature and pressures of the electrochemical reac-
tion. For SOFCs working at temperature from 800 ◦ C–1000 ◦ C and pressures up
to 15 atm, an i0 of 5300–6300 A/m2 for the anode and 2000–3000 A/m2 for the
cathode are recommended from the study by the present authors [23].
3.2 Ohmic loss
The ohmic loss comes from the electric resistances of the electrodes and the current
collecting components, as well as the ionic conduction resistance of the electrolyte
layer. Therefore, the conductivity of the materials for the cell components and the
current collecting pathway are the two factors most influential to the overall ohmic
loss of a SOFC.
In state-of-the-art SOFC technology, lanthanum manganite suitably doped with
alkaline and rare earth elements is used for the cathode (air electrode) [20, 27], yttria
stabilized zirconia (YSZ) has been most successfully employed for electrolyte,
and nickel/YSZ is applied over the electrolyte to form the anode. Temperature
could significantly affect the conductivity of SOFC materials. Especially for the
electrolyte, for example, the resistivity could be two orders of magnitude smaller if
its temperature increases from 600 ◦ C to 1000 ◦ C. The equations of resistivity for
SOFC components suggested in literature are collected in Table 2.
Table 2: Data and equations for resistivity of SOFC components.

Cathode Electrolyte Anode Interconnect
( · cm) ( · cm) ( · cm) ( · cm)
Bessette 0.008114e500/T 0.00294e10350/T 0.00298e−1392/T –

et al. [29]
Ahmed ∗ 0.0014 0.3685 + 0.002838e10300/T ∗ 0.0186 ∗ 0.5
et al. [30]
Nagata ∗ 0.1 10.0e[10092(1/T −1/1273)] ∗ 0.013 ∗ 0.5
et al. [18]
Ferguson T e1200/T 1 e10300/T T e1150/T T e1100/T
4.2×105 3.34×102 9.5×105 9.3×104
et al. [31]
∗At temperature of 1000 ◦ C.

A careful check for the equations in Table 2 was conducted. The expressions by
Bessette et al. [29] were found to be reliable, and to give nearly identical predictions
as those by Ahmed et al. [30] and Nagata et al. [18]. The predicted data for anode
resistivity by the equation of Ferguson et al. [31] shows significant discrepancies
with the predictions by other equations.
It is rational to assume that the passage of the charge-carrying species through
the electrolyte, or the ion conduction through the electrolyte, is a charge transfer,
like a current flow. In a planar type SOFC, as shown in Fig. 3, the current collects
through the channel walls, also called ribs, after it moves perpendicularly across
the electrolyte layer. The network circuit for current flow modeled by Iwata et al.
[19] considers the channel walls as current collection pathways in a planar SOFC.
However, the height and the width of the gas channel are both small (less than
3 mm), and the electric resistance through the channel wall might be negligible [30].
This simplification leads to the consideration that the current is almost exclusively
perpendicularly collected, which means that the current flows normally to the tri-
layer of the cathode, electrolyte and anode. When calculating the local current
density, the ohmic loss is thus simply accounted for in the following way [30]:
(E − ηaAct − ηcAct ) − Vcell
I = A · , (40)
(δa ρea + δe ρee + δc ρec )
where A is a unit area on the anode/electrolyte/cathode tri-layer, through which
the current I passes; δ is the thickness of the individual layers; ρe is the resistivity of
the electrodes and electrolyte; Vcell is the cell terminal voltage; and the denominator
of the right-hand side term is the summation of the resistance of the tri-layer. The
Joule heating due to current flow in the volume of A×δ is expressed for the anode
in the form of
a
QOhmic = I 2 · (δa ρea /A). (41)
This is also applicable to the electrolyte and cathode by replacing the thickness and
resistivity accordingly.
Figure 3: Schematic of a planar type SOFC.

Figure 4: Schematic of a tubular SOFC.
Figure 5: Ion/electron conduction network in a tubular SOFC.
In case the current pathway is relatively long in a fuel cell, as, for example, in a
tubular type SOFC (shown in Fig. 4), the current collects circumferentially, which
leads to a much longer pathway [32] compared to that of a planar type SOFC.
In order to account for the ohmic loss and the Joule heating of the current flow
in the circumferential pathway, a network circuit [23, 25, 33, 34] for current flow
may be adopted, as shown in Fig. 5. Because the current collection is symmetric
in the peripheral direction in the cell components, only half of the tube shell is
deployed with a mesh in the analysis. The local current routing from the anode to
cathode through the electrolyte is determinable based on the local electromotive
force, EMF, the local potentials in the anode and cathode, and the ionic resistance
of the electrolyte layer, which yields the expression:
E − ηaAct − ηcAct − (V c − V a )
I= , (42)
Re
where V a and V c are the potentials in the anode and cathode, respectively. Re
is the ionic resistance of the electrolyte layer given a thickness of δe and a unit
area of A:
Re = ρee · δe /A, (43)
ρee
where the is the ionic resistivity of the electrolyte.
In order to obtain the local current across the electrolyte by using eqn (35),
supplemental equations for V a and V c are necessary. Applying Kirchhoff’s law of
current to any grid located in the anode, the equation associating the potential of
the central grid point P with the potentials of its neighboring points (east, west,
north, south) and the corresponding grid P in the cathode can be obtained:
a
VEa − VPa VWa − VPa VN − VPa VSa − VPa
+ + +
Rae Raw Ran Ras

VPc − VPa − (EP − ηPAct )
+ = 0. (44)
ReP
In the same way for a grid point P in the cathode:

c
VEc − VPc VWc − VPc VN − VPc VSc − VPc
+ + +
Rce Rcw Rcn Rcs

VPa − VPc + (EP − ηPAct )
+ = 0, (45)
ReP
where Ra and Rc are the discretized resistances in the anode and cathode respec-
tively, which are determined according to the resistivity, the length of the current
path, and the area upon which the current acts; ηPAct is the total activation polariza-
tion, including from both the anode side and the cathode side.
With all of the equations for the discretized grids in both the cathode and anode
given, a matrix representing the pair of eqns (37) and (38) can be created. When
finding a solution for such a matrix equation for the potentials, the following approx-
imations are useful:
1. At the two ends of the cell tube there is no longitudinal current flow, and,
therefore, an insulation condition is applicable.
2. At the symmetric plane A–A, as shown in Figs 4 and 5, there is no peripheral
current in the cathode and anode, unless the cathode or anode is in contact with
nickel felt, through which the current flows in or out.
3. The potentials of the nickel felts are assumed to be uniform due to their high
electric conductivities.
4. Since the potential difference between the two nickel felts is the cell terminal
voltage, the potential at the nickel felt in contact with the anode layer can be
assumed to be zero. Thus, the potential at the nickel felt in contact with the
cathode will be the terminal voltage of the fuel cell.
Once all the local electromotive forces are obtained, the only unknown condition
for the equation matrix is either the total current flowing out from the cell or
the potential at the nickel felt in contact with the cathode. This highlights two
approaches that can be taken when predicting the performance of a SOFC. If the
total current taken out from the cell is prescribed as the initial condition, the terminal
voltage can be predicted. On the other hand, one can prescribe the terminal voltage
and predict the total current, i.e., the summation of the local current I across the
entire electrolyte layer.
Once the potentials are obtained in the electrode layer, the volumetric Joule
heating in the electrode for a volume centered about P will be:

1 (VEa − VPa )2 (VWa − VPa )2 (VNa − VPa )2
q̇P =
a
+ +
2 Rae Raw Ran

(V a − V a )2
+ S a P (xP · r a · θP · δa ), (46)
Rs

1 (VEc − VPc )2 (VWc − VPc )2 (VNc − VPc )2
q̇P =
c
+ +
2 Rce Rcw Rcn

(VSc − VPc )2
+ (xP · r c · θP · δc ), (47)
Rcs

(E − η P − V c + V a )2
P P P
q̇Pe = Act
(xP · r e · θP · δe ), (48)
ReP
where the r and δ with the corresponding superscripts of a , c , and e are the average,
radius and thickness, respectively, for the anode, cathode and electrolyte, and xP
and θP are the P-controlled mesh size in the axial and peripheral directions, as
shown in Fig. 5. The volumetric heat induced from the activation polarization in
the anode and cathode is:
P,a
q̇Act = IP · ηP,a
Act /(xP · r · θP · δ ),
a a
(49)
P,c
q̇Act = IP · ηP,c
Act /(xP · r · θP · δ ).
c c
(50)
The thermodynamic heat generation occurring at the anode/electrolyte interface in
the area around P is:
QPR = (H − G) · IP /(2F). (51)
3.3 Mass transport and concentration polarization
Due to their gradual consumption, the fractions of the reactants and oxidant will
decrease, in the fuel and air streams, respectively, which will cause the electromotive
force to decrease gradually along the flow stream. On the other hand, due to the mass
transport resistance, the concentration of the gas species will encounter polarization
in between the core flow region and the electrode surface, which will result in lower
partial pressures for the reactants, but higher partial pressures for the products at the
Figure 6: The interrelation amongst concentration and other parameters.
electrode surfaces. Therefore, the fuel cell terminal voltage will be lower than the
ideal value that is indicated by the Nernst equation. At high cell current density, the
increased requirements for the feeding of the reactants and removal of the products
can make the concentration polarization higher, and, thus, the cell output potential
will sharply decrease.
In order to take the concentration polarization into account when calculating the
electromotive force, the local partial pressures of the reactants and products at the
electrode surface are used. However, this requires the solution of the concentration
fields for the gas species in the fuel and oxidizer channels, which might be either
simply based on a one-dimensional [35–37] or else based on a complicated two-
or three-dimensional solution for the mass conservation governing equations [23,
38–40]. In fact, the concentration fields are strongly coupled with the gas flow,
temperature, and the distribution of the electromotive force in the ways indicated
in Fig. 6. First, the gas species mass fraction determines the gas properties in the
flow field, while the flow fields affect the gas species concentration distribution
and temperature. Second, the gas species concentration field and temperature dis-
tribution determines the electromotive forces, while the ion/electron conduction
due to the electromotive force determines the mass variation and heat generations
in the fuel cell. The inter-dependency of these parameters will be discussed in
detail in the following section when modeling a SOFC in order to predict both the
fuel cell performance and the detailed distributions of the temperature, gas species
concentration, and flow fields.
4 Computer modeling of a tubular SOFC

An operation curve for a SOFC that characterizes the average current density versus
the terminal voltage is very important when designing a SOFC system or a hybrid
SOFC/GT system [41–44]. Other information, like the temperature and concen-
tration fields in a SOFC, is also of high concern for the safe operation of both the
SOFC itself and the downstream facilities if a hybrid system is under consideration.
Although there have been some experimental data generated about the operational
performance and temperature of SOFCs [25–28], rigorous experimental testing for
a SOFC is still rather tough because of its high operating temperature. Therefore,
numerical modeling of SOFCs is very necessary.
The purpose of computer simulation for a SOFC is to predict the operational
characteristics in terms of the average current density versus the terminal voltage
(based on prescribed operating conditions). The operating conditions of a SOFC
are solely determined by fixing the flow rates and the thermodynamic state of
the fuel and oxidant, as well as a load condition such as terminal voltage, the
current being withdrawn, and external load [45]. The flow rates and thermodynamic
conditions of the fuel and oxidant may be called internal conditions, and the terminal
voltage, current to be withdrawn, and external load may be designated external
conditions. Like any kind of “battery,” the external load condition of a SOFC
determines the consumption of the fuel/oxidant and the generation of products in
the electrochemical reaction [46]; the only difference in a fuel cell is its continuous
feeding of fuel/oxidant and removal of products and waste species.
According to the different ways of prescribing the external parameters, the fol-
lowing three schemes might be designed in order to predict the other unknown
parameters when constructing a numerical model for a SOFC: (1) Use the internal
conditions and terminal voltage to predict the total current to be withdrawn. (2) Use
the internal conditions and current to be withdrawn to predict the terminal voltage.
(3) Use the internal conditions and external load to predict the terminal voltage and
current density.
The cost of iterative computation using the three schemes is quite different. In
the first scheme, the cell terminal voltage is known, and thus the local current can
be obtained, for example, by using eqn (42) and solving eqns (44) and (45), for
a planar and tubular type SOFC, respectively, once the temperature and partial
pressure fields of the gas species are available. The integrated value from the local
current will be the total current to be withdrawn from the SOFC. In the second
scheme, however, the terminal voltage needs to be assumed, and then checked by
integrating the total current from the local current until the calculated total current
agrees with the prescribed value. In this computation process, a proper method is
needed to find the best-fit terminal voltage iteratively. The third scheme resembles
the second scheme, in that one needs to assume a terminal voltage to find the total
current. The computation will be stopped only when the voltage-current ratio equals
the prescribed load.
With an understanding of the principles of the energy conversion, chemical equi-
librium, potential loss, and the operation of a SOFC, a computer model for a SOFC
can now be constructed. Generally speaking, the modeling and computation for
a tubular SOFC and a planar SOFC share rather common features except for the
Ohmic losses and Joule heating, for which differences result from the different
structures variation in the current pathway in the electrodes. Relatively speaking,
the tubular SOFC has a more complex current pathway in the electrodes [27] and
will be discussed in the following analysis. Modeling works on planar type SOFCs
are available in both the current author’s work and in the literature [17, 19, 30, 31,
39, 47–49].
In the following subsections, there are three issues that address the construction
of a numerical model.
4.1 Outline of a computation domain
In a practical tubular SOFC stack, multiple tubular cells are mounted in a container
to form a cell bundle, as shown in Fig. 7. A pre-reformer might be put adjacent to
the cell bundles [50, 51]. In order to conduct a modeling study with relatively less
complexity, it is assumed that most of the single tubular SOFCs operate under the
same environment of temperature and concentrations of gas species. This allows
the definition of a controllable domain in the cross-section, which pertains to one
single cell, as outlined by the dashed-line square in Fig. 7. It is then specified that
there must be no flow velocity and fluxes of heat and mass across the outline.
This will significantly simplify the analysis for a cell stack. Through analysis of
the heat/mass transfer and the chemical/electrochemical performance for the single
cell and its controllable area, one can obtain results very useful for evaluating the
performance of an entire cell stack.
Also considering the longitudinal direction, the heat and mass transfer in the
above outlined square area enclosing the tubular SOFC are three-dimensional in
nature. For a solution of the three-dimensional governing equations of momentum,
energy, and species conservation, a large number of discretized mesh points are
necessary, which results in an unacceptably heavy computational load. In order to
reduce computational cost, the square area enclosing the tubular SOFC is approxi-
mated to be an equivalent circular area; therefore, the domain enclosing the single
tubular SOFC is viewed as a 2-dimensional axi-symmetric one, as seen in Fig. 7.
Figure 7: Schematic of a tubular SOFC in a cell stack.

Figure 8: Computation domain for a tubular SOFC.
It should be noted, though, that the zero-flux, or insulation of heat and mass transfer
at the boundary remains unchanged, even given this geometric approximation.
From the preceding discussions, an axi-symmetrical two-dimensional (x − r)
computation domain is profiled as shown in Fig. 8, which includes two flow streams
and a solid area of the cell tube and air-inducing tube.
4.2 Governing equations and boundary conditions
Since the mass fractions of the species vary in the flow field, all of the thermal
and transport properties of the fluids are functions of the local species concentra-
tion, temperature, and pressure; therefore, the governing equations for momentum,
energy, and species conservation (based on mass fraction) have variable thermal
and transport properties:
∂(ρu) 1 ∂(rρv)
+ = 0, (52)
∂x r ∂r

∂(ρuu) 1 ∂(rρvu) ∂p ∂ ∂u 1 ∂ ∂u
+ =− + µ + rµ
∂x r ∂r ∂x ∂x ∂x r ∂r ∂r

∂ ∂u 1 ∂ ∂v
+ µ + rµ , (53)
∂x ∂x r ∂r ∂x

∂(ρuv) 1 ∂(rρvv) ∂p ∂ ∂v 1 ∂ ∂v
+ =− + µ + rµ
∂x r ∂r ∂r ∂x ∂x r ∂r ∂r

∂ ∂u 1 ∂ ∂v 2µv
+ µ + rµ − 2 , (54)
∂x ∂r r ∂r ∂r r

∂(ρCpuT ) 1 ∂(rρCpvT ) ∂ ∂T 1 ∂ ∂T
+ = λ + rλ + q̇, (55)
∂x r ∂r ∂x ∂x r ∂r ∂r

∂(ρuYJ ) 1 ∂(rρvYJ ) ∂ ∂YJ 1 ∂ ∂YJ
+ = ρDJ ,m + rρDJ ,m + SJ . (56)
∂x r ∂r ∂x ∂x r ∂r ∂r
These equations are applied universally to the entire computation domain; how-
ever, zero velocities will be assigned to the solid area in the numerical computation.
In the energy conservation equation, thermal energy from the chemical and elec-
trochemical reactions (expressed by eqns (33), (34), (51)) and the Joule heating
in electrodes and electrolyte (expressed by eqns (46)–(50)), represented by q̇, are
introduced as source terms in the proper locations in the fuel cell. Some terms due
to energy diffusion driven by the concentration diffusion of the gas species are very
small, and thus neglected [52, 53]. The boundary conditions for the momentum,
heat and mass conservation equations are as follows:
1. On the symmetrical axis, or at r = 0: v = 0, and ∂φ/∂r = 0, where φ represents
general variables except for v.
2. At the outmost boundary of r = rfo : there are thermally adiabatic conditions;
impermeability for species and non-chemical reaction are also assumed, which
gives v = 0, and ∂φ/∂r = 0, where φ represents general variables except for v.
3. At x = 0: the fuel inlet has a prescribed uniform velocity, temperature, and
species mass fraction; the solid part has u = 0, v = 0, ∂T /∂x = 0, and
∂YJ /∂x = 0.
4. At x = L: the air inlet has a prescribed uniform velocity, temperature and species
mass fraction; the gas exit has v = 0, ∂u/∂x = 0, ∂T /∂x = 0, and ∂YJ /∂x = 0;
the tube-end solid part has u = 0, v = 0, ∂T /∂x = 0, and ∂YJ /∂x = 0.
5. At the interfaces of the air/solid, r = rair , and fuel/anode, r = rf : u = 0 is
assumed.
In the fuel flow passage, the mass flow rate increases along the x direction due
to the transferring in of oxide ions. Similarly, a reduction of the air flow rate occurs
in the air flow passage, due to the ionization of oxygen and the transferring of the
oxide ions to the fuel side. Therefore, radial velocities at r = rair and r = rf are:
fuel,species
ṁx
vf = fuel
r=rf , (57)
ρx
air,species
ṁx
vair = r=r , (58)
air air
ρx
where ṁ [kg/(m2 s)] is mass flux of the gas species at the interface of the electrodes
and fluid, which arises from the electrochemical reaction in the fuel cell. The mass
fractions of all participating chemical components at the boundaries of r = rair and
r = rf are calculated with consideration of both diffusion and convection effects
[54, 55]:
∂YJ
ṁJx ,air = −DJ ,air ρxair + ρxair YJ vair , (59)
∂r
∂YJ
ṁJx , fuel = −DJ , fuel ρxfuel + ρxfuel YJ vf . (60)
∂r
Table 3: Properties of SOFC materials.
Thermal conductivity Cp Density

(W/(m K)) (J/(kg K)) (kg/m3 )
Cathode d 11; c 2.0; b 2.0 b 623 a 4930
Electrolyte d 2.7; c 2.7; b 2.0 b 623 a 5710
Anode c 11.0; d 6.0; b 2.0 b 623 a 4460
Support tube c 1.0
Air-inducing tube c 1.0
Interconnector b 13; c 2.0; d 6.0 b 800 a 6320; b 7700
aAhmed et al. [30]; b Recknagle et al. [39]; c Nagata et al. [18]; d Iwata et al. [19].
It is worth noting that the mass fluxes for the species in the above equations, eqns
(57)–(60), strongly relate to the ion/electron conduction; the determination of mass
variation and related mass flux that arise from the electrochemical reaction has
been discussed (as expressed by eqns (30)–(32)) in Section 2. As a consequence,
the mass/mole fraction at the solid/fluid interface, derived from eqns (59) and (60),
will be used for the determination of the partial pressures and, thereof, the local
electromotive forces by eqn (11).
The properties of solid materials in a SOFC are given in Table 3, which show
some variation based on the different literature sources. The single gas properties
are available from references [11] and [56]. For gas mixtures, equations from ref-
erences [11, 57] are available, and some selected equations from reference [11] for
calculating the properties are listed in the following section.
The mixing rule for the viscosity is:
1/2 1/4 2
n
Xi µ i 1 Mi −1/2 µi Mj
µm = n ; φij = 1/2 1 + 1+ ,
j=1 X j φ ij 8 M j µj Mi
i=1
(61)
where µm (Pa · sec) is the viscosity for the mixture, and µi or µj are the viscosities
of individual species (Pa · sec); Mi or Mj is the molecular weight of a species; Xi or
Xj is the mole fraction; and when i = j, φij = 1.
The mixing rule for the thermal conductivity of gases at atmospheric pressure or
less is:
n
Xi ki
km = n ;
i=1 j=1 Xj Aij
  (62)


3/4 1/2 2
1 µi Mj T + Si T + Sij
Aij = 1+ ,
4 µj Mi T + Sj  T + Si
where km [W/(m · K)] and ki [W/(m · K)] are the thermal conductivities of the mix-
ture and species; Sij = C(Si Sj )1/2 , and C = 1.0, but when either or both components
i and j are very polar, C = 0.73; for helium, hydrogen, and neon, Si or Sj is 79 K;
otherwise, Si = 1.5Tbi and Sj = 1.5Tbj , where Tb is the boiling point temperature
of species; and the unit of T is K.
When the gas mixture is above atmospheric pressure, the following correction
is applied to the km obtained above:
A × 10−4 (eBρr + C)
k = k + , (63)
1/6
Tc M 1/2 5
2/3 Z c
Pc
ρr < 0.5, A = 2.702, B = 0.535, C = −1.000,

0.5 < ρr < 2.0, A = 2.528, B = 0.670, C = −1.069,
2.0 < ρr < 2.8, A = 0.574, B = 1.155, C = 2.016,
where k [W/(m · K)] is the gas thermal conductivity at the temperature T (K) and
pressure P of interest in the mixture; k [W/(m · K)] is the thermal conductivity at T
and atmospheric pressure obtained by eqn (62); ρr = Vc /V is the reduced density;
Vc (m3 /kmol) is the critical molar volume; V (m3 /kmol) is the molar volume at T
and P; Tc (K) is the critical temperature; M is the molecular weight; Pc (MPa) is
the critical pressure; Zc = Pc Vc /(RTc ) is the critical compressibility factor; and
R is the gas constant, which is 0.008314 MPa · m3 /(kmol · K). The mixture critical
properties are obtained via the following equations:
n
Tcm − Tpc
Pcm = Ppc + Ppc 5.808 + 4.93 Xi ωi , (64)
Tpc
i=1
n

Xj Vcj
Tcm = n Tcj , (65)
j=1 i=1 Xi Vci

Vcm = φi φj vij (i = j), (66)
i j
where
2/3
Xj Vcj Vij (Vci + Vcj ) Vci − Vcj
φj = ; vij = ; Vij = −1.4684 + C,
n 2/3 2.0 V +V
i=1 Xi Vci
ci cj
and C is zero for hydrocarbon systems and is 0.1559 for systems containing a
non-hydrocarbon gas. In all the above equations, Xi or Xj is the mole fraction of a
species in the mixture; ωi is the acentric factor of a species; Pcm , Tcm and Vcm are
the mixture critical properties; and Ppc and Tpc are the pseudocritical properties of
the mixture, which are expressed as:

n
n
Tpc = Xi Tci ; Ppc = Xi Pci . (67)
i=1 i=1
Table 4: Atomic diffusion volumes for use in eqn (68).
Atomic and structural diffusion-volume increments v [11]
C 16.50 H2 7.07
H 1.98 N2 17.9
O 5.481 O2 16.6
N 5.69 CO 18.9
Aromatic ring −20.2 CO2 26.9
Heterocyclic ring −20.2 H2 O 12.7
The gas diffusivity of one species against the remaining species of a mixture is
expressed in the form of:
0.5
1 − Xi 0.01013T 1.75 M1i + M1j
Dim = , Dij = , (68)
jj=i Xj /Dij P[( vi )1/3 + ( vj )1/3 ]2
where units of T , P, and D are K, Pa, and m2 /sec, respectively; Mi or Mj is the

molecular weight; and all vi or vj are group contribution values for the subscript
component summed over atoms, groups and structural features, which are listed in
Table 4.
4.3 Numerical computation
In order to conduct a numerical computation for flow, temperature, and concentra-

tion fields in a SOFC, a mesh system with a sufficient grid number both in the r
and x directions must be deployed at the computational domain. All the governing
equations may be discretized by using the finite volume approach, and the SIMPLE
algorithm can be adopted to treat the coupling of the velocity and pressure fields
[58, 59].
The temperature difference between the cell tube and the air-inducing tube might
be large enough to have radiation heat transfer; therefore, a numerical treatment
based on the method introduced in the literature [60] can be used to consider the
radiation heat exchange.
As has been discussed at the beginning of Section 4, the computation may be
based on the internal conditions and the current to be withdrawn; and, as a conse-
quence of the simulation, the terminal voltage will be given as an output along with
other operational details. The convenience of using this procedure in the simulation
is discussed next.
It is quite common in practice that the total current is prescribed in terms of the
average current density of the fuel cell. Also, instead of the flow rates of fuel and
air, the stoichiometric data are prescribed in terms of the utilization percentage of
hydrogen and oxygen. This kind of designation of the operating conditions results
in a convenient comparison of the fuel cell performance based on the same level of
average current density and the hydrogen and oxygen utilization percentage. The
inlet velocities of fuel and air are, then, obtainable in the forms of:

Acell icell RTf
ufuel = , (69)
2FUH2 XH2 Afuel Pf

Acell icell RTair
uair = , (70)
4FUO2 XO2 Aair Pair
where icell is the cell current density; Acell is the outside surface area of the fuel cell;
Afuel and Aair are the cross-sectional inlet areas of the fuel and air; Pf , Pair and Tf ,
Tair are the inlet pressure and temperature of the fuel and air flows respectively; XH2
and XO2 are the mole fractions of hydrogen in the fuel and oxygen in the air, respec-
tively; and UH2 and UO2 are the utilization percentage for hydrogen and oxygen.
The computation process is highly iterative and coupled in nature. As the first
step, the latest local temperature, pressure, and species’ mass fractions are used in
the network circuit analysis to obtain the cell terminal voltage and local current
across the electrolyte, and thus the local species’ transfer fluxes and local heat
sources. In the second step, the local temperature, pressure and species’ mass frac-
tions are, in turn, obtained through solution of the governing equations under the
new boundary conditions determined by the latest-available species’fluxes and heat
sources. The two steps iterate until convergence is obtained.
4.4 Typical results from numerical computation for tubular SOFCs
The present authors have conducted numerical computations for three different sin-
gle tubular SOFCs [23], which have been tested by Hagiwara et al. [26], Hirano
et al. [25], Singhal [27], and Tomlins et al. [28]. The fuel tested by Hirano et al.
[25] had components of H2 , H2 O, CO and CO2 ; therefore, there is a water-shift
reaction of the carbon monoxide in the fuel cell to be considered together with
the electrochemical reaction. The fuel used by the other researchers [26–28] had
components of H2 and H2 O, where there is no chemical reaction except for the
electrochemical reaction in the fuel channel. The dimensions of the three different
solid oxide fuel cells tested in their studies are summarized in Table 5, in which
the mesh size adopted in our numerical computation is also given. The operating
conditions are listed in Table 6, including the species mole fractions and the tem-
perature of the fuel and air in those tests, which are the prescribed conditions for
the numerical computation. In the experimental work by Singhal [27], a test of the
pressure effect was also conducted by varying the fuel and air pressure from 1 atm
to 15 atm. It is expected that the experimental data for these SOFCs in different
dimensions and operating conditions will facilitate a wide benchmark range for
validation of the numerical modeling work.
4.4.1 The SOFC terminal voltage

The computer calculated and the experimentally obtained cell terminal voltages
under different cell current densities are shown in Fig. 9. The relative deviation of
Table 5: Example SOFCs with test data available.
Data sequence: Outer diameter (mm)/Thickness (mm)/Length (mm)
Singhal [27]
Hagiwara et al. [26] Hirano et al. [25] Tomlins et al. [28]
Air-inducing tube 7.00/1.00/485 6.00/1.00/290 12.00/1.00/1450

Support tube – 13.00/1.50/300 –
Cathode 15.72/2.20/500 14.40/0.70/300 21.72/2.20/1500
Electrolyte 15.80/0.04/500 14.48/0.04/300 21.80/0.04/1500
Anode 16.00/0.10/500 14.68/0.10/300 22.00/0.10/1500
Fuel boundary 18.10/ – /500 16.61/ – /300 24.87/ – /1500
Grid number (r × x) 66×602 66×602 66×1602
Table 6: Species’ mole fractions, utilization percentages, and temperatures.
Air fuel
O2 %−UO2 /N2 %/T (◦ C) H2 %–UH2 /H2 O%/CH4 %/CO%/CO2 %/T (◦ C)
I 21.00–17.00/79.00/600.0 98.64–85.00/1.36 /0/ 0 /0 /900.0

II ∗ 21.00–25.00/79.00/600.055.70–80.00/27.70/0/10.80/5.80/800.0
∗∗ 21.00–25.00/79.00/400.0 55.70–80.00/27.70/0/10.80/5.80/800.0
III 21.00–17.00/79.00/600.0 98.64–85.00/1.36 /0/ 0 /0 /800.0

∗ Current density = 185 mA/cm2 ; ∗∗ Current density = 370 mA/cm2 .
I: Tested by Hagiwara et al. [26].
II: Tested by Hirano et al. [25].
III: Tested by Singhal [27] and Tomlins et al.[28].
the model-predicted data from the experimental data is no larger than 1.0% for the
SOFC tested by Hirano et al. [25], 5.6% for that by Hagiwara et al. [26], and 6.0%
for that by Tomlins et al. [28].
It is interesting to observe from Fig. 9 that, under the same cell current density,
the cell voltage of the SOFC tested by Hagiwara et al. [26] is the highest and that
by Hirano et al. [25] is the lowest. The mole fraction of hydrogen in the fuel for the
SOFC tested by Hirano et al. [25] is low, which might be the major reason that this
cell has the lowest cell voltage. Because the current must be collected circumfer-
entially in a tubular type fuel cell, the large diameter of the cell tube investigated
by Singhal [27] and Tomlins et al. [28] will lead to a longer current pathway. Thus,
the cell voltages of these cells are lower than those found by Hagiwara et al. [26],
even though the former investigators tested the SOFCs at a pressurized operation
of 5 atm, which, in fact, helps to improve the cell voltage.
Under a current density of 300 mA/cm2 , the cell voltage and power increase with
the increasing operating pressure, as seen in Fig. 10. The agreement between our
Figure 9: Results of prediction and testing for cell voltage versus current density.
(The operating pressure of the cell tested by Hagiwara et al. [26] and
Hirano et al. [25] is 1.0 atm, and that by Tomlins et al. [28] is 5 atm.)
Figure 10: Effect of operating pressure on the terminal voltage and power.
model-predicted results and the experimental ones by Singhal [27] is quite good,
showing a maximum deviation of 7.4% at a low operating pressure. When the
operating pressure increases from 1 atm to 5 atm, the cell output power shows a
significant improvement of 9%. However, raising the operating pressure becomes
less effective for improving the output power when the operating pressure is high.
For example, the cell output power shows an increase of only 6% when the operat-
ing pressure increases from 5 atm to 15 atm. The reason for this is that the operat-
ing pressure contributes to the cell voltage in a logarithmic manner. Nevertheless,
pressurized operation of the fuel cell can improve the output power significantly.
For example, when increasing the operating pressure from 1 atm to 15 atm, the
cell output power can have an increment of 15.8%. There is no doubt from the
above investigation that the investigators can satisfactorily predict the overall per-
formance of a SOFC through numerical modeling and computation. On the basis
of this good agreement with the overall fuel cell performance, the internal details
of the flow, temperature, and concentration fields from numerical prediction can
also be reliably presented.
4.4.2 Cell temperature distribution

Because the measurement of temperature in a SOFC is very difficult, only three
experimental data points, the temperature at the two ends and in the middle of the
cell tube, were available from the work on Hirano et al. [25]. Figure 11 shows the
simulated cell temperature distribution for the SOFC, for which Hirano et al. [25]
provided the test data. The agreement of the simulated data and the experimental
results is good in the middle, where the hotspot is located; relatively larger devia-
tions between the predicted and experimental values appear at the two ends of the
cell. Nevertheless, such a discrepancy is acceptable when designing a SOFC with
respect to concerns about the prevention of excessive heat in the cell materials.
The predicted temperature distributions for the fuel cells tested by Hagiwara
et al. [26] and Tomlins et al. [28] are given in Fig. 12. Unfortunately, there was
no experimental data on the cell temperature. Generally, the two ends of the cell
tube have lower temperatures than the middle of the cell tube. However, at low
current densities, the hotspot is located closer to the closed end of the cell. With an
increase in current density, the hotspot shifts to the open-end side, and the hotspot
temperature also decreases, which improves the uniformity of the temperature dis-
tribution along the fuel cell. It should be observed that the heat transfer between
the cooling air and the cell tube at the closed-end region is dominated by laminar
jet impingement, since the exit velocity from the air-inducing tube is quite low.
However, the velocity of the exit air from the air-inducing tube affects the heat
Figure 11: Longitudinal temperature distribution in the fuel cell.

Figure 12: Predicted longitudinal temperature distribution for two SOFCs.
transfer coefficient significantly. For the high current density case, the flow rate of
air also becomes large accordingly. Thus, the heat transfer coefficient between the
air and the fuel cell closed-end region is increased. This can suppress the tempera-
ture level of the closed-end region of the fuel cell significantly. Since the air receives
a large amount of heat at the closed-end region, its cooling to the fuel cell in the
downstream region becomes weak, and the uniformity of the cell temperature dis-
tribution becomes much better when the fuel cell operates at high current densities.
4.4.3 Flow, temperature and concentration fields

Figure 13 shows the flow and temperature fields for the SOFC tested by Hirano
et al. [25] at a current density of 185 mA/cm2 . The air speed in the air-inducing
tube has a slight acceleration because the air absorbs heat and expands in this flow
passage. After leaving the air-inducing tube, the air impinges on the closed end
of the fuel cell, and then flows backwards to the outside. In this pathway, the air
obtains heat from the heat-generating fuel cell tube and transfers the heat to the
cold air in the air-feeding tube. It is easy to understand that the electrochemical
reaction at the closed end of the fuel cell is strong because the concentrations of
fuel and air are both high there. Therefore, the heat generation due to Joule heating
and the entropy change of the electrochemical reaction is high at the upstream area
of the fuel path. However, it is known from both experiments and computation that
the closed-end region of the fuel cell does not demonstrate the highest temperature;
therefore, it is believed that the cooling of the air in the closed-end area of the fuel
cell is responsible for this. After being heated at the closed-end region, air exhibits
a higher temperature, and its cooling ability to the cell tube is low when it is in
the annulus between the air-inducing tube and the cell tube. At the cell open-end
region, the air in the annulus can transfer heat to the incoming cold fresh air in the
air-inducing tube, and this will help it to cool the fuel cell tube.
Figure 13: Predicted flow and temperature fields for the SOFC reported by Hirano
et al. [25] at a current density of 185 mA/cm2 .
From this airflow arrangement, the hotspot temperature of the cell tube may
mostly occur in the center region in the longitudinal direction of the cell tube.
The airflow has two passes, incoming in the air-inducing tube and outgoing in
the annulus between the air-inducing tube and the cell tube. The heat exchange in
between the two passes allows the air to mitigate its temperature fluctuation in the
whole air path, and thus the temperature field in the fuel cell might be maintained
as relatively uniform. Nevertheless, the heat generation, air and fuel temperature,
and air-cooling to the fuel cell will collectively affect the temperature field in the
fuel cell. Therefore, the hot spot position in a cell tube might shift more or less
away from the center region depending on the operating condition of the fuel cell.
Figure 14 shows the gas species’mole fraction contours for the same SOFC under
the same operating conditions as discussed with respect to Fig. 13. In the air path,
oxygen consumption at the closed-end region is relatively large, which leads to
more densely distributed contour lines. The contour shape of oxygen also indicates
a relatively larger difference of the mole fraction between the bulk flow and the
wall of the cathode/air interface. This implies that the mass transport resistance on
the air side might be dominant in lowering the cell performance if the stoichiometry
of the oxygen is low. Feeding more air than is needed is already well applied in
operational fuel cell technology.
Figure 14: Predicted fields of the mole fraction of the species for the SOFC reported
by Hirano et al. [25] at a current density of 185 mA/cm2 .
The hydrogen budget is collectively determined by the consumption by the

electrochemical reaction and the generation from the water-shift reaction of CO.
Since the consumption dominates, the hydrogen mole fraction decreases along
the fuel stream. Corresponding to this hydrogen variation, consumption due to the
water-shift reaction and production due to the electrochemical reaction cause the
water vapor to increase gradually along the fuel stream. The water-shift of CO
proceeds gradually in the fuel path, and thus the mole fraction of CO decreases but
the CO2 increases. The shape of the contour lines of the species in the fuel path is
Figure 15: Predicted streamwise molar flow rate variation for species in the fuel
channel for the SOFC reported by Hirano et al. [25].
relatively flat from the cell wall to the bulk flow. This indicates that mass diffusion
in the fuel channel is relatively stronger than that in the airflow.
For a further illustration of the variation of the gas species, Fig. 15 shows the
molar flow rate variation along the fuel path. In one third of the length from the fuel
inlet, the hydrogen flow rate shows a faster decrease and the water flow rate shows
a faster increase, indicating a strong reaction in the upstream region. The flow rate
of CO and CO2 vary roughly in a linear style, and a small amount of CO still exists
in the waste gas.
5 Concluding remarks
Fuel cell technology is currently under rapid development. To improve SOFC per-
formance, for high power density and efficiency, efforts have been made to reduce
the three over-potentials: activation polarization, ohmic loss, and concentration
polarization. Better understanding of these three over-potentials is also very impor-
tant in developing accurate computer models for predicting the overall performance
and internal details of a SOFC.
The activation polarization relates to the porous structure of the electrode and
electrocatalyst materials. The state-of-the-art in material and manufacturing pro-
cesses for the electrodes and electrolyte has been reported by Singhal [27]. The
reduction of ohmic losses also heavily relies on the reduction of electronic and ionic
resistances in the electrodes and electrolyte. A shorter current collection pathway
also helps to reduce ohmic loss. A new design, referred to as a high power density
solid oxide fuel cell (HPD-SOFC), has been developed by Siemens Westinghouse
Power Corporation [27, 32], and has a significantly shorter current pathway, and
thus improves the power density significantly. A planar structure also promises
to have a shorter current pathway and thus a higher power density, and measures
for reducing the ohmic loss in a planar type SOFC have been reported by Tanner
and Virkar [61]. The reduction of mass transport resistance, or the concentration
polarization, has not been given much attention. Mass transfer enhancement has
been reported to be effective in polymer electrolyte membrane fuel cells (PEMFCs)
for obtaining a higher cell current density [62] before a sharp drop in cell voltage
(which is due to excessive concentration polarization). It might also be possible for
SOFCs to obtain a higher current density by means of mass transfer enhancement.
In a numerical model of a SOFC, the precise calculation of the over-potentials
is very important in order to accurately predict the overall current-voltage perfor-
mance. The heat generation from the over-potentials is also significant in com-
puting the temperature, flow, and species concentration fields. With respect to the
activation polarization, studies elucidating the data and equations for the exchange
current density are still needed. For the prediction of ohmic losses, reliable property
data for electrodes are required. Additionally, a method for analyzing a complex
network circuit in a SOFC needs to be developed. The concentration polarization
is considered in the numerical computation by using the local mole fractions of
the species at the interface of the electrode and fluid when calculating the electro-
motive force by the Nernst equation. Because the porous electrodes also serve as
the reaction site, there is no well-described model for the mass transport resistance
in the electrodes. Adopting a lower exchange current density, which induces a larger
over-potential of the activation polarization, may be a way to incorporate the mass
transport resistance in the electrodes into the activation polarization. The method
given by Hirano et al. [25] for the consideration of mass transport resistances in
the electrodes is convenient, but may be too simple and needs more investigation.
With the progress being made in computer modeling of SOFCs, it is expected
that costs for research and development of SOFCs will be significantly reduced by
using computer simulations in the future.
References
[1] Srinivasan, S., Mosdale, R., Stevens, P. & Yang, C., Fuel cells: reaching the
era of clear and efficient power generation in the twenty-first century. Annual
Review Energy Environment, 24, pp. 281–328, 1999.
[2] Gardner, F.J., Thermodynamic processes in solid oxide and other fuel cells.
Proc. Instn. Mech. Engrs., 211, Part A, pp. 367–380, 1997.
[3] Suzuki, K., Iwai, H., Kim, J.H., Li, P.W. & Teshima, K., Solid oxide fuel cell
and micro gas turbine hybrid cycle and related fluid flow and heat transfer.
The 12th International Heat Transfer Conference, Grenoble, August 18–23,
1, pp. 403–414, 2002.
[4] Harvey, S.P. & Richter, H.J., Gas turbine cycles with solid oxide fuel cells,
Part I: Improved gas turbine power plant efficiency by use of recycled exhaust
gases and fuel cell technology. ASME Journal of Energy Resources Technol-
ogy, 116, pp. 305–311, 1994.
[5] Layne,A., Samuelsen, S., Williams, M. & Hoffman, P., Developmental status
of hybrids. International Gas Turbine & Aeroengine Congress & Exhibition,
Indianapolis, Indiana, USA, No. 99-GT-400, 1999.
[6] Ali, S.A. & Moritz, R.R., The hybrid cycle: integration of turbomachinery
with a fuel cell. International Gas Turbine & Aeroengine Congress & Exhi-
bition, Indianapolis, Indiana, USA, No. 99-GT-361, 1999.
[7] Bevc, F., Advances in solid oxide fuel cells and integrated power plants.
Proc. Instn. Mech. Engrs., 211, Part A, pp. 359–366, 1997.
[8] Freni, S. & Maggio, G., Energy balance of different internal reforming MCFC
configurations. International Journal of Energy Research, 21, pp. 253–264,
1997.
[9] Wojcik, A., Middleton, H., Damopoulos, I. & Herle, J.V., Ammonia as a
fuel in solid oxide fuel cells. Journal of Power Sources, 118, pp. 342–348,
2003.
[10] Wark, K., Thermodynamics, Second Edition, McGraw-Hill Book Company:
New York, pp. 536–561, 1971.
[11] Perry, R.H. & Green, D.W., Perry’s Chemical Engineer’s Handbook, Seventh
Edition, McGraw-Hill: New York, 1986.
[12] Onuma, S., Kaimai,A., Kawamura, K., Nigara, Y., Kawada, T., Mizusaki, J. &
Tagawa, H., Influence of the coexisting gases on the electrochemical reaction
rates between 873 and 1173 K in a CH4 -H2 O/Pt/YSZ system. Solid State
Ionics, 132, pp. 309–331, 2000.
[13] Li, P.W., Schaefer, L., Wang, Q.M. & Chyu, M.K., Computation of the con-
jugating heat transfer of fuel and oxidant separated by a heat-generating
cell tube in a solid oxide fuel cell, Paper No. IMECE2002-32564. Proc.
of ASME International Mechanical Engineering Congress and Exposition,
Nov. 17–22, New Orleans, Louisiana, USA, 2002.
[14] Hartvigsen, J., Khandkar, A. & Elangovan, S., Development of an SOFC
stack performance map for natural gas operation. Proc. of the Sixth Interna-
tional Symposium on Solid Oxide Fuel Cells. Honolulu, Hawaii, USA, 1999;
on Web site www.mtiresearch.com/pubs.html No. MTI00-07.
[15] Massardo, A.F. & Lubelli, F., Internal reforming solid oxide fuel cell-gas
turbine combined cycles (IRSOFC-GT), Part A: Cell model and cycle ther-
modynamic analysis. International Gas Turbine & Aeroengine Congress &
Exhibition, Stockholm, Sweden, No. 98-GT-577, 1998.
[16] Chan, S.H., Khor, K.A. & Xia, Z.T., A complete polarization model of a solid
oxide fuel cell and its sensitivity to the change of cell component thickness.
Journal of Power Sources, 93, pp. 130–140, 2001.
[17] Keegan, K., Khaleel, M., Chick, L., Recknagle, K., Simner, S. & Deibler, J.,
Analysis of a planar solid oxide fuel cell based automotive auxiliary
power unit. SAE 2002 World Congress, Detroit, Michigan, No. 2002-01-
0413, 2002.
[18] Nagata, S., Momma, A., Kato, T. & Kasuga, Y., Numerical analysis of output
characteristics of tubular SOFC with internal reformer. Journal of Power
Resources, 101, pp. 60–71, 2001.
[19] Iwata, M., Hikosaka, T., Morita, M., Iwanari, T., Ito, K., Onda, K., Esaki,
Y., Sakaki, Y. & Nagata, S., Performance analysis of planar-type unit SOFC
considering current and temperature distributions. Solid State Ionics, 132,
pp. 297–308, 2000.
[20] Minh, N.Q. & Takahashi, T., Science and Technology of Ceramic Fuel Cells,
Elsevier: New York, 1995.
[21] Ota, T., Koyama, M., Wen, C.J., Yamada, K. & Takahashi, H., Object-based
modeling of SOFC system: dynamic behavior of micro-tube SOFC. Journal
of Power Sources, 118, pp. 430–439, 2003.
[22] Srikar, V.T., Turner, K.T., Andrew Ie, T.Y. & Spearing, S.M., Structural
design consideration for micromachined solid-oxide fuel cells. Journal of
Power Sources, 125, pp. 62–69, 2004.
[23] Li, P.W. & Chyu, M.K., Simulation of the chemical/electrochemical reaction
and heat/mass transfer for a tubular SOFC working in a stack. Journal of
Power Sources, 124, pp. 487–498, 2003.
[24] Li, P.W., Schaefer, L. & Chyu, M.K., Investigation of the energy budget in
an internal reforming tubular type solid oxide fuel cell through numerical
computation, Paper No. IJPGC2003-40126. Proc. of the International Joint
Power Conference, June 16–19, Atlanta, Georgia, USA, 2003.
[25] Hirano, A., Suzuki, M. & Ippommatsu, M., Evaluation of a new solid oxide
fuel cell system by non-isothermal modeling. J. Electrochem. Soc., 139(10),
pp. 2744–2751, 1992.
[26] Hagiwara, A., Michibata, H., Kimura, A., Jaszcar, M.P., Tomlins, G.W. &
Veyo, S.E., Tubular solid oxide fuel cell life tests. Proc. of the Third Inter-
national Fuel Cell Conference, Nagoya, Japan, pp. 365–368, 1999.
[27] Singhal, S.C.,Advances in solid oxide fuel cell technology. Solid State Ionics,
135, pp. 305–313, 2000.
[28] Tomlins, G.W. & Jaszcar, M.P., Elevated pressure testing of the Simens West-
inghouse tubular solid oxide fuel cell. Proc. of the Third International Fuel
Cell Conference, Nagoya, Japan, pp. 369–372, 1999.
[29] Bessette N.F., Modeling and simulation for solid oxide fuel cell power system,
Georgia Institute of Technology, Ph.D Thesis, 1994.
[30] Ahmed, S., McPheeters, C. & Kumar, R., Thermal-hydraulic model of a mon-
olithic solid oxide fuel cell. J. Electrochem. Soc., 138, pp. 2712–2718, 1991.
[31] Ferguson, J.R., Fiard J.M. & Herbin, R., Three-dimensional numerical sim-
ulation for various geometries of solid oxide fuel cells. Journal of Power
Sources, 58, pp. 109–122, 1996.
[32] Singhal, S.C., Progress in tubular solid oxide fuel cell technology. Electro-
chemical Society Proceedings, 99–19, pp. 40–50, 2001.
[33] Sverdrup, E.F., Warde, C.J. & Eback, R.L., Design of high temperature solid-
electrolyte fuel cell batteries for maximum power output per unit volume.
Energy Conversion, 13, pp. 129–136, 1973.
[34] Li, P.W., Suzuki, K., Komori, H. & Kim, J.H., Numerical simulation of
heat and mass transfer in a tubular solid oxide fuel cell. Thermal Science &
Engineering, 9(4), pp. 13–14, 2001.
[35] Aguiar, P., Chadwick, D. & Kershenbaum, L., Modeling of an indirect

internal reforming solid oxide fuel cell. Chemical Engineering Science, 57,
pp. 1665–1677, 2002.
[36] Bessette, N.F. & Wepfer W.J., A mathematical model of a tubular solid
oxide fuel cell. Journal of Energy Resources Technology, 117, pp. 43–49,
1995.
[37] Campanari, S., Thermodynamic model and parametric analysis of a tubular
SOFC module. Journal of Power Sources, 92, pp. 26–34, 2001.
[38] Haynes, C. & Wepfer, W.J., Characterizing heat transfer within a commercial-
grade tubular solid oxide fuel cell for enhanced thermal management. Int. J.
of Hydrogen Energy, 26, pp. 369–379, 2001.
[39] Recknagle, K.P., Williford, R.E., Chick, L.A., Rector, D.R. & Khaleel, M.A.,
Three-dimensional thermo-fluid electrochemical modeling of planar SOFC
stacks. Journal of Power Sources, 113, pp. 109–114, 2003.
[40] Li, P.W., Schaefer, L. & Chyu, M.K., Interdigitated heat/mass transfer and
chemical/electrochemical reactions in a planar type solid oxide fuel cell.
Proc. of ASME 2003 Summer Heat Transfer Conference, Las Vegas, Paper
No. HT2003-47436, 2003.
[41] Joon, K., Fuel cells – a 21st century power system. Journal of Power Sources,
71, pp. 12–18, 1998.
[42] Watanabe, T., Fuel cell power system applications in Japan. Proc. Instn.
Mech. Engrs., 211, Part A, pp. 113–119, 1997.
[43] Veyo, S.E. & Lundberg, W.L., Solid oxide fuel cell power system cycles.
Proc. of the International Gas Turbine & Aeroengine Congress & Exhibi-
tion, Indianapolis, Indiana, USA, Paper No. 99-GT-365, 1999.
[44] White, D.J., Hybrid gas turbine and fuel cell systems in perspective review.
Proc. of the International Gas Turbine & Aeroengine Congress & Exhibition,
Indianapolis, Indiana, USA, Paper No. 99-GT-419, 1999.
[45] Li, P.W., Schaefer, L., Wang, Q.M., Zhang, T. & Chyu, M.K., Multi-gas
transportation and electrochemical performance of a polymer electrolyte
fuel cell with complex flow channels. Journal of Power Sources, 115
pp. 90–100, 2003.
[46] Cahoon, N.C. & Weise, G. (eds) The Primary Battery, Vol. II, John Wiley &
Sons: New York, 1976.
[47] Li, P.W., Schaefer, L. & Chyu, M.K., Three-dimensional model for the con-
jugate processes of heat and gas species transportation in a flat plate solid
oxide fuel cell. 14th International Symposium of Transport Phenomenon,
Bali, Indonesia, June 6–9, pp. 305–312, 2003.
[48] Li, P.W., Schaefer, L. & Chyu, M.K., The energy budget in tubular and
planar type solid oxide fuel cells studied through numerical simulation.
International Mechanical Engineering Congress and Exposition 2003,
(IMECE2003-42426), Washington, D.C., Nov. 16–21, 2003.
[49] Burt, A.C., Celik, I.B., Gemmen, R.S. & Smirnov, A.V., A numerical study
of cell-to-cell variations in a SOFC stack. Journal of Power Sources, 126,
pp. 76–87, 2004.
[50] George, R.A. & Bessette, N.F., Reducing the manufacturing cost of
tubular SOFC technology. Journal of Power Sources, 71, pp. 131–137,
1998.
[51] Krumdieck, S., Page, S. & Round S., Solid oxide fuel cell architecture
and system design for secure power on an unstable grid. Journal of Power
Sources, 125, pp. 189–198, 2004.
[52] Bird, R.B., Stewart, W.E. & Lightfoot, E.N., Transport Phenomena,
John Wiley & Sons: New York, 1960.
[53] Williams, F.A., Combustion Theory, Benjamin/Cummings Publishing Co.,
1985.
[54] Eckert, E.R.G. & Drake, R.M., Heat and Mass Transfer, 2nd Edition of Intro-
duction to the Transfer of Heat and Mass, McGraw-Hill Book Company, Inc.,
1966.
[55] Turns, S.R., Introduction to Combustion: Concept and Application, Second
Edition, McGraw-Hill Higher Education: New York, 1999.
[56] Incropera, F.P. & DeWitt, D.P., Introduction to Heat Transfer, Third Edition,
John Wiley & Sons, 1996.
[57] Todd, B. & Young, J.B., Thermodynamic and transport properties of gases
for use in solid oxide fuel cell modeling. Journal of Power Sources, 110,
pp. 186–200, 2002.
[58] Patankar, S.V., Numerical Heat Transfer and Fluid Flow, McGraw-Hill,
1980.
[59] Karki, K.C. & Patankar, S.H., Pressure based calculation procedure for vis-
cous flows at all speed in arbitrary configurations. AIAA Journal, 27(9),
pp. 1167–1174, 1989.
[60] Beckermann, C. & Smith, T.F., Incorporation of internal surface radiation
exchange in the finite-volume method. Numerical Heat Transfer, Part B, 23,
pp. 127–133, 1993.
[61] Tanner, C.W. & Virkar A.V., A simple model for interconnect design of planar
solid oxide fuel cells. Journal of Power Sources, 113, pp. 44–56, 2003.
[62] Nguyen T.V., A gas distributor design for proton-exchange-membrane fuel
cells. J. Electrochem. Soc., 413(5), pp. L103–L105, 1996.
Nomenclature
a Stoichiometric coefficient of chemical species.
A Chemical species. Area (m2 ). General variable.
Acell Outer surface area of fuel cell (m2 ).
Aair , Afuel Inlet flow area of air and fuel, respectively (m2 ).
b Stoichiometric coefficient of chemical species.
B Chemical species. General variable.
C General variable.
Cp Specific heat capacity at constant pressure [J/(kg K)].
DJ ,m Diffusion coefficient of jth species into the left gases
of a mixture (m2 /s).
E Electromotive force or electric potential (V).

F Faraday’s constant [96486.7 (C/mol)].
g Gibbs free energy (J/mol).
h Chemical enthalpy (kJ/kmol) or (J/mol).
H Height (m).
i Current density (A/m2 ).
i0 Exchange current density (A/m2 ).
I Current (A).
k Thermal conductivity (W/m K).
KPR , KPS Chemical equilibrium constant for reforming and shift
reactions, respectively.
L Length (m).
m Mass transfer rate or mass consumption/production rate (mol/s).
ṁ Mass flux [mol/(m2 s)].
M Molecular weight (g/mol).
Mf Total mole rate of fuel flow (mol/s).
ne Number of electrons involved in per fuel molecule in oxidation
reaction.
p, P Pressure (Pa) or position.
q̇ Volumetric heat source ( W/m3 ).
Q Heat energy (W).
r Radial coordinate (m).
r a , rc , re Average radius of anode, cathode, and electrolyte layers (m).
R Universal gas constant [8.31434 J/(mol K)].
Ra , Rc , Re Discretized resistance in anode, cathode, and electrolyte ().
S Source term of gas species (kg/m3 ); General variable.
T Temperature (K).
u Velocity in axial direction (m/s).
U Utilization percentage (0–1).
v Velocity in radial direction (m/s); Diffusion volume in eqn (68).
V Specific volume (m3 /kmol).
Vcell Cell terminal voltage (V).
V a, V c Potentials in anode and cathode, respectively (V).
W Width (m).
x Stoichiometric coefficient of chemical species;
Axial coordinate (m).
X Chemical species. Mole fraction.
x̄, ȳ, z̄ Reacted mole rate of CH4 , CO and H2 , respectively in a section
of interest in flow channel (mol/s).
y Stoichiometric coefficient of chemical species; Coordinate (m).
Y Chemical species. Mass fraction.
z Coordinate (m).
Z Compressibility factor.
Greek symbols
θ Circumferential position. Angle.

δ Thickness of electrodes and electrolyte layers (m).
G Gibbs free energy change of a chemical reaction (J/mol).
G 0 Standard state Gibbs free energy change of a chemical
reaction (J/mol).
H Enthalpy change of a chemical reaction (J/mol).
S Entropy production [J/(mol K)].
x One axial section of fuel cell centered at x position (m).
λ Thermal conductivity [W/(m ◦ C)].
µ Dynamic viscosity (Pa s).
ρ Density (kg/m3 ).
ρea , ρec Electronic resistivity of anode and cathode respectively ( · cm).
ρee Ionic resistivity of electrolyte ( · cm).
ρr Reduced density.
ηAct Activation polarization (V ).
Subscripts
a Anode.
c Cathode.
cell Overall parameter of fuel cell.
e, w, n, s East, west, north, and south interfaces between grid P and it
neighboring grids.
E, W , N , S East, west, north, and south neighboring grids of grid P.
f Fuel.
i Subscript variable.
j Gas species; Subscript variable.
m Mixture.
P Variables at grid P.
R Electrochemical reaction.
x Axial position.
X,Y Chemical species.
x Variation in the channel section of x.
Superscripts
a Anode. Sequence.
b Sequence.
c Cathode. Sequence.
e Electrolyte.
in Inlet of a channel section of interest.
out Outlet of a channel section of interest.
P Variables at grid P.
R Reaction.
x Axial position.
CHAPTER 2
Numerical models for planar solid oxide

fuel cells
S.B. Beale
National Research Council, Ottawa, Canada.
Abstract
This article discusses various numerical techniques used to model single-cells and
stacks of planar solid oxide fuel cells. A brief history of the solid oxide fuel cell
(SOFC), and a survey of modeling efforts to-date are presented. The fundamental
thermodynamics of the SOFC are introduced, together with the equations govern-
ing the ideal (Nernst) potential. Factors affecting operating cell voltages are then
discussed. A simple iterative calculation procedure is described, whereby inlet mass
factions, flow rates, and cell voltage are known, but outlet values, and current are
required. This provides a paradigm for more complex algorithms set out in the
remainder of the text. The next level of complexity is provided for by numeri-
cal integration schemes based on ‘presumed-flow’ methodologies, where the inlet
flow rates of oxidant and fuel are assumed to be uniform. Local mass sources and
sinks are computed from Faraday’s equation. These are then used to correct the
continuity and mass transfer equations. Temperature variations are also computed;
because heat and mass transfer effects affect the output of SOFCs, significantly.
Some more advanced chemistry, heat and mass transfer issues are further detailed.
The most detailed cell models are obtained using computational fluid dynamics
codes, based on finite-volume and other techniques. Additional improvements to
these (and other) codes involve the detailed modeling of the electric potential (in
place of the Nernst equation formulation), and the analysis of the combined kinet-
ics and mass transfer problem in the porous electrode media. Finally stack models
are introduced. These may be full-scale computational fluid dynamics models or
simpler models based on volume-averaging techniques.
Figure 1: Schematic of typical planar SOFC.
1 Introduction
1.1 History and types of solid oxide fuel cell
There are many types of fuel cell under development. Most popular among these
are the proton exchange membrane fuel cell (PEMFC) and the solid oxide fuel
cell (SOFC) [1–3]. The SOFC is a solid-state device that utilizes the oxygen ion
conducting property of the ceramic zirconia. The operation of a SOFC differs from
that of a PEMFC or a conventional battery, in that both electronic and ionic circuits
are due to the flux of negatively charged particles; electrons and O2− ions. The
positive ions or ‘holes’are immobilised in the solid matrix, at the time of fabrication.
The book by Williams [4] details developments in SOFC technology prior to the
1960’s.
A SOFC consists of an anode or fuel electrode, an electrolyte, and a cathode or
air electrode. The anode is typically a porous cermet, for example the combination
of NiO and Yttria-stabilised Zirconia (YSZ). A typical material for the cathode is
Sr doped LaMnO3 , a perovskite. The electrolyte consists of a layer of YSZ, and is
fabricated as thin as possible in order to reduce Ohmic losses. Fuel and air passages
are required to supply the products and remove the reactants; the gas passages are
typically rectangular micro-channels, though planar channels may also be utilized
occasionally. In addition, porous media inserted between the fluid passages and
electrodes may serve as gas diffusion layers.
Both planar and cylindrical SOFC geometries are to be found; the latter avoid
mechanical sealing problems, whereas the former allow for relatively large elec-
trical power densities to be achieved. The discussion in this paper focuses on the
planar design. In order to reach the necessary operating voltage, fuel cells are con-
nected together in a stack by situating a metallic or ceramic interconnect between
each cell. These are often made of stainless steel. Cells are connected electrically
Figure 2: Planar SOFC stack. Courtesy Global Thermoelectric Inc.
in series, but hydraulically in parallel. The interconnects also function as a hous-

ing for the flow channels for the air and fuel which are supplied via manifolds.
The balance-of-plant makes sure that the working fluids are supplied under the
desired operating conditions. Fuel cells may be operated in co-flow, counter-flow
and cross-flow. The industrial SOFC designs encountered by the present author were
for cross-flow. Hydrogen is normally provided by reforming methane or methanol
in a reactor. The electrical efficiency of a system running on natural gas can exceed
50%. The remaining energy is released as heat.
The basic operation of a SOFC is as follows: reduction takes place at the cathode.
O2 + 4e− → 2O2− . (1)

At the anode, if the fuel is H2 , the oxygen ion combines with hydrogen to form
water, electricity and heat.
H2 + O2− → H2 O + 2e− . (2)
Thus the overall reaction is
H2 + 12 O2 → H2 O.
(3)
1 kg + 8 kg 9 kg
1.2 Survey of modeling techniques
Mathematical models of heat and mass transfer and electrochemistry provide tools
for performance prediction of single-cells and stacks of fuel cells. Thermo-fluid
simulations can predict the thermal and electrochemical performance, and struc-
tural analysis models can predict the mechanical behavior. Our concern here is
with the former. Among the first to perform calculations on SOFC performance
were Vayenas and Hegedus [5]. Since then, the scope and detail of modeling have
increased. Models have been developed at various scales: micro-scale or nano-scale
studies aim at the development of better electrodes through mathematical analy-
sis. Single-cell models are often the first consideration for fuel cell manufacturers
embarking on a new product design, and are currently a target application for the
vendors of commercial CFD codes. Stack modeling is necessary for the overall
design and often involves parametric studies. Among the most salient issues that
need to be addressed are the electric potential field, porous media transport phenom-
ena, and chemical kinetics (shift reactions, internal reforming). The development
of a model for conservation of heat, mass, and charge in SOFCs is detailed in [6–8].
Such models have been applied to SOFC stacks [9–12]. Shift and reforming reac-
tions have also been considered [13, 14] as well as heat and mass transfer issues
[15, 16]. The size and complexity of SOFC stacks requires the use of very large com-
puters, if a conventional CFD code is employed. Because of this, simpler models
based on engineering assumptions have sometimes been devised [10, 12, 17, 18]
in order that personal computers may perform such calculations. The speed and
memory requirements of a computer code will be limited by the application. For
detailed SOFC design large-scale CFD codes will be required, whereas for real-time
control, rapid response will preclude all but the simplest of schemes from being
employed.
1.3 Thermodynamics of solid oxide fuel cells
Newman [19] states that thermodynamics, kinetics, and transport phenomena are
fundamental to understanding electrochemical systems. It is important to realise
that transport phenomena are affected by thermodynamic properties and chemical
kinetics. Physicochemical hydrodynamics [20, 21] is the study of such mutual
interactions. It is timely to start with the thermodynamics of fuel cells. Figure 3
illustrates the notion of the fuel cell as a thermodynamic device. It is postulated
Figure 3: Fuel cell as a thermodynamic system.

that the fundamental relation [22] for the internal energy, U , may be expressed as
U = U (S, V , Nj , Q), (4)
where S is entropy, V is volume, Nj is mole number of species j, and Q is charge

number. Conservation of energy requires that,
dU = dWHeat + dWMech + dWChem + dWElec , (5)

dU = TdS − PdV + µj dNj + EdQ, (6)
j
where T = (∂U /∂S)V ,Nj ,Q , P = − (∂U /∂V )S,Nj ,Q , µj = (∂U /∂Ni )S,V ,Q , and
E = (∂U /∂Q)S,V ,Nj are temperature, pressure, chemical potential, and electric poten-
tial, respectively.
Texts in physical chemistry invariably introduce Gibbs free energy, G, which for
a chemical reaction which does not involve charge transfer may be written:

dG = dU − TdS + PdV = µj dNj . (7)
j
The free energy is useful for analysis of situations where temperature and pres-
sure are constant (e.g.
at ambient). Note that for the situation under consideration
dU − TdS + PdV = j µj dNj .
For an electrochemical process, where temperature and pressure are maintained
constant and there is no change in internal energy, chemical energy is converted to
electrical energy:

µj dNj = −EdQ, (8)
j
where it is understood that Nj is positive for inflow and negative for outflow, i.e.,
the sum of chemical energy of the reactants less that of the products.
The molar enthalpy of formation of the chemical reaction defined by eqn (3) is
H 0 = −247.3 kJ/mol at 1000 ◦ C, whereas the molar Gibbs energy of formation
(i.e. the chemical energy of formation) is G 0 = −177.4 kJ/mol, the negative sign
indicating that heat will be released at a rate of 69.9 kJ/mol. The electrochemical
reaction is thus exothermic. The difference between these terms H 0 −G 0 is
frequently referred to in a somewhat cavalier fashion as the TS 0 term. However
there are other heat sources in a fuel cell, which can change the entropy (and
internal energy) of the system. It is also worth noting that a fuel cell is not open
to the atmosphere, so pressure and temperature can and do vary. It is therefore
prudent to appreciate that eqn (8) is just a simplified form of eqn (6). Nonetheless
it is eqn (8) that is invariably the point of departure for a fuel cell analysis, with
irreversibilities being added to modify the basic postulate that ideally all chemical
energy would be converted to electrical energy.
Since the air and fuel in a SOFC are high-temperature gases, we may treat them
as multi-component perfect gases, such that
PV = NRT , (9)
where R = 8.3144 × 103 J/mol.K is the gas constant, and

N= Nj (10)
j
for the air-side and fuel-side gases. The chemical potentials are given by [22]:
µj = µ0j (T ) + RT ln P + RT ln xj = µ0j (T ) + RT ln Pj , (11)
where
xj = Nj /N = Nj Nk (12)
k
is mole fraction and Pj = xj P is the partial pressure of species j.
Faraday’s law gives the number of moles of oxygen consumed by the electro-
chemical reaction as
Q
Ñ = , (13)
nF
where F = 96 485 Coulomb/mol is Faraday’s constant and n = 2 is the valence of
oxygen.
Thus the change in chemical energy is

µj Nj = νj µj Products − νj µj Reactants Ñ (14)
and νi are stoichiometric coefficients. For example,
ν1 A1 + ν2 A2 → ν3 A3 , (15)

µi Ni = (ν1 µ1 + ν2 µ2 − ν3 µ3 )Ñ = (G 0 + RT ln Kp )Ñ , (16)

where G 0 = [(νj µj )Products − (νj µj )Reactants ] is referred to as the molar Gibbs
free energy for the chemical reaction. (NB: The author denotes G = G/Ñ J/mol
for consistency with the literature).
The equilibrium constant, KP , is given by
P1ν1 P2ν2
Kp = ν . (17)
P3 3
This may be re-written in terms of mole fractions, xj , e.g., for a SOFC with
hydrogen as fuel:

1 xH2 xO0.5
0
µi Ni = G + RT ln Pa + RT ln 2
Ñ , (18)
2 xH2 O
where Pa is air pressure. Substituting molar flow rate for current using Faraday’s
law we immediately obtain

µi Ni
E=− . (19)
nF
Since the valence of oxygen is 2, we have,
0.5

RT xH2 xO RT
E=E + 0
ln 2
+ ln Pa (20)
2F xH2 O 4F
which is the Nernst equation for a solid oxide fuel cell. The quantity E is the
reversible or Nernst potential. In what follows below, we make the assumption
that we may associate the thermodynamic pressure with the hydrodynamic pres-
sure, and the thermodynamic potential with the electrostatic potential. For practical
reasons, E 0 is redefined such that Pa is gauge (relative) pressure rather than abso-
lute air pressure, in which case the latter term in eqn (20) is replaced by the form
(RT /4F) ln (1+Pgauge /P0 ), and P0 is a reference operating pressure. The latter form
is more useful when considering detailed transport models where values of pressure
and temperature vary. Generally speaking Pgauge /P0 is very small so changes in E
are largely due to variations in mole fraction and temperature.
Values of E 0 are typically enumerated as polynomials, e.g.,
E 0 = a − bT . (21)
Typical values would be a = 1.214 V and b = 3.2668 V/K. Higher order

polynomials may also be found. The physical significance of the terms on the right-
side of eqn (21) are a ≈ G 0 /nF and b ≈ T S 0 /nF. Although these terms are
themselves temperature dependent, this is a quite reasonable approximation in many
situations. The process illustrated in Fig. 3 is an open thermodynamic cycle. Thus
when a load is connected, a finite current i = Q̇ flows. Under these circumstances
there will be irreversible changes, and the actual cell voltage will be lower than the
reversible Nernst potential.
1.4 Cell voltage and current
The conduction mechanism in the ionic electrolyte differs fundamentally from that
in the neighbouring electrodes and interconnects (electronic conduction). Therefore
the electrolyte polarises at the edges and a double charge layer is established at the
interfaces. The change in chemical potential across the interface, µj , is balanced
by the opposing change in the electric potential for electrochemical equilibrium,
nFφ. The sum of these terms is referred to as the electrochemical potential.
Figure 4 illustrates the double charge layer in a fuel cell, while Fig. 5 shows the
potential distribution. The Nernst potential is just the sum of the anodic and cathodic
potentials, E=φa +φc . It is not possible to measure the potential, φ, at a single
electrode, though it is possible to measure changes from equilibrium, η = δφ,
Figure 4: Double charge layer.
Figure 5: Electric potential at open circuit and with a current flowing.
when a current is drawn. The double-layer is just a few nanometres thick. There is
no net change in potential, in the absence of an electric current, within the bulk of
the electrolyte. In other words, there are step changes in potential at the anode and
the cathode, but zero change within the electrolyte, as shown in Fig. 5.
Whenever a current, i, is drawn, the voltage drops and the actual cell voltage, V ,
is given by
V = E − ηe − ηa − ηc = iRl , (22)
where the terms ηe , ηa and ηc are electrolytic, anodic, and cathodic losses, respec-
tively. These have often been referred to as overpotentials or polarisations in the
electrochemistry and chemical engineering literature. Rl is the resistance of the
external load.
The potential difference across the electrolyte
ηe = iRe = E − ηa − ηc − V (23)
is sometimes referred to as the Ohmic overpotential. Here Re is the internal resis-

tance of the electrolyte. The resistivity of the electrolyte is highly temperature
dependent, and decreases with increasing temperature. There will also be a simi-
lar (usually smaller) Ohmic loss associated with metallic interconnect(s), and for
convenience that latter will be ignored for now. Thus work is being done by the
electric field on the ions in the electrolyte and interconnects, whereas work is being
done against the field in crossing the electrodes. Although the anodic and cathodic
overpotentials are themselves a function of the current, it is sometimes conve-
nient for numerical purposes to lump all the losses into a single linearised internal
resistance, R̄
ηe + ηa + ηc = iR̄, (24)
where it is understood that R̄ = R̄(i), i.e. the slope of the V-i curve is not constant.
The V-i curve or, more often the voltage-current-density, V-i , curve is the single
most important characteristic to the electrochemical scientist. N.B. The reader will
note that in this text, a ‘dot’ is used to denote a time derivative, and a ‘dash’ to
denote a length derivative. Thus if q has units [J], then q̇ has units [W/m2 ].
While mole fractions are often found in electrochemical texts, in engineering
problems it is much more common to utilize mass fractions, mj , which may be
converted to/from molar fractions, xj , as follows:
Mj
mj = xj , (25)
M
where the mixture molecular weight is just
1
M = x j Mj = mj . (26)
j j Mj
The reactions at the electrode surfaces lead to continuity and species sources/sinks
in the working fluids. The mass fluxes, ṁj (in kg/m2 s), are related to the current
density, i = i/A, by Faraday’s law
Mj i
ṁj = ± , (27)
1000nj F
where Mj is the molecular weight of chemical species j (in g/mol), and n is the
valence. It is trivial to re-write the Nernst equation in terms of mass fractions.
0.5

RT m H 2 m O 1 RT
E = E0 + ln 2
+ ln (Ma ) + 0.4643 + ln Pa , (28)
2F mH2 O 2 4F
where Ma is the mixture molecular weight, eqn (26), for air, and

MH2 O
ln ≈ ln (18) − ln (2) − 12 ln (32) = 0.4643. (29)
MH2 MO0.5
2
In electrochemistry, molar-based units are frequently encountered, but in transport

phenomena, for a variety of reasons, mass-based units are preferred. This duality
is a fact-of-life when dealing with fuel cells. Both have their uses and limitations.
1.5 Activation losses
Electrochemical reactions can only occur at finite rates. For a charged particle to
pass across a double charge layer, it must possess more free energy than a cer-
tain minimum. From an engineering perspective the activation energy is similar
to bulk-to-interphase energy, required for example in multi-phase systems involv-
ing phase-change. Figure 6 illustrates the basic principle for a single-step single-
electron electrochemical reaction A → B. The sharp sign, #, indicates the activated
state. Eyring’s analysis [23] is useful for such rate-limiting situations. The analysis
proceeds as follows: the probability density function for ions crossing the electrode
boundary follows a Boltzmann distribution, and it can readily be shown that the
forward rate constant, kf mol/sec, is given by
kf = k0 exp (−Gf /RT ). (30)
A backward rate constant, kb , is similarly defined. For an open circuit (equilibrium),
kf = kb = k0 . If a current flows, the potential difference deviates from the open
circuit value, and the required free energy will also change as a result of the electric
field:
G = Gf − βFφ, (31)
where β is a symmetry coefficient, 0 ≤ β ≤ 1, which is a measure by which a
change in electric potential affects the required activation energy, see Fig. 6. Under
this condition, the forward rate will not equal the backward rate. The former is
given by,
if = FNA k0 exp (−Gf /RT ) exp (−ηFφ/RT ) = i0 exp (−βFη/RT ), (32)
where xA is the mole fraction of compound A, and M is the mixture molecular
weight. The quantity η = δφ = φ − φ0 is the so-called activation overpo-
tential. Similarly, for the backwards reaction B → A the charge flux of charge is
ib = i0 exp (−(1 − β)Fη/RT ) (33)
Figure 6: Schematic of concept of activation energy for a single-step single-electron

transfer reaction.
and the form of the Butler-Volmer equation, appropriate for a single-step single-
electron transfer process, is obtained,
i = i0 [ exp (−βnFη/RT ) − exp ((1 − β)nFη/RT )]. (34)
The quantity i0 is the exchange current. In general i0 will be a function of molar

concentration. That dependence can be theoretically determined by ensuring com-
patibility with the thermodynamic requirement E = E 0 + (RT /F) log (PB /PA ) for
equilibrium, where PA and PB are partial pressures. Setting if = ib = i0 it follows
that the exchange current density is just i0 = Fk0 PA PB .
1−β β
The two terms in eqn (34) are essentially similar to Arrhenius-type expressions
occurring in chemically reacting systems. The following points are noted: for small
currents, eqn (34) reduces to a linear form, obtained by expanding it in a Taylor series
about η = 0. At higher overpotentials (currents) an exponential form is obtained;
for the latter situation one of the two terms in eqn (34) is negligibly small, depending
on the direction of flux of charged particles. The exponential form is referred to as
a Tafel equation. The ratio i/i0 determines if one is in the linear or Tafel regions.
For a multi-step electrochemical reaction, the Butler-Volmer equation is generally
written,
i = i0 exp (nβf Fη/RT ) − exp (−nβb Fη/RT )
. (35)
The quantities βf and βb are referred to as transfer coefficients. In deriving
eqn (35) for a multi-step reaction [24, 25] it is generally assumed that at any elec-
trode there is only one rate limiting step, with all other steps being considered at
equilibrium. In view of these restrictions, it is probably wise to regard eqn (35) as
being a semi-empirical formulation, for which βf and βb are determined experi-
mentally. Equation (35) may be considered as an implicit form of the relationship
of η = η(i, T ) which must be solved iteratively, for any given i0 and β. Typical
values for planar SOFCs are i0 = 1 − 5 000 A/m2 . In general two sets of eqn (35)
must be solved at both the anode and cathode to yield ηa and ηc for use in eqn (22),
i.e. there are two values of i0 and four of β (N.B. β is often assumed to be ½). The
overpotential at one electrode may be negligible compared to the other. Following
the argument above for a single-step reaction for two electrodes, and imposing the
requirement that for equilibrium, the sum of the potentials must equal the ther-
modynamic Nernst potential, eqn (20), and with the additional assumption that
βf = βb = 1/2, it may be reasonably concluded that the concentration dependence
of the exchange current is i0 ∝ PO0.252
at the cathode and i0 ∝ PH0.5
2
PH0.5
2O
at the anode
for a SOFC.
Simplifying the Butler-Volmer equation has the advantage of rendering an explicit
solution for η(i, T ); however the calculation procedure for the cell/stack is iterative
by nature, and therefore there is little problem in using the implicit form, above.
Charge transfer losses are particularly important at low current densities. The SOFC
is a high-temperature device, and thus activation losses are generally less signifi-
cant than in other fuel cells such as proton exchange membrane fuel cells, due to
the presence of the exponent (RT )−1 . The subject of electrode kinetics, or electrod-
ics, is an important one and the reader interested in more than the brief overview
provided here is referred to specialised texts such as Gileadi [24], Bockris et al.
[25] and Bard and Falkner [26].
1.6 Diffusion losses
Mass transfer losses are due to diffusion becoming a rate limiting factor in providing
O2 and H2 to the cathode and anode, and removing H2 O at the anode. Thus transport
phenomena in the fuel and air channels can and do affect the electrical performance
of the fuel cell. This becomes increasingly important at high current densities where
concentration gradients are high. Mass transfer effects are sometimes handled in
an analogous manner to charge transfer (activation) by introducing a concentration
overpotential, also known as a concentration polarisation. The origin of these terms
is historical. They may be written in terms of either molar or mass fractions. The
latter is adopted here, namely

RT mb
η=± ln (36)
nF mw
for inclusion in eqn (22). The symbols mb and mw denote bulk and wall values
of mass fraction (in the interests of readability the indicial notation mj,b etc., is
dropped) and the sign is positive for products and negative for reactants. There
is no real advantage to defining overpotentials in the form of eqn (22). Rather,
all that is necessary is the recognition that the Nernst equation be based on wall
values, instead of those in the bulk of the fluid. Some authors [3, 27] suggest that
diffusion effects in fuel cells be modelled using so-called equivalent film theory,
introducing a Nernst length scale. That approach is not recommended here. The
suggested approach for mass transfer calculations in SOFCs is outlined in Beale
[28] and follows standard approaches [29–32] to the generalized engineering mass
transfer problem.
In fuel cells, heterogeneous chemical reactions on the electrode surfaces lead to
sources and sinks in the continuity and species (mass fraction) equations. Let it be
supposed that the mass flux, ṁ , is given by
ṁ = gB, (37)
where ṁ = ±Mi /1000nF is negative for O2 and H2 and positive for H2 O, and g is a
mass transfer conductance. For historical reasons the symbol g is used to represent a
generalised conductance, whereas the symbols h (Europe) and α (N. America) are
reserved for heat transfer, with temperature as independent variable. There is clearly
an element of redundancy in this convention.
The driving force, B, is defined by
mb − mw
B= . (38)
mw − m t
By convention, both ṁ and B are positive for injection (H2 O at the anode)
and negative for suction (O2 and H2 at the cathode and anode, respectively),
while the conductance g is always positive. The subscript, t, refers to the so-called
transferred-substance state (T-state) [29, 32, 33]. For multi-component mixtures,
T-state values are given by,
ṁj
mj ,t = , (39)
ṁ

where ṁ = j ṁ j . Thus it can be seen from the stoichiometry of eqn (3) that for
the air-side mt = 1 for O2 , whereas mt = −1/8 for H2 and mt = +9/8 for H2 O on
the fuel-side. Equation (37) may be conveniently rewritten in dimensionless form
g∗
B= b, (40)
g
where g ∗ denotes the value of g in the limit ṁ → 0, and the blowing parameter,
b, is given by
ṁ
b= ∗. (41)
g
Equation (38) may be rearranged to obtain the required form
mw 1 + mmbt B
= . (42)
mb 1+B
Ideally actual mass transfer data for B(b) for the particular channel geome-
try under consideration are available; if not, the solution to the 1-D convection-
diffusion equation,
B = exp (b) − 1 (43)
may be used as an approximation. Figure 7 shows data from numerical simulations
of fully-developed mass transfer in ducts of various geometries [34]. It can be
Figure 7: Mass transfer driving force as a function of blowing parameter, from [28].
seen that data normalized in the form of B vs. b (or equivalent) conforms well
to the curve corresponding to eqn (43). The choice of non-dimensional number is
important; for example use of a wall Reynolds number, Re, or Sherwood number,
Sh, for the abscissa in place of the normalised convection flux, b, will not result in
the data reducing to (roughly) a single curve. Equations (42) and (43) or equivalent,
are sufficient to compute mass transfer losses in the form of eqn (22) in a manner
which is reasonably rigorous while at the same time being reasonably simple to
implement.
Thus knowledge of g ∗ for air-side and fuel-side geometries together with exper-
imental data or an analytical expression for B(b) is sufficient to obtain a reasonable
estimate of diffusion losses in a fuel cell. Some deviations from Fig. 7 are to be
anticipated for a variety of reasons (entry-length considerations, Schmidt number
variations, etc.). If experimental data are available, these should always be used. It
is to be noted that for many situations the mass flow rates are small, and under the
circumstances g = g ∗ and B = b. Values of g ∗ may be obtained from the zero-mass
flux Sherwood number, Sh∗ ,
g ∗ Dh
Sh∗ = . (44)

Note that diffusion control is important in fuel cells at high current densities/mass
fluxes, not in the limit Sh → Sh∗ . Figure 8 shows values of Sh∗ for fully-developed
mass flow and scalar transport in rectangular ducts of various aspect ratios, as given
in Table 43 of ref. [35]. It is to be noted that suction (e.g. at the cathode) reduces
mass transfer, i.e. g/g ∗ < 1, whereas blowing (e.g. at the anode) has the opposite
effect g/g ∗ > 1. Knowledge of the T-state value is also useful when performing
detailed CFD simulations. These may be used to prescribe boundary conditions
in the species equations (see Section 2.5). Although this introductory analysis is
Figure 8: Sh∗ (Nu∗ ) for rectangular ducts in the limit of zero blowing. From [34],
data of Schmidt [35].
Figure 9: Schematic representation of a SOFC with hydrogen as fuel.
appropriate for mass transfer in the gas channels, Section 2.9 shows that it may be
readily extended to consideration of more complex geometries involving combined
mass transfer in channels and porous diffusion layers. Section 2.11 discusses some
further considerations for mass transfer analysis in fuel cells.
1.7 Basic computational algorithm
We are now in a position to construct a simple algorithm for computation of the

performance for a fuel cell. Assume the following are known: cell voltage, V ,
temperature, T , air pressure Pa , inlet flow rates for air and fuel, ṁa,in , ṁf ,in , and
mass fractions of the component species in the air, mO2 ,in , mN2 ,in , and the fuel,
mH2 ,in , mH2 O,in , respectively.
1. Guess an initial value for the current i.
2. Calculate mass sources/sinks, ṡO2 , ṡH2 , ṡH2 O , from Faraday’s law. Compute
air and fuel outlet mass flow rates, ṁa,out , ṁf ,out , and mass fractions, mO2 ,out ,
mH2 ,out , mH2 O,out , mN2 ,out .
3. Calculate average of inlet and outlet bulk mass fractions and then compute wall
mass fractions, using eqn (42).
4. Compute the anode and cathode overpotentials, ηa , ηc , from the Butler-Volmer
equation eqn (35), based on wall values.
5. Compute the Nernst potential, E, from eqn (20), based on wall values.
6. Compute the current, i, from
E − ηa − ηc − V
i= . (45)
Re
7. Repeat steps 2–6 until convergence is obtained.
This simple procedure will form the basis for subsequent more complex proce-
dures described further below. The continuity relationships in step (5) are easily
computed. For example; ṁa,out = ṁa,in + ṡO2 and ṁf ,out = ṁf ,in + ṡH2 + ṡH2 O
while the mass fraction equations are just mk,out = mk,in + ṡk /(ṡk + ṁk,in ), where
k = O2 , H2 , H2 O, as appropriate.
One important note regards the requirement of prescribed current vs. prescribed
voltage. Traditionally physicists work with prescribed voltage, and electrochemists
with prescribed current, i (or current density, i ). If it is required that the galvano-
static condition, as opposed to the potentiostatic condition, be prescribed, then it
is necessary to adjust the prescribed voltage until the required current is obtained.
This is straightforward and two methods for achieving the same end are discussed
in Section 2.7.
2 Computer schemes
The numerical scheme described above is of limited practical utility, though it does
set the stage for what follows. Numerical schemes vary from very simple schemes
requiring a few seconds to run on a personal computer, to complex models which
require large computers and long run times. The latter might be used for detailed
cell design whereas the former could be employed in a system model including
balance-of-plant or for real-time control.
If a SOFC or stack is well-designed, the inlet flow streams will be uniform.
However, there may be very substantial variations of mass fraction, temperature,
and current density across the fuel cell. Also the flow velocities along the cell
passages vary due to mass sources/sinks, and as a result, the pressure may also
change in a non-linear fashion. Thus there are many circumstances where it is
necessary to solve only the continuity and scalar transport equations, employing the
previously-described algorithm, and not involve the pressure-corrected momentum
equations in the solution. In other situations, for example when investigating a
particular configuration of flow channels, solutions to the momentum equations are
required.
It is important to distinguish between situations where the domain is tessellated
so finely that diffusive components (viscous effects, heat conduction, and gas dif-
fusion) are captured, and situations where it is necessary to invoke a rate equation
for the diffusion flux of some general field variable, .

∂
− = g. (46)
∂y w
The local gradient of at the wall is thus replaced by the bulk-to-wall (or in some
cases bulk-to-bulk) difference, , thus invoking the concept of a conductance,
g, discussed in detail above in Section 1.6. This notion is central to the classical
subject of convective heat and mass transfer [36].
In general, when discretising a SOFC, a mesh will be constructed. It may be
structured or unstructured, and it may or may not be boundary fitted (so that each
cell corresponds to a given fluid/solid material). For detailed simulations where
the diffusion terms are evaluated directly, a fine mesh, concentrated in the near-
wall fluid boundaries, will be required as shown in Fig. 10(a). If a rate-equation
assumption is invoked, a coarse body-fitted mesh, (b), may be employed, or (c) the
mesh may be arbitrary corresponding to local volume averaging. In this case no
Figure 10: Some possible grids used to discretise a planar SOFC (a) Detailed
boundary-fitted (b) Boundary-fitted rate-based formulation (c) Volume-
averaged (non-boundary-fitted).
spatial distinction is made between the different materials in the fuel cell. All of
these schemes have their advantages and disadvantages. For the latter case storage
allocation may be required for more than one phase at each computational cell,
whereas for case (b) it is necessary to keep track of all material properties and
inter-phase coefficients (conductances), which may be numerous. For case (a), the
grid may be so large as to render calculations intractable for large SOFC stacks.
Thus there is a trade-off between detail and speed. The code designer is obliged
to make some decisions at the outset regarding the nature of the mesh and the
corresponding numerical scheme.
Let it be supposed that the SOFC is composed of four distinct media: (1) air,
(2) fuel, (3) electrolyte, and (4) interconnect. Let it further be assumed that
the electrodes are sufficiently thin that they may be considered as forming the
electrolyte surfaces. We shall relax these assumptions in due course. In the next
section, we shall consider how to solve such a system using a finite-volume
procedure [37].
2.1 General scalar equation
Let is be proposed that a prototype equation having the form
∂(ρj εj j )
+ ∇ · ρεj u j j = εj αjk (k − j ) + ∇ · (εj ∇j ) + εj ṡ (47)
∂t
j
(i) (ii) (iii) (iv) (v)

be adopted, where is a generic scalar (continuity, mass fraction, enthalpy, etc.)

α is a volumetric inter-phase transfer coefficient and is an exchange coeffi-
cient. Let the terms in eqn (47) be referred to as, from left to right: (i) transient,
(ii) convection, (iii) inter-phase transfer, (iv) diffusion or within phase transfer,
and (v) source.
If the volume fractions, εj , are constant they make no contribution to the overall
balance, and can be eliminated prior to integrating the equations. Moreover, if a
boundary-fitted mesh is employed, εj = 0 for all but one ‘phase’ for which εj = 1.
For simplicity they will therefore be removed in the analysis. Generally-speaking,
terms (iii) and (iv) are mutually exclusive. Also, although (iii) is listed here as a
distinct term, it is almost always coded in the form of a linearised source term (v).
For steady co-flow, counter-flow and cross-flow, the equations simplify further and
in some situations, reduce to ordinary differential equations.
Equation (47) may readily be converted to linear algebraic equations having the
form,
aW (W − P ) + aE (E − P ) + aS (S − P ) + aN (N − P )

+ CP (VP − P ) + SP = 0, (48)
where in the interests of brevity, the steady 2-D form has been adopted. Equa-
tion (48) is the form appropriate to the finite-volume method. The compass notation
E = East, W = West, S = South, N = North, T = Previous time step, etc. has
been adopted [37]. The terms SP and CP (VP − P ) are referred to as fixed source
and linearised source terms, respectively.
2.2 Continuity
It is usual to solve the phase continuity equations

∂ρj
· ρj u j = ṁ
+∇ j . (49)
∂t
The source terms in the air and fuel passages are due to the electrochemical
reactions at the electrodes, namely:

ṁj = ± Mk i /1000nF, (50)
k
where Mk is the molecular weight of chemical element k (O2 , H2 , H2 O) in fluid

j, (air or fuel). Since the continuum equations are integrated, it is equivalent to
prescribe source terms as being per unit volume or per unit area.
2.3 Momentum
The decision to solve the momentum equations depends on the problem at hand.
For single SOFCs where it is known, a priori, that the inlet flows are uniform
and steady, there is little point to adding additional complexity to the code. For a
3-D SOFC stack problem, or for a single cell which has yet to be designed, the flow
may vary substantially, and it is therefore necessary to solve a generic equation of
the form:
∂(ρj u j )
j − Fj εj u j + ∇
· (ρj u j ; u j ) = −∇P
+∇ · (µ∇
u j ), (51)
∂t
where the subscript j refers to either air or fuel as distinct phases. Since the phases do
not intermingle, the inter-phase terms are of the form Fj εj (0 − u j ). The second term
on the right-side of eqn (51) is normally absent for a detailed numerical analysis,
Fig. 10(a). Conversely, for cases Figs 10(b) and (c), it is assumed that if transient
and inertial effects were negligible, the overall pressure drop would be entirely due
to fluid drag or resistance,
j = −Fj U
∇P j = −εj Fj u j , (52)
where U is a local bulk superficial velocity, u is a local bulk interstitial velocity,

and the quantity Fj has been referred to as a ‘distributed resistance’[38], which for
a homogeneous porous media of permeability, kDarcy , has the significance
Fj = µ/kDarcy . (53)
Generally-speaking the dependent variable for case Fig. 10(a) is the actual local
velocity, u ; for (b) u is the local bulk interstitial velocity, and for (c) it may be
chosen to be either u or U. For fully-developed laminar duct flows with negligible
mass transfer,
τw C
f∗= 1 = . (54)
2 Re
2 ρu
Shah and London [35] suggest that for rectangular ducts, dimensions L × H :
C = 24(1−1.3553α∗ +1.9467α∗2 −1.7012α∗3 +0.9564α∗4 −0.2537α∗5 ), (55)
where α∗ = H /L or L/H , whichever is a minimum. The Reynolds number, Re =

Dh ρu/µ, is based on a hydraulic diameter Dh = 4A/P, where A is the flow area
and P is the wetted perimeter. It can readily be shown that
2C µ
Fj = . (56)
ε Dh2
For a detailed numerical simulation, Fig. 10(a), knowledge of eqns (52) to (56)
is irrelevant. The drag is directly obtained from the viscous term in eqn (51). Con-
versely, for a rate-based formulation, the viscous term is discarded in favour of
eqns (52) to (56). In reality, F will also vary due to mass transfer effects, f/f ∗ = 1, as
τ = µ|∂u/∂y| is a function of the shape of the velocity profile which depends on the
rate of mass transfer. This profile is not necessarily similar to the scalar mass fraction
profile [39]; eqn (43) or Fig. 7 cannot be used to correct F. Detailed experimental
or numerical data of f/f ∗ as a function of the momentum blowing parameter are

required for the particular geometry under consideration. At present, this consider-
ation is often simply ignored. Reference [18] contains a comparison of a distributed
resistance formulation with a detailed CFD methodology for a SOFC stack includ-
ing heat and mass transfer effects.
2.4 Heat transfer
Equation (47) is considered appropriate for the analysis of heat transfer, with the
solved-for variable generally being written in terms of temperature (not enthalpy
or internal energy) as

∂ ρj cj Tj
+ ∇¯ · (ρcj u j Tj ) = · (k ∇T
αjk (Tk − Tj ) + ∇ j) + ∇
· q̇ + q̇ . (57)
∂t
j
A volumetric heat source occurs in the electrolyte due to irreversible Joule heat-
ing. This is given by
i (E − V )
q̇ = , (58)
He
where He is the thickness of the electrolyte. For thin electrodes the energy of
activation at the electrodes may be prescribed as per unit area with a magnitude
q̇ = i η (59)
normal to the electrode surface. An additional heat source is due to the entropy
change as discussed in Section 1.3 and can be expressed as
i
q̇ = (G0 − H 0 ). (60)
2F
Technically this should be split into anodic and cathodic components. For SOFCs
this source term is usually prescribed at the anode. Convective sources exist also
by virtue of mass transfer at the walls and these must also be accounted for (see
below). The reader will note that there is inter-phase heat transfer between solids
and fluids. At the same time there is within-phase metallic conduction, for example
in the metallic interconnects. This is one case where both terms (iii) or (iv) in
eqn (47) are present. The volumetric heat transfer coefficients are computed as
αV = Ū A, (61)
where A is the area for heat transfer, V is cell volume, and Ū is an overall heat
transfer coefficient, obtained using harmonic averaging, for example;
1 1 H
= + , (62)
Ū A hA kAS cond
Figure 11: Shape factor for square ducts for the case, L/H = Lchannel /Hchannel = 2.
where Scond is a conduction shape factor. (N.B. the standard definition in heat
transfer texts includes the term H /A, where A is the maximum surface area for heat
transfer and H is the thickness of the solid region). Figure 11 shows an example of
Scond for square geometry. These data were obtained by the present author by means
of numerical simulation. The fuel cell designer should obtain Scond from a numerical
simulation for the Laplacian system, ∇ 2 φ = 0, for the actual geometry under
consideration. Fin theory may be employed, if there are significant temperature
variations within the interconnects. Conduction shape factors may also be used
when considering the potential distribution in the interconnects where the electric
resistance is given by
H
r= . (63)
σAScond
It is to be noted that for both electric and thermal conduction, the shape function
will vary as a function of the direction of the local current density or heat flux
vector. It is thus an engineering approximation.
Temperature variations in fuel cells are inevitable, even when the heat sources
are entirely uniform (corresponding to uniform current density and resistance).
Figure 12 shows typical simulated temperature variations for a cross-flow SOFC;
Achenbach [10] compares the performance for co-flow and counter-flow as
well. Temperature variations impact in a number of ways; in particular the elec-
trolyte resistance, re , is extremely temperature dependent. This in turn influences the
current density and hence the internal power dissipation. Temperature also affects
the local Nernst potential, E0 . Generally speaking lower temperatures are required
for mechanical stability, however if methane is used as fuel, higher temperatures
are needed to assist the reforming reaction. The thermally conducting interconnect
is beneficial; were it not present, the temperature gradients in Fig. 12 would be
much greater.
Figure 12: Temperature variation, deg. C, in a SOFC for cross-flow.
2.5 Mass transfer
Treatment for mass transfer within the gas phases is particularly simple, namely:
∂(ρmk ) · (ρ uj mk ) = ∇ · (∇mk ) + ∇
· j ,
+∇ k (64)
∂t
where the diffusion term is absent for cases Fig. 10(b)(c). The reader will note that
there are no inter-phase mass transfer terms. Mass sources/sinks at the wall are of
the form
jk = ṁ mk ,t . (65)
Thus mass transfer values are just the T-state values. The wall values, needed
for the Nernst equation, are obtained directly by the calculation procedure, for case
Fig. 10(a). For cases Fig. 10(b) and (c) these must be computed using eqns (42)–
(43), or equivalent, from the local bulk mass fractions. This is particularly simple
with the mass-based formulation, given above.
2.6 Numerical integration schemes
The generalised coupled system of equations is quite complex, and we shall illus-
trate by considering the simplified case of the energy equation for the case of steady
2-D cross-flow, for which numerous terms may be eliminated. Specifically, in the
fluids the inter-phase terms (Fig. 10(a)), and/or diffusion terms (Figs 10(b) and (c))
are absent or negligible; in the solids terms (i) and (ii) are absent. Thus we can write
eqn (47) for the air and electrolyte in the following reduced forms,
dTa
(ρuc)a = αae (Te − Ta ) + αai (Ti − Ta ), (66)
dx
· ke ∇T
0=∇ e + αae (Ta − Te ) + αfe (Tf − Te ) + q̇ , (67)
where the subscripts a, f , e, i refer to air, fuel, electrolyte, and interconnect phases.
Similar expressions may easily be written for the fuel and interconnect. These may
readily be discretised as follows:
∗
ċa (TaW − TaP ) + αae V (TeP − TaP ) + αai V (TfP − TaP ) + aF (TaP − TaP ) = 0,
(68)
dew (TeW − TeP ) + dee (TeE − TeP ) + des (TeS − TeP ) + den (TeN − TeP )
∗
+ αae V (TaP − TeP ) + αfe V (TfP − TeP ) + aF (TeP − TeP ) + q̇P = 0, (69)
where ċ = ṁc = ρuAc and de = kA/|P − E| etc., are convection and diffusion
terms, respectively. This is a set of simplified 2-D finite-volume equations (eqn (48))
with = T . For air: convection dominates and the linear coefficients are aW =
ċa , aE = aS = aN = 0; the inter-phase terms may be linearised with coefficients
Ce = αae V , Ci = αai V , and values Ve = TeP , Vi = TiP , and there are no internal
sources of heat S = 0. For the electrolyte: the linkages are diffusive, aE = dew ,
etc., and the inter-phase terms are as above, namely Ca = αae V , Cf = αfe V , and
Va = TaP , Vf = tfP . This time, however, there is a heat source S = Q̇P . Thus
by careful consideration of the governing equations, it is possible to effect great
savings in computer resources, when generating source code in-house. The symbol
TP∗ denotes the value of TP at the previous iteration, and aF is an inertial or false
time step coefficient. This mechanism is used to relax the scheme in order to procure
convergence.
2.7 Iterative procedure
The calculation proceeds precisely as in Section 1.7, except that calculations are
performed on a per-unit-cell basis rather than at the mean of the inlet and outlet
values. A local Nernst potential may be computed:
E(x, y) = V + i r + ηa + ηc = V + ri i , (70)
where ri is a local internal resistance. Average values of Nernst potential, Ē, current
density, i¯ , electrolyte resistance, r̄ e etc. are obtained by summation.
The galvanostatic (constant current) condition is usually implemented for stack
models to ensure overall charge conservation. When the overall current or mean
current density, ī , is prescribed, some form of ‘voltage correction’, V , is required.
For example a correction may be applied in the form
V = −ri ī , (71)
where V = V ∗ + V is the desired cell voltage, and V ∗ is the value of V at the

previous iteration; similarly ī = ī + ī∗ is the desired value and ī∗ is the previous
computed value. Thus the mean current density, ī ∗ , is computed at the end of
each iterative cycle compared to the desired value ī, and the voltage is corrected
accordingly. r̄i need not be exact; any reasonable value will procure conversion.
An alternative approach taken in many codes is to directly set
V = Ē − r̄i ī . (72)
If eqn (72) is employed, r̄i must be carefully computed using numerical inte-
gration over the surface of the SOFC. Because very small changes in V can effect
large changes in i , relaxation is generally employed. We have used both eqns (71)
and (72) with success in our numerical codes and schemes. The reader will note
that if the full potential field is solved-for as a scalar variable, as discussed below in
Section 2.10, there is no need to adjust the voltage. Another adjustment often made
while performing fuel cell calculations is to change the inlet flow rates until partic-
ular values of the utilisation for fuel and air are obtained. This is straightforward.
2.8 Additional chemistry and electrochemistry
One advantage of the SOFC is that CO and CH4 may be used as fuels in addition to
pure H2 . This introduces additional kinetics and further complicates mass transfer
calculations. If CO is employed as fuel, the anode surface reaction is
CO + O2− → CO2 + 2e− . (73)
If both CO and H2 are present, the two reactions at the anode are considered to
be in parallel, but the net anode and cathode reactions are in series, so that
i = iO

2

= iH 2

+ iCO . (74)
The Nernst potentials are not the same, ECO/O2 = EH2 /O2 , however charge trans-
fer (activation) and mass transfer mechanisms will ensure that the two parallel reac-
and i ) adjust until a single cell potential, V , is obtained, such that
tion rates (iH 2 CO
V = ECO/O2 − ηCO,a − ηc − i R = EH2 /O2 − ηH2 ,a − ηc − i R, (75)

where ηCO,a and ηH2 ,a are evaluated iteratively so that ECO/O2 − ηCO,a = EH2 /O2 −
ηH2 ,a where ηCO,a is a function of iCO , and η
H2 ,a is a function of iH2 such that
eqn (74) is identically satisfied.Although this adds complexity, nothing fundamental
has changed in the basic algorithm. No further assumption need be made regarding
the reaction kinetics, other than estimates for i0 and β for each anodic reaction, as
discussed in Section 1.5.
The CO and H2 may be produced by reforming methane
CH4 + H2 O → CO + 3H2 . (76)
The rate of reaction is considered to be kinetically controlled [10]. The water-gas

shift-reaction is
H2 O + CO → CO2 + H2 . (77)
This process is generally considered to be thermodynamically controlled.
Figure 13: Detail of gas diffusion layer.
2.9 Porous media flow
Up to the present point, the electrodes have been assumed to be sufficiently thin as to
be treated as 2-D objects. There are many situations where this is not true, and these
should be treated as electro-chemically reacting porous media. Moreover, SOFCs
are usually designed with gas diffusion layers, in the form of porous media, as
illustrated in Fig. 13. Lchannel and Lrib are the length of the gas channel and the
solid ‘rib’. Under the circumstances, Darcy’s law may be used to compute the flow
field, as detailed in Section 2.3. Here is a situation where the volume fraction, ε,
should be retained in eqn (47). Reynolds number based on a pore diameter is used
to evaluate whether the Forcheimer-modified form [40] is required to account for
inertial effects. Many codes compute within-phase mass transfer by assuming an
effective diffusion coefficient

eff = , (78)
τ
where τ is a ‘tortuosity’ factor, which is a measure of the mean interstitial path
length of the flow lanes per unit overall length. In general, it is to be noted that the
rib design of SOFCs is one of competing interests [41]: a high ratio of Lchannel /Lrib ,
will increase Ohmic losses in the narrow ribs, whereas a low aspect ratio may result
in non-uniform concentrations along the widely spaced gas diffusion layer. This is,
of course, a function of the diffusivity (and permeability) of the gas diffusion layer
as well as the current density.
Recent work [34] has shown that the analysis developed in Section 1.6 may be
extended to the combined or conjugate problem of simultaneous mass transfer in
both the channels and diffusion layers, provided the driving force, b, is based on the
overall or average conductance for the entire geometry, ḡ ∗ , obtained by harmonic
averaging of the combined conductances of the gas channel and diffusion layer.
The relative magnitude of mass transfer in the channels and diffusion layers is
determined not only by the values Sh for the gas channel and diffusion layer, but
also by the ratio eff /.
It is to be noted that within the electrodes the mean bulk mass fractions must
differ from the mean wall values for a current to flow. Since it is not feasible to dis-
cretise the actual porous matrix for any practical geometry, the rate of mass transfer
will need to be estimated, again using the methodology outlined in Section 1.6, in
order that the Nernst equation be correctly prescribed. This will in general require
knowledge of Sh∗ , for the particular porous media employed. Such data are rarely
available. However bulk-to-wall mass transfer will be important at high current
densities when the so-called ‘diffusion limit’ is reached, and should therefore be
considered in future codes and methodologies.
Heat transfer in porous media is often analyzed by assuming thermal equilibrium
between the two phases, with algebraic averaging employed to compute an effective
thermal conductivity
keff = εkg + (1 − ε)ks , (79)
where kg and ks are the gas-phase and solid-phase conductivities of the medium.
When modeling fluid flow, heat and mass transfer in porous media, certain clo-
sure assumptions are made by virtue of the local volume-averaging procedure.
These closures are quite reasonable and may be considered reliable. The question
remains: is it possible to lump the analyses to develop a simpler, less computation-
ally complex model? The answer would appear to be in the affirmative: certainly
the techniques applied to heat and mass transfer predictions may readily be applied
to the electric field, with volumetric sources and sinks of potential being applied
across phase boundaries. Moreover, mass transfer in gas diffusion layers may be
amenable to closed form analytic solutions.
2.10 Current and voltage distribution
For planar geometries it is generally sufficient to compute the potential field using
the Nernst equation (eqn (20)). This is a local 1-D model. Moreover it is often
sufficient to neglect Ohmic losses in the metallic interconnects. We shall now
relax these assumptions. The Nernst potential may be considered to be the sum
of the potential differences across the anode and cathode, E = φa + φc ,
where
RTf
φa =φa0 + ln xH2 − ln xH2 O − ηa , (80)
2F
RTa
φc =φc0 − ln Pa + ln xO2 − ηc , (81)
4F
where Tf and Ta are the fuel-side and air-side temperatures (not necessarily the
same). As has been discussed, there is an element of arbitrariness in the choice of
φa0 and φc0 .
A number of codes now involve the solution of the Poisson equations for the
electric field over the entire region (both ionic and electronic)
· (σ ∇φ)
∇ = ṡ , (82)
where σ is the ionic conductivity in the electrolyte or the electronic conductivity

in the interconnects. Equation (82) is to be considered a statement of the principle
of conservation of charge, i = −σ ∇φ. This is a simplified version of the scalar
equation, eqn (47), and presents no difficulty.
If the electrodes are sufficiently thin, source terms prescribed according to eqn (80)
and (81) account for step changes in potential across the anode and cathode. These
may be applied as sources of equal and opposite magnitude on either side of the
electrode boundary (a ‘planar’ dipole). Elsewhere in the interconnect and the elec-
trolyte ṡ is zero, i.e., the Poisson system reduces to Laplace’s equation. Since this
is a conjugate diffusion problem, harmonic averaging of the conductance, σ, should
be employed. When the electrodes are of finite thickness, some care is required as
electronic and ionic conduction zones overlap. Under the circumstances, storage is
generally assigned for two different scalar potentials. The principle of superposition
can be applied to the electric field potential, and other definitions of potential may
be encountered.
A variety of external boundary conditions may be applied to obtain the solution
for the electric field potential. For example, (i) two voltages, V1 , and V2 may be
applied at the boundaries of the interconnects (Dirichlet problem), alternatively
(ii) a uniform current density may be prescribed at both boundaries (Von Neumann
problem), with a single point within the assembly fixed to a reference potential,
or (iii) the current density may be fixed at one interconnect and the voltage at the
other (mixed boundary value problem).
Figure 14 shows flux lines of current density computed from the electric potential.
In this case the field is quite uniform, though in situations where mass fractions
of the products/reactants vary substantially at high i , the iso-galvanic lines can
be highly non-uniform. The additional effort required to obtain a solution for the
electric potential should only be undertaken if significant non-uniformities in the
current distribution are anticipated. This approach is generally combined with an
analysis of flow and mass transfer in the porous gas diffusion layers on either side
of the electrodes.
Figure 14: Particle traces along galvanostatic lines. From [34].

It is worthy of note that the Poisson system, eqn (82), can be considered to be a
simplified version of the Nernst-Planck equation,
i = 2F (ρ u − ∇m)
− (σ ∇φ),
(83)
M
where for convenience we assume the presence of only a single charge carrier, O2− ,
in the electrolyte. Diffusion and convection are generally considered subordinate
to migration in the SOFC electrolyte.
2.11 Advanced diffusion models
The development derived in Section 1.6 is appropriate to so-called ‘ordinary’

diffusion, for which the diffusion flux is given by Fick’s law:

j = −ρD∇m. (84)
Such an analysis is appropriate [42] for non-dilute, multi-component mixtures

only if

D12 = D13 = · · · Djk = D = . (85)
ρ
Kays et al. [31] suggest that this is indeed approximately the case for high
temperature gases, as occur in SOFCs. If it is not the case, the simplest procedure
is to calculate an effective diffusion component according to Wilkes [43] as
1 − xk
Dk ,eff = xj (86)
j =k Djk
which is exact, only for dilute mixtures in a concentrated carrier fluid. A more
complete analysis involves the solution of the Stefan-Maxwell equations [44] for
the diffusion fluxes
2
n
mk jj − mj jk
kM ) = M
∇(m . (87)
ρ Mj Djk
j=1
The set of eqns (87) may be regarded as implicit expressions for the diffusion
fluxes, jk . These may then be substituted into eqn (64) in the usual fashion, to
obtain the mass fractions, mk . Typically, the linear coefficients in the finite-volume
equation (eqn (48)) are caste in the same form as for ‘ordinary’ diffusion and an
additional source term added [45]. This has the effect of preserving the correct form
of non-linearity in the transport equations, for the case of binary diffusion. If the
pore size is sufficiently small that it approaches the mean free path of the colliding
particles, Knudsen diffusion must also be considered. Finally surface diffusion may
also occur. Multi-component diffusion in reacting porous media is a highly complex
subject, considered beyond the scope of this brief introduction.
2.12 Thermal radiation
A SOFC is a high temperature device, and even though temperature differences

within the passages may be small, they may nonetheless be such that radiative
heat transfer is present. The relative importance of radiation in SOFCs is currently
the subject of debate. Radiation may be important within the gas channels, within
the electrolyte-electrode assembly, and also when thermal insulation layers are
employed. Reference [46] reviews some of these issues.
Surface-to-surface radiation in the micro-channels may be accounted-for by the
network method [47],
σTk4 − q0k q0j − q0k

N
q̇k = = , (88)
(1 − εk )/εk Ak 1/Fk−j Ak
j=1
where εk and Ak are the emissivity and area of the kth element, q0k is the radiosity
(the total of emitted and reflected radiation), Fk−j is a configuration factor [48],
and σ = 5.67 × 10−8 W · m 2 K4 is the Stefan-Boltzmann constant. The terms
(1 − εk )/εk Ak and 1/Fk−j Ak in the denominators are referred to as ‘surface’ and
‘space’ resistances respectively. By the construction a network of resistances, the
unknown radiosities may be eliminated and the system of equations solved. Since
the micro-channels are very narrow, only a few immediate neighbours (opposite
and sides) in any given channel will be of any consequence, and the computation
can then be much simplified. The additional work involved in computing surface-
to-surface radiation along the length of the channel is probably not justified. Yakabe
et al. [49] describe a procedure for calculating surface-to-surface radiation in gas
micro-channels, which differed from the standard analysis, eqn (88). VanderSteen
et al. [50] suggest that surface-to-surface radiation could affect the temperature by
as much as 25–50 K.
Although the gas channels are amenable to a simplified analysis, the electrolyte,
and possibly the electrodes of a SOFC constitute radiatively participating media.
Murthy and Fedorov [51] suggest the exclusion of radiation effects in the partic-
ipating electrolyte could result in the over-prediction of temperatures by as much
as 100–200K for the case of an electrolyte-supported SOFC. By contrast, refs. [52,
53] suggest very modest differences in electrolyte temperature; less than 1 K. These
disparities indicate that the need to include radiation effects is, at present, still a
moot point. Reference [51] suggests that the electrolyte may be treated as being
optically thin, while the electrodes may be considered as optically thick. A very
limited amount of experimental data has been gathered in support of this, to date;
however [52, 53] suggest that in fact, the electrodes are so optically thick as to be
entirely opaque for all practical purposes.
The general radiation heat transfer problem in a participating medium involves
heat transfer ‘at a distance’; described mathematically by an integro-differential
equation, referred to as the radiative transfer equation (RTE). For a grey
isotropically-scattering medium, this may be written in the simplified form,

!4π
di eb σs
= −(a + σs ) i + a + i d ω. (89)
ds π 4π
ω=0
(i) (ii) (iii) (iv)

In this section only, the symbol i denotes radiant intensity (not electrical current),
a and σs are absorption and scattering coefficients, s is displacement, ω is solid
angle, and eb = σT 4 is black-body emissive power. The terms in eqn (89) represent
(i) changes in intensity, i, due to: (ii) absorption and out-scattering, (iii) emission,
and (iv) in-scattering, respectively. Detailed numerical solutions to the RTE are
typically conducted using a discrete ordinate method or a Monte Carlo method.
The former is more popular, at the present time, but suffers from ‘ray effects’
which can lead to numerical errors. The RTE solution results in a field of values for
the radiation intensity from which a radiant flux vector, q̇ , can be calculated for
substitution in the energy equation. However, both the discrete ordinate, and Monte
Carlo methods are very compute-intensive. Therefore some additional engineering
assumptions are required if it is possible to invoke simpler radiation models within
multi-channel fuel cells and stacks.
For optically thin regions a multi-flux model may be considered, for which the
1-D two-flux model is the simplest. It is referred to, in the literature, as the Schuster-
Schwarzschild approximation. If it is assumed that the intensity is composed of two
homogeneous components such that q̇+ = πi+ and q̇− = πi− in the +x and −x
directions, then it can readily be shown that eqn (89) may be written as,

d 1 d q̄
= 4a q̄ − eb , (90)
dx a + σs dx

where q̄ = q̇+ + q̇− /2, and for which (the x-component of ) the radiant flux
vector, q̇ = q̇+ − q̇− , may be computed as
1 d q̄
q̇ = − . (91)
a + σs dx
Equation (90) is a diffusion-source equation, and is thus suitable for discretisation
in the form eqn (48). It has been presented in the form developed by Spalding [54],
from which it differs slightly mathematically, being consistent with the formulation
in Siegel and Howell [48]. A six-flux model may similarly be constructed in terms
of three field variables, q̄x , q̄y , q̄z in 3-D, and has the advantage of being in the
same form as the general scalar equation (eqn (47)). However for thin plane layers,
a two-flux model may be sufficient. The discrete ordinate method may be regarded
as a generalised multi-flux method [48].
For optically thick regions, a diffusion approximation is considered appropriate,
with radiative conductivity defined by,
16n2 σT 3
kR = , (92)
3a
a is a (Rosseland mean) absorption coefficient. The term, kR , may simply be added

to the normal conduction term, k, in eqn (57), i.e., keff = k +kR . There are however
difficulties associated with use of the diffusion approximation near boundaries. This
may be remedied by the use of a temperature modified slip boundary condition [48].
Thermal radiation is one mechanism by which it may be possible to actually
control stack temperature and temperature gradients, by manipulating the optical
properties of the component materials or employing radiation shields. Spinnler et al.
[55, 56] propose the use of highly-reflective thermal insulation to control fuel cell
temperatures by minimising losses to the environment.
The above analysis adds considerably to the complexity of the calculation
procedure, even so for situations where radiation is important; it is probably over-
simplistic: In reality, the electrolyte assembly is a porous media with non-
homogeneous optical properties, including dependent scattering, while the
interconnect is typically metallic and hence there may be both specular and diffuse
components to the surface radiosity. The spectral nature of the radiative properties
of the SOFC components also need to be determined. The subject has received little
attention to-date and should be considered a topic for future research by scientists
and engineers seeking to design high-temperature fuel cells.
The four outstanding technical issues in numerical analysis of themo-fluids are;
chemical kinetics multi-phase flow, radiative heat transfer and turbulence. For
SOFCs, the chemical kinetics is a complex issue that needs to be addressed; multi-
phase flow is superficially not a problem, but the presence of porous media and
the need for volume-averaging raises many of the same issues. Thermal radiation
may be a matter for concern, as discussed above. Turbulence is not generally gen-
erated at the Reynolds numbers encountered within the passages of planar SOFCs,
however turbulence may be encountered within the inlets and outlets and manifold
passages of SOFC stacks.
3 Stack models
SOFCs are stacked to increase the working voltage to a reasonable value. Fuel and
air are typically introduced via manifolds (risers and downcomers). Various design
configurations are possible. The main concerns of the stack designer are uniformity
of flow and pressure gradients. If SOFCs were not operated in stacks, it probably
would be feasible to conduct detailed numerical analysis using CFD codes without
invoking a rate equation assumption, at least for simple geometries. The presence
of numerous fuel cells, each with multiple channels, renders this situation unlikely
for large stacks for the foreseeable future.
A rough estimate of the relative magnitude of parasitic losses in the cells and
manifolds can be made by assuming the flow is uniform, then analysing to see if
this is a reasonable assumption. It may readily be shown from eqns (54) and (55), that
for planar passages neglecting injection/suction that the pressure drop across the
stack is just,
48Lµū 48LµQ̇
P̄ = = , (93)
H2 nfc WH 3
Figure 15: Pressure (Pa) in a 50-cell SOFC stack assembly. From [57].
where nfc is the number of fuel cells in the stack, 2H is the height of the (air
or fuel) channels, W is the width of the cell, L is the length of the fuel cells in the
flow direction, and Q̇ = nfc BH ū is the volumetric discharge, assuming uniform flow
conditions. A similar, though more complex, expression may be readily derived for
rectangular passages. The basic principle of stack design is that the losses across the
cells should be large in comparison to those in the inlet manifolds. Large changes in
pressure drop are effected by changing the height of the passages, since p̄ ∝ 1/H 3 .
Equation (93) is to be considered only approximate. Mass transfer will increase
or decrease pressure for the cases of injection and suction, respectively. This purely
inertial effect is in addition to the frictional f/f ∗ effects discussed earlier (recall
Bernoulli’s law P + 12 ρu2 = constant). In the manifolds suction/blowing leads to
all fluid being entirely evacuated over the stack height, H , and the latter must be
considered. Berman [58] was the first to derive an analytical solution for the pres-
sure distribution for a plane duct with mass transfer at both walls; analyses for
suction/blowing at only one wall also exist [59, 60]. For rectangular geometry it
is simple to utilize a numerical calculation (CFD) procedure. Figure 15 shows the
pressure distribution in a 50-cell stack assembly; it can be seen that (1) in the inlet
manifold the pressure gradient decreases due to suction, (2) there is relatively uni-
form pressure within the stack, and (3) an increasing pressure gradient is observed
in the outlet manifold due to side-injection. The reader will note that different con-
tour scales have been used in the stack and manifold regions and that the overall
(parasitic) pressure losses are much larger in the former than the latter. As the stack
height increases, the pressure difference between the manifolds at the top of the
stack becomes much less than at the bottom and flow maldistributions may occur.
Another interesting feature of fuel cell stacks is that secondary temperature distri-
butions in the vertical direction arise, even when the flow and heat source term (due
to Joule heating) are entirely uniform. Figure 16 shows the temperature distribution
in a 10-cell stack obtained using a detailed numerical simulation. The oscillations
in the temperature contours are due to the finite differences in temperature between
Figure 16: Temperature distribution (deg. C) in a 10-cell SOFC stack with uniform
heat sources. Adapted from [34].
the air, fuel, electrolyte and interconnects for each cell in the stack. The secondary
temperature gradients are caused by these materials being ordered, creating a net
heat flux in the vertical direction. The stack designer cannot necessarily presume
that thermo-mechanical behavior is the same for a stack as for a single cell, as has
been presumed by many researchers in the past.
4 Closure
Transport phenomena in SOFCs influence every aspect of their design and

operation. Fluid flow, heat conduction, convection, and radiation, mass transfer,
electrochemistry, charge transfer kinetics and thermodynamics are all important
mechanisms. In the last 5–10 years much progress has been made in the devel-
opment of robust engineering computer codes to model transport phenomena and
physico-chemical hydrodynamics in these devices. This is significant progress,
however further developments are needed both in models and on input data of
properties.
Although it is currently fashionable to state that growth in this area is ‘rapid’, it
remains to be seen whether recent events will continue to be sustainable. On the one
hand, design methodologies may continue to evolve exponentially, as the product
unfolds and becomes commercially viable in the years to come. On the other, it
could be that analysis tools will plateau and languish, if the technology proves to
be uncompetitive, and promises of economic viability prove to be a ‘bubble’. The
author’s hope is that the former will prove to be the case, and that years from now,
we will see improvements, not only in analysis methods, but also as a result, in the
products.
It is clear that if SOFCs, or indeed any other new technologies, are to supplant
existing products and processes, they must be designed properly. Every available
analysis tool and advanced scientific method needs to be brought to bear upon the
problem, if good engineering is to replace good will and sway the public to buy
new and untested products over established and functioning ones, in a competitive
market place.
In many ways the success or failure of fuel cells will ultimately depend on factors
completely unrelated to the thermodynamic efficiency of the electrode-electrolyte
assembly, so popular among research proposals. Probably reliability and safety
rank even higher than cost-per-unit energy, and while mathematical modeling alone
cannot safeguard these attributes, careful engineering using established techniques
such as those described in the chapters in this book (and already widely employed
in design of conventional power-generation equipment) will likely prove a good
investment, and aid in better understanding of equipment performance. To be com-
petitive, fuel cells need to be at least as well or better designed than existing products.
Finally, for a variety of reasons, numerical models of physico-chemical pro-
cesses can now be developed in much shorter time-scales (months), and for much
less cost than it takes to build complex experimental test rigs (years). While this
situation cannot be blamed upon those who seek to construct models and codes, it is
nonetheless an unhealthy situation; a balanced program of research should always
involve the continuous feedback of experimental data to analysis tools, as well as
the abstraction of that which is being modelled to the pragmatist. It is therefore
imperative to establish reliable data bases of empirical data for code evaluation
purposes.
Acknowledgements
There are many people who have assisted the author in this endeavour, both directly
and indirectly. Thanks are due to (in alphabetical order): Katherine Cook, Kyle
Daun, Wei Dong, Eliezer Gileadi, Ron Jerome, Yongming Lin, Fengshan Liu, and
Rod McMillan.
References
[1] Appleby, A.J. & Foulkes, F.R., Fuel Cell Handbook, Van Nostrand Reinhold:
New York, 1989.
[2] Kordesh, K. & Simader, G., Fuel Cells and their Applications, VCH: New
York, 1996.
[3] Larminie, J. & Dicks, A., Fuel Cell Systems Explained, Wiley: Chichester,
2000.
[4] Williams, K.R., An Introduction to Fuel Cells, Elsevier: Amsterdam, 1966.
[5] Vayenas, C.G. & Hegedus, L.L., Cross-flow, solid-state electrochemical
reactors: A steady-state analysis. Ind. & Eng. Chem. Fundamentals, 24,
pp. 316–314, 1985.
[6] Fiard, J.M. & Herbin, R., Comparison between finite volume and finite ele-
ment methods for the numerical simulation of an elliptic problem arising in
electrochemical engineering. Computational Methods in Applied Mechani-
cal Engineering, 115, pp. 315–338, 1994.
[7] Ferguson, J.R., Analysis of temperature and current distributions in planar
SOFC designs. Proceedings Second International Symposium on Solid Oxide
Fuel Cells, Athens, Greece, pp. 273–280, 1991.
[8] Herbin, R., Fiard, J.M. & Ferguson, J.R., Three-dimensional numerical sim-
ulation of the temperature, potential and concentration distributions of a unit
cell for various geometries of SOFCs. Proceedings, European Solid Oxide

Fuel Cell Forum l, Luzern, Switzerland, 1994.
[9] Karoliussen, H., Nisancioglu, K., Solheim, A., Odegard, R., Singhal, S.C.
& Iwahara, H., Mathematical modeling of cross plane SOFC with internal
reforming. Proceedings, Third International Symposium on Solid Oxide Fuel
Cells, Honolulu, Hawaii, 1993.
[10] Achenbach, E., Three-dimensional modeling and time-dependent simulation
of a planar solid oxide fuel cell stack. Journal of Power Sources, 73, pp. 333–
348, 1994.
[11] Bessette, N.F. & Wepfer, W.J., Electochemical and thermal simulation of a
solid oxide fuel cell. Chemical Engineering Communications, 147, pp. 1–15,
1996.
[12] Bernier, M., Ferguson, J. & Herbin, R., A 3-dimensional planar SOFC stack
model. Proceedings, Third European Solid Oxide Fuel Cell Forum, Nantes,
France, pp. 483–495, 1998.
[13] Ahmed, S., McPheeters, C. & Kumar, R., Thermal-Hydraulic model of a
monolithic solid oxide fuel cell. Journal of the Electrochemical Society,
138, pp. 2712–2718, 1991.
[14] Sira, T. & Ostenstad, M., Temperature and flow distributions in planar SOFC
stacks. Third International Symposium on Solid Oxide Fuel Cells, Honolulu,
Hawaii, pp. 851–860, 1993.
[15] Costamagna, P. & Honegger, K., J. Electrochem. Soc., 145(11), pp. 2712–
2718, 1998.
[16] Chan, S.H., Khor, K.A. & Xia, Z.T., Journal of Power Sources, 93,
pp. 130–140, 2001.
[17] Beale, S.B., Lin, Y., Zhubrin, S.V. & Dong, W., Computer Methods for
Performance Prediction in Fuel Cells. Journal of Power Sources, 11(1–2),
pp. 79–85, 2003.
[18] Beale, S.B. & Zhubrin, S.V., A distributed resistance analogy for solid oxide
fuel cells. Numerical Heat Transfer, Part B, in press.
[19] Newman, J.S., Electrochemical Systems, Prentice-Hall Inc.: Englewood Cliffs,
N.J., 1973.
[20] Levich, V.G., Physicochemical Hydrodynamics, Prentice-Hall: Englewood
Cliffs, N.J., 1962.
[21] Probstein, R.F., Physicochemical Hydrodynamics: An Introduction, Butter-
worths: Boston, 1989.
[22] Callen, Thermodynamics, John Wiley & Sons Inc.: New York, 1960.
[23] Glasstone, S., Laidler, K.J. & Eyring, H., The Theory of Rate Processes:
The Kinetics of Chemical Reactions, Viscosity, Diffusion and Electrochem-
ical Phenomena. International Chemical Series, McGraw-Hill: New York,
1941.
[24] Gileadi, E., Electrode Kinetics for Chemists, Chemical Engineers, and Mate-
rial Scientists, Wiley-VCH: New York, 1993.
[25] Bockris, J.O.M., Reddy, A.K.N. & Gamboa-Aldeco, M., Modern Electro-
chemistry. Vol. 2, Plenum: New York, 2000.
[26] Bard, A.J. & Faulkner, L.R., Electrochemical Methods, 2nd edn, John Wiley
& Sons: New York, 2001.
[27] Fuel Cell Handbook, 5th edn, U.S. Department of Energy, National Energy
Technology Laboratory: Morgantown/Pittsburgh, 2000.
[28] Beale, S.B., Calculation procedure for mass transfer in fuel cells. Journal of
Power Sources, 128(2), pp. 185–192, 2004.
[29] Spalding, D.B., Convective Mass Transfer: An Introduction, Edward Arnold:
London, 1963.
[30] Spalding, D.B., A standard formulation of the steady convective mass trans-
fer problem. International Journal of Heat and Mass Transfer, 1, pp. 192–
207, 1960.
[31] Kays, W.M., Crawford, M.E. & Weigand, B., Convective Heat and Mass
Transfer, 4th edn, McGraw-Hill: New York, 2005.
[32] Mills, A.F., Mass Transfer, Prentice Hall: Upper Saddle River, N.J., 2001.
[33] Kays, W.M. & Crawford, M.E., Convective Heat and Mass Transfer, 2nd
edn, McGraw-Hill: New York, 1980.
[34] Beale, S.B., Conjugate mass transfer in gas channels and diffusion layers
of fuel Cells. Proceedings 3rd International Conference on Fuel Cell Sci-
ence, Engineering and Technology, eds R.K. Shah & S.G. Kandlikar, ASME:
Ypsilanti, Michigan, 2005.
[35] Shah, R.K. & London, A.L., Laminar flow forced convection in ducts.
Advances in Heat Transfer, eds T.F. Irvine & J.P. Hartnett, Academic Press:
New York, 1978.
[36] Jacob, M., Heat Transfer, Wiley: New York, 1949.
[37] Patankar, S.V., Numerical Heat Transfer and Fluid Flow, Hemisphere:
New York, 1980.
[38] Patankar, S.V. & Spalding, D.B., A Calculation Procedure for the
Transient and Steady-state Behavior of Shell-and-tube Heat Exchangers.
Heat Exchangers: Design and Therory Sourcebook, eds N. Afgan &
E.U. Schlünder, Scripta Book Company: Washington, D.C., 1974.
[39] Beale, S.B., Mass transfer in plane and square ducts. International Journal
of Heat and Mass Transfer, in press.
[40] Vafai, K. & Tien, C.L., Boundary and Inertia Effects on Flow and Heat
Transfer in Porous Media. International Journal of Heat and Mass Transfer,
24(2), pp. 195–203, 1981.
[41] Ferguson, J.R., Fiard, J.M. & Herbin, R., Three-dimensional numerical sim-
ulation for various geometries of solid oxide fuel cells. Journal of Power
Sources, 58, pp. 109–122, 1996.
[42] Knuth, E.L., Multicomponent diffusion and Fick’s law. Physics of Fluids, 2,
pp. 339–340, 1959.
[43] Wilke, C.R., Diffusional properties of multicomponent gases. Chemical
Engineering Progress, 46(2), pp. 95–104, 1950.
[44] Taylor, R. & Krishna, R., Multicomponent Mass Transfer, Wiley-Interscience:
New York, 1993.
[45] Kleijn, C.R., van der Meer, T.H. & Hoogendoorn, C.J., A mathematical
model for LPCVD in a single wafer reactor. Journal of the Electrochem-
ical Society, 136(11), pp. 3423–3433, 1989.
[46] Damm, D.L. & Fedorov, A.G., Radiation heat transfer in SOFC materials
and components. Journal of Power Sources, in press.
[47] Oppenheim, A.K., Radiation analysis by the network method. Transactions
of the ASME, 78, pp. 725–735, 1956.
[48] Siegel, R. & Howell, J.R., Thermal Radiation Heat Transfer, 4th edn,
Hemisphere: Washington, 2002.
[49] Yakabe, H., Ogiwara, T., Hishinuma, I. & Yasuda, I., 3-D model calculation
for planar SOFC. Journal of Power Sources, 102, pp. 144–154, 2001.
[50] VanderSteen, J.D.J., Austin, M.E. & Pharaoh, J.G., The role of radiative heat
transfer with participating gases on the temperature distribution in solid oxide
fuel cells. Proceedings 2nd International Conference on Fuel Cell Science,
Engineering and Technology, eds R.K. Shah & S.G. Kandlikar, Rochester,
NY, 2004.
[51] Murthy, S. & Federov, A.G., Radiation heat transfer analysis of the monolith
type solid oxide fuel cell. Journal of Power Sources, 124, pp. 453–458, 2003.
[52] Damm, D.L. & Fedorov, A.G., Spectral radiative heat transfer of the planar
SOFC. Proceedings International Mechanical Engineering Congress and
Exposition, ASME: Anaheim, California, 2004.
[53] Daun, K., J., Beale, S.B., Liu, G. & Smallwood, G.J., Radiation heat trans-
fer in solid oxide fuel cells. Proceedings of 2005 Summer Heat Transfer
Conference, ASME: San Francisco, California, 2005.
[54] Spalding, D.B., Mathematical Modelling of Fluid-mechanics, Heat-transfer
and Chemical-reaction Processes: A Lecture Course. 1980, Computational
Fluid Dynamics Unit, Imperial College, University of London: London.
[55] Spinnler, M., Winter, E.R.F. & Viskanta, R., Studies on high-temperature
multilayer thermal insulations. International Journal of Heat and Mass
Transfer, 47, pp. 1305–1312, 2004.
[56] Spinnler, M., Winter, E.R.F., Viskanta, R. & Sattelmayer, T., Theoretical
studies of high-temperature multilayer thermal insulations using radiation
scaling. Journal of Quantitative Spectroscopy and Radiative Transfer, 84,
pp. 477–491, 2004.
[57] Beale, S.B., Ginolin, A., Jerome, R., Perry, M. & Ghosh, D., Towards a vir-
tual reality prototype for fuel cells. PHOENICS Journal of Computational
Fluid Dynamics and its Applications, 13(3), pp. 287–295, 2000.
[58] Berman, A.S., Laminar flow in channels with porous walls. Journal of
Applied Physics, 24(9), pp. 1232–1235, 1953.
[59] Jorne, J., Mass transfer in laminar flow channel with porous wall. Journal
of the Electrochemical Society, 129(8), pp. 1727–1733, 1982.
[60] Lessner, P. & Newman, J.S., Hydrodynamics and Mass-Transfer in a Porous-
Wall Channel. Journal of The Electrochemical Society, 131(8), pp. 1828–
1831, 1984.
Nomenclature
A Area (m2 )
a Absorption coefficient (m−1 ), coefficient in finite-volume equations,
length (m)
B Width (m)
b Length (m)
c Specific heat (J/kgK)
D Diffusion coefficient (m2 /s)
Dh Hydraulic diameter (m)
E Nernst potential (V), electric potential (V)
eb Black-body emissive power (W/m2 )
F Distributed resistance (kg/m2 s),
Faraday’s constant, 96.48 (Coulomb/mol),
Radiation configuration factor ()
G Gibbs free energy (J)
g Mass transfer conductance (kg/s)
H Height (m), enthalpy (J)
h Heat transfer coefficient (W/m2 K), Planck constant 6.626 × 10−34 (J.s)
i Current (A), radiant intensity (W/sr.m2 )
j Diffusion source
KP Equilibrium constant
k Thermal conductivity (W/mK), rate constant (mol/m2 )
kDarcy Permeability (m2 )
k Thermal conductivity (W/mK)
kr Radiative conductance (W/mK)
L Length (m)
M Molecular weight (kg/mol)
m Mass fraction (kg/kg), mass (kg)
N Number of cells in stack (), mole number
n Charge number (), valence ()
nfc Number of fuel cells in SOFC stack
P Pressure (Pa), partial pressure (Pa)
Q Charge (Coulomb)
Q̇ Volumetric discharge (m3 /s)
q Heat source term (W)
q0 Radiosity (W)
R Gas constant, 8.314 × 103 (J/molK), resistance (Ohm)
r Resistance (Ohm/m2 )
S Entropy (J/K)
Scond Conduction shape factor ()
s Source term, specific heat (J/kgK)
T Temperature (◦ C)
U Internal energy (J), superficial velocity (m/s)
Ū Overall heat transfer coefficient (W/m2 K)
u Interstitial velocity (m/s)

V Volume (m3 ), voltage (V)
W Work (J)
T Temperature (K)
xi Mole fraction (mol/mol)
Greek symbols
α Volumetric transfer coefficient (W/m3 K)

α∗ Aspect ratio ()
β Transfer coefficient, symmetry coefficient (), Blockage factor ()
ε Volume fraction, void fraction (), emmissivity (m−1 )
Exchange coefficient (kg/ms)
General scalar
φ General scalar, electric field potential (V)
η Overpotential, polarisation (V)
ν Stoichiometric coefficient
ρ Density (kg/m3 )
µ Chemical potential (J/mol), viscosity (W/mK)
σ Electric conductivity (Ohm−1 m), Stefan-Boltzmann
constant 5.67 × 10−8 (W/m2 K4 )
σS Scattering coefficient (m−1 )
τ Shear stress (N/m2 ), tortuosity ()
Non-dimensional numbers
B Driving force
b Blowing parameter
f Friction coefficient
Nu Nusselt number
Sh Sherwood number
Sc Schmidt number
Re Reynolds number
Subscripts
0 Reference state
a Air, anode
b Bulk, backward
c Cathode
e Electrolyte
f Fuel, forward
i Interconnect, internal
l Load
t total, transferred substance state
w Wall
Superscripts
0 Reference state, equilibrium state

∗ Zero mass transfer
. Per unit time
’ Per unit length
+ Positive direction
− Negative direction
CHAPTER 3
Electrochemical and thermo-fluid modeling of a

tubular solid oxide fuel cell with accompanying
indirect internal fuel reforming
K. Suzuki1 , H. Iwai2 & T. Nishino2
1 Department of Machinery and Control Systems,
Shibaura Institute of Technology, Japan.
2 Department of Mechanical Engineering, Kyoto University, Japan.
Abstract
Development of fuel cells has been boosted by global and regional environmental
issues. Among others, the solid oxide fuel cell (SOFC) has been drawing much
attention as a unit for distributed energy generation. Numerical analysis is used as a
powerful tool in research and development of fuel cells, especially of the SOFC for
which important and necessary thermal management information has scarcely been
supplied experimentally. A central issue for achieving maximum benefit from the
numerical analysis is how to properly model the complex phenomena occurring in
the fuel cells. This chapter presents a numerical model for a tubular SOFC including
a case with indirect internal reforming. In this model, the velocity field in the air and
fuel passages and heat and mass transfer in and around a tubular cell are calculated
with a two-dimensional cylindrical coordinate system adopting the axisymmetric
assumption. Internal reforming and electro-chemical reactions are both taken into
account in the model. Electric potential field and electric current in the cell are also
calculated simultaneously allowing their nonuniformity in the peripheral direction.
A previously developed quasi-two-dimensional model was adopted to combine the
assumed axisymmetry of velocity, heat and mass transfer fields and the periph-
eral nonuniformity of electric potential field and electric current. Details of those
numerical procedures are described and examples of the calculated results are dis-
cussed. After presenting a few fundamental results, several strategies to reduce the
maximum temperature and temperature gradient of the cell are examined for a case
with indirect internal fuel reforming.
1 Introduction
One of the most important global issues in the present world is the suppression
of global warming. Global warming is proceeding related to the increase of atmo-
spheric concentration of carbon dioxide. Discharge of carbon dioxide into the atmo-
sphere occurs in the eruption of volcanoes but the largest emission rate of carbon
dioxide originates from power plants and from vehicles both of which support the
modern world. Since drastic depression of energy demand and usage of automo-
biles cannot be accepted easily by people living in the modern society, development
of energy conversion systems being free from the emission of carbon dioxide or,
at least, of a system having a lower emission rate of carbon dioxide is important.
Fuel cells are one of the prospective power generation systems for this purpose.
There are five types of fuel cells characterized by the difference in materials for the
electrolyte and in this relation by the difference in operation temperature. They are
alkaline fuel cells, phosphoric acid fuel cells, polymer electrolyte fuel cells, molten
carbonate fuel cells and solid oxide fuel cells (SOFCs). In this chapter, attention is
paid to the last one, the SOFC.
Solid oxide fuel cells (SOFCs) have an advantage in that not only hydrogen but
also a variety of hydrocarbons can be used as fuels. This is because solid oxides used
for the electrolyte, are oxygen-ion conducting materials. In the case of fuel cells
using hydrogen-ion conducting or proton conducting materials as the electrolyte, the
fuel should in principle be hydrogen only but, in the case of the SOFC, a variety of
fuels including carbon monoxide in addition to hydrocarbons can react with oxygen-
ions conducted across the electrolyte. High operation temperature provides another
advantage of SOFCs. No catalyst is needed for the electrochemical reaction if they
are operated at a reasonably high temperature except in the case of low temperature
SOFCs [1]. In addition to this, high quality thermal energy of the effluent from
the SOFCs can be recovered in various ways. Construction of a hybrid system that
fuses the SOFC to a gas turbine to obtain high electricity generation efficiency is one
of them [2–4]. In addition, steam reforming of hydrocarbon fuels is possible in or
around the cells with the aid of high operation temperature. A variety of commercial
materials can be used as a catalyst for the reforming reaction. Fuel reforming
supplies hydrogen and carbon monoxide from methane as fuel, which makes the
fuel cell operation simpler and additionally provides a means to cool and therefore
to control the high temperature fuel cell. When the reforming reaction takes place on
the anode with heat supplied directly from the electrochemical reaction, it is referred
to as direct internal reforming (DIR). In this case, the cell structure is simple because
no additional catalyst except the anode is needed. However, there are still difficulties
in proper control both of reforming and electrochemical reactions within the anode
itself [5–7]. On the other hand, the case in which the reforming reaction takes place
on catalysts that are positioned apart from the anode is referred to as indirect internal
reforming (IIR). This is the case treated in this chapter as will be discussed later.
Another advantage of SOFCs is that the electrolyte is solid. This results in geomet-
rical flexibility of the cell design. Several types of SOFCs having different geometry
have been proposed, tested and developed so far. They are tubular SOFCs, planar
Transport Phenomena in Fuel Cells 85
Figure 1: Schematics of three types of SOFC having different geometry.
SOFCs and disk type SOFCs. Schematics of the three types of SOFC are illus-
trated in Fig. 1. Planar SOFCs and disk type SOFCs are prospective in a point
that they can be stacked in a compact way and that they can have higher electric
power density. Therefore, they are the subject of keen research and development
activities [8–11]. On the other hand, tubular SOFCs are one of the most developed
towards practical use. The two major advantages of the tubular cells are that they
have relatively good thermal shock resistance and that there is no need for special
sealing treatment. Owing to these advantages, a thousands-hour-long operation test
of an atmospheric pressure 100-kW tubular SOFC system has successfully been
performed [12]. In this system, cathode-supported SOFCs are used, in which air
flows inside the tubular cell structure and fuel flows around the tubular cell along
its axis. Natural gas is used as fuel and it is fully reformed in “in-stack reformers”,
which are placed between the rows of cell bundles composed of many tubular cells.
Reformers are heated indirectly by the cells. The cell structure is simpler in this
case but the stack structure is rather complicated. On the other hand, in the case
of anode-supported SOFCs, fuel flows inside the tubular cell and air flows out-
side the tubular cell. In this case, a fuel feed tube mounted inside the tubular cell,
through which fuel is injected into the tubular cell, can be used as a reformer. The
feed tube is filled with the catalyst material and indirect but internal fuel reforming
(IIR) proceeds while fuel flows through it. This is the concept of indirect internal
reforming type tubular SOFCs (IIR-T-SOFCs) to be discussed in this chapter. In this
case, although the cell structure is a little more complicated due to the catalyst to
be embedded in it, the stack structure becomes fairly simple. There is no in-stack
reformer so that the size and geometry of the stack can be designed more easily
and more flexibly for various SOFC electricity generation systems having different
power output capacity. In addition, there is a chance to control the local thermal
field in the cell by adjusting the distribution density of the fuel reforming catalyst.
This is one of the achievements of earlier numerical analyses [13, 14]. At present,
development of this type of SOFC is just in a conceptual design stage. So it is mean-
ingful to study its feasibility right now and numerical simulation is considered to
be an effective tool for this purpose.
Numerical simulation has become a useful means in research and development
in various engineering fields. Fuel cells are no exception and various numerical
simulations of fuel cells have been conducted by several research groups. Numer-
ical modeling of fuel cells is complicated due to a variety of phenomena occurring
in the cells such as multi-component gas flows with heat and mass transfer, elec-
trochemical reactions between fuel and air, and electric potential field and ionic
and electric current in the cell. Numerical simulation is especially important for
SOFCs. They need good thermal management because the operation temperature
is high. However, the detailed information necessary for management is not eas-
ily supplied from measurement. Numerical simulation is a sole means to supply
such information. There are two types of numerical simulations applied to SOFCs.
One is for a single cell [15–20] and another for a stacked-cell module [21–24]. In
this chapter, the former type of numerical simulation is treated and in particular
description is given to a numerical model developed for a tubular SOFC including
a case of IIR-T-SOFC based on the model for the cathode-supported tubular SOFC
reported in [15, 16]. Finally, some results are also presented as examples of the
simulations made with the described model.
2 General remarks on the mechanism of IIR-T-SOFC

Prior to describing the numerical modeling and simulated results, general remarks
will be given to the phenomena in the indirect internal reforming tubular solid oxide
fuel cells, the IIR-T-SOFCs, in this section. Except for the part related to internal fuel
reforming, description to follow can be applied to a general case without internal
fuel reforming including the case of a conventional cathode-supported tubular solid
oxide fuel cell.
2.1 Tubular cell
Figure 2 shows schematic views of two types of a tubular SOFC. One is a conven-
tional Siemens-Westinghouse type cathode-supported tubular cell [25] and another
is a single cell to be used in the IIR-T-SOFC stack. The latter type of fuel cell
consists of a tubular cell and a feed tube inserted in it, basically similar to a design of
Figure 2: Schematic view of a conventional tubular SOFC and IIR-T-SOFC.
Figure 3: Schematic view of a cell stack.
conventional tubular SOFCs. The cell tube is composed of two electrodes sand-
wiching an electrolyte layer. Electrochemical reactions take place at both these
electrodes, anode and cathode. The difference between the two cases illustrated in
Fig. 2 is that the hydrocarbon to be used as fuel flows inside the feed tube and that
the feed tube is filled with porous catalysts in the case of the IIR-T-SOFC. So fuel is
reformed “indirectly” in each tubular cell. That is to say, fuel is reformed inside the
feed tube, changes flow direction at the closed end of the cell tube, reacts electro-
chemically on the anode and then flows out of the cell. On the other hand, air flows
outside the cell tube and reacts on the cathode. One drawback of this cell design is
that the interconnects are exposed to the high-temperature air so that serious oxi-
dization of the interconnects may occur. However, this problem is being resolved
due to the development of new interconnect materials such as the doped-lanthanum
chromite series in recent years [26, 27].
The tubular cell can be stacked either electrically in series or electrically in
parallel as shown in Fig. 3. The number of the tubular cells in the stack is adjusted so
as to fit the required electricity capacity. When the number of the cells is sufficiently
large, most of the cells in the core region of the stack should be in the same thermal
and operating conditions. Therefore, attention is paid in the present study to a single
representative tubular cell in the core region of the stack.
2.2 Internal reforming process
One of the important phenomena occurring in the IIR-T-SOFCs is the internal

reforming of fuel and it takes place, as mentioned above, inside the feed tube.
Although a variety of hydrocarbons can be used as fuel, description is given here
as an example for the case where methane, a major component of natural gas, is
used as fuel.
For methane, the steam reforming process is widely known as a conventional
process for producing hydrogen [28, 29] and it proceeds on catalysts such as nickel-
alumina through the following two chemical reactions:
Reforming reaction:
CH4 + H2 O ←→ 3H2 + CO. (1)
Shift Reaction:
CO + H2 O ←→ H2 + CO2 . (2)
The equilibrium constant of the steam reforming reaction described by eqn (1) is
reasonably large so that methane can almost totally be reformed if the system is kept
at the operation temperature 700 ◦ C or higher. However, a supply of thermal energy
is needed for the reforming reaction to proceed since it is strongly endothermic.
Therefore, how methane is reformed inside the feed tube is highly dependent on
the local conditions including not only temperature but also partial pressure of each
chemical species and density of the catalysts. On the other hand, the water-gas shift
reaction described by eqn (2) is a weak exothermic reaction. The important point
is that the shift reaction is almost in equilibrium in the entire fuel passage because
its reaction rate is much faster than that of the steam reforming reaction.
2.3 Electrochemical process
The most important phenomenon occurring in the tubular cell is a series of events
associated with the electrochemical reaction. In the case where methane is used as
fuel and is totally reformed inside the feed tube, the electrochemical processes on
the anode and cathode can be described by the following reactions:
H2 + O2− → H2 O + 2e− (anode), (3)

CO + O2− → CO2 + 2e− (anode), (4)
1
2 O2 + 2e− → O2− (cathode). (5)
Figure 4: Basic concept of the electrochemical process.
Overall reactions are equivalent to the following:
H2 + 12 O2 → H2 O, (6)
CO + 12 O2 → CO2 . (7)
Figure 4 illustrates the mechanism of the electrochemical processes in the tubular

cell. On the cathode side, oxygen ions are produced by the combination of oxygen
molecules in the air flow with electrons. The produced oxygen ions move through
the electrolyte to the anode side. Electrons are released from the oxygen ions at the
anode and the produced oxygen molecules react with hydrogen or carbon monox-
ide molecules there. The electrons released through the anode reactions move to
the cathode of the neighbouring cell through the interconnect or return to the orig-
inal cathode via an external circuit – i.e. electric current is generated through the
tubular cells.
The magnitude of the electric current directly relates to the rates of the electro-
chemical reactions and depends on a variety of factors. These factors can be clas-
sified into three groups: electromotive force, internal resistance, and external load
of the cell. The electromotive force is an ideal cell terminal voltage to be achieved
at zero electric current or when the electrochemical reaction proceeds very slowly.
The internal resistance causes the loss of terminal voltage and includes three differ-
ent types of overpotentials; namely ohmic overpotential, activation overpotential
and concentration overpotential. Finally, the external load of the cell regulates the
electric current to be realized with the fuel cell under operation and is actually equal
to the cell terminal voltage divided by the magnitude of electric current.
The electromotive force is largely determined by the concentration of fuel. In

general, it decreases as hydrogen and carbon monoxide are consumed and therefore
as their concentrations become lower. As for the ohmic overpotential, ionic conduc-
tivity of the electrolyte is one of the important factors. It has a tendency to decrease
as the temperature rises. This is the main reason why operation temperature must
be high in the case of SOFCs. Electric conductivity of the electrodes also affects the
magnitude of the ohmic loss of terminal voltage. Electric current flows through the
electrodes basically in the circumferential direction as shown in Fig. 4. Activation
overpotential occurs related to the sluggishness of electrochemical reaction and
tends to decrease as the cell temperature rises. Concentration overpotential is also
an important factor affecting the cell performance and is related to the radial con-
centration nonuniformity of participating chemical species, i.e. fuel and oxygen.
At the reaction sites, their concentrations are lower so that the real electromotive
force to be attained under such concentrations is lower than the counterpart that
would be achieved using the concentration in the core flow regions. This difference
is the concentration overpotential. These electromotive forces and internal resis-
tances vary from place to place because temperature and gas composition are not
uniform in the cell. Finally, as for the external load, electric current becomes larger
and terminal voltage becomes lower as it decreases. Therefore, the output power
of the cell, which is the product of the electric current and the terminal voltage,
reaches a peak at a certain value of the electric current or of the external load.
2.4 Purpose and key points of the analysis
In the development of good SOFCs, in particular, good IIR-T-SOFCs, there are three
fundamental problems to be solved: how to improve the power generation perfor-
mance, how to prevent thermal crack failure, and how to reduce the cost of produc-
tion. Numerical analysis is expected to contribute to the solution of these issues.
Among them, the thermal crack issue is an especially suitable one to be tackled with
numerical analysis. Thermal crack failure is caused by a combination of several ther-
mal conditions including the appearance of a hot spot or of excessively high temper-
ature, the generation of excessively large spatial temperature gradients, the time
changing rate of temperature in start-up of the fuel cells and the frequency of load
change or of thermal conditions in practical operation. Cell temperature or electroc-
hemical and other chemical reactions governing the cell temperature cannot be mea-
sured and such detailed information can only be supplied from numerical analysis.
The thermal field in the cell is determined by the balance of several fundamental
phenomena: heat generation accompanied with the electric current and the electro-
chemical processes, heat transfer among different parts of the cell having different
temperature, heat absorption by the endothermic internal fuel reforming process,
and heat removal by the air flow. Therefore, we must integrate all of the calcula-
tion for the gas flow fields, temperature fields both in gas flow passages and solid
parts, electrochemical and fuel reforming processes and related mass transfer fields
of participating chemical species, and ionic and electric current fields. This must
be done with some assumptions to keep the computation load within the reach of
engineering workstations but certainly must be done as accurately as possible in
order to establish reasonable databases for the thermal field of the cell. As for the
gas flow velocity, temperature and mass transfer fields, at least two-dimensional
axisymmetric analysis would be required to study heat and mass transfer at a rea-
sonably accurate level. Meanwhile, consideration of peripheral nonuniformity is
requisite for the electric potential field. This is because, although ionic current
occurs in radial direction across the electrolyte, electric current flows basically in
the circumferential direction in the electrodes as shown in Fig. 4. Of course it is bet-
ter if three-dimensionality is taken into account for all related fields, but it requires
an extremely large CPU time. Hence what to assume and how to integrate those
calculations are the key points of the numerical analysis on tubular SOFCs includ-
ing the IIR-T-SOFCs. In this study, a previously developed quasi-two dimensional
treatment [15, 16] is adopted as a base to develop the new model. This quasi-two
dimensional model combines the assumed axisymmetry of the velocity, heat and
mass transfer fields with the peripheral nonuniformity of the electric potential and
electric current fields.
3 Numerical modeling
This section presents the details of the numerical modeling of a tubular SOFC,
particularly of an IIR-T-SOFC. This model is based on a previously developed
model for a cathode-supported tubular SOFC to be operated with reformed fuel
or especially with hydrogen [15, 16]. Therefore, the main parts of the model are
common to both cases with and without accompanying internal fuel reforming.
However, some general parts common to both cases will be outlined for simplicity
referring to the references and other parts specific to the case with indirect internal
reforming will be emphasized.
As mentioned in the preceding section, various phenomena occur in the tubular
cell such as gas flows, heat and mass transfer, internal reforming, electrochemical
reaction, non-uniform electric potential field and ionic and electric currents and gen-
eration of heat and electricity. These phenomena are closely related to one another,
thus they should be considered simultaneously in the numerical analysis. For the
sake of the reader’s convenience, geometry of the cell and general assumptions
adopted in the present model are described as a preparation to read the following
parts. After this, modeling for the most basic phenomena in fuel cells, i.e. electro-
chemical reactions, will be described first. Next, reforming and shift reactions are
described. Then, the modeling for the velocity, temperature and concentration fields
will be discussed. Description will be made of the modeling for electric potential
fields, electric current and ohmic heating. Finally, an overall picture of the model
is given at the end of the section.
3.1 Computational domain and general assumptions for heat and

mass transfer
A longitudinal sectional view of the tubular cell is shown in Fig. 5 with the dimen-
sions and the chemical reaction equations to be considered in this study. In the
examples for IIR-T-SOFCs, of which results will be discussed later, the cell tube
Figure 5: Longitudinal cross sectional view of a tubular fuel cell.
geometry is assumed to be 500 mm long and to have a 6.9 mm inner radius and a
9.6 mm outer radius. For computational convenience, air is taken to flow through
an annular space between a tubular fuel cell outer surface and a co-axial confining
cylindrical wall of which the radius is 14.6 mm. A square drawn with a fine dot-
ted line in Fig. 3 illustrates the average air flow space in a stack. The surface of
this artificial confining wall is assumed to be thermally adiabatic and the area of
the annular space is equated with the average air flow space for a single cell in a
stack. This treatment is reasonable for a tubular cell located in the core of stack. The
hemi-spherical tube end is replaced by a flat end in the model and the computational
domain is indicated by the broken line in Fig. 5. The cell tube and fuel feed tube
are solid and the feed tube is assumed to be filled with porous catalytic material,
and the other areas are gas flow passages.
In this study, except in a case of a hydrogen-fuelled cathode-supported tubular
SOFC, fuel is assumed to be a mixture of fresh methane and recirculated effluent
exhausted from the fuel cell, namely a mixture of hydrogen, steam, CO, CO2 and
methane, and air is treated as a mixture of oxygen and nitrogen. All gas components
are treated as ideal gases in the calculation of density and electromotive force. In
the calculation of thermal fields, temperature is adopted as a variable to solve but
the thermodynamic properties like specific heat at constant pressure, enthalpy and
Gibb’s free energy and transport properties like viscosity and heat conductivity
are treated variable with temperature. Their local values are evaluated at local
temperature by consulting the tables of properties included in the program [30, 31].
Properties of mixture are evaluated with mixing laws by making use of the partial
pressure of each chemical species [32]. Gaseous fluids are treated to be Newtonian
and Reynolds number is lower than two hundred so that both air and fuel flows
are assumed to be laminar, steady and axisymmetric. The effect of gravity is not
considered.
3.2 Model for electrochemical reactions
The most basic phenomena in the fuel cell are the electrochemical processes by
which chemical energy of the fuel is directly converted into electricity. In this rela-
tion, consumption of fuel and oxygen occurs. In addition to electricity generation,
heat is generated as an ineffective part of the used chemical energy of the fuel. Here
it is first discussed how they are modelled.
As mentioned in the preceding section and shown in Fig. 5, the electro-chemical
reactions of hydrogen and CO, described by eqns (6) and (7), are considered in
this study. The local electromotive forces to be generated by these reactions are
calculated based on the following Nernst equation:

R0 T PH2 O
EH2 /O2 = EH2 /O2 −
0
ln 0.5
, (8)
2F PH2 PO 2

0 R 0 T P CO
ECO/O2 = ECO/O 2
− ln 2
0.5
, (9)
2F PCO2 PO 2
where

−GH
0
2 /O2
EH0 2 /O2 = , (10)
2F

−GCO/O
0
0
ECO/O 2
= 2
. (11)
2F
The subscripts “H2 /O2 ” and “CO/O2 ” indicate the reactions of eqns (6) and (7),
respectively. GH 0 0
and GCO/O are the changes of the standard Gibbs free
2 /O2 2
energy accompanied with the reactions, and F is the Faraday constant. It is noted
here that the local partial pressures of chemical species at the surface of each
electrode are assigned to the partial pressures in the above equations. This means
that the concentration overpotential related to the concentration non-uniformity
existing in the fuel and air flow passages is automatically taken into account in the
electromotive forces described by eqns (8) and (9).
The overall electrochemical reactions described by eqns (6) and (7) can be divided
into the electrode reactions described by eqns (3), (4) and (5). It is necessary to
consider the activation overpotential for each electrode reaction. In this study,
the following equations suggested by Achenbach [18] are used to describe the
activation overpotentials:
−1
2F PH2 0.25 Aa
ηH2 = iH2 /O2 kH exp − , (12)
R0 T 2 Pfuel R0 T
−1
2F PCO 0.25 Aa
ηCO = iCO/O2 kCO exp − , (13)
R0 T Pfuel R0 T
−1
4F PO2 0.25 Ac
ηO 2 =i kO exp − , (14)
R0 T 2 Pair R0 T
where kH2 , kCO and kO2 are the coefficients of each equation: kH2 = 2.13 × 108 ,
kCO = 2.98 × 108 and kO2 = 1.49 × 1010 A/m2 . Aa and Ac are the activation energy:
Aa = 1.1 × 105 and Ac = 1.6 × 105 J/mol. In addition, iH2 /O2 and iCO/O2 are the
current densities in the electrolyte arising from the reactions of eqns (6) and (7),
respectively, which are related to the consumption rates of hydrogen and carbon
monoxide as will be discussed later. They are also related to the local current density
i as follows:
i = iH2 /O2 + iCO/O2 . (15)
Here the following relations exist between the electromotive force, activation
overpotential, ionic current density across the electrolyte, and the electric potential
of the anode and cathode, Va and Vc :
EH2 /O2 − (ηH2 + ηO2 ) − iRe he = Vc − Va , (16)

ECO/O2 − (ηCO + ηO2 ) − iRe he = Vc − Va , (17)
where Re is the ionic resistivity of the electrolyte, which can be estimated consid-
ering temperature dependency as suggested by Bessette et al. [19], and he is the
thickness of the electrolyte. From the above relationships, the local current density
i is obtained if the local potential difference Vc − Va is given, or vice versa. In the
present model, interconnects are assumed to work ideally and potential difference
Vc − Va is therefore assumed to be constant along the tube axis. It is therefore
equal to the cell terminal voltage. As will be explained later, the cell terminal volt-
age is given to start the computation and both the current density and potential
fields are calculated by making use of an equivalent electrical circuit model, details
of which will be discussed later. In eqns (16) and (17), note that concentration
overpotential to occur in relation with the nonuniform distributions of participat-
ing chemical species in air and fuel flow passages has been taken into account in
eqns (8) and (9). However, the concentration overpotential to be generated by the
diffusion resistance of participating chemical species inside thin anode and cathode
layers has been ignored. An approximate method to treat this type of concentration
overpotential has been proposed in [33]. However, microstructure of the porous
electrodes, or more specifically tortuosity as well as porosity and permeability of
the porous electrodes, is not normally available. The authors found that the magni-
tude of the ignored concentration overpotential is not large for a case without the
indirect internal fuel reforming but in future this must be revisited more carefully
when appropriate data for the structure of the cell supporting material and porous
electrodes become available.
3.3 Model for internal fuel reforming
Steam reforming of methane proceeds in the fuel feed tube preceding the electro-
chemical reactions at the electrodes in the case of the IIR-T-SOFC. Therefore the
modeling of internal fuel reforming is now discussed. As shown in Fig. 5, the steam
reforming reaction described by eqn (1) and water-gas shift reaction described by
eqn (2) are considered inside the feed tube, and the shift reaction is considered to
further proceed outside the feed tube too. In this study, the reaction rates of the two
reactions, Rst and Rsh , are locally calculated as follows:

−57840
Rst = 1.75PCH 1.2
4
W cat exp , (18)
R0 T
+ −
Rsh = ksh PCO PH2 O − ksh PH2 PCO2 . (19)
Equation (18) is based on an empirical formula suggested by Odegard et al. which

is referred to in reference [34]. Wcat is the filling mass density of the catalyst for
the steam reforming and it is an important parameter in a sense that control of
its distribution can be used as a means to change the distribution pattern of cell
temperature. In some examples to be discussed later, it is actually demonstrated
+ −
how this control is effective. R0 is the universal gas constant. ksh and ksh denote
the rate constants of forward and backward water-gas shift reactions and the value
of Rsh is determined following the method suggested by Lehnert et al. [35]. Their
values vary from one place to another because of the nonuniform distribution of
cell temperature. An important point is that these constants are so large that the shift
reaction is always close to its equilibrium. Chemical equilibrium is represented by
the equilibrium constant which is a function of temperature and is equal to the ratio
between the reactant partial pressures and the product partial pressures as follows:

+
ksh pCO2 pH2 −Gsh 0
Ksh = − = = exp , (20)
ksh pCO PH2 O R0 T
where Gsh 0 is the change of the standard Gibbs free energy accompanied with the
shift reaction. This equilibrium constant is introduced into eqn (19) to calculate the
rate of the shift reaction.
3.4 Governing equations of velocity, temperature and concentration

fields and boundary conditions
Air and fuel are continuously supplied to a fuel cell to keep its appropriate operation
and they flow through their respective flow passages in and around the fuel cell.
Electrochemical reactions are supplied by diffusion of the participating chemical
species from the core flow region toward the reaction sites. Generated heat in the cell
is partly used as a source to supply the energy to support the endothermic reforming
reaction proceeding in the fuel feed tube but its main portion is transferred to and
removed by fuel and air flows. Heat generation is an ineffective part of chemical
energy converted in the fuel cell and therefore occurs at a rate related to the rate
of electricity generation. Electricity generation results from the electrochemical
reaction of the fuel. The fuel consumption rate is related to the electric charge or
ionic transfer rate across the electrolyte. Electric current occurs in the electrodes
where the transferred electric charge is collected. Electric current occurs in a pattern
consistent with the electric potential fields to be established in the electrodes. Ohmic
heat generation occurs in accordance with the electric current in the electrodes
and ionic current across the electrolyte. All these phenomena are interrelated to
each other in a complicated manner. Therefore, all of the equations governing
the phenomena must be integrated in a consistent manner and must be solved
simultaneously in an iterative procedure.
The most basic governing equations to be solved in the numerical model are
the ones for the velocity field, temperature field and concentration field in the fuel
cell. They are different among the three kinds of areas: gas area, solid area and
porous area. For the gas area, the following two-dimensional continuity, momen-
tum, energy and mass transfer equations for laminar flows apply:
∂ρUx 1 ∂rρUr
+ = 0, (21)
∂x r ∂r

∂Ux ∂Ux ∂P ∂ ∂Ux 1 ∂ ∂Ux
ρUx + ρUr =− + µ + rµ , (22)
∂x ∂r ∂x ∂x ∂x r ∂r ∂r

∂Ur ∂Ur ∂P ∂ ∂Ur 1 ∂ ∂Ur µUr
ρUx + ρUr =− + µ + rµ − 2 , (23)
∂x ∂r ∂r ∂x ∂x r ∂r ∂r r

∂T ∂T ∂ ∂T 1 ∂ ∂T
ρCp Ux + ρCp Ur = λ + rλ + Q, (24)
∂x ∂r ∂x ∂x r ∂r ∂r

∂Yj ∂Yj ∂ ∂Yj 1 ∂ ∂Yj
ρUx + ρUr = ρDjm + rρDjm + Sj . (25)
∂x ∂r ∂x ∂x r ∂r ∂r
In the above equations, Ux and Ur are the x- and r-components of the velocity
and T is the temperature. Yj is the mass fraction of chemical species j and Djm is the
mass diffusivity of species j in the multicomponent mixture of gases. As described
above, all of the thermodynamic and transport properties of each chemical species
are treated as local variables varying with temperature and the properties of the
mixture are evaluated based on mixing laws by making use of the mass fraction
of each chemical species. The terms Q and Sj in eqns (24) and (25) denote the
source terms to be described later. The effects of the electrochemical and reforming
reactions on the thermal and concentration fields are included in these source terms.
The effect of radiative heat transfer appears in the matching conditions at gas-
solid interfaces. However, viscous dissipation in the energy equation is neglected.
Thermal diffusion in mass transfer and enthalpy transfer due to species mass transfer
are not considered either in the present model.
Table 1: Thermal conductivity of the solid area.
Anode Cathode Electrolyte Support tube Feed tube
λs [W/mK] 11.0 6.23 2.7 1.1 1.1
For the solid part, the governing equation to be solved is the following heat
conduction equation:

∂ ∂T 1 ∂ ∂T
0= λs + rλs + Q, (26)
∂x ∂x r ∂r ∂r
where λs is the thermal conductivity of the solid. Values of λs of each material are
treated to be constant and specific values used in the present study are shown in
Table 1 [34].
For the porous area, the governing equations derived by the method of volume-
averaging are applied because the microstructure of porous media is generally
too complicated to be considered directly in computations. In the adopted method,
physical values are locally-averaged for a representative elementary volume, which
is sufficiently larger than the scale of the fine structure of the porous medium and
is sufficiently smaller than the scale of the porous body itself [36]. Consequently,
the following transport equations of the averaged physical values are used in the
present model:
∂ρUx 1 ∂rρUr
+ = 0, (27)
∂x r ∂r

1 ∂Ux ∂Ux ∂P 1 ∂ ∂Ux 1 ∂ ∂Ux
ρUx + ρUr =− + µ + rµ
ε2 ∂x ∂r ∂x ε ∂x ∂x r ∂r ∂r
"
µ ρf
− Ux − √ Ux Ux2 + Ur2 , (28)
K K

1 ∂Ur ∂Ur ∂P 1 ∂ ∂Ur 1 ∂ ∂Ur µUr
ρU x + ρU r = − + µ + rµ −
ε2 ∂x ∂r ∂r ε ∂x ∂x r ∂r ∂r r2
"
µ ρf
− Ur − √ Ur Ux2 + Ur2 , (29)
K K

∂T ∂T ∂ ∂T 1 ∂ ∂T
ρCp Ux + ρCp Ur = λeff + rλeff + Q, (30)
∂x ∂r ∂x ∂x r ∂r ∂r

∂Yj ∂Yj ∂ ∂Yj 1 ∂ ∂Yj
ρUx + ρUr = ρDjm,eff + rρDjm,eff + Sj .
∂x ∂r ∂x ∂x r ∂r ∂r
(31)
In principle, the physical values in the above equations represent the local
“intrinsic” phase average of the gas (average over the gas volume, rather than the
Table 2: Parameters of the porous area.
ε [–] K [m2 ] f [–] λs [W/mK] λeff [W/mK] Djm,eff [m2 /s]

√
0.9 1.0 × 10−7 0.088 10.0 ελ + (1 − ε)λs (1 − 1 − ε)Djm
total volume). This is the case for Ux and Ur which denote the gas phase average
of the local gas velocity components. However, this is not the case for temperature.
A two-equation model for the thermal field [37] is not adopted in this study so that
local thermal equilibrium between the two phases is assumed. Therefore, T denotes
the local average both over the gas and solid phases. In addition, ε and K are the
porosity and permeability of the porous medium, respectively. f is the inertia coef-
ficient that depends on the Reynolds number and the microstructure of the porous
medium [38], while λeff and Djm,eff are the effective thermal conductivity and the
effective mass diffusivity of species j, respectively. Values of these parameters used
in this study are shown in Table 2.
The above three pairs of governing equations are properly conjugated to match at
the boundaries between different areas. No-slip condition is applied for the velocity
field at solid surfaces and gas-electrode interfaces. Impermeable condition is also
applied to the artificial confining wall of air flow and the inner and outer surfaces of
the fuel feed tube. At the gas-electrode interfaces, the normal velocity component
is set equal to a value incurred by non-zero mass flow rate at the electrode due to the
electrochemical reaction [16]. The artificial air-flow-confining wall is assumed to be
thermally adiabatic and continuity of heat flux is applied at all of other solid surfaces
and gas-electrode interfaces. At the gas-electrode interfaces, mass production or
consumption rate of each chemical species is equated to a value regulated by the
electrochemical reaction, which will be briefly discussed later. Other solid surfaces
are treated as unreactive. In starting computation, inlet conditions must be given
for air and fuel flows. In addition, the cell terminal voltage is given as an external
condition to start the computation and electric current density to be achieved under
the given terminal voltage is calculated. Through iteration, velocity, temperature
and concentration fields are determined as their converged solutions.
3.5 Discretization scheme
All the above governing equations can be solved only numerically and introduction
of some sort of discretization is indispensable. So, a short description is given here
about the discretization scheme adopted in the present model just as a hint for some
readers who may want to develop their own home-made program. However, it may
be worth noting that a number of commercial codes available for general use in a
variety of thermo-fluid problems may be transformed into one applicable for fuel
cell simulation.
All the above governing equationsl listed in section 3.4, eqn (21) through
eqn (31), and eqns (38) and (39) to appear in section 3.6 also can be expressed in
Table 3: Variables and coefficients of each governing equation.
Eqn φ ρ̂ Ŝ
∂P
(22) Ux ρ µ −
∂x
∂P µUr
(23) Ur ρ µ − − 2
∂r r
(24) T ρCp λ Q
(25) Yj ρ ρDjm Sj
(26) T 0 λs Q
"
ρ µ ∂P µ ρf
(28) Ux − − Ux − √ Ux Ux2 + Ur2
ε2 ε ∂x K K "
ρ µ ∂P µUr µ ρf
(29) Ur − − 2 − Ur − √ Ur Ux2 + Ur2
ε2 ε ∂r εr K K
(30) T ρCp λeff Q
(31) Yj ρ ρDjm,eff Sj
the following general form:

∂φ ∂φ ∂ ∂φ 1 ∂ ∂φ
ρ̂Ux + ρ̂Ur = + r + Ŝ, (32)
∂x ∂r ∂x ∂x r ∂r ∂r
where φ represents the variable of interest. ρ̂ and are the coefficients of the
convection term and the diffusion term, respectively, while Ŝ denotes the source
term. Specific values of these factors are summarized in Table 3 for several equa-
tions. Although there are a variety of discretization schemes available, the finite
volume method is adopted in the present model with the staggered grid system. In
the staggered grid system, two types of grid arrangements are provided: the velocity
grids to store the velocity components and the normal grids for the other scalars
such as pressure, temperature and concentrations. The velocity grids are shifted in
position from the normal grids as illustrated in Fig. 6 to stably couple the velocity
and pressure fields [39].
Integrating over each control volume for the normal grids, painted grey in Fig. 6,
eqn (32) can be discretized into the following form:
{(ρ̂Ux φ)e − (ρ̂Ux φ)w }rm r + {(r ρ̂Ur φ)n − (r ρ̂Ur φ)s }x

∂φ ∂φ ∂φ ∂φ
= − rm r + r − r x + Ŝ · rm rx
∂x e ∂x w ∂r n ∂r s
(33)
Figure 6: Staggered grid for the Finite Volume Method.
in which the subscripts e, w, n and s indicate the positions of four faces of the control
volume under consideration, and rm denotes the arithmetic mean of rn and rs . Note
that the control volume for velocity components is different from the counterpart
for a scalar because the staggered grids are employed. In order to interpolate the
fluxes of convection and diffusion at the interfaces expressed by e, w, n and s,
the Power-Law scheme suggested by Patankar [39] is adopted here. As a result,
the fluxes on the interfaces can be represented by the values on the grid points as
follows:

∂φ
ρ̂Ux φ − rm r = aE (φP − φE ), (34)
∂x e
where aE is a coefficient determined by the Power-Law scheme. This is an exam-

ple for the interface e, and similar expression can be adopted on the other inter-
faces. Consequently, eqn (33) can be transformed into the following algebraic
equation:
aP φP = aE φE + aW φW + aN φN + aS φS + b, (35)
where
aP = aE + aW + aN + aS , (36)
b = Ŝ · rm rx. (37)
The above discretized equations for velocity, temperature and concentration have
to be solved simultaneously since gas properties are dependent on temperature
and concentration of each gas component. To solve for velocity, the pressure field
must be determined because the pressure gradient is included in the momentum
equations. In order to solve both the velocity and pressure fields concurrently,
the SIMPLE algorithm put forward by Patankar et al. [39] is adopted in this model.
Other details of discretization may be found in [39].
3.6 Equations for electric potential and electric circuit
Here is described the method to calculate the electric potential fields to be generated
in electrodes and accompanying flow of electric current. The mathematical descrip-
tion in this section is given directly in discretized form for the reader’s convenience.
In the present model, it is assumed that ionic current flows only radially across the
electrolyte but electric current through electrode layers only in the circumferential
(θ) direction as shown in Fig. 4 based on the fact that the electric resistivities of the
electrodes are much smaller than the ionic resistivity of the electrolyte. In addition,
the interconnect is treated to act ideally so that the potential difference between
its position attaching the anode, θ = 0◦ , and its position attaching the cathode,
θ = 180◦ , is taken to be equal to the cell terminal voltage.
Figure 7 illustrates the grid system adopted in the calculation of the electric
potential field in the cell tube and accompanying electric current. According to
Kirchhoff’s law, the following discretized equations can be derived for both of the
two electrodes:
(iu − id )ha − iPP · re θ = 0, (38)
(iu − id )hc + iPP · re θ = 0, (39)
where iPP is the current density given by eqn (15) along a particular line connecting
the grid points P and P located across the electrolyte as illustrated in Fig. 7. iu ,
id , iu and id are the values of current density to occur respectively along the
four lines connecting the two grid points separated in the circumferential direction,
U and P and P and D in the anode layer and U and P and P and D in the
cathode layer. re is the radius of the mid surface of the cylindrical electrolyte shell
layer.
Figure 7: Grid for the equivalent electrical circuit.

According to Ohm’s law, the electric current in the electrodes can be related to the
difference in the electric potential between the related grid points:
ha ha
iu h a = (VU − VP ), id ha = (VP − VD ), (40)
Ra ra δθu Ra ra δθd
hc hc
iu hc = (VU − VP ), id hc = (VP − VD ), (41)
Rc rc δθu Rc rc δθd
where Ra and Rc are the electric resistivities of the anode and cathode, respectively.
They are estimated by the empirical formula suggested by Bessette et al. [19] in
this model. Rewriting eqn (15) with Ẽ defined by the following equation:
Ẽ = EH2 /O2 − (ηH2 + ηO2 ) = ECO/O2 − (ηCO + ηO2 ), (42)
the electric current across the electrolyte iPP can be represented by the potential
difference and Ẽ as follows:
re θ
iPP re θ = {ẼPP − (VP − VP )}. (43)
Re he
Hence eqns (40) through (43) can be rewritten into the following algebraic equa-
tions:
aP VP = aU VU + aD VD + aPP VP − b, (44)
aP VP = aU VU + aD VD + aPP VP + b, (45)
where
ha ha hc hc
aU = , aD = , aU = , aD = , (46)
Ra ra δθu Ra ra δθd Rc rc δθu Rc rc δθd
re θ re θ
aPP = , b= Ẽ, (47)
Re h e Re h e
aP = aU + aD + aPP , (48)
aP = aU + aD + aPP , (49)
eqns (44) and (45) can be solved numerically with a scheme similar to the one to
solve momentum, heat and mass transfer equations described in the Section 3.5.
3.7 Mass production or consumption rate of each chemical species through

electrochemical and reforming reactions
The mass production rate of each chemical species by the electrochemical reactions
is directly related to the electric current density caused by each of the electrochemi-
cal reactions, eqns (6) and (7), and its value, ṡj , is tabulated in Table 4. For example,
the mass consumption rate of hydrogen, ṡH2 , is equal to the product of the molecular
Table 4: Mass production or consumption rate of each species by the reactions

(6) and (7).
Eqn ṡH2 ṡH2 O ṡCO ṡCO2 ṡO2

iH2 /O2 iH2 /O2 iH2 /O2
(6) − MH2 MH2 O 0 0 − MO2
2F 2F 4F
iCO/O2 iCO/O2 iH /O
(7) 0 0 − MCO MCO2 − 2 2 MO2
2F 2F 4F
Table 5: Mass production or consumption rate of each species by the reactions

(1) and (2).
Eqn SH2 SH2 O SCO SCO2 SCH4
(1) 3Rst MH2 −Rst MH2 O Rst MCO 0 −Rst MCH4

(2) Rsh MH2 −Rsh MH2 O −Rsh MCO Rsh MCO2 0
iH2 2/O
weight MH2 and the molar consumption rate of hydrogen 2F , where the Faraday
constant F designates the electric charge of one mole of electrons. It must be noted
here that two moles of electrons participate into the electrochemical reaction of
one mole of hydrogen as seen in Fig. 5. These species mass production rates have
nonuniform distributions in the circumferential direction because the current den-
sity is nonuniform in that direction. In the present quasi-two-dimensional model,
therefore, they are peripherally averaged before being introduced into axisymmetric
two dimensional mass transfer equations for the concentration fields.
The production or consumption rate of each chemical species by the reforming
and shift reactions is also calculated as shown in Table 5. The value of mass produc-
tion or consumption rate for each chemical species is introduced into the species
mass transfer equation as a part of its source term.
3.8 Model for thermodynamic heat generation rates
The thermodynamic heat generation rates related to the electrochemical reactions

are equal to the generation rate of the ineffective part of the input energy, which
cannot be converted into electricity, and are calculated as follows:

(−HH2 /O2 )
q̇H2 /O2 = − Ẽ iH2 /O2 , (50)
2F

(−HCO/O2 )
q̇CO/O2 = − Ẽ iCO/O2 . (51)
2F
These heat generation rates are included in the calculations of the energy equation as
a part of its source term. The thermodynamic heat generation rate by the reforming
reactions, are calculated based on the reaction rates discussed in the Sections 3.3
and 3.7, as follows:
Qst = −Hst Rst , (52)
Qsh = −Hsh Rsh , (53)
where Hst and Hsh are the enthalpy change accompanied with each reaction.
3.9 Ohmic heat generation
With the current density and electric potential fields obtained in Section 3.6, ohmic
heat generation rates in the electrodes and electrolyte are calculated as follows:
iu2 δθu + id2 δθd iu2 δθu + id2 δθd

QP = Ra , QP = Rc , (54)
δθu + δθd δθu + δθd
QPP = iPP
2
Re . (55)
These heat generation rates are nonuniform in the θ-direction. In the present quasi-
two-dimensional method, they are peripherally averaged before they are introduced
into the source term of the axisymmetric two-dimensional energy equation for the
thermal field [15, 16]. Averaged results for the heat generation rates read as follows:
π π
θ=0 qP θ qP θ
Qa = , Qc = θ=0 , (56)
π π
π
qPP θ
Qe = θ=0 . (57)
π
3.10 Radiation model
Now modeling of radiation heat transfer will be discussed. In the present model,
the radiation heat transfer between the inner surface of the cell tube and the outer
surface of the feed tube is considered. In the case of the IIR-T-SOFC, this plays an
important role in transferring the heat generated by the electrochemical reactions
at the tubular cell to the feed tube, and therefore, to the reaction site of endothermic
fuel reforming.
In the case of a tubular SOFC, the cell tube is slender in geometry. For example, in
the case of a tubular IIR-T-SOFC to be studied in this chapter, the fuel flow passage
is an annular space between two cylindrical surfaces facing to each other, i.e. anode
surface and fuel feed tube outer surface. The width of the annular space is 2.4 mm.
This is quite small compared to the cell tube length, 500 mm. This slenderness of
the tubular cell has important implications for thermal radiation.
The first of the implications to note here is that the optical length of the gas
media flowing through the annular space is small. Emittance of the gas layer is
given as the product of the optical length and absorption coefficient of gas which is
an increasing function of the partial pressure of each chemical species participating
in radiation. So absorption into or emission from gaseous media can be neglected

or the annular space can be treated transparent to radiation so long as the operation
pressure is not very high.
Description of the modeling of radiation heat transfer will begin with an expla-
nation about the computation method of radiation heat transfer for general use.
Certainly, the first feature of thermal radiation in a tubular fuel cell just noted
above is taken into account. For more details of this part of discussion, refer to the
reference [40]. Now consideration is given to an enclosure covered by the inner
surface of the anode and the outer surface of the fuel feed tube and this enclosure
is considered to consist of N diffusely emitting and diffusely reflecting gray sur-
faces. The surfaces are each isothermal. Surface j has temperature Tj and emittance
εj . The net rate of heat loss flux qi from a surface i is equal to the difference between
the emitted radiation and absorbed portion of the incident radiation. Therefore, it
is given as:
qi = εi σTi4 − εi Ii , (58)
where Ii is the radiation incident on surface i per unit time and unit area and
absorptance of the surface i is replaced by the emittance εi taking into account
Kirchhoff’s law of radiation.
Here is introduced the radiosity Bj of a surface j. The radiosity is the sum of the
reflected and emitted radiant fluxes. So the following relationship holds:
Bj = εj σTj4 + (1 − εj )Ij . (59)
A fraction of Bj is directed to the surface i and contributes to the incident radiation

on the surface i. Such contribution from the surface j to the incident radiation on
surface i per unit area is now calculated as:
Iji = Fji Bj Aj /Ai = Fij Bj , (60)
where Fji and Fij are the shape factors and the reciprocity rule Aj Fji = Ai Fij has
been used. Now similar contributions come from every surface so that Ii is finally
expressed as follows:

N
N
Ii = Fij Bj = Fij (εj σTj4 + (1 − εj )Ij ). (61)
j=1 j=1
Iterative calculation is needed to determine the value of Ii or the value of qi with

the determined value of Ii . However, this iterative procedure needs a large com-
putational load if a large number is assigned to N . Actually the computation is
excessively large for the present purpose of using the result for optimum design
or thermal management of a tubular SOFC. Therefore, introduction of a certain
degree of approximation into the modeling is desirable. The procedure followed is
to reduce the number of surfaces considered in the computation.
The approximation adopted in the present model is again based on the slenderness
of the tubular cell. For just an example, attention is paid to a geometrical situation
Figure 8: Geometry of two facing surfaces.
illustrated in Fig. 8. Two small elementary surfaces having 20 mm length, surfaces

1 and 2, are positioned in parallel to each other at a distance of 2.4 mm in this
figure. An angle viewing one of them, say surface 1, from a point M on another
surface 2 is about 153 degrees. The viewing angle of other parts, say surface 3,
outside the surface 1 from the point M is only 27 degrees. So, the value of F21 is
close to 1 and both of F23 and F32 are close to zero. Naming the surfaces situated
on both sides of the surface 2 as surface 4, F24 = 0 and F22 = 0 hold. From the
same reasoning as above both of F14 and F41 are close to zero too. So, based on
the slender geometry of the tubular cell, radiation heat transfer between the two
surfaces 1 and 2, directly facing each other, is considered. Then, it can be assumed
that only two shape factors F21 and F12 take a non-zero value. Under this treatment,
eqn (61) can be very much simplified. Certainly, by increasing the number of the
surfaces, a better approximation is obtained but at the cost of computational time.
Now assigning the suffix 1 to the anode inner surface and 2 to the fuel feed tube
outer surface, F21 = 1, F12 = A2 /A1 and F11 = 1 − A2 /A1 hold under the adopted
assumption. Thus eqn (61) can be approximated as:

2
I1 = F1j Bj = F11 B1 + F12 B2 , (62)
j=1

2
I2 = F2j Bj = F21 B1 (63)
j=1
and eqn (60) as:

B1 = ε1 σT14 + (1 − ε1 )I1 , (64)
B2 = ε2 σT24 + (1 − ε2 )I2 . (65)
Similarly, eqn (58) can finally be approximated as:
q1 = ε1 σT14 − ε1 I1 , (66)
q2 = ε2 σT24 − ε2 I2 . (67)
In the present model, these equations are locally used at each axial position and
these radiant fluxes are included in the calculation of heat transfer as a part of the
heat flux both at the inner surface of the anode and at the outer surface of the fuel
feed tube.
Monochromatic radiation properties of ceramics are highly dependent on wave-

length [41, 42]. Pore size of the electrodes and support materials is close to the
wavelength of thermal radiation. Thinning the electrodes creates a similar situa-
tion that the electrode thickness comes closer to the wavelength too. This situation
makes the problem complicated. Pore microstructure affects the wavelength depen-
dency of the radiation properties of the fuel cell element materials and thin element
layer can become semi-opaque. Therefore, sophisticated treatment of the radiation
properties of the electrode and support materials is desirable. However, radiation
properties of such materials have not been thoroughly studied yet. Therefore, for
simplicity, the absorption coefficient of the solid materials of the fuel cell is assumed
to be constant in the present model. This assumption however must be revisited
when the micro structure and radiation properties of the fuel cell element materials
have been thoroughly studied.
Radiation heat transfer among the tubular cells in the module can be another
factor affecting the performance of the fuel cell. This is very true in the case of the
module having an “in-stack reformer”. However, in the case of the present IIR-
T-SOFC or a tubular SOFC using fuel reformed outside the module, every single
tubular fuel cell located in the core of the module is surrounded by other tubular
fuel cells similar in conditions including the temperature distribution on the outer
surface of the cell tube. Again, the space between neighbouring tubular cells is
small since tubular cells are stacked in a compact way in the module. So neglecting
radiation heat transfer between the outer surfaces of neighbouring tubular cells is
less serious in such cases of a tubular SOFC. However, it must be studied more
seriously in the case of the module having an in-stack reformer and this heat transfer
is a problem to be carefully studied in the modeling of cell stack modules.
3.11 Overall picture of the model
In closing this section, an overall picture of the model is presented in Fig. 9. One of
the main parts of this computation is the calculation of the temperature and concen-
tration fields assuming their two-dimensionality. Based on the calculated tempera-
ture and species concentrations, the electromotive force and several resistances are
calculated for the calculation of electric current and potential fields allowing their
circumferential nonuniformity, which is another main part of this model. Based on
the results of the electric fields, the peripheral average of the ohmic heat generation
is calculated. Local current density is related to the mass production rate of partici-
pating chemical species. The heat generation rates and species mass production rates
thus calculated are fed back to the calculation of the temperature and concentration.
The gas properties and the reforming reaction rate are also interrelated with the
temperature and concentration fields. Therefore, all of the above calculations are
iterated simultaneously. As a result of iteration, all of the temperature, concen-
tration and electric fields of the cell are obtained. Finally the cell performance,
such as average current density, output power or energy conversion efficiency, is
obtained. A more detailed description of this overall scheme can be found in the
references [8, 15, 16].
Figure 9: Overall picture of the IIR-T-SOFC model.
4 Results and discussion

In this section, some calculation results obtained by making use of the numerical
model described in the preceding sections are discussed. In section 4.1, discussion
is first given to some results of the model applied to a conventional hydrogen-
fuelled tubular SOFC and its validation is presented by comparing some results with
published experimental data. How this model is powerful as a tool for optimum
design and thermal management of a tubular SOFC is also demonstrated in this
section. In sections 4.2 and 4.3, discussion will be given to some results obtained
by applying the model to an IIR-T-SOFC. In section 4.2, the results for a “Base
case” will be presented so as to illustrate the essence of the phenomena in a tubular
cell accompanying the indirect internal fuel reforming. Then, methods to lower
and/or to make more uniform the temperature of the cell will be discussed along
with other results in section 4.3.
4.1 Results for a cathode-supported tubular SOFC without accompanying

indirect internal reforming [16, 43, 44]
In this type of tubular SOFC, as shown in Fig. 2, air is supplied inside the tubular cell
through the air feed tube and fuel flows around the tubular cell in the axial direction.
So the innermost layer of the cell is the cathode and outermost layer is the anode.
Electrolyte material is assumed to be Yttrium-stabilized Zirconia (YSZ) and the
fuel is pure hydrogen. So the parts of the present model concerning fuel reforming
and electrochemical reaction of carbon monoxide are inactivated. In this case, heat
absorption by the fuel reforming does not exist either and heat transfer from the
cathode inner surface to the outer surface of the air feed tube is less important.
For this reason, radiation heat transfer between the two surfaces is ignored in this
Table 6: (a) Inlet conditions.
Inlet species and their

weight concentrations Inlet temperature Inlet pressures
Fuel side H2 89% 870 ◦ C 1.013 × 105 Pa

H2 O 11% 870 ◦ C 1.013 × 105 Pa
Air side O2 21% 600 ◦ C 1.013 × 105 Pa
N2 79% 600 ◦ C 1.013 × 105 Pa
Table 6: (b) Geometry of cathode supported cell.
Components’ Thickness
Supporting tube 1500 µm

Cathode 1000 µm
Electrolyte 50 µm
Anode 150 µm
Diameter
Inner side of air-inducing tube 8.0 mm

Outer side of air-inducing tube 9.0 mm
Inner side of supporting tube 13.8 mm
Outer side of anode 19.2 mm
Outer diameter of fuel channel 29.2 mmφ
Length of cell unit 500 mm
particular example. Inlet conditions of air and fuel flows are tabulated in Table 6(a).
Geometry of a tubular cell is kept to be the one given in Table 6(b) unless otherwise
stated. Fuel and air utilization factors are set to be 0.85 and 0.167 respectively.
Illustration of local quantities shown in Figs 14–16 is for the case of an electric
current density of 3500 A/m2 . Other details are found in reference [16].
Figure 10 illustrates the current-voltage (I-V) diagram of the studied tubular
SOFC. Calculated results agree fairly well with the experiments performed by
Hagiwara et al. [45]. For small current density conditions, simulated results are a
little higher than the experiments. This over-prediction may be related to a simple
treatment of activation overpotential. It is assumed in this particular computation
that the activation polarization is proportional to current density and the value
of the proportionality constant is adjusted at an optimum current density around
3000 A/m2 . However, at smaller current density, the cell temperature is lower, and
a larger activation overpotential or a larger proportionality constant should have
been adopted. In the examples to be discussed in the Section 4.2 and 4.3, more
sophisticated treatment of activation overpotential is adopted as already discussed
Figure 10: Cell voltage-current density diagram.
Figure 11: Conversion efficiency and output power.
in Section 3.2. Except for this point, obtained results look reasonable and the present
model is basically validated.
Figure 11 illustrates the cell electricity generation efficiency and the power to
be achieved with a single tubular cell. With an increase of current density, power
to be obtained increases up to a certain value but efficiency decreases. This is
the normal tendency and optimum current density is determined by a trade-off
between the two quantities having opposite tendency. Figure 12 shows the cell
terminal voltage for four cases of tubular SOFCs having different tube length.
All the calculated results almost overlap. In this computation, the fuel utilization
factor is kept constant. This means that both air and fuel are supplied at a rate
proportional to the cell tube length. Larger air flow rates enhance the cooling of
the cell but this is countered by larger ohmic heating at lower temperatures. Under
the same electric current density, therefore, both the concentrations of fuel and
oxygen decrease at a rate proportional to the tube length almost keeping similarity
Figure 12: Effect of cell length on output voltage at constant fuel utilization factor.
Figure 13: Effects of tube diameter on energy efficiency.
in the temperature field among the three cases. This results in such a condition that
both electrochemical reaction and heat generation proceed similarly or in a manner
proportional to the cell tube length. Figure 13 shows the effects of the cell tube
diameter. In the case of a cell tube having larger diameter, the length not only of
electrolyte but also of electrodes is larger. Then larger ohmic loss occurs, reducing
the terminal voltage, and ohmic heat generation also becomes larger because of
the larger collection of electric current per unit length of interconnect. Therefore,
a thinner tubular cell can have a higher performance. Figure 14 shows the axial
distributions of the local EMF and current density. Both quantities show the same
tendency to decrease in the downstream direction. Concentrations of hydrogen and
oxygen in the first half of the cell are still high so that the EMF in that region is high
too. Similarity between the current density distribution and the EMF distribution
indicates that the axial electric current is minor compared to the peripheral one or
that interconnect works well.
Figure 14: Distribution of EMF and current density.
Figure 15: Temperature distribution in a cell.
Figure 15 shows the temperature distribution in the tubular SOFC unit. As for the
fuel flow temperature, it rises downstream of the inlet and approaches 1000 ◦ C, then
keeps an almost uniform value in the middle part of the cell and slightly decreases
toward the exit end. In the air feed tube, air is heated after it flows into the tube.
A noticeable transverse gradient of air temperature is found to exist in the annular
space over the whole length of the air feed tube. This gradient drives heat transfer
from the cell structure, where heat generation occurs, to the air flow in the annular
space. This ensures effective cooling of the cell. Radiative heat transfer, ignored in
the present computation, may flatten the streamwise temperature non-uniformity
but is not expected to change the trend of the temperature distribution significantly.
This is because the convective heat transfer considered in the present study is effec-
tive already to reduce the differences in surface temperature among various parts of
the cell. So air acts as a coolant of the cell. One point related to the convective heat
transfer should be noted here. In various numerical models proposed so far, a con-
stant value for the Nusselt number is used in evaluating the local heat removal rate
to the air flow. However, this is not a good assumption and actually the heat transfer
coefficient is significantly nonuniform along the axial direction [46]. In relation to
the discussion of the concentration polarization, distributions of hydrogen and oxy-
gen concentrations are presented in Fig. 16. In the fuel passage, the concentration of
hydrogen decreases downstream, especially steeply in the first part after the inlet.
However, the radial non-uniformity of the hydrogen concentration is not significant
Figure 16: Hydrogen and oxygen concentrations.
so that the related concentration polarization is judged to be minor. In contrast to

this, radial non-uniformity of oxygen concentration is noticeable. This difference is
attributed to the difference in the diffusivity, i.e. the diffusivity of hydrogen is much
higher than that of oxygen. Therefore, concentration polarization induced by the
nonuniform distribution of species concentration in the flow space is not negligible
at the cathode side but it can be ignored on the anode side.
4.2 Results for the Base case with accompanying indirect

internal reforming
In the examples for which results will be discussed here, cell tube geometry is
assumed as is shown in Fig. 5 to be of 500 mm length, 6.9 mm inner radius and
9.6 mm outer radius. An adiabatic cylindrical wall artificially introduced to confine
air flow for computational convenience has 14.6 mm in radius. Other details of cell
geometry are tabulated in Table 7.
The inlet and outlet conditions of the fuel and air are summarized in Table 8 and
computational conditions for the Base case are shown in Table 9. In the present
study, fuel mean velocity at the inlet is set equal to such a value that the limiting
value of the average current density is about 5000 A/m2 (e.g. 0.923 m/s when the
inlet temperature is 800 ◦ C). In contrast, the air mean velocity at the inlet is set
to be 2.0 m/s except for some special cases to be specified later. The fuel and
air temperatures at the inlets are changed from one case to another as will be
described in a later section but unless otherwise stated they are the values given
in the table. The temperature at the closed end of the cell tube is set equal to the
air inlet temperature. Catalyst distribution “U-20” means that the catalyst mass
density, Wcat , is set at 2.0 × 106 g/m3 uniformly inside the feed tube. This is such a
value that the supplied methane fuel can be fully reformed even when the electrical
circuit is open. In this case, fuel reforming is thermally supported just by the heat
transferred from hot air and fuel flows.
The case number for the Base case 2-800U20 indicates that the inlet velocity of
the air is 2 m/s, the inlet temperature of the fuel and air is 800 ◦ C, and the catalyst
distribution is the pattern U-20. “Neuman” appearing in the table means that zero
axial gradient is assumed at the outlet for the variable under concern as its boundary
condition. As mentioned before, the cell terminal voltage is given at the start of the
computation as a condition to calculate the electric current and potential fields in
Table 7: Size of fuel cell elements.
Components’ Thickness
Supporting tube 1500 µm

Anode 150 µm
Electrolyte 50 µm
Cathode 1000 µm
Diameter
Inner side of feed tube 8.0 mm

Outer side of feed tube 9.0 mm
Inner side of supporting tube 13.8 mm
Outer side of cathode 19.2 mm
Outer diameter of fuel channel 29.2 mm
Length of cell unit 500 mm
Table 8: Gas inlet and outlet conditions.
Velocity Temperature Molar fraction [−] Total pressure [Pa]
Fuel inlet Poiseuille Constant H2 0.200 110000

flow H2 O 0.500
CO 0.020
CO2 0.030
CH4 0.250
Air inlet Plug flow Constant O2 0.209 110000
N2 0.791
Outlet Neuman Neuman Neuman Neuman
Table 9: Computational conditions of the Base case.
Velocity Velocity Temperature Temperature

(Fuel inlet) (Air inlet) (Fuel inlet) (Air inlet) Catalyst
[m/s] [m/s] [◦ C] [◦ C] distribution
Base case 0.923 2.0 800 800 U-20

2-800U20
order to identify the given external load of the cell. Unless otherwise stated, the
results of the Base case are shown with particular emphasis on the condition in
which the cell terminal voltage is set at 0.55 V and accordingly the average current
density becomes 3926 A/m2 .
4.2.1 Thermal and concentration fields

Figure 17 shows velocity vectors inside and outside the cell tube. Note that this
figure is not to scale: the radial (r) direction is magnified ten times larger than
the axial (x) direction. The fuel flow is accelerated inside the feed tube due to an
increase in the total number of moles caused by the steam reforming. Fuel flows
into the cell tube from the open end of feed tube and flows back to the tube outlet
through annular fuel flow passage. Fuel flow inside the tube is almost in a stagnant
condition at the bottom end.
Temperature contours for the same case are shown in Fig. 18. Note that the darker
tone corresponds to the higher local temperature. The temperature in and around
the cell basically becomes higher downstream in relation to the air flow because
of the heat generation accompanying the electrochemical process. However, the
temperature is noticeably lower near the inlet of the fuel due to the endothermic
effect of the steam reforming reaction.
Here the x-direction distributions of the temperature inside the electrolyte and the
feed tube are shown in Fig. 19. Conspicuous non-uniformity of temperature exists
as seen in the figure. This is not the case of Siemens-Westinghouse type hydrogen-
fuelled tubular fuel cell [16] and must be avoided because it can lead to thermal
crack failures of the cell. Furthermore, the maximum temperature of the electrolyte
Figure 17: Velocity vectors around the cell for an average current density of
3926 A/m2 in the Base case (2-800U20).
Figure 18: Temperature field in and around the cell for an average current density
of 3926 A/m2 (2-800U20).
Figure 19: Local temperature of the electrolyte and the feed tube for an average
current density of 3926 A/m2 (2-800U20) and that in a similar case but
not considering the effect of radiation.
reaches about 1050 ◦ C, which is the hot spot temperature to be avoided from the
material view point. By the way, the broken lines in Fig. 19 show the result in which
the effect of radiation described in section 3.4 is excluded. From this result, it is
confirmed that the radiation contributes noticeably to the heat transfer between the
cell tube and the feed tube in this case with accompanying indirect internal fuel
reforming.
Figure 20 shows mole fraction contours of hydrogen, steam, CO, CO2 , methane
and O2 , respectively. Note that the grey tone levels differ among each chemical
species. Inside the feed tube, hydrogen and CO are produced while steam and
methane are consumed by the steam reforming. Methane is rapidly reformed near
the feed tube inlet because the amount of the catalyst is excessive for the operating
condition of the cell under study. Meanwhile, outside the feed tube, hydrogen and
CO are consumed while steam and CO2 are produced mainly by the electrochemi-
cal reactions. The reason for the considerable change of the mole fraction occurring
at the bottom of the cell is that the fuel flow stagnates while the electrochemical
reactions take place there. As for the air side, oxygen is consumed by the electro-
chemical reaction. However, variation of its mole fraction is not so large because
the flow rate of air is adjusted so as to effectively remove the generated heat and
is normally in excess of that required to complete the electrochemical reactions. In
other words, the oxygen utilization factor is set normally as low as 0.2 or 0.3.
The molar flow rates of each chemical species inside and outside the feed tube
are shown in Figs 21(a) and (b), respectively. Characteristics of the electrochemical
and reforming reaction described above can be confirmed in these figures.
4.2.2 Electric potential and current fields

Distribution of electromotive force (EMF) is shown in Fig. 21. Although two dif-
ferent values of EMF, EH2 /O2 and ECO/O2 , are calculated in this study as described
Figure 20: Mole fraction distributions of each chemical species for an average
current density of 3926 A/m2 (2-800U20).
in section 3.6, these two differ very slightly from each other by the reason that
the shift reaction is very close to equilibrium everywhere in the cell. Hence only
EH2 /O2 is plotted here as the EMF. As seen in this figure, the EMF strongly depends
on the mole fraction of hydrogen at the anode surface: the EMF has a peak near the
outlet of the feed tube and becomes lower toward the cell tube outlet. However, it
becomes a little higher near the cell tube outlet again due to the larger change of
Gibb’s free energy, G, at lower temperature.
Figure 21: Molar flow rate of each chemical species of fuel for an average current
density of 3926 A/m2 (2-800U20).
Figure 22: Electromotive force and current density distributions for an average
Distribution of current density is also shown in Fig. 22. Note that the distribution
plotted in this figure is the peripheral average of local current density at each
axial position. This figure indicates that the distribution of current density is not
similar in shape to that of EMF. This is because the current density is also related
to the activation overpotential and the ohmic loss which depend strongly on the
temperature. As shown in Fig. 23, both the activation and ohmic overpotentials are
significantly reduced around x = 0.4, where the cell temperature takes a peak value
(see Figs 18 and 19). That is to say, the current density distribution depends on the
thermal field as well as, or more than, on the concentration field. At the same time,
it is also true that the temperature tends to become higher where the large current
is generated, so that a kind of enhancement due to a mutually assisting interaction
exists between the cell temperature and the electric current.
Incidentally, the ohmic loss in the anode and cathode is closely related to the dis-
tribution of the current density in the θ-direction, which is essentially non-uniform.
Details can be found in the references [13, 14].
Figure 23: Activation overpotential and ohmic loss in the electrolyte for an average
Figure 24: Temperature of the electrolyte for several different average current
densities (2-800U20).
4.2.3 Power generation characteristics

Before moving into the discussion on the power generation characteristics of the
fuel cell, results of the electrolyte temperature for several cases of different average
current densities are given here. Figure 24(a) shows the x-direction distribution of
electrolyte temperature for the cases of average current density 2191, 3081 and
3926 A/m2 . All the three distributions have similar features as those described in
section 4.1.1, but temperature becomes higher on the whole as the average current
density becomes larger. The relation between the average current density and the
maximum and average temperatures of the electrolyte is shown in Fig. 24(b).
The power generation characteristics of the cell are significantly affected by the
cell temperature. Figure 25 shows the cell-averaged EMF, activation overpotential
and ohmic loss in the electrolyte for various values of average current density. The
cell terminal voltage, which is practically given as the external condition to start
the calculation in this study, is also plotted here. From this figure, the following
Figure 25: Average EMF and overpotentials vs. average current density
(2-800U20).
Figure 26: Output power and energy conversion efficiency vs. average current
density (2-800U20).
things are confirmed. First, the cell-averaged EMF goes down at a nearly con-
stant rate as the average current density increases. This is largely because the local
concentrations of hydrogen and CO decrease due to their larger consumption rates.
The activation overpotential does not increase proportionally to the average current
density countered by the rise in temperature as mentioned above. The ohmic loss
in the electrolyte also does not increase for the same reason. However, the ohmic
loss in the anode and cathode, which is basically larger than that in the electrolyte
because the electric current flows long distances there in the θ-direction as men-
tioned in section 3.7, increases nearly in proportion to the average current density.
From these results, the output power and energy conversion efficiency of the
cell are also obtained as shown in Fig. 26. Here the efficiency is calculated based
on the lower heating value (LHV) of the fuel consumed in the cell. Judging from
Table 10: Computational conditions for the additional cases.
Velocity Velocity Temperature Temperature

(Fuel inlet) (Air inlet) (Fuel inlet) (Air inlet) Catalyst
[m/s] [m/s] [◦ C] [◦ C] distribution
2-750U20 0.880 2.0 750 750 U-20

4-800U20 0.923 4.0 800 800 U-20
2-800U02 0.923 2.0 800 800 U-02
2-800L02 0.923 2.0 800 800 L-02
these calculated results, fairly high performance seems to be achieved under the
conditions for the Base case.
4.3 Strategies for the ideal thermal field
As described in section 4.1.1, the thermal field in the cell obtained for the Base
case is undesirable in a sense that the thermal field is seriously non-uniform and the
maximum temperature is rather high. Hence the strategies to flatten and/or lower
the distribution of cell temperature are discussed here.
Decreasing the inlet temperatures of the gases is a primary method to lower the
cell temperature. Increasing the air flow rate also seems to be effective when the
cell temperature is higher than the inlet temperature of the air. To achieve a near-
uniform thermal field, controlling the density distribution of catalyst inside the feed
tube should be useful. In order to examine those effects, calculations for the four
cases summarized in Table 10 have been conducted.
In Case 2–750U20, the inlet temperatures of the gases are decreased down to
750 ◦ C and accordingly the fuel inlet velocity is reduced to 0.880 m/s to keep its
molar flow rate at the same value as the Base case. In Case 4-800U20, the air flow
rate is doubled from the Base case. In Case 2-800U02 and 2-800L02, the density
distribution of catalyst is changed to “U-02” and “L-02”, which are described
in Fig. 27. The amounts of the catalyst in U-02 and L-02 are optimised for the
conditions of average current density about 4000 A/m2 . That is, fuel is completely
reformed just near the end of the feed tube under such conditions. The calculation
results for the above four cases are discussed below in sequence.
4.3.1 Effect of gas inlet temperature

The temperature contours obtained at the average current density 4043 A/m2 in
Case 2-750U20 are shown in Fig. 28, and the electrolyte temperature distribution
in the x-direction is shown in Fig. 29. Owing to the lower gas inlet temperatures, the
cell temperature decreases on the whole. However, the maximum temperature is not
so reduced, and thus the non-uniformity of the temperature becomes more serious.
This is because the activation overpotential becomes drastically larger where the cell
temperature drops below 850 ◦ C and accordingly the current density distribution
Figure 27: Mass density distribution of the catalyst inside the feed tube.
of 4043 A/m2 (2-750U20).
Figure 29: Effect of the gas inlet temperature on the electrolyte temperature dis-
tribution.
becomes more noticeably non-uniform as shown in Fig. 30. Additionally, electricity

conversion efficiency decreases by more than 5% compared to the Base case (see
Table 11) due to the increase in activation overpotential caused by the decrease in
temperature.
Figure 30: Effect of the gas inlet temperature on the local current density dis-
tribution.
Table 11: Calculation results for the Base case and the additional cases.
Case & Maximum Average

Average current Terminal Output Efficiency temperature temperature
density voltage power (LHV) (electrolyte) (electrolyte)
[A/m 2 ] [V] [W] [%] [◦ C] [◦ C]
2-80-U20: 3926 0.55 58.2 41.6 1047.0 957.6

2-750U20: 4043 0.48 52.3 36.3 1023.3 913.2
4-800U20: 3920 0.50 52.8 37.8 968.0 909.5
2-800U02: 3911 0.54 56.9 41.0 995.4 942.6
2-800L02: 4051 0.52 56.8 39.5 974.2 939.2
4.3.2 Effect of air flow rate

The temperature contours obtained for the average current density 3920 A/m2 in
Case 4-800U20 are shown in Fig. 31, and the electrolyte temperature distribution
in the x-direction is shown in Fig. 32. Owing to the increased flow rate of air,
the coolant, the cell temperature, especially the maximum, is fairly reduced in
comparison with the Base case, and accordingly the temperature gradient is also
reduced. Therefore, to increase the air flow rate is quite effective for flattening the
cell temperature distribution, though often it should be adjusted so as to fulfil the
requirements to optimize the performance of a topping or bottoming device.
4.3.3 Effect of density distribution of catalyst

The temperature contours obtained for the average current density 3911 A/m2 of
the Case 2-800U02 are shown in Fig. 33. Since the amount of the catalyst inside the
feed tube is optimised, the maximum temperature becomes lower and a sharp fall
of the cell temperature near the fuel inlet observed in the Base case is mitigated.
of 3920 A/m2 (4-800U20).
Figure 32: Effect of the air flow rate on the electrolyte temperature distribution.
of 3911 A/m2 (2-800U02).
The contours for the average current density 4051 A/m2 of the Case 2-800L02
are shown in Fig. 34. The thermal field becomes almost uniform in a large area
of the cell by arranging the catalyst density in a linear distribution rather than in
the uniform distribution. These effects of the density distribution of the catalyst
on the thermal field are seen more clearly in Fig. 35, which shows the electrolyte
temperature distribution in the x-direction.
of 4051 A/m2 (2-800L02).
Figure 35: Effect of the catalyst distribution on the electrolyte temperature distri-
bution.
Figure 36 shows the mole flow rate of methane inside the feed tube for the above
cases. As shown in this figure, the difference in the catalyst density distribution
greatly affects how the internal reforming takes place inside the feed tube, which
means how the heat generated by the electrochemical process is absorbed there.
However, the total heat absorption rate inside the feed tube is not so changed
accompanying the change in the catalyst density distribution because reforming
is almost 100% accomplished in all of the above cases. Therefore, the average
electrolyte temperature scarcely decreases and accordingly the power generation
performance is little affected by the change of the catalyst density distribution
(see Table 11).
5 Conclusions
This chapter has presented a numerical model for a single tubular SOFC including a
case with accompanying indirect internal fuel reforming or for an IIR-T-SOFC. The
model developed is useful to analyze not only its electricity generation performance
but also the details of the phenomena inside the cell. A variety of phenomena such as
Figure 36: Effect of the catalyst distribution on the molar flow rate of methane
inside the feed tube.
heat and mass transfer, internal reforming, electrochemical reaction, heat generation
and production of chemical species, and generation of electric field proceed inside
the cell in a three dimensional manner. All of them are taken into consideration in the
developed model but in a quasi-two dimensional way. All aspects of the numerical
model and numerical procedures have been described in the first half of this chapter.
Some numerical results obtained by using the developed numerical model have
been presented, first for a conventional type tubular SOFC and then for an IIR-
T-SOFC. In the first part, it was demonstrated how the developed model works
in analyzing not only the fuel cell performance but also the details of the phe-
nomena in and around the tubular cell. In the second part, how to reduce the cell
temperature gradient has been discussed. The thermal field of the cell is largely
determined by the balance of three fundamental phenomena: heat generation with
the electrochemical process, heat absorption with the internal reforming process,
and heat removal by the air flow. In the electrochemical process, a kind of mutually
enhancing interaction has been observed between the rise in cell temperature and
increase in electric current. It has been confirmed that increasing the air flow rate
or decreasing the air utilization factor is effective to flatten the temperature distri-
bution. It has been also demonstrated that changing the density distribution of the
catalyst inside the feed tube greatly affects how the reforming reaction takes place
and therefore how the cell is cooled.
Acknowledgements
This study has been carried out as one of the CREST projects titled “Micro Gas
Turbine and Solid Oxide Fuel Cell Hybrid Cycle for Distributed Energy System”
under the support of the Japan Science and Technology Agency (JST). The authors
would like to extend their appreciation also to Dr. Peter Woodfield, Saga University,
who gave us valuable discussions and critiques in finishing this chapter.
References
[1] Zhu, B., Advanced ceramic fuel cell R&D. Proc. 2nd Int. Conf. Fuel Cell
Science, Engineering and Technology, Rochester, pp. 409–417, 2004.
[2] Suzuki, K., Teshima, K. & Kim, J.-H., Solid oxide fuel cell and micro gas
turbine hybrid cycle for a distributed energy generation system. Proc. 4th
JSME-KSME Thermal Engineering Conference, 3, pp. 1–8, 2000.
[3] Suzuki, K., Iwai, H., Kim, J.-H., Li, P.-W. & Teshima, K., Keynote paper:
solid oxide fuel cell and micro gas turbine hybrid cycle and related fluid
flow and geat transfer. Proc. 12th International Heat Transfer Conference,
Grenoble, pp. 403–414, 2002.
[4] Suzuki, K., Song, T.W. & Iwai, H., Micro gas turbine – solid oxide fuel cell
hybrid system for distributed energy generation. Proc. Int. Symposium on
Distributed Energy Systems in the 21st Century, Tokyo, 2002.
[5] Dicks, A.L., Advances in catalyst for internal reforming in high temperature
fuel cells. J. Power Sources, 71, pp. 111–122, 1998.
[6] Ahmed, K. & Foger, K., Kinetics of internal reforming of methane on
Ni/YSZ-based anodes for solid oxide fuel cells. Catalysis Today, 63, pp.
479–487, 2000.
[7] Takeguchi, T., Kikuchi, R., Yano, T., Eguchi, K. & Murata, K., Effect of pre-
cious metal addition to Ni-YSZ cermet on reforming of CH4 and
electrochemical activity as SOFC anode. Catalysis Today, 84, pp. 217–222,
2003.
[8] Ahmed, S., McPheeters, C. & Kumar, R., Thermal-hydraulic model of a
monolithic solid oxide fuel cell. Journal of Electrochemistry, 138(9),
pp. 2712–2718, 1994.
[9] Nabielek, H., GT-SOFC coupling work at julich tesearch venter. Proc. Work-
shop on Development of High Efficiency Gas Turbine/Fuel Cell Hybrid
Power Generation System, pp. 111–138, 2004.
[10] Amaha, S., Baba, Y., Yakabe, H. & Sakurai, T., Improvement of the perfor-
mance for the anode-supported SOFCs. Proc. of 2nd Int. Symposium on Fuel
Cell Science, Engineering and Technology, Rochester, pp. 45–48, 2004.
[11] Dang, Z., Iwai, H. & Suzuki, K., Numerical modelling of a disk shape planar
SOFC. Proc. of 2nd Int. Symposium on Fuel Cell Science, Engineering and
Technology, Rochester, pp. 361–368, 2004.
[12] George, R.A., Status of tubular SOFC field unit demonstrations. J. Power
Sources, 86, pp. 134–139, 2000.
[13] Nishino, T., Komori, H., Iwai, H. & Suzuki, K., Development of a com-
prehensive numerical model for analyzing a tubular-type indirect internal
reforming SOFC. Proc. of the 1st Int. Conf. on Fuel Cell Science, Engineer-
ing and Technology, eds R.K. Shah & S.G. Kandlikar, ASME: Rochester, pp.
521–528, 2003.
[14] Nishino, T., Development and numerical prediction of a comprehensive ana-
lytical model of an indirect–internal–reforming tubular SOFC. Master Thesis
in Dept. Mechanical Engineering, Kyoto University, 2004.
[15] Suzuki, K., Li, P.W., Komori, H. & Nishino, T., Internal Report at Heat
Transfer Laboratory. Kyoto University, HT2002-01-01, 2002.
[16] Li, P.-W. & Suzuki, K., Numerical model and performance study of a tubular
oxide fuel cell. Journal of Electrochemistry, 151(4), A548–557.
[17] Aguiar, P., Chadwick, D. & Kershenbaum, L., Modelling of an indirect
internal reforming solid oxide fuel cell. Chemical Engineering Science, 57,
pp. 1665–1677, 2002.
[18] Achenbach, E., Three-dimensional and time-dependent simulation of a pla-
nar solid oxide fuel cell stack. J. Power Sources, 49, pp. 333–348, 1994.
[19] Bessette II, N.F., Wepfer, W.J. & Winnick, J., A mathematical model of a
solid oxide fuel cell. J. Electrochem. Soc., 142(11), pp. 3792–3800, 1995.
[20] Ferguson, J.R., Fiard, J.M. & Herbin R., Three-dimensional numerical sim-
ulation for various geometries of solid oxide fuel cells. J. Power Sources,
58, pp. 109–122 , 1996.
[21] Bessette II, N.F. & Wepfer, W.J., Prediction of solid oxide fuel cell power sys-
tem performance through multi-level modeling. Journal of Energy Resources
Technology, 117, pp. 307–317, 1995.
[22] Song, T.W., Kim, J.H., Ro, S.T. & Suzuki, K., Quasi-2D model of IIR-SOFC
stack for SOFC/MGT hybrid system. Proc. ICOPE – 3, Kobe, 2003.
[23] Song, T.W., Sohn, J.L., Kim, J.H., Kim, T.S., Ro, S.T. & Suzuki, K., Para-
metric studies for a performance analysis of a SOFC/MGT hybrid power
system based on a quasi-2D model. Proc. ASME Turbo Expo 2004, 2004.
[24] Song, T.W., Sohn, J.L., Kim, J.H., Kim, T.S., Ro, S.T. & Suzuki, K., Detailed
performance analysis of a solid oxide fuel cell – micro gas turbine hybrid
power system. ASME Heat Transfer/Fluid Engineering Summer Conference,
North Carolina, 2004.
[25] Singhal, S.C.,Advances in solid oxide fuel cell technology. Solid State Ionics,
135, pp. 305–313, 2000.
[26] Hatchwell, C., Sammes, N.M., Brown, I.W.M. & Kendall, K., Current col-
lectors for a novel tubular design of solid oxide fuel cell. J. Power Sources,
77, pp. 64–68, 1999.
[27] Mori, M. & Sammes, N.M., Sintering and thermal expansion characterization
of Al-doped and Co-doped lanthanum strontium chromites synthesized by
the Pechini method. Solid State Ionics, 146, pp. 301–312, 2002.
[28] Xu, J. & Froment, G.F., Methane steam reforming, methanation and water-
gas shift: I. intrinsic kinetics. AIChE Journal, 35(1), pp. 88–96, 1989.
[29] Xu, J. & Froment, G.F., Methane steam reforming: II. Diffusional limitations
and reactor simulation. AIChE Journal, 35(1), pp. 97–103, 1989.
[30] Sonntag, R.E. & Van Wylen, G.J., Introduction of Thermodynamics, 3rd edn,
John Wiley & Sons, 1991.
[31] Poling, B.E., Prausnitz, J.M. & O’connell, J.P., The Properties of Liquids &
Gases, 5th edn, McGraw-Hill: New York, 2000.
[32] Wilke, C.R., Diffusional properties of multicomponent gases. Chemical
Engineering Progress, 46(2), pp. 95–104, 2000.
[33] Kim, J.W. et al., Polarization effects in intermediate temperature, anode-

supported solid oxide fuel cells. Journal of the Electrochemical Society,
146(1), pp. 69–78, 1999.
[34] Nagata, S., Momma, A., Kato, T. & Kasuga, Y., Numerical analysis of output
characteristics of tubular SOFC with internal reformer. J. Power Sources,
101, pp. 60–71, 2001.
[35] Lehnert, W., Meusinger, J. & Thom, F., Modelling of gas transport phenom-
ena in SOFC anodes. J. Power Sources, 87, pp. 57–63, 2000.
[36] Carbonell, R.G. & Whitaker, S., Heat and mass transfer in porous media.
Fundamentals of Transport Phenomena in Porous Media, eds J. Bear &
M.Y. Corapcioglu, Martinus Nijhoff Publishers: Dordrecht, pp. 123–198,
1984.
[37] Muralidhar, K. & Suzuki, K., Analysis of flow and heat transfer in a rect-
angular mesh using a non-Darcy thermally non-equilibrium model. Int. J.
Heat Mass Transfer, 44, pp. 2493–2504, 2001.
[38] Vafai, K. & Tien, C.L., Boundary and inertia effects on flow and heat transfer
in porous media. Int. J. Heat Mass Transfer, 24, pp. 195–203, 1981.
[39] Patankar, S.V., Numerical Heat Transfer and Fluid Flow, Hemisphere Pub-
lishing Corporation: Washington, 1980.
[40] Sparrow, E.M. & Cess, R.D., 2nd edn., Thermal Science Series. Radiation
Heat Transfer, ed. E.R.G. Eckert, Wadsworth Publishing Company, Inc.:
Belmont, 1970.
[41] Makino, T., Kunitomo, T., Sakai, I. & Kinoshita, H., Thermal radiation
properties of ceramic materials. Heat Transfer-Japanese Research, 13(4),
pp. 33–50, 1984.
[42] Makino, T., Present research on thermal radiation properties and charac-
teristics of materials. International Journal of Thermophysics, 11(2), pp.
339–352, 1990.
[43] Li, P.W. & Suzuki, K., Numerical study of the performance of a tubular solid
oxide fuel cell – effects of geometric parameters and thermal conditions.
Proceedings of Symposium on 2000/2001 Achievement of CREST Project
on SOFC – MGT Hybrid System for a Distributed Energy Generation, pp.
15–20, 2002.
[44] Dang, Z., Nishino, T., Iwai, H. & Suzuki, K., Progress on numerical simu-
lation of heat and mass transfer within tubular and planar type solid oxide
fuel cells. Proceedings of Symposium on 2002/2003 Achievement of CREST
Project on SOFC – MGT Hybrid System for a Distributed Energy Genera-
tion, pp. 53–64, 2003.
[45] Hagiwara, A., Michibata, H., Kimura, A., Jaszcar, M.P., Tomlins, G.W. &
Veyo, S.E., Proceedings of the 3rd International Fuel Cell Conference,
p. 365, 1999.
[46] Suzuki, K., Kimijima, S., Dang, Z. & Iwai, H., Heat and mass transfer in a
tubular solid oxide fuel cell. 6th KSME-JSME Thermal and Fluid Engineer-
ing Joint Conference, Jeju, CD-ROM, KJ04, 2005.
Nomenclature
a coefficient in the discretized equations
A activation energy [J/mol]
b source term in the discretized equations
Cp specific heat at constant pressure [J/kg K]
Djm mass diffusivity of species j in
multicomponent gas [m2 /s]
e emissivity [–]
E electromotive force [V]
E0 standard electromotive force [V]
Ẽ substantial electromotive force defined
by eqn (42) [V]
f inertia coefficient [–]
F Faraday constant [C/mol]
F11 , F12 , F21 , F22 configuration factors [–]
h thickness [m]
i electric current density [A/m2 ]
J radiosity [W/m2 ]
kH2 , kCO , kO2 coefficients in eqns (12), (13) and (14) [A/m2 ]
+ −
ksh , ksh velocity constants of the water-gas
shift reaction [mol/m3 Pa2 s]
K permeability [m2 ]
Ksh equilibrium constant of the water-gas
shift reaction [–]
M molecular weight [kg/mol]
P pressure [Pa]
q radiant flux [W/m2 ]
q̇ heat generation per unit area of the
electrolyte [W/m2 ]
Q heat generation [W/m3 ]
r radial coordinate [m]
R resistivity [ m]
Rst , Rsh reaction rates of the reforming reactions [mol/m3 s]
R0 universal gas constant [Pa m3 /mol K]
ṡ mass generation per unit area of the
electrolyte [kg/m2 s]
Sj mass generation of species j [kg/m3 s]
Ŝ source term in the general transport equation
T temperature [K]
Ux , Ur velocity components [m/s]
V electric potential [V]
Wcat catalyst mass density [g/m3 ]
x axial coordinate [m]
Yj mass fraction of species j [–]
Subscripts
a anode
c cathode
cat catalyst
e electrolyte
eff effective
s solid
st steam reforming reaction: eqn (1)
sh water-gas shift reaction: eqn (2)
1 cell tube side in the radiation model
2 feed tube side in the radiation model
H2 /O2 electrochemical reaction of hydrogen: eqn (6)
CO/O2 electrochemical reaction of CO: eqn (7)
e, w, n, s interfaces on a control volume shown in Fig. 6
E, W , N , S grid points in a control volumes shown in Fig. 6
u, u , d , d interfaces on a control volume shown in Fig. 7
U , U , D, D grid points in a control volumes shown in Fig. 7
P, P grid points on the intended control volumes
PP electrolyte between the grid points P and P
Greek symbols
diffusivity in the general transport equation

G 0 standard Gibbs free energy change [J/mol]
H enthalpy change [J/mol]
r control volume size for r direction [m]
x control volume size for x direction [m]
θ control volume size for θ direction [rad]
δr distance between grid points for r direction [m]
δx distance between grid points for x direction [m]
δθ distance between grid points for θ direction [rad]
ε porosity [–]
η activation overpotential [V]
θ circumferential coordinate [rad]
λ thermal conductivity [W/m K]
µ viscosity [Pa s]
ρ density [kg/m3 ]
ρ̂ density in the general transport equation
σ Stefan-Boltzmann constant [W/m2 K4 ]
φ variable in the general transport equation
CHAPTER 4
On heat and mass transfer phenomena in

PEMFC and SOFC and modeling approaches
J. Yuan1 , M. Faghri2 & B. Sundén1
1 Divisionof Heat Transfer, Lund Institute of Technology, Sweden.
2 Department of Mechanical Engineering and Applied Mechanics,
University of Rhode Island, USA.
Abstract
There are similarities among various transport processes in solid oxide fuel cells
(SOFCs) and proton exchange membrane fuel cells (PEMFCs), but also some dif-
ferences. The present work concerns modeling and numerical analysis of heat,
mass transfer/species flow, two-phase transport and effects on the cell performance
in SOFCs and PEMFCs. Numerical calculation methods are further developed to
enable predictions of convective heat transfer and pressure drop in flow ducts of the
fuel and the oxidant. The unique boundary conditions (thermal, mass) for the flow
ducts in fuel cells are identified and implemented. The composite duct consists of a
porous layer, gas flow duct and/or solid current inter-connector (or -collector). The
results from this study are applicable for other investigations considering overall
fuel cell modeling and system studies, as well as the emerging field of micro-reactor
engineering.
1 Introduction
In a fuel cell, electrical energy is generated directly through the electrochemical
reaction of oxidant (oxygen from air) and fuels (such as natural gas, methanol, or
pure hydrogen) at two electrodes separated by an electrolyte. When pure hydrogen
is used, the only products of this process are heat, electricity and water. Unlike a
battery, fuel cells do not store energy. The energy conversion is achieved without
making use of the materials that constitute an integral part of the fuel cell structure.
It should be noted that fuel cells convert chemical energy directly into electricity
without an intermediate combustion process.
One of the main factors that have influenced the development of fuel cells has
been the increasing concern about the environmental consequences of fossil fuel
utilization in the production of electricity and for the propulsion of vehicles. More
importantly is the increasing global awareness of how industry activities influ-
ence the environment and how a sustainable energy development can be achieved
with a tremendously increasing world population. Fuel cells may help to reduce
our dependence on fossil fuels and diminish poisonous emissions into the
atmosphere.
The operation of a fuel cell requires a fuel electrode (Anode), oxidant elec-
trode (Cathode), electrically-insulating ionic conductor (Electrolyte), and external
electric circuit. In general, fuel cells can be classified according to the type of
ionic conductor (Electrolyte) they use and the temperature range at which they
operate.
Several types of fuel cells are currently under development. Alkali fuel cells
(AFCs) use alkaline potassium hydroxide as the electrolyte, and have been used for
a long time by NASA on space missions. In proton exchange membrane fuel cells
(PEMFCs) and phosphoric acid fuel cells (PAFCs), hydrogen fuel dissociates into
free electrons and protons (positive hydrogen ions). The hydrogen protons migrate
through the electrolyte to the cathode. The liquid-fed direct methanol fuel cell
(DMFC) feeds a solution of methanol and water to the anode.At the cathode, oxygen
from air, electrons from the external circuit and protons combine to form pure water
and heat. All these three types are low temperature fuel cells. High temperature
fuel cells, such as solid oxide fuel cells (SOFCs) and molten carbonate fuel cells
(MCFCs) are of particular interest, because their high temperature operation allows
natural gas to be used as a fuel, and the hybrid concept involving a combination of
a fuel cell and a gas turbine becomes feasable. The overall system efficiency can
be significantly increased in a hybrid system. Operation at a temperature of about
1000 ◦ C (conventional, electrolyte-supported planar design), 700 ◦ C (intermediate
temperature, anode-supported design) and pressures greater than one atmosphere
leads to solid oxide fuel cells (SOFCs) as one of the choices. It has been found that
the proton exchange membrane fuel cell (PEMFC) system has some advantages,
such as its relative simplicity of design and operation, low cost construction and
self-starting at low temperatures. Both SOFC and PEMFC systems are expected
to play a significant role in the next generation of primary or auxiliary power for
stationary, portable, and automotive systems.
During the last decades, a large amount of research activities have been carried
out on fuel cells worldwide, with particular interest and focus on SOFC and PEMFC
systems. High performance, low cost and high reliability have been considered as
the primary aspects and concerns for fuel cells to compete with well-developed fos-
sil fuel power technology, such as the internal combustion engine. Most of the work
has focused on creating new materials and material processes for the manufacturing
of specific systems so as to achieve good cell/stack performance while minimizing
the final system cost and size. To help expand future market opportunities for fuel
cells, additional fundamental understanding and research work are needed. More
attention needs to be focused on detailed analysis of transport processes, even at
micro scale level. Because of the particular sensitivities and complexities of fuel
cells, analysis and optimization of fuel cell components/system can effectively be
performed by numerical modeling and simulation.
The present work concerns numerical analysis of heat and mass transport pro-
cesses, and fluid flow in ducts of both planar type solid oxide fuel cells and proton
exchange membrane fuel cells. Numerical models have been developed to enable
predictions of convective heat transfer and pressure drop in flow ducts of the fuel
and the oxidant. The composite duct consists of a porous layer, gas flow duct and/or
solid current inter-connector (or collector). The results from this study are appli-
cable for related investigations considering overall fuel cell modeling and system
studies, e.g., by providing heat transfer coefficients for various conditions, as well
as in the emerging field of micro-reactor engineering. In general, the work also
contributes to the understanding of duct flow.
2 Fuel cell modeling development

Despite the differences in terms cell structure/materials employed and operating
conditions (such as temperature, pressure), supply of species to an active surface of
cells in the stack is a common feature of fuel cells, such as SOFCs and PEMFCs. It
is so because the performance degradation for the cells operating within the stack
results from the unequal distribution of reactant mass flow among the cells. Under-
standing the various gas and heat transport processes is crucial for increasing the
power density, reducing manufacturing costs and accelerating commercialisation
of fuel cell systems. To this end, detailed modeling and improved simulation tools
are required to fully characterize the complex multi-dimensional, multi-phase and
multi-component transport processes on media that are both porous and electro-
chemically reactive.
2.1 Basics of SOFCs and PEMFCs
There are various similarities in the transport processes occurring in SOFCs and
PEMFCs. In the anode duct, the fuel (e.g., H2 ) is supplied and air (O2 + N2 ) is intro-
duced in the cathode duct, and these ducts are separated by the electrolyte/electrode
assembly. Reactants are transported by diffusion and/or convection to the elec-
trode/electrolyte (SOFC) or catalyst/electrolyte (PEMFC) interfaces, where elec-
trochemical reactions take place.An electrochemical oxidation reaction at the anode
produces electrons that flow through the inter-collector (bipolar plate, for PEMFC)
or -connector (for SOFC) to the external circuit, while the oxide ions (in SOFCs)
or protons (in PEMFCs) pass through the electrolyte to the opposing electrode.
The electrons return from the external circuit to participate in the electrochemical
reaction at the cathode. In the electrochemical reaction process, part of the oxygen
is consumed in the cathode duct, while the hydrogen is consumed in the anode
duct. Heat and water (H2 O) are the only by-products during the process. The water
generated is injected into the anode duct further along the duct in SOFCs, while in
PEMFCs, it enters into the cathode duct. The electrochemical reactions in SOFCs
can be written as:
Cathode reaction: 1
2 O2 + 2e− → O2− . (1a)
Anode reaction: H2 + O2− → H2 O + 2e− . (1b)
and for PEMFCs:
Cathode reaction: 1
2 O2 + 2e− + 2H+ → H2 O. (2a)
Anode reaction: H2 → 2H+ + 2e− . (2b)
The overall reaction is as follows:

1
2 O2 + H2 → H2 O. (3)
Due to the flow resistance in the fuel cells, the pressure drop (P) along the
ducts and in the manifolds can cause non-uniform flow distribution. Furthermore,
the output of electrical energy will differ in terms of voltage potential and in some
cases even gas re-circulation occurs. At some severe conditions, the lack of gas
in some channels can cause the irreversible damage to the fuel cell components.
The pressure drop depends on the channel and manifold structures, flow streams
etc. However, the temperature is always non-uniform even when there is a constant
mass flow rate in the ducts. This is caused by the heat transfer and phase change
(in PEMFCs), which in turn causes fluctuation in the available T. Heat transfer
occurs in the following manner:
• Between the cell component layers and the flowing air and fuel streams. This can
be described in terms of heat transfer coefficients ha (for air channel), hf (for
fuel channel) due to forced convective heat transfer with or without natural
convection;
• Between the fuel and air streams across the interconnect layer in terms of the
overall heat transfer coefficient, U ;
• In solid structures in terms of heat conduction with different thermal conduc-
tivities, ki (i = electrolyte, electrodes and current interconnect layers).
For the electrolyte and porous layers, often referred to as the membrane electrode
assembly (MEA) in PEMFCs, the overall principal energy balance can be written as:
Qc + hf Af (Te − Tf,av ) + ha Aa (Te − Ta,av ) = Qs , (4)

where Qc is the heat conduction in the solid structure, Qs is the heat source to account
for the electrochemical heat generation, ohmic heating caused by the electrical
resistance due to the current flow; h is the convective heat transfer coefficient; T is
the temperature. Equation (4) shows that the heat transfer coefficients in the fuel
and oxidant ducts are important.
There are certainly some specific aspects and phenomena which need to be care-
fully investigated for different applications.As an example, Nafion® membranes are
often employed in the electrolytes in PEMFCs. These membranes possess high ion
conductivity by selecting perfluorosulphonic acid copolymers with a short pendant
group. However, the performance of the membranes, in terms of electrical conduc-
tivity, strongly depends on the water content. Water management in the membrane
is one of the major issues in PEMFCs. As discussed later in this chapter, there are
several factors affecting the water content in the membrane, such as the water drag
through the electrolyte (electro-osmotic), back diffusion of product water from the
cathode to the anode. For the case of excessive accumulation of water vapor in the
cathode, condensation may occur. Consequently, different methods of water man-
agement have been proposed and investigated. Another issue is the temperature
range in which the membranes are stable. Therefore, both water and thermal man-
agement are coupled and need to be carefully balanced, as will be discussed later.
SOFCs employ solid oxide material as electrolyte and are, therefore, more stable.
There are no problems with water management, liquid water flooding in the cathodes
or slow oxygen reduction kinetics in SOFCs. On the other hand it is difficult to find
suitable materials for operation at high temperatures.
There are other processes which only occur in SOFCs, such as internal reforming
of fuels as pure hydrogen is not used and co-generation of heat/electricity with other
power systems (such as gas turbines). For PEMFCs, external reforming is needed
to handle hydrocarbon fuels. It should be mentioned that the reforming issues are
not treated in this chapter but instead focus is on the composite duct flows in SOFCs
and PEMFCs.
2.2 Modeling development
Modeling has already played an important role in fuel cell development since it
facilitates a better understanding of parameters affecting the performance of fuel
cells and fuel cell systems. Moreover, water management within the cathode is
a key consideration in the design of the PEMFCs. However, knowledge of the
behavior of liquid water in electrodes is limited by the inability to make in situ
measurements. Better understanding of the transport of water in the PEMFC elec-
trode can be obtained from models that capture the important physical processes.
This is particularly evident in works published in the open literature during recent
years. There has been a range of models developed, from simple lumped models
of individual cell channels to more complex three-dimensional detailed ones of
complete stacks.
2.2.1 Modeling approaches

There are several issues which affect the choice of modeling strategies, and should
be considered before selecting a fuel cell modeling approach. The most important
factors are the objectives and features of the model, which should be clearly defined
and specified. These include the following issues, such as steady-state/transient,
theoretical/semi-empirical, components/system study, lumped/multidimensional,
accuracy/time/flexibility, validation/documentations and so on. The development
of modeling and simulation tools is a cost and time consuming process. The level of
user knowledge and available resources (such as personnel and computer facilities
etc.) are also constraints to include in the decision process.
Based on detailed electrochemical, fluid dynamics, species/current transport
and heat transfer relationships, a theoretical fuel cell modeling approach usually
employs the basic equations, such as the Stefan-Maxwell equation for gas-phase
transport, and the Butler-Volmer equation for cell voltage [1]. The electrochemical
and transport processes are tightly coupled. For proper water and thermal manage-
ment, this approach includes not only the electrochemical reaction but also thermal-
and fluid-dynamic equations. Multi-component species transport and heat transfer
are important for providing a detailed picture of all processes in the fuel cell and the
system. The output of the study can provide details of the processes, such as fuel cell
species distribution/flow pattern, current density/temperature distribution, voltage
and pressure drop, etc. On the other hand, based on experimental data specific to
the applications and operating conditions, semi-empirical fuel cell models have
also been developed during the recent years. Both approaches have advantages and
disadvantages. The theoretical modeling approach is flexible to applications and
operating conditions, and may be appreciated when detailed studies are desired.
However, development and implementation of this approach takes a longer time,
and it is difficult to validate due to lack of detailed data in the open literature. At
present, the most readily available data are simply the overall I -V characteristics
for a cell or stack. While the semi-empirical approach is already validated to some
extent, it does not provide sufficient details. It should also be noted that it must be
modified for each new application or operating conditions, and may not be suitable
in some cases.
During recent years, reliable commercial fuel cell codes (or in some cases, a
fuel cell module added to an already existing software) have been developed. For
instance, detailed SOFC and PEMFC models have been included in the Chemical
Engineering Module of FEMLAB, based on the platform of the MathWorks simula-
tion code MATLAB [2]. More detailed modules associated with other commercial
codes are being further developed and will be incorporated into standard cell and
stack modeling when they have been validated.
2.2.2 Various existing models

In the last few years, attempts to simulate the velocities, pressures, temperatures,
mass fractions, electric currents and potentials in fuel cells with given boundary
conditions have been presented. This section only covers literature dealing with the
relevant problems of heat transfer and gas flow modeling in SOFCs and PEMFCs.
2.2.2.1 SOFC modeling For conventional electrolyte-supported planar struc-

ture, Vayenas [3] created a two-dimensional mixing cell model for cross-flow to
simulate the distribution of gas species, temperature, and current density. For a
unit cell with square and rectangular ducts, constant temperature was assumed for
the solid walls, and the gas-phase composition was uniform. Another model was
developed by Ahmed et al. [4]. This could be used to simulate the electrochemistry
and thermal hydraulics in a cross-flow monolithic SOFC with alternating layers
of anode, electrolyte, cathode, and interconnect. Based on the average thermal and
compositional conditions, a Nusselt number (= hd /k) of 3.0 based on the equivalent
diameter of the rectangular flow ducts was used for the convective heat transfer
between the gases and the solid surfaces. In this model, the generated heat was
released in the electrolyte, and the pressure drop in the channels was modeled by
assuming fully developed laminar flow.
Bernier et al. [5] proposed a three-dimensional mathematical model to compute
the local distribution of the electrical potential, temperature, and concentration
of the chemical species. In the gas ducts, the thermal flux is mainly convective
and conductive from the duct walls to the other solid parts. A computer code was
developed by Melhus and Ratkje [6], which was used to find simultaneous solu-
tions of all conservation equations for mass, energy and momentum in a quasi
three dimensional simulation for single flat SOFCs. The reduction of the mass of
O2 in the cathode chamber was assumed to be regained in the anode as H2 O for the
mass balance. Regarding the distribution of the gas flow, Boersma and Sammes [7]
developed a model to simulate the non-uniform gas flow distribution along the
height of a fuel cell system.
For the anode-supported design, the porous anode usually has a thickness of
1.5–2 mm [8–11], which is the thickest component and supporting structure, while
a thin layer of electrolyte (∼10 µm) is deposited on its surface. The transport path
length for the fuel gases from the flow duct to the anode/electrolyte interface where
the electrochemical reaction happens, is at least equal to the porous anode thickness.
The transport rate of fuel gases is controlled by the porous layer microstructure (e.g.,
pore size, permeability, and porosity), pressure gradient between the flow duct and
porous layer, gas composition, etc. There are various polarizations or losses (such as
ohmic, activation and concentration) affecting the overall performance of SOFCs.
It has been revealed that one of the principal losses in this design is attributed to
concentration polarization caused by limitation of gaseous species transport through
the porous anode [8], because the size of the porous anode might be bigger than
that of the flow duct.
2.2.2.2 PEMFC modeling It has been found that the proton exchange membrane
fuel cell system has some advantages, such as its relative simplicity of design and
operation, low cost construction materials and self-starting at low temperatures.
However, water management is a critical issue for PEMFC, i.e., a high water-content
must be maintained in the membrane in order to obtain acceptable ion conductiv-
ity. Water molecules are transferred from the anode duct to the cathode duct of
the membrane by electro-osmosis during PEMFC operations. If this transport rate
of water is higher than that of back diffusion, the anode gases will dry out, and
the membrane becomes dehydrated and too resistive to conduct current. On the
other hand, cathode flooding occurs when the water removal rate fails to reach
its minimum transport rate, which is caused by both the transport from the anode
duct mentioned above and the generation of water by the electrochemical reaction
H+ /O2 at the catalyst surface [12]. Consequently, a sufficient amount of water must
be supplied to the anode duct to make up for the loss due to net water transfer from
the anode, and water should be removed at a sufficient rate from the cathode duct
to keep an active catalyst surface for reaction.
Another challenging issue in PEMFCs is thermal management. In order to pre-
vent excessive operating temperature and drying out of the membrane, the heat
generated by the electrochemical reaction should be removed properly. The ther-
mal management has a strong impact on the fuel cell performance, by affecting the
transport of water and gaseous species as well as the electrochemical reactions in
the cells. Both thermal and water management of PEMFCs are unique compared
to other types of fuel cells.
Various water and/or thermal management systems have been claimed to be
efficient by different approaches of humidification designs (injection of liquid/vapor
water) and operating conditions [13–17] and anode water removal [18]. It should be
noted that most of the models are one- or two-dimensional, and are limited by the
isothermal assumptions and also neglect the potentially significant gas pressure drop
within the electrodes. However, these studies provided a fundamental framework
to build multi-dimensional and multi-component models, such as a fully three-
dimensional [19–20], some of which include the heat transfer effects on the overall
performance of fuel cells [20]. More recently, researchers have started to investigate
two-phase water flow, by taking into account effects of heat generation/transfer
and/or pressure drop [21–25].
2.2.2.3 Other processes relevant to fuel cell modeling In fuel cells, the gaseous
reactant flows at both the cathode and anode are subject to fluid injection and
suction along the porous interface to the electrolyte. However, such processes are
often simulated as flow in porous wall ducts at constant heat flux boundary condition
[26]. Because both heat transfer and pressure distributions are significantly affected,
fluid flow and heat transfer in ducts with mass transfer in porous walls have received
a great deal of attention in the past decades [26–28]. Hwang et al. [26] simulated
flow and heat transfer in a square duct in the range of −20.0 < Rem < 20.0, with
boundary conditions of one porous wall subjected to a constant heat flux, while the
other three walls were assumed to be adiabatic and impermeable.
A better understanding of thermal engineering applications is required when
porous materials are present in the duct flow. Because of its simplicity and rea-
sonable accuracy within a certain range of applications, the Darcy model has been
used for the majority of the existing studies on gas flow and heat transfer in porous
media. It has been found that the Darcy model has some limitations, and inertial
forces should be taken into account when the interstitial flow velocity (i.e., the
flow through pores of a porous medium) is not small [29]. It was reported in [30]
that heat transfer can be significantly affected by Darcy number and the thermal
conductivity ratio (between thermal conductivity of the porous media and that of
the fluid). Various types of interfacial conditions between a porous medium and a
gas flow duct were analyzed in details for both gas flow and heat transfer in [31]. It
was found that as some of the gas flow penetrates sideways into the porous layer,
the remaining gas flows downstream at decreasing flow rates. The static pressure
in such a duct changes along the main flow stream due to the following reasons:
the friction between the gas flow and the internal surfaces of the duct creates pres-
sure losses, and the mass permeation across the interface between the flow duct
and the porous layer implies that mass and momentum are transferred to/from the
porous layer [32]. The permeation process in the porous media is usually considered
as an overall mass transport with a constant permeability.
2.2.2.4 Further challenges As discussed above, various models have been

developed and improved for both SOFCs and PEMFCs. However, the modeling
improvements are still limited due to, among other things, the limited material data
available in the open literature. At a minimum, temperature and/or concentration
dependent electrical resistivity and electrochemical activation energies, as well as
the diffusion characteristics of the porous materials are pre-requisites to model
development. Another critical factor limiting modeling development is the lack of
test data that can be used to validate models.
3 Main processes in SOFCs and PEMFCs

The major processes significant to fuel cell characteristics are similar in SOFCs
and PEMFCs. These processes are the species transport, electrochemical reactions,
electronic and ionic transport, and heat transfer and distribution. Figure 1 shows
Figure 1: Schematic sketch of a unit cell for: (a) PEMFC and (b) SOFC.
a unit cell structure of fuel cells (representing PEMFCs in this case). It includes
various components, such as fuel and oxidant gas ducts, electrolyte (polymer elec-
trolyte membrane for PEMFCs), anode and cathode diffusion layers, catalyst layers
in between them, and current inter-conductors.
3.1 Gas transport
In a fuel cell stack, the gas transport processes consist of:

• The fuel and oxidant gases flow separately through the gas manifolds where no
electrochemical reactions occur;
• The fuel and oxidant gases flow along cell ducts where there is absorption of
the reactants and the injection of reactive products from/to active surface;
• In the porous layers (electrodes), transport of the reactant gases occur towards
the catalyst layer (in PEMFCs), or active surfaces (in SOFCs), and the exhaust
gases are rejected to the cell ducts through the open pores;
• The exhaust gases from each cell are discharged through the gas output mani-
folds.
Fuel cell ducts and manifolds should be designed/configured to have appropriate
gas flow rate and flow uniformity to the reactive surface. An important concern is
the net pressure loss, which should be as low as possible to reduce parasitic power
needed to operate pumps or compressors. Consequently, a laminar flow regime is
effected in most of the fuel cells by employing small velocity and cross-sections in
the manifolds and ducts [33].
The appropriate mass flow rate of reactants (fuels and oxidants) is determined
by various factors, such as the requirement for the electrochemical reaction, proper
thermal and water management, and fuel reforming reaction (SOFCs) etc. In
PEMFCs, water management is critical to avoid the membrane dry out and cathode
flooding. To deal with this concern, the oxidant flow rate may be increased to reduce
the excess water generation.
3.2 Electrochemical reactions
At the active surface, the electrochemical reactions take place as described in eqns
(1) and (2) for SOFCs and PEMFCs, respectively. The overall cell reaction is shown
as eqn (3). The impacts of the electrochemical reactions on the gas mass balance are
represented by the absorption of reactants and generation of products at the active
surfaces, in terms of mass flux rate J (kg/m2 s). The mass flux rate is related to local
current density I (A/m2 ) and reads as follows when pure hydrogen is used as fuel.
• SOFC anode
I
JH2 = − MH2 , (5)
2F
I
JH 2 O = MH2 O . (6)
2F
• PEMFC anode
I
JH2 = − MH2 , (7)
2F
α
JH 2 O = − I · MH2 O . (8)
F
• SOFC cathode
I
JO2 = − MO2 . (9)
4F
• PEMFC cathode
I
JO2 = − MO2 , (10)
4F
1 + 2α
JH 2 O = I · MH2 O . (11)
2F
In the equations above, α is the net water transport coefficient, for the case of PEM-
FCs, which represents the net water transport through the membrane by electro-
osmotic drag and back diffusion due to the water concentration difference, and
hydraulic permeation due to the pressure difference between the two sides. Other
symbols can be found in the nomenclature. It should be noted that negative sign
(−) in the above equations represents gas consumption, while plus (+) means gas
generation.
3.3 Heat transfer
As mentioned before, the heat transfer processes include various aspects, such as
the convective heat transfer between the solid surface and the gas streams, conduc-
tive heat transfer in the solid and/or porous structures. Furthermore, heat genera-
tion occurs at the active surface in association with the electrochemical reactions
and cell losses. The heat generation at the active surface, qb , can be expressed as
follows [20]:
I
qb = − HH2 O MH2 O − I · Vcell , (12)
2F
where H is the enthalpy change of formation of water. The first term on the RHS
in the above equation accounts for the amount of heat energy released by the water
formation, while the second one accounts for the current density generated by the
electrochemical reaction. When the inhomogeneous current density is taken into
consideration, the total local heat generation must be defined due to local joule
heating.
3.4 Various transport processes in the electrodes (porous layers)
Electrodes for fuel cells are generally porous to ensure maxium active surfaces,
and to allow the injection of the generated products to the ducts. The mass transfer
is dominated by the gas diffusion and/or convection, as discussed later in this chap-
ter. This is ensured by the open pores of the electrodes, in terms of permeability
and/or porosity. Another requirement of a porous electrode is to also have a good

ionic conductivity, because the ionic particles are transported via the solid matrix
of the porous layer. In general, the electrodes should have a balanced performance
and long-time stability. In PEMFCs, a catalyst material is frequently employed,
such as platinum or platinum/ruthenium, whereas SOFCs utilize much cheaper
catalyst materials such as nickel due to reduced activation polarization at higher
temperature.
3.5 Other processes appearing in fuel cell components
The electrolyte of fuel cells transport ions created by the electrochemical reaction
at one electrode to the other. In PEMFCs the proton is transported through the
electrolyte, while in SOFCs the oxygen ion is transferred. Reducing the electrolyte
thickness and internal ohmic losses is a major requirement. On the other hand, the
electrolyte should be impermeable to gases (fuel and oxidant) for the purpose of
minimizing reactant crossover.
The cell inter-connectors or -collectors involve heat transfer by thermal conduc-
tion and current collection. Consequently, high electrical conductivity and thermal
conductivity are the basic requirements. The materials having the following features
are often employed, i.e., impermeability to reactants and chemically stable in oxi-
dizing and reducing environments. In general, PEMFCs need more expensive mate-
rials, individually machined graphite or even gold-plated stainless steel materials.
It is well-known that the polarization curve, which represents the cell voltage
behaviour against operating current density (V -I curve), is the standard measure of
the performance for fuel cells, and depends on both the operating conditions and
the component design. The operating conditions include the working temperature,
partial pressures of fuel and oxidant and their utilization rate, and/or the water
concentration in the components. On the other hand, the design parameters could be
the porosity, tortuosity, thickness of the electrodes (concentration loss), thickness of
the electrolyte (ohmic loss), and the electrode/electrolyte interface (activation loss).
4 Processes and issues in SOFC and PEMFC

Despite various similar processes discussed above, there are certainly processes
and/or phenomena which are connected to a specific fuel cell type and should
be carefully considered. Among these, water management is a challenging issue
for PEMFC, while internal fuel reforming at the anode side and high operating
temperature have unique features relating to mass transport and heat transfer if
fuels other than pure hydrogen are used in SOFCs. Moreover, thermal radiation
may be important in SOFCs while it is not in PEMFCs.
4.1 Water management in PEMFCs
Polymer electrolyte membranes in PEMFCs are basically water filled to have high
proton conductivity, as mentioned earlier. Factors influencing the water content in
the electrolyte are generally two transport processes, i.e., water drag through the
electrolyte membrane (a shell of H2 O is transported via the electrolyte for every
proton transported), and back diffusion of generated water from the cathode into the
anode through the electrolyte. The first one is often referred to as electro-osmotic
transport in the literature, and the latter one is due to the gradient of water content in
the electrolyte. The effective electro-osmotic coefficient α is an important parameter
to represent water transport between the anode and the cathode. It includes the
effects of both electro-osmosis and water back diffusion.
Water management in the electrolyte is one of the major issues in PEMFCs. This
is because during PEMFC operation anode gases can be dried out if the electro-
osmosis transport rate is higher than that of back diffusion, which consequently
causes the electrolyte membrane to become dehydrated and too resistive to conduct
current. On the other hand, water is generated at the cathode active surface and trans-
ported to the cathode duct. Cathode flooding may occur when the water removal
rate fails to reach its threshold transport/generation rate. Both dry-out and water
flooding should be avoided, and various water management schemes have been pro-
posed. More detailed discussions on these issues can be found in [13–18]. It should
also be noted that condensation can occur in the cathode duct when local vapor sat-
uration condition occurs in the duct. As discussed later in this work, this case mainly
happens at high current densities and low operating temperatures of fuel cells.
When such condensation occurs the transport process becomes two-phase, which
besides flooding the cathode layer can considerably complicate the modeling pro-
cedure as no experimental results are available for two-phase flow in PEMFCs.
Instead, much attention has been paid on numerical investigations to reveal the
relationships between the water saturation, proton conductivity (ohmic loss), the
level of catalyst flooding (activation loss), and the effective diffusivity of the porous
layer (concentration loss); see [21–26].
4.2 Fuel reforming issues in SOFC
Because of the high operating temperature, an SOFC can convert not only hydrogen
into electricity, but can also reform hydrocarbon compounds into reactant fuels. For
instance, methane can be converted to H2 and CO2 in a steam reforming process
within the anode of SOFCs. This reforming process takes place at the surface and
in a very thin layer of the anode porous nickel cermets (ceramic metal) [10]. It is
often referred to as internal reforming in the literature.
The methane reforming reaction in this case can be written as follows:
CH4 + H2 O → CO + 3H2 , H = 206 kJ/mol, (13a)

CO + H2 O → CO2 + H2 , H = −41 kJ/mol. (13b)
Equation (13b) is usually referred to as water gas shift reaction. The overall
reforming reaction is:
CH4 + 2H2 O → CO2 + 4H2 . (13c)

It should be mentioned that the above processes in eqn (13c) are net endothermic
and the overall balance of the reaction requires external heat input. This heat can
be supplied by the exothermic electrochemical reaction, as given in eqns (1) and
(2). Due to the fast reforming reaction compared to the electrochemical reaction,
the endothermic steam reforming process may lead to local sub-cooling, and/or
mechanical failure due to thermally induced stresses [10].
5 Modeling methodologies for transport processes in

SOFC and PEMFC
5.1 General considerations
A typical configuration of a simulated fuel cell duct is shown in Fig. 2. This

section focuses on the establishment of equations for the analysis of the fuel
cell ducts appearing in SOFCs and PEMFCs. The duct under study includes the
gas flow duct, porous layer (anode/cathode) and solid current inter-connector or
collector.
The variables to be solved for are the gas transport velocities in the x, y and z
directions, the gas mass concentration/distribution, pressure drop, temperature dis-
tribution and convective heat transfer coefficient (in term of Nusselt number, Nu).
The liquid water saturation and its effects on the current density distribution are
modeled for PEMFCs.
The governing equations consisting of species (H2 , H2 O, O2 , etc.), momentum
and energy equations, are solved for various sub-domains. Auxiliary equations
are employed to calculate the source terms in the governing equations. A unified
framework is developed for both SOFC and PEMFC, by implementing specific
source terms to account for different transport processes and boundary conditions.
Figure 2: Schematic drawing of a composite fuel cell duct (Anode-supported planar

SOFC) under consideration.
5.2 Assumptions
Steady laminar flow of incompressible fluid is considered. The appropriate mass

flow rate of reactants is determined by several factors related to various require-
ments, e.g., maintaining proper water balance and thermal management. Water
management concerns in PEMFCs may need an increased flow rate, which can be
described by a parameter named the stoichiometric ratio for an electrode reaction.
In general, the gas flow duct should be designed to minimize pressure drop, while
providing adequate and evenly distributed mass transfer through the porous layer
for the electrochemical reaction. Thus, typically laminar flow with Re numbers of
the order of 100–1000 have been employed for fuel cells [33]. For simplicity, the
following additional assumptions are applied:
• The inlet-velocity and axial-temperature distributions of the species are assumed
uniform, and the inlet Reynolds number is assumed to be Rein (=Uin Dh /ν) 1;
• An electrochemical reaction is assumed to occur at the interface between elec-
trodes and electrolyte. The released heat is simulated as a wall heat flux qb at
one duct wall (the bottom wall in Fig. 2);
• When the porous layer is thick and needs to be considered (as for the ducts
in PEMFCs and anode-supported SOFCs), it is assumed to be homogeneous
and characterized by effective parameters, such as porosity, permeability and
thermal conductivity, and the fluid in the porous layer is in thermal equilibrium
with the solid matrix;
• Mass consumption and generation are simulated by a mass transfer flux (see
following sections for more details);
• In PEMFCs, liquid water appears in the form of small droplets in the gas species.
A multi-phase mixture model is employed to describe two-phase flow and heat
transfer in the composite duct. This model enables the calculation of the indi-
vidual phase velocities from the mixture flow field when the capillary flow due
to the pressure gradient and gravity-induced phase migration are considered.
For simplicity, the interfacial shear force and surface tension force between the
liquid water and the gas phase are neglected; liquid water has the same pressure
as the gas species.
5.3 Governing equations
The governing equations to be solved are the continuity, momentum, energy and
species equations. The mass continuity equation is written as
∇ · (ρeff V) = Sm . (14)
The source term Sm in the above equation accounts for the mass balance caused
by the reaction from/to the active surface Aactive (bottom surface in Fig. 2).
For PEMFCs, it corresponds to the hydrogen and water consumption on anode
side, oxygen consumption and water generation on the cathode side, respectively.
These are given by [19, 20],

I α Aactive
Sm = SH2 + Sa,H2 O = − MH2 − I · MH2 O , (15)
2F 2F V

I (1 + 2α)I Aactive
Sm = SO2 + Sc,H2 O = − MO2 + MH2 O , (16)
4F 2F V
where V refers to control volume at the active site.
The momentum equation reads
∇ · (ρeff VV) = −∇P + ∇ · (µeff ∇V) + Sdi . (17)
The inclusion of the source term Sdi allows eqn (17) to be valid for both the
porous layer and the flow duct
Sdi = −(µeff V/β) − ρeff BVi |V|. (18)
The first term on the right hand side of the above equation accounts for the linear
relationship between the pressure gradient and flow rate according to Darcy’s law.
The second term is the Forchheimer term which takes into account the inertial force
effects, i.e., the non-linear relationship between pressure drop and flow rate. In
eqn (18), β is the porous layer permeability, and V represents the volume-averaged
velocity vector of the species mixture. For example, the volume-averaged velocity
component U in the x direction is equal to εUp , where ε is the porosity, Up the
average pore velocity (or interstitial velocity). This source term accounts for the
linear relationship between the pressure gradient and flow rate by the Darcy law.
It should be noted that eqn (17) is formulated to be generally valid for both the
flow duct and the porous layer. The source term is zero in the flow duct, because
the permeability β is infinite. Equation (17) then reduces to the regular Navier-
Stokes equation. For the porous layer, the source term (eqn (18)) is not zero, and
the momentum eqn (17) with the non-zero source term in eqn (18) can be regarded
as a generalized Darcy model.
The energy equation can be expressed as

keff
∇ · (ρeff VT ) = ∇ · ∇T + Swp , (19)
cpeff
where Swp is the heat source associated with the water phase change (condensation/
vaporization) for the case of PEMFC (see [20]).
Swp = Jwl × hwl , (20)
where Jwl is the mass flux of liquid water by phase change, and hwl is the water
latent heat.
The species conservation equations are formulated in the general equation,
∇ · (ρeff Vφ) = ∇ · (ρeff Dφ,eff ∇φ) + Sφ , (21)

where φ is the mass fraction. The above equation is solved for the mass fraction
of H2 in the case of an SOFC anode, and O2 , H2 O(v) and H2 O(l) in the case of
a PEMFC cathode, respectively. The concentration of the inert species, nitrogen,
is determined from a summation of the mass fractions of the other species. For
PEMFC, the source term in eqn (21) includes water vapor and liquid water caused
by the phase change. It is written as follows [20]

Pw,sat − Pwv massi
Swv = −Swl = MH2 O . (22)
P − Pw,sat Mi
i
In eqn (22), massi refers to mass of species i; Pwv refers to partial pressure of
water vapor, Pw,sat is the saturation pressure at the local temperature, while P is
local pressure. If the partial pressure of water vapor is greater than the saturation
pressure, water vapor will condense, and a corresponding amount of liquid water
is formed. The water vapor partial pressure Pwv in the above equation is calculated
based on its concentration and local pressure of the gas mixture, while the saturated
pressure Pw,sat at the local temperature reads [21],
log10 Pw,sat = −2.179 + 0.029T − 9.183 × 10−5 T 2 + 1.445 × 10−7 T 3 . (23)
5.4 Boundary and interfacial conditions
A constant flow rate U = Uin is specified at the inlet of the gas flow duct, while
U = 0 is specified at the inlet and the outlet of the inter-connector (or collector)
and porous layer. The other boundary conditions employed can be written as:
∂T
U = V − Vm = W = 0, −keff = qb ,
∂y
∂φ
−ρeff Dφ,eff = Jφ at bottom wall (y = 0), (24)
∂y
U = V = W = 0, q = 0 (orT = Tw ), Jφ = 0
at top and side walls, (25)
∂U ∂V ∂T ∂φ
= =W = = =0
∂z ∂z ∂z ∂z
at mid-plane (z = a/2). (26)
It should be noted that all the walls for the above boundary conditions are on
the external surfaces of the solid layer and porous layer. In eqn (24), Vm is the
wall velocity of mass transfer caused by the electrochemical reaction (see eqns (1)
and (2)). The detailed procedure to obtain Vm was presented in [35], and the final
form of the source term in eqn (14) is given by
ν a
Sm = ρeff Rem , (27)
Dh A
where Rem = Vm Dh /ν is the wall Reynolds number caused by the electrochemical

reaction. The other variables can be found in the nomenclature list. qb in eqn (24)
is the heat source caused by the reaction and was given in eqn (12) (see [20]).
Among various interfacial conditions between the porous layer and gas flow
region, the continuity of velocity, shear stress, temperature, heat flux, mass fraction
and flux of species (for the oxygen, water vapor and liquid water, respectively) are
adopted, i.e.,
U− = U+ , (µeff ∂U /∂y)− = (µf ∂U /∂y)+ , (28)
T− = T+ , (keff ∂T /∂y)− = (kf ∂T /∂y)+ , (29)
φ− = φ+ , (ρeff Dφ,eff ∂φ/∂y)− = (ρeff Dφ,eff ∂φ/∂y)+ . (30)
Here the subscript + (plus) is for the fluid side, while − (minus) is for the porous
layer side. Moreover, the thermal interfacial condition eqn (29) is also applied at
the interface between the porous layer and solid layer with ks instead of keff .
5.5 Additional equations
It should be noted that the properties in the above equations with subscript ‘eff’
are effective ones. For the flow duct, the effective properties are reduced to regular
values of the species mixture based on the species composition, or regarded as con-
stant values in some cases. In the porous layer, there are many factors affecting the
effective properties such as the microstructure or nanostructure of the porous layer,
species composition and local temperature etc. It is not easy to obtain more accurate
values at this moment because the available data of the porous layer structure are
still limited. It has been found that setting µeff = µf and ρeff = ρf provide good
agreement with experimental data [43]. For the sake of simplicity, this approach is
adopted here as well.
To reveal the porous layer effects, parameter studies can be carried out for the
conductivity keff and species diffusion coefficients Dφ,eff by employing the ratios, θ,
θk = keff /kf , (31)
θD = Dφ,eff /Dφ . (32)
In eqn (31), kf is the species mixture conductivity in the porous layer, and is
estimated by [39],
−1
1 xi
kf = · xi kfi + , (33)
2 kfi
i
where xi is the mole fraction, and kfi conductivity of the species component. The
diffusion coefficients Dφ in eqn (32) are the values of the species components in the
species mixture, i.e., DO2 , DH2 O (v) and DH2 O (l) for oxygen, water vapor and liquid
water, respectively. However, the binary diffusion coefficients of the components

in pure air are used as estimations of Dφ in the calculations [23].
For the case of PEMFC, the effective diffusivity ratios are corrected by applying
the so-called Bruggemann correction [23, 24] to account for the effects of porosity
in the porous layer,
θD = ε1.5 . (34)
It should be noted that the thermal physical properties of the species mixture, such
as the density ρf , viscosity νf are estimated as functions of the local concentration.
According to Dalton’s law, the relative humidity of the species mixture is def-
ined as
Pwv P
η= = xwv , (35)
Pw,sat Pw,sat
where P is the pressure, Pwv the water vapor partial pressure, Pw,sat the saturation
pressure identified in eqn (23), xwv water vapor molar fraction. The liquid phase
saturation s is employed to describe the liquid water volume fraction in the species
mixture. It reads [21]
ρφw − ρg φwv
s= , (36)
ρwl − ρg φwv
where φ is mass fraction, ρg gas phase density, ρwl liquid phase density. The density
of the two-phase species mixture is
ρ = ρg (1 − s) + ρwl s. (37)
5.6 Solution methodology
Due to the similarity of the conservation equations for SOFC and PEMFC, eqns
(14, 17, 19, 21) can be written in the general form as

∇ · (ρeff V) = ∇ · ,eff ∇ + S , (38)
where denotes any of the dependent variables, is the diffusivity and S is a

source term or sink term.
Because there is no analytical solution to eqn (38), computational fluid dynamics
(CFD) methods have to be employed to obtain the discrete solutions. Once in this
form, the equations are integrated over control volumes. The boundary conditions
are introduced as source terms in control volumes neighboring boundaries whenever
appropriate. Details of various numerical schemes can be found in [34].
The numerical analysis, performed in this chapter, has been conducted by devel-
oping various source terms/boundary conditions related to the fuel cell transport
processes, and implementation into general purpose and in-house developed CFD
codes. The finite-volume method is used in the codes. More detailed discussion on
the codes can be found in [35].
It should be noted that the source term in eqn (14) (accounting for mass transfer
effects) is zero in most of the regions, and non-zero only in the regions neighboring
boundaries, where mass transfer caused by the electrochemical reaction occurs
(bottom wall in Fig. 2). The CFD procedures are modified accordingly and the
source term Sm is implemented in the pressure correction equation to adjust the
mass balance due to mass transfer. For the composite duct, it is clear that no gas
flow is present in the solid inter-connector (or -collector). Equations (17) and (21)
are solved by applying high viscosity values and only the heat conduction equation,
derived from the energy eqn (19), is solved for this domain.
In order to evaluate the performance of the numerical method and codes, test cal-
culations considering grid sensitivity, code performance and validation have been
carried out. These results can be found in [35] and the literature cited there. Due to
the lack of experimental data for fuel cells, it should be mentioned that the computa-
tional codes have been validated by comparisons with fully developed/developing
conditions in pure duct flow with mass transfer and ducts for various thicknesses
of the porous layer.
6 Results and discussions

In the first section, dimensionless pressure differences and convective heat transfer
coefficients, represented by friction factors and Nusselt numbers, respectively, are
presented for conventional SOFC design, i.e., the porous electrodes (anode/cathode)
are very thin and negligible, compared to the flow ducts. For the PEMFC and anode-
supported SOFC design, the porous layer is thick and should be included in the
analysis. The importance of gas flow and heat transfer in the porous layer and the
effects on the transport processes in the flow duct are presented and discussed. For
simplicity, the thermal boundary condition TBC-I refers to that of the constant heat
flux at one wall, and constant temperature at the other three walls. TBC-II refers
to the combination of constant heat flux at one wall, and thermal insulation at the
remaining walls.
6.1 Mass transfer effects on the gas flow and heat transfer
For electrolyte-supported SOFC ducts, an overall mass balance is considered for

the flow duct when the species consumption and generation happen in a fully
developed flow.
ṁin + ṁm = ṁout , (39)
where ṁm is the mass flow rate from the active wall caused by the electrochemical
reaction. Because Rem ( = Vm Dh /v) is assumed to be very small compared to the
main flow Reynolds number, the mass transfer rate ṁm is given by:
d ṁm = ρaVm dx (40)
in which, dx is the increment in the main flow direction. The change of mass flow
rate due to the mass transfer (suction/injection) reads:

∂Ubulk ∂Ubulk
d ṁ = ρA Ubulk + dx − ρAUbulk = ρA dx. (41)
∂x ∂x
By combining eqns (40) and (41), the mass flow rate change can be written [35].
∂Ubulk v a
Sm = ρ = ρRem , (42)
∂x Dh A
where ∂Ubulk /∂x is the velocity gradient in the main flow direction induced by
the mass transfer. To characterize the overall pressure difference between inlet and
outlet, either a pressure coefficient Cp or an apparent friction factor fapp of the gas
flow in a duct can be employed as
(Pin − P)
Cp = 2 /2)
, (43a)
(ρUbulk
Dh dP
4fapp = 2
, (43b)
(ρUbulk /2) dx
where Ubulk is the mean velocity of the main flow, Dh is the hydraulic diameter
defined in the conventional manner, dP/dx the pressure gradient along the main
flow direction.
The bulk velocity is calculated as:
#
UdA
Ubulk = # (44)
dA
and the hydraulic diameter is defined as:
4A
Dh = , (45)
P∗
A is the cross-sectional area and P ∗ is the wetted perimeter. It is clear that the mass
transfer contributes to a change of the main flow velocity, thus the local Reynolds
number varies as a function of x and mass flow rate even when the flow is fully-
developed. It should be noted that the fluid pressure is affected by inertia forces
and wall friction. As an example, the pressure difference increases for the case
of injection, and it is larger than for the case without mass transfer. The apparent
friction factor fapp is employed in this study because it incorporates the combined
effect of wall shear and the change in momentum flow rate due to the effects of
mass generation and consumption by the electrochemical reaction.
For fully developed flow with mass transfer, the Nusselt number can be derived
from an energy balance in the duct. As mentioned earlier, the mass flow through
the porous wall is much smaller than the main flow, and the total heat flow rate qb
(per unit length of the duct) can then be calculated by (Fig. 2):
dT
qb = ρUbulk cp A − ρcp Vm a (Tbulk − Tw ) . (46)
dx
This equation can be rearranged for a rectangular duct as:
qb Dh ρUbulk cp A dT
dx Dh Dh
∗
= ∗
− ρcp Vm a ∗ . (47)
Tbulk − Tw kP (Tbulk − Tw ) kP kP
The left hand side is the definition for the Nu and the first part on the right hand
side is the definition of the Nusselt number without mass transfer (Nuf ).
qb Dh
Nu = , (48)
(Tbulk − Tw )kP ∗
ρUbulk cp A dT
dx Dh
Nuf = , (49)
(Tbulk − Tw ) kP ∗
where Tw , the wall temperature, is the same on three walls but varies on the fourth
wall for TBC-I. Tbulk is the bulk flow mean temperature in the cross section, and is
calculated as #
T |U | dA
Tbulk = # . (50)
|U | dA
The second part of eqn (47) is the contribution to the Nusselt number by mass
transfer. This term is rewritten as
b
Num = −PrRem . (51)
P∗
Equation (47) is then rewritten as:
Nu/Nuf = 1 + Num /Nuf . (52)
The Nusselt number Nuw can be defined as:

hw Dh qw Dh
Nuw = = , (53)
k k(Tw − Tbulk )
hwDh q w Dh
Nuw = = , (54)
k k(T w − Tbulk )
where Nuw and Nuw are spanwise variable and average Nusselt numbers of the
heated wall at location x, respectively. qw is the wall heat flux; Tw and T̄w are
spanwise variable and average temperature of the heated wall, respectively. The
dimensionless axial distance x∗ in the flow direction for the hydrodynamic entrance
region is defined as:
x∗ = x/(Dh Re). (55)
Mass transfer effects on the friction factor and Nusselt number are shown in Fig. 3
for a rectangular duct with TBC-I. For the case of mass injection from the porous
wall, additional mass is induced to the duct and thus the axial velocity increases. As
clarified earlier, the fapp Re is related to the pressure gradient as well as changes in
the momentum flux in the main flow direction. As can be seen from Fig. 3(a), fapp Re
always increases for mass injection (Rem > 0), while it decreases for mass suction
(Rem < 0). On the other hand for heat transfer, the temperature of the fluid will
increase due to the heat induced by mass injection into the fluid, while a decrease
Figure 3: Mass transfer (Rem ) effects on the fully developed convective flow: (a) the
apparent friction factor and (b) Nusselt number in a rectangular duct at
TBC-I, SOFC.
appears for the case of mass suction. The Nu/Nuf is thus reduced by mass injection,
which can be seen in Fig. 3. A large aspect ratio has a significant effect on both
fapp Re and Nu/Nuf , while a small aspect ratio gives less effect. Both fapp Re and
Nu/Nuf approach the values for the case without mass transfer (Rem = 0), if the
aspect ratio becomes about 0.1. The figures show also that the fapp Re and Nu/Nuf
has a minimum when the aspect ratio is unity, i.e., a square duct. More discussion
can be found in [35].
Comparisons of fapp Re and Nu for a developing flow with and without mass
transfer in a rectangular duct are shown in Fig. 4 with TBC-II. It is clear that
the apparent friction factor fapp Re is decreased while the Nusselt number Nu is
increased. More discussion can be found in [35, 45].
6.2 Porous layer effects on the transport processes
6.2.1 Transport processes in PEMFCs

In this section, the gas flow and heat transport in a porous layer are included for
PEMFCs. The geometry of the duct considered is similar to that in [19, 20], i.e.,
overall channel dimension is 10 cm ×0.20 cm ×0.16 cm (x ×y ×z), gas flow duct is
10 cm×0.12 cm×0.08 cm (x×y×z), while the diffusion layer is 10 cm×0.04 cm×
0.16 cm (x × y × z). The thickness ratio hr (thickness of porous layer hdiff over total
height h of the channel) is 20%. The CFD procedure used is modified accordingly
and the source term Sm is implemented in the pressure correction equation to adjust
the mass balance due to mass transfer. The conditions and parameters considered
as the base case are shown in Table 1, along with the references for the various
parameters used in the model.
Figure 4: Mass transfer effects on fapp Re and Nu for developing flow and heat
transfer in a rectangular duct (aspect ratio 2:1) at Re = 250, Rem = −1,
TBC-II, SOFC.
Table 1: Parameters implemented as the base case in the PEMFC study.
Cases Parameter Value Reference
Anode inlet Mole fraction of H2 /Water vapor 0.53/0.47/80 [20]

H2 O/Tin ,◦ C
Cathode inlet Air with Tin ,◦ C 80 [20]
Operating Permeability of diffusion layer β, m2 2 × 10−10 [20]
condition Porosity ε 0.7 [19]
Cell voltage Vcell , V 0.53 [20]
Net water transport coefficient α 0.3 [19]
Current collector thermal 5.7 [20]
conductivity k, W/mK
The local current density I of the cell is essential for source term calculations.
Because the gas flow and heat transfer in the porous layer and flow ducts are of major
interest initially, the approach used in [21] is adopted by prescribing a local current
density at the porous layer surface close to the catalyst layer (bottom surface in the
study). It should be mentioned that this limitation is abandoned in the following
section, where the local current density will be calculated. For a similar geometry,
the width-average local current density profiles along the duct length in [19] are
adapted to account for various operating conditions (co-/counter-flow, humid/dry
air, membrane thickness and cell voltage). The current density of the base case in this
study refers to the case by Yi and Nguyen [19], while the current densities of cases
1, 2 and 3 correspond to cases 1.0, 1.1 and 1.2 from [19], respectively. The current
Figure 5: Effects of: (a) thickness and (b) permeability of porous layer on axial
velocity profiles in the cathode duct at θk (= keff /kf ) = 1, PEMFC.
density of the base case is the largest, and case 3 represents the smallest one while
the remaining cases have current densities in between the others. A parameter study
with various constant current densities has also been performed.
Figure 5(a) presents porous layer thickness effects on the main flow velocity
profile (close to the exit) for a fixed value of the permeability (βi = 2.0 × 10−10 ).
The heights of the gas flow duct and solid current collector are kept constant, and
the thickness ratio hr is approached by varying the porous layer thickness and the
total height of the duct. It is obvious that the fluid velocity in the porous layer
decreases significantly by increasing the thickness of the porous layer. Figure 5(b)
corresponds to the case where the thickness of the porous region equals 20 percent
of the duct height and the effect of permeability is shown. Clearly, in the porous
layer the fluid flow rate is low, and it becomes significantly reduced when the
permeability is low (e.g., βi = 2.0 × 10−11 ). However, the corresponding velocity
profiles in the gas flow duct look very similar (see Fig. 5(b)).
In Fig. 5(a), it is noticed that the velocity gradient at the interface region between
the porous layer and gas flow duct becomes sharper as the thickness ratio of the
diffusion layer increases. As a result, a larger fapp Re can be found in Fig. 6(a) with
a larger thickness ratio of the porous layer (i.e., 40%). Compared to Fig. 6(a), it
is found that the permeability at constant thickness ratio has less effect on fapp Re,
which is consistent with the small effect of permeability on the velocity gradient at
the interface region, shown in Fig. 6(b).
The effects of various current densities on the cross-sectional averaged Nu are
shown in Fig. 7(a). As seen here, Nu of the base case is smaller than those of other
cases. A large current density (i.e., base case) can contribute to a large net mass
injection to the duct, and to a bigger temperature difference between the heated
surface and fluid because more mass and thermal energy are inserted into the duct
by the mass injection. A parameter study for various overall current densities with
Figure 6: Effects of: (a) thickness and (b) permeability of the porous layer on fapp Re
of cathode duct at θk = 1, PEMFC.
Figure 7: Effects of: (a) current density and (b) permeability on Nu along the main
flow direction of the cathode duct at Re = 300, θk = 1.0, hr = 20%,
PEMFC.
constant values has been performed, and the results are also shown in Fig. 7(a). The
conclusion is that a large current density (1.5 A/cm2 in Fig. 7(a) affects the decrease
of Nub . Figure 7(b) shows Nu with various permeabilities at a fixed current density
(base case) and other porous layer parameters for the cathode duct. It is found that
by increasing the permeability (βI = 2.0 × 10−9 ) Nu will increase, otherwise Nu
will decrease [45].
As mentioned earlier, there exist some studies in the available literature modi-
fying the Darcy model limitations. For example, the term accounting for friction
due to macroscopic shear was included in the model. This model is usually referred
Table 2: The inertial coefficient B in eqn (18).
Model The inertial coefficient References
BFD-1 B = εF/(K)0.5 [29, 40, 41, 42]

BFD-2 B = 1.75(1 − ε)/(ε3 d ) [43]
Table 3: The Forchheimer coefficient F in BFD-1 model.
Model The Forchheimer coefficient References
BFD-1a F = 1.8/(180ε5 )0.5 [29]

BFD-1b F = 0.143ε−1.5 [40]
Figure 8: (a) fapp Re and (b) Nu of a cathode duct by generalized BD and BFD
models.
to as the Brinkman-extended Darcy (BD) model. It was also suggested that a term
representing the inertial energy of the fluid should be included in the Darcy’s model
and this is often referred to as the Forchheimer term [29, 40]. The inertial coeffi-
cient B in eqn (18) depends very much on the microstructure or nanostructure of
the porous medium, and theoretical determination of it is not easy. As an example,
two models from the literature concerning B are given in Table 2. It is clear that
model 2 in the table needs more detailed information about the porous medium
microstructure or nanostructure in a PEMFC, e.g., particle diameter d , which is
not available at the present moment. Therefore, only model 1 is used here. Table 3
shows methods how to determine the Forchheimer coefficient F in model 1. Both
methods have been adopted in this study, together with a parametric study.
Figures 8(a) and (b) show the variations of the apparent friction factor ratio
fapp Re/fd Re (where fd is the friction factor for fully developed flow) and the Nusselt
number Nu along the axial direction with various models. It is clear that both
Figure 9: Effects of the inertial energy on: (a) fapp Re and (b) Nu of a cathode duct
predicted by the generalized BFD models, PEMFC.
fapp Re/f d Re and Nu rapidly decay due to both the hydrodynamic and thermody-
namic boundary layer development. Compared to the BD model, the generalized
BFD models predict a larger apparent friction coefficient and a smaller Nu for the
PEMFC cathode duct, as seen in Figs 8(a) and (b). This is because the additional
inertial force due to the Forchheimer term (B) retards the side-flow through the
porous layer thereby allowing the axial velocity and its gradient to increase. A
similar effect occurs due to mass injection into the duct.
On the other hand, the retarded flow from the porous layer induces more heat
into the duct. It is worth noting that the BFD model has significant effects on the
decrease of Nu, but has small effects on the increase of fapp Re. It is also clear
that both BFD-1a and 1b models provide similar results for Nu, which are close to
that at F = 104 in the figure. From Table 3, it can be verified that the Forchheimer
coefficient F from both models is in the order of 104 for this specific PEMFC case
(ε = 0.7) [46, 47].
Parametric studies of the inertial effects, in terms of the Forchheimer coefficient
F, have been conducted for a wide range, covering PEMFC cases. Results are shown
in Figs 9(a) and (b). With reference to fapp Re in Fig. 9(a), it is noticed that an increase
of F (BFD Models vs BD Model) can increase fapp Re, but all the cases provide
similar fapp Re, i.e., the Forchheimer coefficient F has a limited effect on fapp Re. On
the other hand for heat transfer, significant changes of Nu are expected by increasing
the Forchheimer coefficient F, i.e., the inertial force with a larger Forchheimer
coefficient forces more gas from the porous layer to the flow duct. Consequently,
more heat induced by the retarded flow can be transferred into the duct from the
heated wall, and the convective heat transfer coefficient, Nu, decreases, see Fig. 9(b).
6.2.2 Transport processes in SOFCs

For anode-supported SOFCs, the following duct geometries are employed [9]:
length of the duct L = 20 mm; width of the porous layer a = 2 mm, and its thickness
Figure 10: (a) Permeation Reynolds number Rep , (b) fapp Re and Nu along the main
flow direction at the base case condition, SOFC.
hp = 2 mm; while the width of the flowing duct b = 1 mm, and its height hd = 1 mm.
Fuel gas is 0.80 in mass fraction of H2 , and 0.20 of water vapor with an inlet tempera-
ture Tin = 750 ◦ C; thermal conductivity ratio kr (= keff /kf ) = 1, effective dynamic
viscosity µeff = µf ; porosity ε = 0.5 and permeability βI = 1.7 × 10−10 m2 ;
Rein = 100, Rem = 1.0, DH2 , f = 3 × 10−4 m2 /s.
Figure 10 shows the permeation Reynolds number Rep , the friction factor fapp Re
and the Nusselt number Nu along the main flow direction of an anode-supported
SOFC duct. Similar to the wall Reynolds number Rem (for mass transfer across
the bottom wall), the permeation Reynolds number is defined as Rep = Vp Dh /ν
for gas permeation across the interface. Here, Vp is the velocity caused by the gas
permeation across the interface. It is found that Rep has a large negative value (i.e.,
permeation into the porous anode layer) at the inlet region, see Fig. 10(a). Due to
the decreasing pressure gradient along the duct, permeation into the porous layer
becomes smaller. On the other hand, H2 O generation caused by the electrochemical
reaction at the bottom wall, together with back permeation described above, con-
tributes to a mass injection into the flow duct. This is confirmed by a small but posi-
tive (i.e., back flow into the flow duct) Rep shortly downstream the inlet in Fig. 10(a).
For a pure flow duct with impermeable walls, fapp Re decays rapidly from the
inlet, and levels off to a constant value as the convective gas flow becomes fully
developed (see Fig. 10(b)). For the anode duct with a porous layer, the friction factor,
fapp Re, is small at the inlet region but increases rapidly at the entrance region and
also levels off to a near-constant value shortly downstream along the main flow
direction. Similar to a suction flow from a duct, there is certainly a decrease in the
friction factor from that of the pure forced convection, as mentioned before.
For increasing x∗ , the Rep becomes smaller and its contribution to the decrease
of fapp Re is less significant. This contribution will be zero when Rep = 0. Along the
flow direction beyond this position, the gas flow is possibly affected by the following
Figure 11: (a) H2 and (b) H2 O mass concentration distribution along the main flow
direction of an SOFC anode duct.
mechanisms: secondary flow and back permeation to increase fapp Re, convective
flow to decrease fapp Re. It can be clearly observed that fapp Re in Fig. 10(b) is nearly
constant downstream from the entrance region. Further downstream in the flow
duct, the secondary flow and back permeation balance each other and the effects
on the main flow disappear. The Nu for the composite duct has a similar behaviour
as that of the pure flow duct. However, a slightly bigger Nu can be observed for the
composite duct (see Fig. 10(b)), due to the mass permeation into the porous layer.
From the discussion above, it is clear that mass permeation across the interface
has more significant effects on the gas flow than on the heat transfer, both in the
entrance region and further downstream.
H2 and H2 O concentration profiles along the main flow direction are shown in
Figs 11(a) and 11(b), respectively. It is found that the H2 concentration decreases,
while H2 O increases along the main flow in the porous layer and the flow duct. This
is due to the consumption of H2 and generation of H2 O during the electrochemical
reaction. Moreover, the gradients of the H2 and H2 O concentrations in the direction
normal to the active surface (the bottom surface in Fig. 11) are larger close to the
reaction sites compared to the interface areas of both porous layer and flow duct [48].
The performance of the anode-supported SOFC duct is also analysed using the
vertical component of the total hydrogen mass flux vector at the active site (bottom
surface), which can be expressed as:

∂mH2
JH2 , y b = ρeff mH2 V + ρeff DH2 ,eff . (56)
∂y b
It should be noted that the first term on the right hand side represents the con-
vection transport, while the second term is the contribution due to the diffusion.
Figure 12 shows a comparison of the hydrogen mass fluxes by convection, diffusion
and the total value for the base condition. It can be seen that convection mass flux
has a large negative value (i.e., fuel species transport is to the reaction sites) at the
Figure 12: Various modes of hydrogen flux at the active surface for the base case
condition, SOFC.
inlet region. Due to the decreasing pressure difference along the duct, convection
becomes weaker downstream. On the other hand, water generation caused by the
reaction at the active sites, together with the back permeation mentioned above,
contributes to species flowing back to the flow duct. This is confirmed by a small
positive value of the convection flux. It is also clear that the hydrogen diffusion
flux is small in the entrance region due to a small hydrogen concentration gradient
as discussed above, but it increases and reaches a peak value at a certain position
of the duct. This is caused by the reaction. The diffusion flux maintains almost a
constant value further downstream.
By comparing the absolute values of convection and diffusion fluxes, it is found
that the convection is stronger in the entrance region; however, the diffusion dom-
inates the species transport downstream. The position for this occurrence is about
1/6 length from the inlet for this specific case. Consequently the total flux from
eqn (56) is controlled by convection in the entrance region, and by diffusion for the
rest of the duct, as seen in Fig. 12.
6.3 Two-phase flow and its effects on the cell performance
As mentioned earlier, the important feature of the two-phase flow model imple-
mented in the work described in this chapter is based on the two-phase mixture
approach to account for the phase change and its effects on the gas flow and
heat transfer. The approach is to model the liquid water transported by the multi-
component gas mixture in terms of the convection and the diffusion. The amount
of water undergoing phase-change is calculated based on the partial pressure of
water vapor and the saturation pressure. It is worthwhile to note that two-phase
flow and heat transfer are concerned to get local pressure, temperature and species
component composition. The model is therefore considered as a non-isothermal
Figure 13: (a) Velocity vectors and (b) contours of temperature T along the main
flow direction of a composite duct, PEMFC.
and -isobaric model. As mentioned above, the source term in eqn (21) is for the
water phase change. When the partial pressure of water vapor is greater than the sat-
uration pressure, water vapor will condense to liquid water. Consequently, the mass
fraction will be reduced in the main gas flow, together with a release of water latent
heat, which continues to occur until the partial pressure equals the local saturation
pressure. On the other hand, if the partial pressure is lower than the saturation pres-
sure, the saturated water, if any, will evaporate. It should be mentioned that the
source term in eqn (22) concerning the water phase-change and the associated heat
source term in eqn (20) correspond to the control volumes where two-phase water
appears, and these are not treated as the boundary conditions.
In this section, the main results of numerical simulations are reported and dis-
cussed for a cathode duct of PEMFCs. Figure 13(a) shows the velocity profile along
the main flow direction, in which the scale of the vector plots (i.e., 2 m/s) is a ref-
erence value of the maximum velocity. As shown in the figure, a parabolic profile
is clearly observed in the flow duct. On the other hand, because the species have
difficulties penetrating into the porous layer, the velocity in the porous layer is very
small except in the region close to the flow duct.
Figure 13(b) shows that the temperature increases along the main flow direction.
The variation in temperature distribution can also be observed in the vertical direc-
tion with a slightly larger value close to the bottom surface. These effects are created
by the heat generation due to both the reaction close to the active surface, and the
latent heat release by water condensation in the two-phase region, caused by the
increase in the water vapor concentration in this area. It is worthwhile to note that
the temperature is non-uniformly distributed. By considering the local temperature
distribution, the effects on the saturation pressure can be found, which is not the
case for the isothermal assumption employed by different authors in literature [49].
Water activities in the duct are shown in Fig. 14. The mass concentration of
the water vapor at the entrance is 23%, which corresponds to the saturated case at the
base condition (Tin = 70 ◦ C). It can be observed that water vapor is generated at the
bottom surface, and is transported back to the flow duct through the porous layer.
Figure 14: Mass concentration profiles for: (a) water vapor and (b) liquid water
along main flow direction of PEMFC cathode duct at the base case.
Figure 15: Liquid saturation profile along the main flow direction in a composite
duct at the base case, PEMFC.
For this reason, larger mass fractions of water vapor can be found in the porous
layer close to the bottom surface. Therefore, the partial pressure of water vapor is
greater in the regions mentioned above, and smaller at the interfacial region and
the flow duct. Based on the calculated partial pressure and the saturation condition,
the liquid water content was predicted and shown in Fig. 14(b). It is found that the
liquid water appears in both the flow duct and the porous layer, with the largest mass
fraction (around 10%) in the porous layer close to the exit. Because the saturation
pressure is proportional to the local temperature, smaller saturated pressures can
be expected for the flow duct compared to the porous layer. This is the reason why
the liquid water can appear in the flow duct as well, but with smaller mass fractions
(less than 5%). A proper gradient of the liquid water concentration should be kept
for the liquid water to be driven out of the porous layer.
Figure 15 shows the corresponding liquid water saturation level along the main
flow direction of the composite duct. As shown in eqn (36), liquid saturation is
represented by the volume occupied by the liquid phase divided by the total flow
Figure 16: Liquid water flux at the active surface for: (a) the main flow direction
and (b) the cross-sections in a composite duct for the base case, PEMFC.
volume of the duct. It is clear that the liquid saturation s is zero in the single-phase
gas flow region. It is so because the species density reduces to the gas phase density,
and the water mass fraction equals the mass fraction of the water vapor as well.
The liquid saturation s is 1.0 for the case of the single-phase liquid flow. As shown
in Fig. 15, the liquid saturation s is zero in the inlet region and increases along the
flow direction. It is also true that the liquid saturation s decreases from the active
site to the flow duct at the same x due to the liquid water transport, as discussed
later in this paper. The liquid water occupies more volume in the porous layer with
the largest value of s appearing in the corner close to the active surface at the exit
of the duct, while the single-phase gas species occupy most of the flow duct except
that small values of s can be observed at the interface region after a certain distance
downstream the inlet.
As discussed above, the liquid water mass composition and saturation level have
highest values at the active surface, consequently larger contribution to the species
flow and heat transfer can be expected. In this section, the liquid water flow in the
porous layer is evaluated by the liquid water mass flux at the active site, as shown
in eqn (56). It should be noted that a positive value represents liquid water flowing
from the active site to the porous layer and then the flow duct.
Figure 16(a) shows the comparison of the liquid water mass fluxes by convection,
diffusion and the total value for the base case. It is found that both convection and
diffusion mass fluxes have very small values at the inlet region.
Due to the phase change, when the water vapor is generated and the satura-
tion condition is reached, the liquid water can be accumulated in this region. Its
mass concentration and gradient along the duct become larger. Bigger values can
therefore be observed for both the convection and diffusion downstream a cer-
tain distance from the inlet. As shown in Fig. 16(a), the saturated water transport
is dominated by diffusion which takes place mainly in the porous layer. In fact,
the convective contribution to the total water flux is estimated to be about 15%
Figure 17: Local current density distribution for: (a) the active surface and (b) the
cross-sections in a cathode duct of a PEMFC at the base case.
or less. Consequently the total flux from eqn (56) can be said to be controlled by
the diffusion for this specific case.
Figure 16(b) shows the predicted total values of the cross-sectional liquid water
mass flux at various locations. As shown, the mass flux is zero at the inlet region.
It is found that the total mass flux has almost uniform values in the cross sections,
except a weak liquid water flow for the site below the solid layer (z/h = 0.2−0.4),
which has a long transport distance from/to the flow duct [50].
As mentioned previously, the local current density (I ) is an important parameter
of the cell, and essential for the source term calculations related to mass injection/
suction and heat generation by the electrochemical reaction. In this section, the local
current density is calculated considering the effects of the liquid water saturation,
and can be expressed as follows,

(1 − s)φO2 O2,trans F
I = Io exp − Vover , (57)
φO2 ,ref b RT
where I is the local current density based on the Tafel equation along the active sur-
face, Io the exchange current density per real catalyst area, φO2 the oxygen species
mass concentration, O2,trans the transfer coefficient for the oxygen reaction, Vover the
cathode over-potential, φO2,ref the oxygen reference concentration (e.g., P = 1 atm)
for the oxygen reaction. Other symbols are given in the nomenclature. The (1 − s)
factor in eqn (57) accounts for the effect of liquid water saturation on the surface
availability of the reaction site.
Figure 17(a) shows a local current density distribution on the active surface (in
the x-z plane). It can be seen that the current density is high near the entrance, and
then decreases along the main flow direction. This is because the oxygen trans-
fer to the reaction site is larger near the entrance region, which is dominated by
the oxygen convection. The reduced current density downstream is due to the
small oxygen transport rate controlled by the diffusion. It is also true that the
current density is non-uniformly distributed in the cross-section with a smaller
Figure 18: Comparison of the calculated results with the experimental data for
polarization curves (V -I curve) of PEMFC, from [52].
value in the corner region beneath the solid current collector. This is due to a
longer transport distance from the flow duct to the active site. With reference to
Fig. 17(b), a similar conclusion can be drawn, i.e., the current density is non-uniform
along the main flow direction and in the cross-section as well. It should be noted
that the liquid water saturation affects this non-uniform distribution as indicated
by eqn (57).
The simulation results from a 3-D PEMFC model developed in [52] are compared
with experimental data available for various parameters. Figure 18 shows a typical
V -I polarization curve at operating and humidification temperatures of 70 ◦ C on
both the anode and cathode sides. From the figure it is found that the comparison
is favorable at low and medium current densities, but this is not true for the case
for the high current densities. It is obvious that the simulation produces a higher
current density. As claimed in [52], the low current density of the experimental
results may be caused by the presence of liquid water content in the catalyst layers
and the gas diffusion layers, which the model employed in [52] neglected.
7 Conclusions
Based on the similarities of major transport processes in SOFCs and PEMFCs, a
unified framework of computational fluid dynamics (CFD) methodology has been
developed by implementing specific source terms to account for transport processes
and boundary conditions. Numerical investigations and analyses for species flow
and heat transfer have been presented for ducts of the type appearing in fuel cells.
Models were developed for ducts with the thermal boundary conditions appearing
in SOFCs and PEMFCs. Also, different duct configurations were considered, such
as rectangular and trapezoidal cross sections, and composite geometries includ-
ing porous layer, flow duct and solid current inter-collector. It was found that the
electrochemical reaction related mass consumption/generation in the electrodes,
duct configuration (such as the characteristics of the porous layer) have major
effects on the gas flow and heat transfer. Furthermore, water management involv-
ing two-phase flow in PEMFCs has been studied, and its importance for the cell
performance has been characterized.
Acknowledgements
Financial support was provided by the Swedish National Programme for Stationary
Fuel Cells. The Swedish Energy Agency (STEM) is the financial supporter and
Elforsk is handling the programme. The Wenner-Gren Foundation supported the
collaboration between Professors Faghri and Sundén.
References
[1] Haraldsson, K. & Vipke, K., Evaluating PEM fuel cell system models.
J. Power Sources, 126, pp. 88–97, 2004.
[2] FEMLAB Chemical Engineering, COMSOL documentation, 2002.
[3] Vayenas, C.G., Debenedettl, P.G., Yentekakls, I. & Hegedus, L.L., Cross-
flow, Solid-state electrochemical reactors: A steady-state analysis. Ind. Eng.
Chem. Fundam., 24, pp. 316–324, 1985.
[4] Ahmed, S., McPheeters, C. & Kumar, R., Thermal-hydraulic model of a
monolithic solid oxide fuel cell. J. Electrochemical Society, 138, pp. 2712–
2718, 1991.
[5] Bernier, M., Ferguson, J. & Herbin, R., A 3-dimensional planar SOFC Stack
Model. Proceedings of Third European Solid Oxide Fuel Cell Forum, Nantes,
France, pp. 471–480, 1998.
[6] Melhus, O. & Ratkje, S.K., A simultaneous solution of all transport processes
in a solid oxide fuel cell. Denki Kagaku, 64, pp. 662–673, 1996.
[7] Boersma, R.J. & Sammes, N.M., Distribution of gas flow in internally man-
ifolded solid oxide fuel-cell stacks. J. Power Sources, 66, pp. 41–45, 1997.
[8] Virkar, A.V., Chen, J., Tanner, C.W. & Kim, J.W., The role of electrode
microstructure on activation and concentration polarizations in solid oxide
fuel cells. Solid State Ionics, 131, pp. 189–198, 2000.
[9] Yakabe, H., Hishinuma, M., Uratani, M., Matsuzaki, Y. & Yasuda, I., Evalu-
ation and modeling of performance of anode-supported solid oxide fuel cell.
[10] Lehnert, W., Meusinger, J. & Thom, F., Modelling of gas transport phenom-
ena in SOFC anodes. J. Power Sources, 87, pp. 57–63, 2000.
[11] Ackmann, T., Haart, L.G.J., Lehnert, W. & Thom, F., Modelling of mass and
heat transport in thick-substrate thin-electrolyte layer SOFCs. Proceedings
of the 4th European Solid Oxide Fuel Cell Forum, Lucerne/Switzerland,
pp. 431–438, July 2000.
[12] Yi, J.S. & Nguyen T.V., An along-the-channel model for proton exchange
membrane fuel cells. J. Electrochem. Soc., 145, pp. 1149–1159, 1998.
[13] Nguyen, T.V. & White, R.E., A water and heat management model for proton
exchange membrane fuel cells. J. Electrochem. Soc., 140, pp. 2178–2186,
1993.
[14] Singh, D., Lu, D.M. & Djilali, N., A two-dimensional analysis of mass
transport in proton exchange membrane fuel cells. Int. J. of Eng. Sci., 37,
pp. 431–42, 1999.
[15] Okada, T., Xie, G. & Meeg, M., Simulation for water management in
membranes for polymer electrolyte fuel cells. Electrochem. Acta, 43,
pp. 2141–2155, 1998.
[16] Kazim, A., Liu, H.T. & Forges, P., Modelling of performance of PEM
fuel cells with conventional and interdigitated flow fields. J. Applied Elec-
trochem., 29, pp. 1409–1416, 1999.
[17] Um, S., Wang, C.Y. & Chen, K.S., Computational fluid dynamics modeling
of proton exchange membrane fuel cells. J. Electrochem. Soc., 147, pp. 4485–
4493, 2000.
[18] Voss, H.H., Wilkinson, D.P., Pickup, P.G., Johnson, M.C. & Basura, V.,
Anode water removal: a water management and diagnostic technique
for solid polymer fuel cells. Electrochimica Acta, 40, pp. 321–328,
1995.
[19] Dutta, S., Shimpalee, S. & Zee, J.W.V., Three-dimensional numerical sim-
ulation of straight channel PEM fuel cells. J. Appl. Electrochemistry, 30,
pp. 135–146, 2000.
[20] Shimpalee, S. & Dutta, S., Numerical prediction of temperature distri-
bution in PEM fuel cells. Num. Heat Transfer (Part A), 38, pp. 111–128,
2000.
[21] You, L. & Liu, H., A Two-phase and multi-component model for the cathode
of PEM fuel cells. ASME IMECE2001/HTD-24273, pp. 1–10, 2001.
[22] He, W., Yi, J.S. & Nguyen, T.V., Two-phase flow model of the cathode of
PEM fuel cells using interdigitated flow fields. AIChE Journal, 46, pp. 2053–
2064, 2000.
[23] Nguyen, T.V., Modeling two-phase flow in the porous electrodes of proton
exchange membrane fuel cells using the interdigitated flow fields. Elec-
trochem. Soc. Proc., 99–14, pp. 222–241, 2000.
[24] Wang, Z.H., Wang, C.Y. & Chen, K.S., Two-phase flow and transport in the
air cathode of proton exchange membrane fuel cells. J. Power Sources, 94,
pp. 40–50, 2001.
[25] Djilali, N. & Lu, D., Influence of heat transfer on gas and water transport in
fuel cells. Int. J. Therm. Sci., 41, pp. 29–40, 2002.
[26] Hwang, G.J., Cheng, Y.C. & Ng, M.L., Developing laminar flow and heat
transfer in a square duct with one-walled injection and suction. Int. J. Heat
Mass Transfer, 36, pp. 2429–2440, 1993.
[27] Kinney, R.B., Fully developed frictional and heat-transfer characteristics of
laminar flow in porous tubes. Int. J. Heat Mass Transfer, 11, pp. 1393–1401,
1968.
[28] Quaile, J.P. & Levy, E.K., Laminar fow in a porous tube with suction. ASME
J. Heat Transfer, 97, pp. 66–71, 1975.
[29] Teng, H. & Zhao, T.S., An extension of darcy’s law to non-stokes flow in
porous media. Chem. Eng. Sci., 55, pp. 2727–2735, 2000.
[30] Alkam, M.K., Al-Nimr, M.A. & Hamdan, M.O., Enhancing heat transfer in
parallel-plate channels by using porous inserts. Int. J. Heat Mass Transfer,
44, pp. 931–938, 2001.
[31] Alazmi, B. & Vafai, K., Analysis of fluid flow and heat transfer interfacial
conditions between a porous medium and a fluid layer. Int. J. Heat Mass
Transfer, 44, pp. 1735–1749, 2001.
[32] Wang, J., Gao, Z., Gan, G. & Wu, D., Analytical solution of flow coefficients
for a uniformly distributed porous channel. Chem. Eng. Journal, 84, pp. 1–6,
2001.
[33] Kee, R.J., Korada, P., Walters, K. & Pavol, M., A generalized model of the
flow distribution in channel networks of planar fuel cells. J. Power Sources,
109, pp. 148–159, 2002.
[34] Versteeg, H.K. & Malalasekera, W., An Introduction to Computational Fluid
Dynamics, the Finite Volume Method, Longman Scientific & Technical:
England, 1995.
[35] Yuan, J., Computational Analysis of Gas Flow and Heat Transport Phe-
nomena in Ducts Relevant for Fuel Cells, Ph.D. Thesis, ISBN 91-628-5540-
9/ISSN 0282-1990, Lund Institute of Technology, Sweden, 2003.
[36] Shah, R.K. & London, A.L., Laminar flow forced convection in ducts (Chap-
ters VII and X). Advances in Heat Transfer, eds T.F. Irvine & J.P. Hartnett,
Academic Press: New York, 1978.
[37] Comiti, J., Sabiri, N.E. & Montillet, A., Experimental characterization of
flow regimes in various porous media-III: limit of darcy’s or creeping flow
regime for Newtonian and purely viscos non-Newtonian fluids. Chem. Eng.
Sci., 55, pp. 3057–3061, 2000.
[38] Natarajan, D. & Nguyen, T.V., A Two-dimensional, two-phase, multicompo-
nent, transient model for the cathode of a proton exchange membrane fuel cell
using conventional gas distributors. J. Electrochem. Soc., 148, pp. A1324–
1335, 2001.
[39] Warnatz, J., Maas, U. & Dibble, R.W., Combustion-Physical and Chemical
Fundamentals, Modeling and Simulation, Experiments, Pollutant Forma-
tion, Springer: Berlin, Germany, 1996.
[40] Chikh, S., Bounedien, A. & K. Bouhadef., Non-darcian forced convection
analysis in an annulus partially filled with a porous material. Num. Heat
Transfer, 28, pp. 707–722, 1995.
[41] Marafie, A., & Vafai, K., Analysis of non-Darcian effects on temperature
differentials in porous media. Int. J. Heat Mass Transfer, 44, pp. 4401–4411,
2001.
[42] Vafai, K. & Kim, S.J., Fluid mechanics of the interface region between a
porous medium and a fluid layer – an exact solution. Int. J. Heat Fluid Flow,
11, pp. 254–256, 1990.
[43] Poulikakos, D. & Renken, K., Forced convection in a channel filled with
porous medium, including the effects of flow inertial, variable porosity, and
Brinkman friction. ASME J. Heat Transfer, 109, pp. 880–888, 1987.
[44] Yuan, J., Rokni, M. & Sundén, B., Simulation of fully developed laminar
heat and mass transfer in fuel cell ducts with different cross sections. Int. J.
Heat Mass Transfer, 44, pp. 4047–4058, 2001.
[45] Yuan, J., Rokni, M. & Sundén, B., A numerical investigation of gas flow and
heat transfer in proton exchange membrane fuel cells. Num. Heat Transfer,
44, pp. 255–280, 2003.
[46] Yuan, J., Rokni, M. & Sundén, B., Three-dimensional computational analysis
of gas and heat transport phenomena in ducts relevant for anode-supported
solid oxide fuel cells. Int. J. Heat Mass Transfer, 46, pp. 809–821, 2003.
[47] Yuan, J. & Sundén, B., A numerical investigation of heat transfer and gas
flow in proton exchange membrane fuel cell ducts by a generalized extended
darcy model. Int. J. Green Energy, 1, pp. 47–63, 2004.
[48] Yuan, J., Rokni, M. & Sundén, B., Gas flow and heat transfer analysis for
an anode duct in reduced temperature SOFCs. Fuel Cell Science, Engi-
neering and Technology, eds R.K. Shah & S.G. Kandlikar, pp. 209–216,
FUELCELL2003-1721, ASME, 2003.
[49] Yuan, J. & Sundén, B., Two-phase flow analysis in a cathode duct of PEFCs.
Electrochemica Acta, 50, pp. 677–683, 2004.
[50] Yuan, J. & Sundén, B., Numerical simulation of two-phase flow and heat
transfer in a composite duct. Proc. of ASME International Mech. Eng. Cong.
and R&D Expo, IMECE 2003-42228, Washington, DC, CD-ROM, 2003.
[51] Beale, S.B., Some aspects of mass transfer within the passages of fuel
cells. Fuel Cell Science, Engineering and Technology, eds R.K. Shah &
S.G. Kandlikar, pp. 293–300, FUELCELL2003-1721, ASME, 2003.
[52] Wang L., Husar, A., Zhou, T. & Liu, H., A parametric study of PEM fuel cell
performance. Int. J. Hydrogen Energy, 28, pp. 1263–1272, 2003.
Nomenclature
A area, m2
a width of porous layer, m
b width of flow duct, m
B microscopic inertial coefficient, 1/m
BD Brinkman-extended Darcy model
BFD Brinkman-Forchheimer-Darcy model
d sphere diameter of porous layer, m
cp specific heat, J/(kg K)
D diffusion coefficient, m2 /s
Dh hydraulic diameter, m
Dhr diameter ratio
fapp apparent friction factor
F Faraday constant (96487 C/mol) or the Forchheimer coefficient
H enthalpy, J/kg
hb heat transfer coefficient, W/(m2 K)
hd height of the duct, m
hp thickness of porous layer, m

hr thickness ratio (hp /hd )
hwl water latent heat, J/kg
I current density, A/m2
J mass flux of species, kg/(m2 s) or kg/(m3 s)
k thermal conductivity, W/(m K)
kr thermal conductivity ratio (keff/kf )
M molecular weight, kg/kmol
MEA membrane electrolyte assembly
Nub spanwise average Nusselt number
O2,trans transfer coefficient for the oxygen reaction
P pressure, Pa
P∗ wetted perimeter, m
q heat flux, W/(m2 )
q heat flow rate per unit length of duct, (W/m)
Re Reynolds number (UDh /ν)
Rem wall Reynolds number (Vm Dh /ν)
Rep permeation Reynolds number (Vp Dh /ν)
S source term
s liquid water saturation
Sdi source term in momentum equations
T temperature, ◦ C
TBC-I thermal boundary condition I
TBC-II thermal boundary condition II
Ui velocity components in x, y and z directions, respectively, m/s
V volume of control volume at active site, m3
v velocity vector, m/s
Vcell cell voltage, V
Vm mass transfer velocity at bottom wall, m/s
Vp permeation velocity across interface, m/s
x, y, z Cartesian coordinates
x∗ hydrodynamic dimensionless axial distance (x/(Dh Re))
x∗∗ thermal dimensionless axial distance (x∗ /Pr)
Greek symbols
α net water transport coefficient

βi permeability of diffusion layer, m2
ε porosity
φ mass fraction
arbitrary variable
η relative humidity, %
µ dynamic viscosity, kg/(m s)
ν kinematic viscosity, m2 /s
ρ density, kg/m3
Subscripts
a anode or air
active at active site
av average
b bottom wall
bulk bulk fluid condition
c cathode
e electrolyte
eff effective parameter
f fluid or fuel
H2 hydrogen
H2 O water vapor
in inlet
m mass transfer
O2 oxygen
out outlet
p porous layer or permeation
s solid layer
w wall
sat saturation
wl water liquid
wp water phase change
wv water vapor
CHAPTER 5
Two-phase transport in porous gas

diffusion electrodes
S. Litster & N. Djilali
Institute for Integrated Energy Systems and
Department of Mechanical Engineering, University of Victoria, Canada.
Abstract
The accumulation of liquid water in electrodes can severely hinder the performance
of PEMFCs. The accumulated water reduces the ability of reactant gas to reach the
reaction zone. Current understanding of the phenomena involved is limited by the
inaccessibility of PEMFC electrodes to in situ experimental measurements, and
numerical models continue to gain acceptance as an essential tool to overcome this
limitation.
This chapter provides a review of the transport phenomena in the electrodes of
PEM fuel cells and of the physical characteristics of such electrodes. The review
draws from the polymer electrolyte membrane fuel cell literature as well as relevant
literature in a variety of fields. The focus is placed on two-phase flow regimes in
porous media, with a discussion of the driving forces and the various flow regimes.
Mathematical models ranging in complexity from multi-fluid, to mixture formula-
tion, to porosity correction are summarized. The key parameters of each model are
identified and, where possible, quantified, and an assessment of the capabilities,
applicability to fuel cell simulations and limitations is provided for each approach.
The needs for experimental characterization of porous electrode materials employed
in PEMFCs are also highlighted.
1 Introduction
1.1 PEM fuel cells
A polymer electrolyte membrane fuel cell (PEMFC) is an electrochemical cell that

is fed hydrogen, which is oxidized at the anode, and oxygen that is reduced at
the cathode. The protons released during the oxidation of hydrogen are conducted
Figure 1: Schematic of a proton exchange membrane fuel cell.
through the proton exchange membrane (PEM) to the cathode. Since the mem-
brane is not electronically conductive, the electrons released from the hydrogen
gas travel along the electrical detour provided and electrical current is generated.
These reactions and pathways are shown schematically in Fig. 1.
At the heart of the PEMFC, is the membrane electrode assembly (MEA). The
MEA is typically sandwiched by two flow field plates (referred herein as the current
collectors) that are often mirrored to make a bipolar plate when cells are stacked
in series for greater voltages. The MEA consists of a proton exchange membrane,
catalyst layers, and gas diffusion layers (GDL). As shown in Fig. 1, the electrode
is considered herein to comprise the components spanning from the surface of the
membrane to the gas channel and current collector. A more detailed schematic of an
electrode (the cathode) is illustrated in Fig. 2. The electrode provides the framework
for the following transport processes:
1. The transport of the reactants and products to and from the catalyst layer,
respectively.
2. The conduction of protons between the membrane and catalyst layer.
3. The conduction of electrons between the current collectors and the catalyst
layer via the gas diffusion layer.
An effective electrode is one that correctly balances each of the transport pro-
cesses. The transport process of primary interest is the efficient delivery of oxygen
to the catalyst layer and adequate expulsion of water from the electrode. The word
Figure 2: Transport of gases, protons, and electrons in a PEM fuel cell cathode.
adequate is used because if water is removed too rapidly from the electrode, fuel
cell performance diminishes due to electrolyte dry out. To efficiently conduct pro-
tons, the membrane and catalyst layers must maintain high levels of humidification.
However, if liquid water is allowed to accumulate in the electrodes, the reactant
delivery is reduced and performance is lost.
An effective electrode is one that correctly balances each of the transport pro-
cesses. The transport process of primary interest is the efficient delivery of oxygen
to the catalyst layer and adequate expulsion of water from the electrode. The word
adequate is used because if water is removed too rapidly from the electrode, fuel
cell performance diminishes due to electrolyte dry out. To efficiently conduct pro-
tons, the membrane and catalyst layers must maintain high levels of humidification.
However, if liquid water is allowed to accumulate in the electrodes, the reactant
delivery is reduced and performance is lost.
1.2 Porous media
Porous media is, by definition, a multiphase system. The solid portion of porous
media is one of the phases. In general, the solid phase of the porous media either
is dispersed within a fluid medium or has a fluid network within a continuous solid
phase. In the case of a continuous solid phase, the fluid medium occupies pores and
the characteristic length is the diameter of the pores. Whereas, if the solid phase
is dispersed, as with a bed of sand, the particle size is the characteristic length.
Many forms of porous media readily deform with the application of internal and
external forces. However, these effects are commonly disregarded because of their
complexity.
In addition to pore diameter, or particle size, there are two more characteristics
of flow paths in porous media: porosity and tortuosity. The porosity is the fraction
of the bulk volume that is accessible by an external fluid. The determination of the
porosity either can neglect inaccessible inclusions in the solid or corrected for the
inclusions. The tortuosity is the characterizing parameter that arises when fluid in
a porous medium cannot travel in a straight path. Instead, the fluid follows through
a tortuous path, which is longer than the point-to-point distance. The indirect fluid
path reduces diffusive transport because of the increased path length and reduced
concentration gradient.
Typically, the solid phase is considered inert (with the exception of heat trans-
fer). As well, deformation is typically neglected. This simplifies the modeling by
neglecting momentum or mass transfer within the solid phase. This simplification
explains why the flow of a single-phase in porous media is generally considered a
single-phase system. Multiphase flow in porous media typically refers to a porous
solid with more than one phase occupying the open volume.
1.3 Porous media in PEMFC electrodes
PEMFC electrodes feature two regions of porous media; the gas diffusion layer
and the comparatively more dense catalyst layer. The gas diffusion layer is much
thicker and open than the catalyst layer. The catalyst layer features significantly
lower void space because of the impregnation of a proton conducting ionomer
(typically Nafion). These contrasts can be seen in Fig. 3, which shows a cross-
section of an entire MEA.
Presently, most catalyst layers are fabricated by applying ink containing Nafion
and carbon-supported catalyst to either the membrane or the gas diffusion layer.
Because the catalyst layer is so thin (∼15 µm) and applied as a Nafion solution,
it can be considered spatially homogeneous. However, the properties of catalyst
can be expected to change with further penetration into the GDL. The magnitude
of this variation depends on whether the catalyst layer is applied to the GDL or
Figure 3: Scanning electron micrography (SEM) image depicting a carbon cloth gas
diffusion layer that features a woven fiber structure. Reprinted from [1],
Copyright (2004), with permission from Elsevier.
the membrane. Catalyst layers typically feature a porosity (or void fraction) of
approximately 5–15%, and pore diameters of roughly 1 µm.
Two materials are typically employed as gas diffusion layers in PEM fuel
cells; carbon cloth and carbon paper. Both materials are fabricated from carbon
fibers. Carbon cloth, visible in Fig. 3, is constructed of woven tows of carbon fibers.
Alternatively, carbon paper (see Fig. 4) is formed from randomly laced carbon
fibers.
Both carbon cloth and carbon paper have approximate pore diameters of 10 µm
and porosities ranging between 40–90%. However, carbon cloth is generally avail-
able in thicknesses between 350–500 µm, whereas carbon paper is available in
thicknesses as low as 90 µm. In addition, the two gas diffusion layer structures
vary by spatial uniformity and degree of anisotropy. Carbon cloth, because of its
woven structure, is spatially heterogeneous on a macroscopic scale, while carbon
paper is spatially homogeneous because of its random lacing. Moreover, the woven
nature of carbon cloth results in three degrees of macroscopic anisotropy. This is
in contrast to the two degrees in carbon paper. All three forms of porous media in
PEM electrodes are summarized in Table 1.
Generally, gas diffusion layers are treated with a PTFE (Teflon) solution to
increase the hydrophobicity of the medium. This is done to aid water manage-
ment in the electrode. The hydrophobicity causes water droplets to agglomerate at
the free surface of the gas diffusion layer. However, the Nafion in catalyst layers is
hydrophilic and will absorb and retain liquid water. Thus, the liquid water produced
travels from a saturated catalyst layer to the free surface of the gas diffusion layer.
It has been theorized that, within the gas diffusion layer, the condensation will
only take place in cracks in the carbon fibers, which are hydrophilic [2]. Thus, it
is likely that the PTFE treatment of gas diffusion layers lessens the condensation
rate. Condensation typically occurs since the fuel and oxidant gases are generally
saturated with water vapour and thus the product water forms as liquid or forms as
Figure 4: Microscope image depicting the random fiber structure of a GDL formed
of Toray carbon paper.
Table 1: Summary of porous media in PEM electrodes.
Porous Spatial Dimensions Pore Thickness

media uniformity of anisotropy Porosity diameter [µm] [µm]
Catalyst Homogeneous Isotropic 5–15% ∼1 ∼15

Layer
Carbon Heterogeneous 3-D 40–60% ∼10 350–500
Cloth
Carbon Homogeneous 2-D 40–90% ∼50 100–300
Paper
Figure 5: Environmental scanning electron micrography (ESEM) image depicting

condensatin on Toray carbon paper. (a) At time = t. (b) At time = t + t.
Reprinted from [2], Copyright (2003), with permission from Elsevier.
vapour that rapidly condenses. Moreover, the depletion of the hydrogen and oxygen
results in the condensation of excess water in the gas streams.
All the traits of liquid in a hydrophobic medium are visible in Fig. 5, which
shows two images of condensation in PTFE-treated carbon paper. First, the liquid
water has formed as droplets instead of a film. As well, it can be seen that over
time, with greater levels of liquid water present, the droplets have connected and
travelled toward areas of greater liquid accumulation. In addition, the droplets are
disperse indicating condensation occurs in randomly oriented cracks in the surface
of the carbon fibers.
The objective of this review is to present methods of modeling multiphase
and multicomponent transport within a porous medium that are applicable to gas
diffusion electrodes. Air is a multicomponent mixture that consists of nitrogen, oxy-
gen, and water vapour. To ensure these modeling approaches are as transparent as
possible, the review will start with single-phase, single-component transport in
porous media. The description will then extend to include multicomponent flow
and heat transfer. Subsequently, the review will proceed to illuminate the modeling
of multiphase systems.
2 Single-phase transport
2.1 Transport of a single-phase with a single component
In the case of a single-phase with a single component in a porous medium under

isothermal conditions, there are two equations required to describe the bulk hydro-
dynamic behavior. These equations are the conservation of mass and the conser-
vation of momentum. The conservation of mass in porous media is expressed as:
∂(ερ)
+ ∇ · (ρu) = 0. (1)
∂t
In this case, u is the superficial velocity that takes into account the facial poros-
ity and is related to the average interstitial velocity (the average velocity in the
pores) by:
u = ui ε. (2)
In an isothermal system, the transport of a single fluid/species in porous media
is driven by the pressure gradient∇P. It is universally accepted when modeling
porous media as a continuum to use the generalized Darcy’s equation form of the
momentum conservation equation:
k
u = − ∇P, (3)
µ
where k is the permeability and k/µ is the viscous resistance.
2.1.1 Permeability
The permeability of a porous medium represents its ability to conduct fluid flow
through its open volume. Permeability can be obtained by applying the conser-
vation of mass and momentum to a pore-scale model. Since most porous media
feature complex geometry and are anisotropic, solutions of the permeability have
only been obtained for idealized conditions. The fibrous media found in the gas
diffusion layer of PEMFC electrodes is very complex and three-dimensional. Two
forms of pore-scale modeling are capillary models and drag models. In capillary
models, the Navier-Stokes equation is applied for ducts in serial, parallel, and
networks. Drag models for approximating permeability are an application of the
Navier-Stokes equation to flow over objects. CFD is now being employed at a pore
level to determine permeability [3, 4]. However, if pore sizes are small enough that
molecular effects require consideration (the flow inside the pore can no longer be
considered a continuum), the CFD approach is inaccurate. Instead, computationally
intensive methods, such as lattice-Boltzmann models, can be employed.
Another approach for evaluating permeability is the semi-heuristic hydraulic
radius method, commonly referred to as the Carman-Kozeny theory. The relation-
ship for the permeability is obtained by applying the equation for Hagen-Poiseuille
flow to a pore with an approximated hydraulic diameter. The derivation results in

the following equation:
2 ε3
dpor
k= , (4)
36kk (1 − ε)2
where dpor is the characteristic pore diameter and kk is Kozeny constant. The Kozeny
constant is evaluated from a shape factor (2 for circular capillaries) multiplied by
the tortuosity factor (roughly 2.5 for a packed bed). Thus, the permeability of a
packed bed is often obtained by using a Kozeny constant of 5. An alternative to
analytical/semi-analytical methods is to rely on empirical data to determine for
the permeability value. An outline for determining permeability experimentally is
given by Biloé and Mauran [5].
2.2 Transport of a single-phase with two components
The presentation of single-phase multicomponent transport will begin with the

equation for the conservation of a single species in a binary mixture. If there
are more than two species that can be considered dilute, their diffusion can be
approximated as binary diffusion with the species of the greatest concentration
(the background species). With the addition of new species and variation in con-
centration, the viscosity and density of the mixture change. These changes can be
accounted for in a variety of manners. The simplest method is volume averaging.
However, Bird et al. [6] offers more theoretical approximations for the mixture vis-
cosity. These values should then be used in the momentum conservation equation
(see eqn (3)). With the exception of the dependence of the viscosity and density on
the concentrations of the species, the hydrodynamic equations are unchanged.
Herein, mass fluxes (nA ), mass concentration (ρA ), and mass fractions (yA ) will
be used to describe velocities and distributions in multicomponent systems. These
variables can be transformed to their molar counterparts by each species’ molar
mass. The mass fraction of A is related to the concentration by yA = ρA /ρ, and the
mass flux of A is related to the velocity (uA ) of A by nA = ρA uA .
The species conservation equation in porous media for binary or dilute multi-
component mixtures is presented in vector form as:
∂(ερA )
+ (∇ · nA ) = SA , (5)
∂t
where SA is the volumetric source/sink of A. The mass flux of A(nA ) is evaluated

as the mass-average flux of A(ρA u) plus the relative mass flux of A( jA ). It should
be remembered that u is the superficial velocity expressed in eqn (2). The relative
mass flux of A( jA ) is the Fickian diffusion term and is a function of the effective
diffusivity of A in the second species B(DABeff ) and the gradient in the concentration
of A(∇ρA ).
jA = −DAB
eff
∇ρA . (6)
Figure 6: Schematic of driving forces of momentum, energy, and mass transfer.

The transport coefficient associated with each phenomenon is listed.
Reproduced from [6].
Subsequently, the mass flux of A (nA ) can be evaluated with the expression:
nA = ρA u − DAB
eff
∇ρA . (7)
By substituting the mass flux of A in eqns (7) and (5), the mass transport equation
for species A is revealed.
∂(ερA )
+ ∇ · (ρA u) = ∇ · DAB
eff
∇ρA + SA . (8)
∂t
However, this is a simplification if the system under consideration is not iso-
thermal. In addition to a mass flux, the gradient in the concentration of A also
drives a flux of energy. This is the Dufour effect. Mass flux of A can also be
attributed to a temperature gradient. This is the Soret effect. The present review
will not investigate these two additional effects, as they are generally negligible in
fuel cell applications. A matrix of fluxes and their driving forces are shown in Fig. 6.
2.2.1 Effective diffusivity in porous media

In order to account for the geometric constraints of porous media, the open space
diffusivity is often corrected with geometric factors. The Bruggemann correction
used by Berning and Djilali [7], and many others modeling gas diffusion layers in
PEMFCs, modifies the diffusivity for porous regions with a function of the porosity:
Deff = ε1.5 D. (9)

However, in many pieces of literature and fundamental studies [5, 8, 9], a function
of both the porosity and the tortuosity factor is adopted. The tortuosity factor is the
square of the tortuosity. The tortuosity is the actual path length over the point-
to-point path length as shown in Fig. 7. The tortuosity factor often varies between
Figure 7: Schematic of tortuosity.
2 and 6, and values as high as 10 have been reported [10]. The effective diffusivity
is obtained with the following relationship:
ε
Deff = D,
τ
2 (10)
Actual path length
τ= .
Point to point path length
The ε/τ term is sometimes referred to as the formation factor. The porosity ε
is a result of the area available for mass transport. The square of the tortuosity is
present because of the extended path length and reduced concentration gradient
that are both represented by the tortuosity. The derivation of this result is presented
by Epstein [9].
The Bruggemann correction and the second expression are ±5% equivalent in
the region of 0.4 < ε < 0.5 with tortuosity equal to a low value of 1.5. Under all
other ranges there is a significant difference between the correlations. It should
be noted that the Bruggemann correction, which is widely employed and quoted,
was obtained from a study on the electrical conductivity of dispersions [10]. The
exponent of 1.5 was an empirically determined factor found for a specific case in
the De La Rue and Tobias paper [10]. Electrical conductivity measurements are
presently one of the only methods of determining the tortuosity. Measurements
of the electrical conductivity are taken when a non-conductive porous media is
saturated with a conductive fluid. However, this method cannot be applied to gas
diffusion layers, which are electrically conductive. Nevertheless, the formation
factor is determined by the ratio of the effective conductivity ke for a porous medium
saturated with a fluid of known conductivity kf .
ε ke
= . (11)
τ kf
In order to use the Bruggemann correction, it is more appropriate to replace the

exponent of 1.5 with the Bruggemann factor α. The Bruggemann factor can then
be presented as a function of the porosity and the tortuosity factor.

ε
εα = ,
τ
(12)
log10 (ε/τ)
α= .
log10 (ε)
In a typical case where the porosity is equal to 0.5 and the tortuosity factor is 3, the
value of the Bruggemann factor α is 2.585. This indicates that commonly employed
exponent of 1.5 may not be applicable to gas diffusion layers. Finally, a good rule
of thumb for the effective diffusivity in typical porous media is a reduction of an
order of magnitude.
2.2.2 Determination of the binary diffusion coefficient

Binary diffusivities DAB are typically calculated based on an empirically developed
formulation presented by Cussler [8]. This is an effective method for determining
the diffusivity in a numerical model, and agrees well with published empirical data.
The empirical method was implemented by Berning and Djilali [7] and is expressed
in Cussler [8] as: √
T 1.75 1/MA + 1/MB
DAB = , (13)
P 1/3 1/3
φ +φ A B
where φ is the diffusion volume and M is the molar mass. Other expressions, such as
the Chapman-Enskog theory [8], require the use of tabulated temperature dependant
values that complicate the procedure for determining the binary diffusivity. With
the knowledge of the diffusivity at given pressure Po and temperature To , the above
expression can be further simplified to:

Po T 1.75
DAB = DAB (To , Po ) . (14)
P To
2.2.3 Transport of a single-phase with more than two components

If considering a system with n components (n > 2) that are not dilute, the evaluation
of the diffusion becomes much more complex. With n components, the diffusive
flux of each species depends on the concentration gradient of the other n−1 species.
This dependence is evident in the Maxwell-Stefan equations for multicomponent
diffusion, which are expressed as:

n
xj Ni − xi Nj
∇ci = . (15)
j=1, j =i
Dijeff
In the above equation, integer subscripts i and j have replaced the letter subscripts
A and B as we are no longer considering a binary system. Above, ci is the molar
concentration, xi, j is the mole fraction, and Ni, j is the molar flux. However, this is a
difficult expression to include in a finite volume CFD code. Berning [11] suggests
the use of an equivalent approach that is termed the generalized Fick’s law.
2.3 Knudsen diffusion
With diffusion in porous media, it is acknowledged that the diffusion mechanism

varies with the length scale of the porous media. The Knudsen number is the non-
dimensional parameter commonly employed to characterize the flow and diffusion
regimes in micro-channels. The Knudsen number is the ratio of mean free path
to pore diameter. When the mean free path is large in comparison to the pore
diameter, the probability of molecule-molecule interaction is small and molecule-
wall collisions dominate. The expression for the Knudsen number is [8]:
λ
Kn =
dpor
kB T
= √ , (16)
dpor 2πσii2 P
where λ is the mean free path, dpor is the pore diameter, kB is the Boltzmann
constant, σii is the collision diameter, and P is the pressure. Flow in porous media
can be categorized into three regimes by the Knudsen number [12]:
1. Continuum Regime, Kn < 0.01;
2. Knudsen Regime, Kn > 1;
3. Knudsen Transition Regime, 0.01 < Kn < 1.
In most of the literature [12, 13], the Knudsen regime is defined by Kn > 1. How-
ever, Karniadakis [14] states the transition region corresponds to 0.1 < Kn < 10,
and the Knudsen regime to Kn > 10. In addition, Karniadakis notes that at Kn > 1
the concept of macroscopic property distribution breaks down. It is also evident in
various plots in Chapter 5 of in Karniadakis [14] that there is a significant differ-
ence in the flow behavior for 0.1 < Kn < 1, and much less variation in the flow
characteristics in the region 1 < Kn < 10. Thus, if the flow is in the upper region
of the transition regime, the flow is still dominated by Knudsen diffusion according
to Karniadakis. It is therefore reasonable to assume that a strictly Knudsen regime
is present when Kn > 1.
In the Knudsen Regime, molecule-wall collisions dominate over molecule-
molecule collisions. Similar to molecular diffusion, flux in the Knudsen regime
is influenced by the gradient of the concentration of a species. The gradient of the
partial pressure is the driving force. The partial pressure gradient is equal to that
of the concentration for constant pressure conditions. However, a new diffusivity
is defined (Knudsen Diffusivity DKn ). Knudsen diffusivity is corrected in the same
manner as the molecular diffusivity in porous media. Since viscous and ordinary
diffusion is negligible in the Knudsen Regime [12, 13], eqn (7) reduces to:
nA = jA
= −DKn
eff
∇ρA (17)
and the mass transport equation (see eqn (8)) is:
∂ερA
= ∇ · DKn
eff
∇ρA + SA . (18)
∂t
2.4 Determination of Knudsen diffusivity
The Knudsen diffusion coefficient is determined from the kinetic theory of gases
and is expressed as [5, 8, 15, 16]:

1 8RT
DKn, A = dp . (19)
3 πMA
It can be inferred from the previous equation that the Knudsen diffusivity is
independent of the other species present in a system. This is because of negligible
collisions between molecules. One species cannot “learn” about the presence of
other species [13]. It follows that no additional considerations are necessary for
systems with more than two components. Quoting from Cunningham [13], “In the
Knudsen regime, there are as many individual fluxes present as there are species (as
in molecular diffusion), and these fluxes are independent of each other (in contrast
to molecular diffusion).”
2.5 Knudsen transition regime
The transition regime is present when 0.01 < Kn < 1. In this regime, both molec-
ular diffusion and Knudsen diffusion (slip flow) are present. A common way to
evaluate this regime is the Dusty Gas Model (DGM) [12, 13, 15, 17, 18]. The DGM
is derived by considering the solid matrix as large stationary spheres suspended
in the gas mixture as one of the species present. The formulation is rigorously
explained by Cunningham [13] and employed for modeling solid oxide fuel cells
by Suwanwarangkul [18]. Often, the effective DGM diffusivity (DDG ) is approxi-
mated in the case of equal molar masses [13, 19]. In these cases, the effective DGM
diffusivity is calculated by:
DKn, A DAB
DDG = . (20)
DKn, A + DAB
This diffusivity can subsequently be corrected for porous media with eqn (10)
and then replace the binary diffusion coefficient in the mass transport equation
(eqn (8)).
2.5.1 Comparison of the diffusivities

The three diffusivities that have been presented (binary molecular diffusivity DAB ,
Knudsen diffusivity DKn, A , and an effective diffusivity retrieved from the dusty-gas
model with equimolar diffusion DDG ) are now compared to a range of Knudsen
numbers. This is done to depict the applicability of each diffusivity to the three
Figure 8: Comparison of the binary molecular diffusivity DAB , Knudsen diffusivity

DKn, A , and an effective diffusivity retrieved from the dusty-gas model
with equimolar diffusion DDG over a range of Knudsen numbers. Oxygen
in nitrogen for a temperature of 350 K and a pressure of 1 atm.
diffusion regimes. The Knudsen number has been varied from 0.01 (continuum
regime) to 10 (Knudsen regime). The diffusivity presented is that of oxygen in
nitrogen at a temperature of 350 K and a pressure of one atmosphere. Figure 8
depicts the three diffusivities.
Figure 8 illustrates the significant difference in approximations of the diffusivities
depending on the regime. Images of Toray carbon paper show voids with widths
between roughly 1 and 100 µm, corresponding to Knudsen numbers between 0.1
and 0.001. A Knudsen number of 0.1 in Fig. 8 is on the boundary of the region where
Knudsen diffusion is shown to dominate. This indicates that Knudsen diffusion
could be of concern, but is not significant for the carbon paper shown in Fig. 2.
Nevertheless, the smaller pore diameters (∼1 µm) in the catalyst layer require the
consideration of Knudsen diffusion.
3 Two-phase systems
A large variety of applications exist for models encompassing multiphase flow,
heat transfer, and multicomponent mass transfer in porous media. These include
thermally enhanced oil recovery, subsurface contamination and remedy, capillary-
assisted thermal technologies, drying processes, thermal insulation materials, trickle
Figure 9: Schematic of the volume fractions.
bed reactors, nuclear reactor safety analysis, high-level radioactive waste reposito-
ries, and geothermal energy exploitation [20]. As well, this combination of transport
phenomena is present when modeling the flooding of PEM fuel cell electrodes.
The phase distribution is potentially the result of viscous, capillary, and gravita-
tional forces. The additional phase can be formed by phase change or is introduced
externally into the system. Each phase can also be a multicomponent mixture and
the components of each phase can in some cases be transported across phase bound-
aries. When modeling the diffusion layer of a PEMFC it is generally accepted that
the second phase, liquid water, is comprised of single component and there is only
transfer of water across the phase boundary.
For porous media in which the void space is occupied by two-phases, the bulk
porosity ε is divided between the liquid εl and gas εg volume fractions. The liquid
saturation sl is the volume occupied by the liquid εl divided by the open pore
volume ε. This relationship is depicted in Fig. 9 and eqn (21).
εs + εl + εg = 1,
εl + εg = ε, (21)
εl
Sl = .
ε
3.1 Two-phase regimes
Two-phase flow exists in three possible regimes; pendular, funicular, and saturated.
The regime present at any time and location depends on the saturation. To some
degree it also depends on the wettability. The aforementioned regimes are illustrated
in Fig. 10. The pendular regime is predominant for low saturations where the liquid
phase is discontinuous. The term “pendular” stems from the pendular rings that form
around sand grains in this regime. The funicular regime occurs when the liquid is
continuous and travels through the pores in a funicular (corkscrew) manner. When
the saturation approaches unity, the liquid saturated regime emerges and the pores
are fully occupied by the liquid.
Figure 10: Schematics of two-phase regimes in porous media. (a) Pendular,

(b) funicular, (c) saturated.
The saturation level at the transition between the funicular and pendular regimes
corresponds roughly to what is termed the immobile saturation (also referred to
as the irreducible saturation). This saturation level is found when no more water
can be removed from a two-phase test sample in a permeation test, often featuring
a centrifuge. The final weight of the sample is compared to the dry weight and
the immobile saturation sim is determined. This immobility is the result of surface
tension. From herein, the saturation s is the reduced saturation, which is the actual
liquid saturation sl scaled as follows [21]:
sl − sim
s= . (22)
1 − sim
The immobile saturation can be expected to be quite high for PEM fuel cell
electrodes. This stems from the results presented by App and Mohanty [22], who
showed the dependence of sim on the capillary number Ca:
uµ
Ca = , (23)
σ
where u and µ are the velocity and viscosity of the invading phase and σ is the
interfacial tension. The capillary number is the ratio of viscous forces to interfacial
tension forces. High capillary numbers arise from a viscous force much greater than
the surface tension forces. This could be the result of a high-velocity. One condition
in which a small capillary number applies is the case of negligible velocity in either
phase.
The capillary number is small in PEMFC electrodes because mass transport is
dominated by diffusion and the velocity term u is quite small. App and Mohanty
stated that in porous cores the immobile saturation for their low capillary number
cases was sim = 0.18, whereas for large capillary numbers the immobile saturation
approached zero. In addition, Kaviany [23] stated that the immobile saturation
increases as the pore size is reduced.
The dependence of the immobile saturation on the capillary number can be
explained by the deformation of droplets under the viscous stress of a high velocity
invading phase. The viscous forces elongate the droplets that eventually bridge
and form a continuous phase. At this moment the phase regime of the displaced
phase transforms from a pendular regime to a funicular regime and capillary flow
Figure 11: Schematic of two processes responsible for surpassing the immobile
saturation: Increasing the capillary number through higher velocities
and increasing the saturation of the liquid.
is initiated. Figure 11 illustrates how an increased capillary number, due to greater

velocities, allows the saturation to surpass the immobile saturation. In addition,
the same figure depicts how increasing liquid saturation causes the transformation
from a pendular regime to a funicular regime (where the liquid is capable of motion)
when the saturation surpasses the immobile saturation.
3.2 Hydrodynamics and capillarity in two-phase systems
The relationships developed for the hydrodynamics of a single-phase in porous

media will now be applied to each phase in the two-phase system (gas and liquid).
The conservation equations for the gas and liquid phases are:
∂(1 − sl )ερg
+ ∇ · ρg ug = Ṡg ,
∂t
(24)
∂sl ερl
+ ∇ · (ρl ul ) = Ṡl ,
∂t
where (1 − sl )ε and sl ε in the first terms in the equations represents the volume of
each phase. Respectively, Ṡg and Ṡl are the volumetric sources of gas and liquid.
These sources, or sinks, can arise from phase change, in which case Ṡg = −Ṡl , or
from an external source.
The single-phase momentum equation (eqn (3)) is adapted to the two-phase
system in a similar fashion. The only notable difference between the single and
two-phase cases is that the permeability is phase specific for the gas (kg ) and liquid
(kl ). These permeabilities are a correction of the bulk permeability (k) for the effect
of the reduced area open to each phase due to the presence of the other phase.
kg
ug = − ∇Pg ,
µg
(25)
kl
ul = − ∇Pl .
µl
Figure 12: Hydrostatic representation of capillary pressure when the liquid is the
non-wetting phase.
It is evident that the system is not fully defined with the general conservation of
mass and momentum equations. The four equations above cannot evaluate the five
variables (sl , ug , ul , Pg , and Pl ) that must be solved. This is because a prominent
phenomenon in multiphase flow in porous media has not been introduced into the
equation set. Expressions for the capillary pressure are employed as the constitutive
relationship that completes the system of equations.
Capillarity and capillary pressure are the result of interfacial tension, which is
the surface free energy between two immiscible phases. The microscopic capillary
pressure is directly proportional to the interfacial tension and inversely proportional
to the radius curvature of the interface. Thus, the lesser the radius of curvature, the
more dominant the effects of capillary pressure. This is the microscopic definition
of the capillary pressure, which is typically formulated as:
σ
Pc ∝ , (26)
r
where r is the characteristic radius of the liquid/gas interface. The macroscopic
definition of the capillary pressure Pc , the pressure difference between the wetting
gas and non-wetting liquid pressures, is included in the two-phase momentum
equations (eqn (25)). This is shown in hydrostatic form in Fig. 12.
Figure 13 is an attempt to use the microscopic and macroscopic definitions of
the capillary pressure to explain capillary motion in a pore. At the end of the pore
where the liquid radius is smaller (lower local saturation), the capillary pressure
is greater than at the end with the larger liquid radius (greater local saturation).
Because the liquid pressure is the sum of the capillary pressure and the gas pressure,
the hydrodynamic pressure of the liquid is greater at the end of the pore with the
smaller radius. Therefore, the bulk motion of the liquid is toward the end with the
greater radius (and local saturation).
Pc = Pl − Pg . (27)
At this point in the discussion, the transport of the liquid water in the elec-
trodes of PEMFCs should be revisited. The transport of liquid water from low
to high saturation, as shown in Fig. 13, is counter-intuitive and could lead to
Figure 13: Schematic of capillary diffusion, in which the liquid is the non-wetting
phase.
incorrect conclusions. In a broad sense, the transport depicted in Fig. 13 illustrates

the penchant for water, in hydrophobic media, to move to ever-increasing pore
diameters according to the capillary pressure’s inverse proportionality to the liquid
radius. Ultimately, the largest radius can be attained when the liquid reaches the gas
channel.
Conversely, the water invades smaller pores in the catalyst layer due to the
hydrophilic nature of the electrolyte. However, the electrolyte phase in the catalyst
layer offers a second mode of water transport. Water can be transported through the
catalyst layer’s electrolyte in a similar fashion to that of the electrolyte membrane.
Though, a review of these transport issues is beyond the scope of the present chapter
and shall be reserved for a separate discussion of transport phenomena in polymer
electrolyte membranes.
Subsequent to the definition of the macroscopic capillary pressure, the momen-
tum equations take the form:
kg
ug = − ∇Pg ,
µg
(28)
kl kl
ul = − ∇Pg − ∇Pc .
µl µl
When interfacial tension is observed as driving mass transport due to a gradient
in the capillary pressure, the phenomena is referred to as capillary diffusion. The
mass flux of the liquid phase due to capillary diffusion can be obtained from the
second term of the liquid momentum equation (eqn 28):
ρkl
ṁl, s = − ∇Pc . (29)
µl
If the gradient of the capillary pressure is assumed to rely only on the saturation
gradient, the liquid transport due to capillary diffusion emerges as:

ρkl dPc
ṁl, s = − ∇sl (30)
µl dsl
and thus the capillary diffusivity is often defined [20]:

kl dPc
D(sl ) = . (31)
µl dsl
Now the momentum equations for the two-phase system (eqn (28)) can be refor-
mulated to eliminate the liquid pressure field from the equation set. Inserting the
definition of capillary diffusivity into the liquid momentum equation yields a func-
tion of the gas pressure and the liquid saturation.
kl
ul = − ∇Pg − D(sl )∇sl . (32)
µl
It is important to note that the interfacial tension, which capillary diffusion is a
result of, is not constant in a non-isothermal and multicomponent system [20]:
σ = σ(T , cA ). (33)
Therefore, diffusion due to interfacial forces can be driven by temperature and

concentration gradients in addition to saturation. These two transport mechanisms
are termed thermal- and solutal-capillary diffusion:

ρkl ∂Pc ∂σ
ṁl, T = ∇T ,
µl ∂σ ∂T
(34)
ρkl ∂Pc ∂σ
ṁl, cA = ∇cA .
µl ∂σ ∂cA
As with capillary diffusion, the thermal- and solutal-capillary diffusivities can
be derived. However, these terms are often not included because of their negligible
contributions to the total mass flux.
Another characteristic of two-phase flow in porous media, which needs to be
introduced before proceeding, is the surface tension between the liquid phase and
solid matrix. The surface tension is dependant on the wettability, or the hydropho-
bicity, of the liquid/solid interface. Figure 14 depicts the effect of hydrophobicity
on the contact angle θ. Hydrophobic interfaces feature a contact angle greater than
90◦ . Teflon (PTFE) features a contact angle of 108◦ . For hydrophobic solids, the gas
is the wetting phase. Hydrophilic interfaces feature a contact angle less than 90◦ .
In this case, the liquid is the wetting phase. The contact angle θ can be calculated
from the gas-liquid σ, gas-solid σgs , and liquid-solid σls interfacial tensions:
σgs − σls
cos (θ) = . (35)
σ
Figure 14: Contact angle for hydrophobic and hydrophilic fluid/solid interfaces.
Figure 15: General form of the capillary pressure curve.
There have been corrections developed for the contact angle in porous media as
an alternatively to using the contact angle on a flat plate. It has been found that the
effective contact angle porous media is less than the actual [24]. This indicates that
the porous form of a material is less hydrophobic than the bulk form. An important
note on surface tension is that it is known that transport of liquid water increases with
the contact angle (more hydrophobic). This trend could be due to the reduced contact
area between the porous media and the liquid (see Fig. 14), which increases the
influence of the viscous forces exerted by the invading phase. This increased liquid
transport is a reason for the impregnation of electrode diffusion layer with PTFE.
3.2.1 Capillary pressure curves

In order to use the capillary pressure as a constitutive relationship, an expression for
the capillary pressure is derived. It would be too difficult to determine the capillary
pressure microscopically as in Fig. 13. Thus, a volume averaging approximation
is utilized. The starting point for derivation this constitutive relationship is the
form of capillary pressure curves obtained in experiment. In these experiments, the
capillary pressure is measured in a sample and compared to the estimated level of
saturation [24]. The form of curves generated is shown in Fig. 15.
It was postulated by Leverett [25] that the capillary pressure versus saturation
relationship could be presented in the non-dimensional form:
1/2
Pc k
J (s) = . (36)
σ ε
This function is typically referred to as the Leverett J-function. Occasionally, a
cosine of the contact angle θ is included in the Leverett J-function [2]:
1/2
Pc k
J (s) = . (37)
σ cos (θ) ε
However, Anderson [24] stated that this is not valid for modeling the effects of
wettability on capillary pressure.
Udell [21] later used data presented by Leverett [25] to determine the J-function
for the porous media in Leverett’s experiments. Udell then compared experimental
results with a one-dimensional steady-state model for packed sand. The porosity
was varied from 0.33–0.39 and permeabilities from 1.39–10.3×10−12 m2 . Good
agreement between the model and the experiment was presented. The J-function
Udell obtained is often referred to as the Udell function, which is expressed as:
J (s) = 1.417(1 − s) − 2.120(1 − s)2 + 1.263(1 − s)3 . (38)
There are also many other J-functions [23], including Scheidegger’s [26]:
0.005
J (s) = 0.364(1 − e−40(1−sl ) ) + 0.221(1 − sl ) + . (39)
sl − 0.08
The above J-function (eqn (39)) is then employed to calculate the capillary pres-
sure with the formula:
σ cos (θ)
Pc = J (s). (40)
(k/ε)1/2
As well, van Genuchten [27] presented a relation for the capillary pressure, which
is a function of saturation and requires empirical constants:
1m 1/n
Pc = s −1 ,
α
(41)
m = 1 − 1/n.
Kaviany [23] stated that the capillary pressure curves must be bounded by the
relationship:
dPc
lim = −∞. (42)
sl →sim ds
Observing the dPc /ds derivative in eqn (30), it could be inferred that Kaviany’s
boundary (eqn (42)) would predict a very large mass flux at sl ≈ sim . However, this
is also the point at which the liquid is in transition between the pendular and
Figure 16: Capillary pressure curves from various relations and experimental
results. For Udell and Scheidegger’s: k ≈ 10−11 m2 , ε = 0.35, and
σ = 0.0644 N/m (air/water). For Anderson’s plot: air and water in an
interfacial Teflon core. For van Genuchten’s: n = 2 and σ ≈ 7 × 10−21 .
funicular regimes. Thus, the high capillary pressure is due to the immobility of the
discontinuous phase. Therefore, the mass flux is limited.
Figure 16 presents these various capillary pressure relationships as a function
of saturation. Experimental results presented by Anderson [24] are also plotted.
It is evident that the experimental results abide by Kaviany’s boundary, whereas
the capillary pressure expressions do not. Another flaw of these expressions is that
the result is the same whether imbibition (increasing sl ) or drainage (decreasing
sl ) is being considered (when it has been clearly shown that there is a significant
difference in reality [23, 24].
Many other expressions for the capillary pressure exist. However, they are not
presented herein. As with Leverett and Udell’s work, these relations are for packed
sand and other representations of soil. There is a significant lack of material eval-
uating capillary pressure in fibrous porous media.
3.3 Relative permeability
When two or more phases occupy the same pores, the amount of pore space avail-
able for each phase is reduced. Referring to eqn (4) it is acknowledged, at least
theoretically, that the permeability is a function of the porosity. Therefore, the per-
meability must be adjusted for the volume fractions occupied by different phases.
The permeabilities of the gas and the liquid are now expressed as:
kg = krg k,
(43)
kl = krl k,
Figure 17: General form of the relative permeability functions.
where k is the single-phase permeability. Figure 17 depicts the general form of

relative permeability functions used for the gas and liquid phases. In addition, the
immobile saturation for the phases is mathematically accounted for in the relative
permeability. The relative permeabilities are commonly expressed as:
krg = (1 − s)3 , (44)

krl = s ,
3
(45)
where s is the reduced saturation (see eqn (22)). Some explanation for the cubic
form of these equations can be found in the Carman-Kozeny equation (see eqn (4)),
where the single-phase permeability is roughly proportional to ε3/(1 − ε)2 .
Recalling that the more dominant the interfacial tension the higher immobile
saturation, it is evident that interfacial tension and the wettability/hydrophobicity
has an effect on the relative permeability. The effect of wettability on the relative
permeabilities was surveyed by Anderson [28]. Figure 18 presents the relative per-
meability of gases for a large range of contact angles. The case of θ = 108◦ is that
of nitrogen displacing water in a Teflonized core. This is a good approximation of
the PEMFC electrode. Equation (44) is plotted to evaluate its validity. Firstly, it is
evident that the gas permeability presented by Anderson is apparently bimodal for
all degrees of wettability. However, the often prescribed cubic function is mono-
tonic. It is also evident in Fig. 18 that eqn (44) would underestimate the relative
permeability at low saturations and would predict values significantly higher than
the experimental results show in the high saturation regions.
Figure 19 depicts the effect of wettability on the displacing phase (water) per-
meability as presented by Anderson [28]. The plots indicate that the immobile
saturation resides between 0.16 and 0.20. The plot also depicts a uniform effect of
wettability on the relative permeability. It is clear in the plot that a hydrophobic
porous structure aids the transport of water by increasing the permeability of the
structure. Again, the cubic relative permeability function (eqn (45)) is plotted for
Figure 18: Relative permeability of displaced phase (gas) for various degrees of
hydrophobicity. θ = 108◦ corresponds to nitrogen displaced by liquid
water in an artificial Teflon core (Anderson, 1987b).
Figure 19: Relative permeability of displaced phase (liquid) for various degrees of
hydrophobicity. θ = 108◦ corresponds to nitrogen displaced by liquid
water in an artificial Teflon core (Anderson, 1987b).
comparison. It can be seen that the cubic function follows the liquid curves to a
higher degree than the gas curves. It is evident that eqn (45) approximately predicts
the relative permeability for interfaces featuring a contact angle of 100◦ , which is
slightly hydrophobic.
It is noted that the relative permeability is seen to increase rapidly at higher
saturations, allowing effective water transport. However, the gas phase is shown
in Fig. 18 to reach its immobile saturation at a liquid saturation of 0.6 for the
hydrophobic cases. At this point, the permeability of the gas reduces to zero.
4 Multiphase flow models

With a clear understanding of the transport phenomena in porous media with non-
isothermal multiphase flows, the models prescribed in literature can be evaluated
for use in PEM fuel cell electrodes. Figure 20 classifies the models by their features.
The distinguishing features include the accounting of liquid water, convection of
the liquid water by the gas, transport of liquid water due to surface tension effects
(capillary diffusion), or whether the liquid is considered to be stationary. The char-
acterization of the various multiphase models that can be applied to the gas diffusion
layer in a PEM will start with most generalized case (multi-fluid) and work toward
the most specific version (porosity correction).
4.1 Multi-fluid model
The multi-fluid model presented herein is the application of the equations developed
previously to a porous medium occupied by air and liquid water. The air is treated as
Figure 20: Classification of PEMFC electrode models.

a multicomponent mixture and the liquid phase is considered as immiscible water.

In the multi-fluid model, as employed for fuel cells by Berning and Djilali [7], each
phase is modelled with its own set of field equations. The two-phases are coupled
by the relative permeabilities, which are sensitive to saturation, and phase change
terms.
The steady state mass and momentum conservation equations governing this
model are:
∇ · (ρg ug ) = ṁp c,
(46)
∇ · (ρl ul ) = ṁp c
and,
kg
ug = − ∇Pg ,
µg
(47)
kl
ul = − ∇Pg − D(sl )∇s.
µl
The species conservation equation in Berning and Djilali [7] was applied only to
the gas phase and is the single-phase species conservation equation (eqn (8)) with
modification of the liquid saturation to account for the reduced volume fraction
open to the gas. The steady state conservation of species A in the gas phase can be
written as:
∇ · (ug ρA ) = ∇ · DAeff ∇ρA + ṠA , (48)
where DAeff is the diffusivity of species A and ṠA is the phase change source term,
which is zero except in the water vapour conservation equation.
4.1.1 Phase change

To complete the multi-fluid model, an expression for the rate of phase change
between the gas and liquid phases must be introduced. The phase change between
the gas and liquid can be either evaporation or condensation, depending on the local
properties. However, the knowledge of molecular dynamics during phase change is
still considered limited [29]. A starting point for the exploration of phase change is
the kinetic theory of gases. The main principle when modeling phase change with
kinetic theory is that there is a maximum amount of vapour that can be accom-
modated at vapour/liquid interface. This maximum accommodation is the mass
transfer limiting characteristic. Thus, through the application of the kinetic theory
of gases, the maximum rate of evaporation for a liquid can be determined.
The goal of recent research in kinetic phase change is the approximation of evapo-
ration and condensation coefficients. The coefficients are ratios of actual mass trans-
fer to the theoretical maximum rate. Eames et al. [29] and Marek and Straub [30]
offer reviews of previously obtained values for a variety of circumstances.
The kinetic theory approximation allows for the consideration of thermal equi-
librium, or differences in the temperature between the gas and liquid phases.
The expression for the rate of mass transfer per unit area of gas/liquid interface
(ṁH2 O ), when the gas and liquid are in thermal equilibrium, was presented by
Eames et al. [29] as:
1/2
MH2 O
ṁH2 O = γk (Ps (T ) − Pv ), (49)
2πR
where Ps (T ) is the water vapour saturation pressure and Pv is the partial pressure of
the water vapour in bulk gas stream. γk is the kinetic evaporation coefficient, which
is equivalent to the condensation coefficient under thermal equilibrium conditions.
The magnitude of evaporation coefficients measured in experiments can range from
0.001 to 1 [30].
A second approach assumes the liquid phase exists in the form of a spherical
droplet. In such a case, the mass transfer is determined from the diffusion rate
between the bulk gas and the surface of the droplet. In addition, a mass transfer
Nusselt number is employed as a dimensionless measure of a droplet’s ability to
exchange mass. In its present form, this method applies only to systems featuring
local thermal equilibrium. The Nusselt number for mass transfer (Num ) from a
liquid droplet to the surrounding gas is [6]:
Num = 2.0 + 0.6Re1/2 Sc1/3 , (50)
where Sc is the Schmidt number (µg /(ρg DH2 Og )) and Re is the Reynolds number
(ρg Dd urel /µg ). Dd is the droplet diameter and urel is the relative velocity between
the gas and the droplet. The droplet’s mass transfer coefficient (γd ) can be obtained
by multiplying the Nusselt number by the diffusivity of water vapour in the gas
(Dvg ) and the inverse of the droplet diameter (1/Dd ).
γd = Dvg Num /Dd . (51)
The mass transfer coefficient is subsequently multiplied by the difference between

the density of water vapour in water-saturated air (ρs ) and the density of water
vapor in the bulk gas (ρ v ) to approximate the mass transfer per unit area of
gas/liquid interface. See Fig. 21 for clarification of these two densities.
ṁH2 O = γd (ρs − ρv ). (52)
This method of determining mass transfer rate was implemented in the modeling
of a PEM fuel cell by Berning [7, 11]. A simplification required to use this model
was to use a mean droplet diameter rather than the actual. Another appropriate
assumption is that since the velocities in the gas diffusion layer are small enough,
the Nusselt number reduces to 2.0 [6]. Therefore, eqn (52) can be expressed as:
2Dvg
ṁH2 O = (ρs − ρv ). (53)
Dd
Figure 21: Schematic of droplet evaporation for thermal equilibrium.
Berning and Djilali [7] also included a correction factor (ω) for reduced phase
change rates in porous media:
2ωDvg
ṁH2 O = (ρs − ρv ). (54)
Dd
When implementing either of the two aforementioned methods of calculating the
rate of phase change, the total interfacial area between the gas and liquid (Agl ) phase
must be determined to calculate the total mass transfer (ṀH2 O ). This is typically
achieved with the area of a spherical droplet of a mean diameter (πDd2 ). The number
of droplets (nd ) in a representative volume (V ) is the volume of liquid (εsl V ) divided
by the volume of a single droplet ( 16 πDd3 ).
6εsl V
nd = . (55)
πDd3
Thus, the interfacial area is:
6εsl V
Agl = (56)
Dd
and the total mass transfer in volume V is:
ṀH2 O = Agl ṁH2 O . (57)
4.1.2 Application
The multi-fluid model is the most general and, conceptually, the most flexible as it
relies on less restrictive assumptions. An example of the application of this model
to fuel cells is shown in Figs 22–24 [11].
Figure 22 shows iso-contours of oxygen and water vapour concentrations. In
these plots, as in subsequent figures, the top and bottom of the vertical axis cor-
respond to the catalyst layer/ GDL and channel/GDL interfaces, respectively, and
the planes correspond to successive locations from the inlet to the outlet of the fuel
Figure 22: Molar oxygen concentration (left) and water vapour distribution
(right) inside the cathodic gas diffusion layer at a current density of
0.8A/cm2 [11].
Figure 23: Rate of phase change [kg/(m3 s)] (left) and liquid water saturation
[-] (right) inside the cathodic gas diffusion layer at a current density
of 0.8A/cm2 [11].
Figure 24: Velocity vectors of the gas phase (left) and the liquid phase (right) inside
the cathodic gas diffusion layer at a current density of 0.8A/cm2 [11].
cell section. The molar oxygen concentration contours are similar to those found
with a single-phase version of the model, with more pronounced oxygen depletion
under the land areas of the collector plate. However, because of phase change, the
concentration of water is relatively uniform. The 0.8% difference can be attributed
to the temperature field’s influence on the saturation pressure of water.
The left-hand side of Fig. 23 presents the phase change in the cathode’s GDL.
Positive values correspond to evaporation, which can be found under the land areas
due to the increased pressure drop. The pressure drop, an artifact of the increased
resistance to gas transport below the land area, reduces the vapour pressure below
the saturation point. Negative values, indicating condensation, are most prevalent
at the catalyst layer/GDL interface because of the oxygen consumption and the
production of water vapour.
On the right-hand side of Fig. 23 the distribution of liquid water saturation in the
cathode’s GDL is shown. For this current density (0.8 A/cm2 ), a maximum satura-
tion of 10% is obtained under the land area at the end of the channel. The gradient
of the saturation is from high levels at the catalyst layer to low levels at the channel
interface. This reflects the implementation of the well-posed hydrophilic formula-
tion of the capillary transport, in which water travels from high to low saturation.
The velocity vectors of the gas and liquid phases are presented in Fig. 24. The
gas phase, shown on the left, indicates the bulk transport of gas to the catalyst layer.
In a single-phase model, the bulk motion of the gas is in the opposite direction due
to the removal of product water vapour. However, when phase change and capillary
transport are accounted for, the removal of product water is, in the most part, by the
liquid phase. This is evident in the plots of the liquid phase velocity vectors on
the right-hand side of Fig. 24.
The broader applicability of the multi-fluid model comes at the cost of solving
for an additional set of field equations and the required coupling of the phases.
This makes the numerical solution much more challenging and computationally
intensive. In particular, the convergence rates and the numerical stability of the
model can be problematic under some operating conditions. Alternative models
based on various levels of simplifications are presented below. Though they are
less general, these models can be more practical and can be effective in simulating
transport in the GDL, provided they are used for appropriate regimes.
4.2 Mixture model
The mixture model has been used to model two-phase flow in PEM electrodes by
several researchers, including Wang et al. [31] and You and Liu [32]. The main
theme of the mixture model is the description water transport, as vapour and liquid,
with traditional mixture theory practices. The resulting equation set is mathemat-
ically equivalent to the multi-fluid model [20]. The reformulation is obtained by
utilizing phase quantities that are relative to that of the mixture. The set equations
employed in the mixture model for steady state conditions are as follows:
The mixture conservation equation is:
∇ · (ρu) = 0. (58)
The mixture momentum equation is:
K
u=− ∇P, (59)
ρν
where ν is kinetic viscosity of the mixture. The mixture species conservation equa-
tion is expressed as:
 

∇ · (γA ρuyA ) = −∇ · (ερD∇yA ) + · · · ∇ · ε ρk sk DAk (∇yAk − ∇yA ) 
k=g, l
 

−∇ ·  yAk j k  , (60)
k=g, l
where the advection correction factors γA in eqn (60) account for the specific veloc-
ity fields encountered by each species. The advection correction factor is formu-
lated as: g
ρ λl yAl + λg yA
γA = g, (61)
ρl sl yAl + ρg sg yA
where the λk terms are the relative mobilities for the gas and liquid phases. They
are expressed as:
krg /νg
λg = ,
krg /νg + krl /νl
(62)
krl /νl
λl = .
krg /νg + krl /νl
The mixture quantities are evaluated as:
ρ = ρl sl + ρg sg ,
ρu = ρl ul + ρg ug ,
g
ρyA = ρl sl yAl + ρg sg yA ,
(63)
g
ρDA = ρl sl DAl + ρg sg DA ,
ρh = ρl sl hl + ρg sg hg ,
1
ν= .
(krl /νl ) + (krg /νg )
The individual phase velocities are extracted from the solution in the post pro-
cessing stage. These velocities are found with the addition of relative velocities
to the mixture velocity (ρg, l ug, l = λg, l ρu + jg, l ). Accounting only for capillary
diffusion, the relative mass flux term j k in eqn (60) is expressed as:

λl λg dPc
jl = k ∇sl ,
ν dsl
(64)
jg = −j l .
The liquid saturation is calculated and updated with each iteration. It is deter-
mined by comparing the total water concentration with the concentration of water
vapour required to saturate the gas phase. If the concentration of water is greater
than saturation concentration, then liquid water must be present. Using the expres-
sion for the total density of water (ρH2 O ) as a function of liquid water saturation (sl ),
and the saturated gas and liquid concentration of water (ρs and ρl respectively),
ρH2 O = sl ρl + (1 − sl )ρv , (65)
the saturation can be determined. It is important to note that ρs is the mass of water
vapour per unit volume of water vapour saturated air (the mass fraction of the water
vapour in the air multiplied by the bulk density of water vapour). ρs can be extracted
from the temperature dependent saturation pressure of air (Ps (T )), the absolute gas
pressure (P), and the bulk density of water vapour (ρv ):
Ps (T )
ρs = ρv . (66)
P
Considering a pure liquid phase, eqn (65) can be rearranged as:
ρH2 O − ρs
sl = . (67)
ρl − ρs
Thus, the saturation can be calculated from the total water concentration, temper-
ature, and pressure.
4.3 Moisture diffusion model
The moisture diffusion model, also referred to as unsaturated flow theory [20], was
developed to determine the transport of liquid water when the only driving force is
capillarity. Luikov [33] and Whitaker [34] are considered pioneers of this formu-
lation. This method was applied to PEM fuel cells by Natarajan and Nguyen [35].
The steady state transport equation for liquid water in the moisture diffusion model
can be expressed as (Wang and Cheng, 1997):

∇ · D(sl )∇sl + Ṡl = 0, (68)
where
ρl kl ∂pc
D(sl ) = , (69)
µl ∂sl
Ṡl = −Ṡg, H2 O (70)
and Ṡ is the mass source due to phase change. The moisture diffusion model could
be incorporated into a CFD code by treating the liquid phase as a scalar species
with no convection terms. This would be a moderately easy method of incorporating
two-phase flow into a single-phase fuel cell model. The mass sources need to be
calculated in the form of a rate as in the multi-fluid model. In addition, the porosity
would require a correction based on the liquid saturation.
4.4 Porosity correction model
The porosity correction model simplifies the present two-phase problem by neglect-
ing the transport of liquid water. The saturation level is computed with each iteration.
In the Kermani et al. [36] model, the temperature and the level of saturation are
calculated iteratively by the internal energy and density of the water in the system.
Subsequently, the volume fraction open to the gas phase is reformulated as:
εg = ε(1 − s). (71)
This model is particularly efficient when saturation levels are low (below the
immobile saturation limit for the liquid water). This model would be the simplest
to append to an existing single-phase fuel cell model.
4.5 Evaluation of the multiphase models in the literature
Table 2 is provided to help determine which model should be employed given

the conditions of the system, the porous media considered, and the computational
resources available. At one extreme is the multi-fluid model. The multi-fluid model
is a strong candidate when an abundance of computational resources is available
and stable phase coupling can be achieved. At the other extreme is the porosity cor-
rection model. This model is an ideal candidate, due to its computational efficiency
when considering saturation levels below the immobile value.
5 Outstanding issues and conclusions

A number of fundamental issues need to be addressed in order to devise reliable
predictive tools for two-phase transport in gas diffusion electrodes. These include:
• The hydrodynamic and diffusive properties of the porous media in the elec-
trodes need to be characterized. The structure of PEM fuel cell gas diffusion
layers is typically fibrous. In the case of carbon cloth layers, the porous matrix
is constructed from woven tows of fibers producing macro- and micro-pores.
Alternatively, carbon paper layers are a formation of randomly laced fibers. It is
obvious that the architecture of gas diffusion layers is significantly different to
that of packed beds, or cylindrical pores in a monolithic structure that are often
the object of porous media studies. It is also clear that these fibrous layers are
anisotropic. Some parameters to be resolved are the three-dimensional tensors
for the area porosity, permeability, and tortuosity. At present, the properties can
only be implemented in an isotropic form in the available commercial CFD
codes.
• The capillarity in the electrode’s porous media requires significant research.
Current expressions used to determine capillary pressure in electrodes are based
on studies of packed beds and rarely include the influence of wettability. Issues
to be resolved include the presentation of a capillary pressure versus saturation
Table 2: Advantages, disadvantages, and areas of application for each of the multi-
phase flow models.
Multiphase Areas of
flow model Advantages Disadvantages application
Multi-fluid – Generalized form. – Highest number of – Best employed for

model – Interphase transfer variables. high saturation
models can be used. – Needs the most conditions because
– Can resolve computational of the need for
complex liquid resources. greater liquid
motion. – Coupling of the resolution.
– Models convection phases can lead to – When the
of liquid by the gas. unstable models. influence of the gas
– Can model species – Requires a on the liquid is
diffusion in liquid. multiphase CFD equivalent to that of
code. the surface tension.
Mixture – Reduced number – May have trouble – Best used when the
model of variables. converging at higher gas pressure is the
– Models the saturations (liquid dominant force on
influence of the gas and gas have the liquid or when
pressure on the significantly capillary forces
liquid. different velocity drive the liquid in
fields). the same direction.
– Cannot employ – High capillary
interphase transfer number (i.e. large
models. pores and high
– Large number of permeability).
mixture quantities to
calculate.
Moisture – One additional – Does not account – When surface
diffusion equation over one- for the influence of tension is the
model phase model. the gas pressure on dominant force on
– Can employ phase the liquid. the liquid.
change models. – Cannot model – Low capillary
interphase transfer numbers (i.e. small
of heat and species. pores and low
permeability).
Porosity – No additional – Does not account – Conditions where
correction transport equations for liquid motion. the liquid saturation
over the one-phase does not exceed the
model. immobile saturation
(i.e. low relative
humidities, very
small pores, and low
current densities).
curve for a gas diffusion layer, the immobile saturation levels for gas and liquid,
and the influence of wettability on those properties.
In closing this discussion, we note that mass transport limitations continue to
be a significant hindrance to achieving higher current densities in PEM fuel cells.
Water management within these fuel cells is a key consideration in their design.
Knowledge of the behavior of liquid water in electrodes is limited by the inability
to make in situ measurements. Better understanding of the transport of water in
the PEMFC electrode can be obtained from models that capture the important
physical processes. Several specific models of two-phase mass transport have been
outlined and discussed in this chapter. These models, once implemented in a CFD
code, will be able to help fuel cell designers improve their understanding of the
transport of liquid water, as well as the transport of reactant and product gases, in
the porous electrodes. It is clear from this review that a critical area to advance
modeling of two-phase transport in gas diffusion electrode is further experimental
characterization of the porous materials employed in PEM fuel cells.
References
[1] Schulze, M., Schneider, A. & Gülzow, E., Alteration of the distribution of the
platinum catalyst in membrane-electrode assemblies during PEFC operation.
J. Power Sources, 127, pp. 213–221
[2] Nam, J.H. & Kaviany, M., Effective diffusivity and water-saturation distri-
bution in single- and two layer PEMFC diffusion medium. Int. J. Heat Mass
Transfer, 46, pp. 4595–4611, 2003.
[3] Ngo, N.D. & Tamma, K.K., Microscale permeability predictions of porous
fibrous media. Int. J. Heat Mass Transfer, 44, pp. 3135–3145, 2001.
[4] Papathanasiou, T.D., Flow across structured fiber bundles: a dimensionless
correlation. J. Multiphase Flow, 27, pp. 1451–1461, 2001.
[5] Biloé, S. & Mauran, S., Gas flow through highly porous graphite matrices.
Carbon, 41, pp. 525–537, 2003.
[6] Bird, R.B., Stewart, W.E. & Lightfoot, E.N., Transport Phenomena, John
Wiley and Sons: New York, 1960.
[7] Berning, T. & Djilali, N., A three-dimensional, multi-phase, multicomponent
model of the cathode and anode of a PEM fuel cell. J. Electrochem. Soc. 150,
pp. A1589–A1598, 2003.
[8] Cussler, E.L., Diffusion–Mass Transfer in Fluid Systems, Cambridge
University Press: New York, 1997.
[9] Epstein, N., On tortuosity and the tortuosity factor in flow and diffusion
through porous media. Chem. Eng. Sci., 44(3), pp. 777–779, 1989.
[10] De La Rue, R.E. & Tobias, C.W., On the conductivity of dispersions.
J. Electrochem. Soc., 106(9), 1959.
[11] Berning, T., Three-Dimensional Computational Analysis of Transport Phe-
nomena in a PEM Fuel Cell. PhD thesis, University of Victoria, 2002.
[12] Kast, W. & Hohenthanner, C.R., Mass transfer within the gas-phase of porous
media. Int. J. Heat Mass Transfer, 43(5), pp. 807–823, 2000.
[13] Cunningham, R.R. & Williams, R.J.J., Diffusion in Gases and Porous Media,
Plenum Press: New York, 1980.
[14] Karniadakis, G.E. & Beskok, A., Micro Flows-Fundamentals and Simula-
tion, Springer-Verlag: New York, 2002.
[15] Feng, C. & Stewart, W.E., Practical models for isothermal diffusion and flow
of gas in porous solids. Ind. Eng. Chem. Fundam., 12(2), 1973.
[16] Taylor, R., Calculation of steady-state multicompnent mass transfer rates in
porous media in the transition region. Ind. Eng. Chem. Fundam., 21, pp. 63–
67, 1982.
[17] Staia, M.H., Cambell, F.R. & Hills, A.W.D., Measurement of gaseous diffu-
sion coefficents in porus reaction products. Ind. Eng. Chem. Res., 26, pp. 438–
446, 1987.
[18] Suwanwarangkul, R., Croiset, E., Fowler, M.W., Douglas, P.L., Entchev, E.
& Douglas, M.A., Performance comparison of Fick’s, dusty-gas and Stefan-
Maxwell models to predict the concentration overpotential of a SOFC anode.
[19] Mezedur, M.M., Kaviany, M. & Moore, W., Effect of pore structure, random-
ness and size on effective mass diffusivity. AIChE Journal, 48(1), pp. 15–24,
2002.
[20] Wang, C.Y. & Cheng, P., Multiphase flow and heat transfer in porous media.
Advances in Heat Transfer, 30, pp. 30–196, 1997.
[21] Udell, K.S., Heat transfer in porous media considering phase change and
capillarity-the heat pipe effect. Int. J. Heat Mass Transfer, 28(2), pp. 485–
495, 1985.
[22] App, J.F. & Mohanty, K.K., Gas and condensate relative permeability at near-
critical conditions: capillary and Reynolds number dependence. J. Petrol.
Sci. Eng., 36, pp. 111–126, 2002.
[23] Kaviany, M., Principles of Heat Transfer in Porous Media, Springer-Verlag:
New York, 1991.
[24] Anderson, W.G., Wettability literature survey – Part 4: Effects of wettability
on capillary pressure, J. Petrol. Technol., pp. 1283–1299, 1987.
[25] Leverett, M.C., Capillary behaviour in porous solids. AIME Trans., 142,
pp. 152–169, 1941.
[26] Scheidegger, A.E., The Physics of Flow through Porous Media, University
of Toronto Press: Toronto, 1974.
[27] van Genuchten, M. Th., A closed-form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Sci. Soc. Amer. J., 44, pp. 892–898,
1980.
[28] Anderson, W.G., Wettability literature survey – Part 5: Effects of wettability
on relative permeability. J. Petrol. Technol., pp. 1453–1468, 1987.
[29] Eames, I.W., Marr, N.J. & Sabir, H., The evaporation coefficient of water: a
review. Int. J. Heat Mass Transfer, 40(1), pp. 2963–2973, 1997.
[30] Marek, R. & Straub, J., Analysis of the evaporation coefficient and the con-
densation coefficent of water. Int. J. Heat Mass Transfer, 44, pp. 39–53,
2001.
pp. 40–50, 2001.
[32] You, L. & Liu, H., A two-phase and transport model for the cathode of PEM
fuel cells. Int. J. Heat Mass Transfer, 45, pp. 2277–2287, 2002.
[33] Luikov, A.V., Systems of differential equations of heat and mass transfer in
capillary-porous bodies (Review). Int. J. Heat Mass Transfer, 18, pp. 1–14,
1975.
[34] Whitaker, S., Simultaneous heat, mass and momentum transfer in porous
media: A theory of drying. Advances in Heat Transfer, 13, pp. 119–203,
1977.
[35] Natarajan, D. & Nguyen, T.V., Three-dimensional effects of liquid water
flooding in the cathode of a PEM fuel cell. J. Power Sources, 115, pp. 66–
80, 2003.
[36] Kermani, M.J., Stockie, J.M. & Gerber, A.G., Condensation in the cathode
of a PEM fuel cell. Proc. of the 11th Annual Conference of the CFD Society
of Canada, 2003.
Nomenclature
AD Droplet area
Agl Area of gas-liquid interface
c Molar concentration
Ca Capillary number
dpor Pore diameter
D Diffusion coefficient
DAB Diffusion coefficient in a binary system
Dij Diffusion coefficient in a multicomponent system
D(sl ) Capillary diffusion coefficient
Dd Droplet diameter
j Relative mass flux
k Permeability
kg, l Relative permeability
kB Boltzmann constant
kk Kozeny constant
Kn Knudsen number
ṁH2 O Phase change mass transfer
M Molecular weight
n Mass flux
N Molar flux
Num Mass transfer Nusselt number
P Pressure
Pc Capillary pressure
Ps Saturation pressure
R Universal gas constant
s Reduced saturation
sim Immobile saturation
sg, l Phase saturation
S Mass source term
Sc Schmidt number
t Time
T Temperature
x Mole fraction
y Mass fraction
u Superficial velocity
ui Average interstitial velocity
V Volume
VD Volume of droplet
Greek symbols
γ Mass transfer coefficient

γA Advection correction factor
ε Porosity
λ Mean free path
λg, l Relative mobility
µ Viscosity
ν Kinematic viscosity
ρA Density of species A
ρ Density
σ Interfacial tension
σii Collision diameter
τ Tortuosity factor
φ Diffusion volume
ω Phase change correction factor
CHAPTER 6
Numerical simulation of proton exchange

membrane fuel cell
T.C. Jen, T.Z. Yan & Q.H. Chen
Department of Mechanical Engineering, University of
Wisconsin-Milwaukee, USA.
Abstract
This chapter presents general mathematical models and numerical simulation
for proton exchange membrane fuel cell (PEMFC) to evaluate the effects of vari-
ous designs and operating parameters on the PEMFC performance. Both one and
two-dimensional models are presented; the advantages and weaknesses of using
one-dimensional and two-dimensional models are discussed in detail. Subsequently,
a general three-dimensional model is developed and described in detail. This three-
dimensional general model accounts for electrochemical kinetics, current density
distribution, hydrodynamics, and multi-component transport. It starts from basic
transport equations including mass conservation, momentum equations, energy bal-
ance, and species concentration in different elements of the fuel cell sandwich, as
well as the equations for the phase potential in the membrane and the catalyst layers.
These governing equations are coupled with chemical reaction kinetics by intro-
ducing various source terms. It is found that all these equations are in a very similar
form except the source terms. Based on this observation, all the governing equations
can be solved using the same numerical formulation in a single domain without
prescribing the boundary conditions at interfaces between different elements of the
fuel cell. Detailed numerical formulations are presented in this chapter. Various
parameters, such as velocity field, local current density distribution, species con-
centration variation along the flow channel, under various operation conditions are
computed by solving these governing equations in a single domain consisting of
gas channel, catalyst layer and membrane. The performance of the PEMFC affected
by various parameters such as temperature, pressure, and the thickness of the mem-
brane is investigated. The numerical results are further validated with experimental
data, which are available in the literature. This general three-dimensional model can
be used for the optimization of PEMFC design and operation. It can also serve as a
building block for the modeling and understanding of PEMFC stacks and systems.
1 Introduction
Although the fuel cell has received much attention in recent years, the underlying
concept is still under heavy investigations [1]. The general concept of fuel cell oper-
ations can be characterized as gas-mixtures transport and transformation of species
by electrochemical reactions. Therefore, one of the most challenging problems in
fuel cell research is to predict the performance of the fuel cell. Using a mathe-
matical model with numerical procedures to simulate the transport phenomena is
a powerful tool to understand the fundamental physical and chemical processes of
a fuel cell. Thus, reliable mathematical models are essential for fuel cell design,
operation and optimization.
2 One-dimensional (1-D) model

The earliest attempts to simulate the phenomena in PEM fuel cells were mainly
using 1-D models. The pioneering work in 1-D modeling includes Savinell and
Fritts [2], Ridge [3], Fritts and Savinell [4], Yang [5], Verbrugge and Hill [6],
Bernerdi and Verbrugge [7], and Springer [8]. These earliest works mostly con-
sidered water and thermal management. During PEM fuel cell operations, water
molecules are carried from the anode side to cathode side of the membrane by
electro-osmosis. If this transport rate of water is higher than that by back diffusion
of water, the membrane will become dehydrated and too resistive to conduct cur-
rent. On the other side, if there is too much water, i.e. cathode side flooding may
occur in the pores of the gas diffuser and hinder the transportation of reactants to the
catalyst side. Consequently, proper water management is required to maintain high
membrane conductivity and prevent flooding. Moreover, the effect of reacting gas
dilution by high vapor pressure must be taken into consideration [1]. Thus, the water
management is critical for efficient performance. Verbrugge and Hill [6] have car-
ried out extensive modeling of transport properties in perfluorosulfonate ionomers
based on dilute solution theory. False [9] reported an isothermal water map based
on hydraulic permeability and electro-osmotic drag data. Fuller and Newman [10]
applied concentrated solution theory and employed literature data on transport prop-
erties to produce a general description of water transport in fuel cell membranes.
Bernardi and Verbrugge [7] took a different approach, in which transport through
the gas diffusion electrodes was considered. They assumed the membrane to be
uniformly hydrated, corresponding to an “ultra-thin membrane” case. Springer [8]
presents an isothermal model, which includes transport through the porous elec-
trodes. Their model required inputs from calculated diffusivities, which are needed
to correct for porosity, and from experimentally determined transport parameters
for the transport through the membrane. Most of these models treated flow chan-
nel as being perfectly well fixed, with no pressure and concentration difference
along the gas channel. These research works were particularly useful in classify-
ing the different models for porous gas diffusion electrodes and providing the key
properties of the membrane required for numerical simulation. These 1-D models
provided the sub-model bases for many following 2-D or 3-D research works.
Here, we introduce a general 1-D PEM fuel cell model based on the research
work by Bernardi and Verbrugge [7].
2.1 General 1-D model
2.1.1 Model description

The fuel cell consists of a membrane sandwiched between two gas-diffusion
electrodes as shown in Fig. 1. The gas-diffusion electrodes are porous compos-
ites made of electronically conductive material mixed with hydrophobic poly-
tetrafluorethylene and supported on carbon cloth. The gaseous reactants can transport
through the electrodes during operation. The electrochemical reactions occur in the
catalyst layer. Humidified hydrogen enters anode gas chamber, transports through
the anode gas diffuser, and dissolves into catalyst layer. Hydrogen molecule is
oxidized and generates protons and electrons. Protons go into membrane and elec-
trons are received by carbon conductor. The overall electrochemical reaction can
be expressed as:
2H ↔ 4H+ + 4e-.
2 (1)
Gaseous reactant O2 from air, usually mixed with H2 O, enters into cathode gas
channel, transports through porous gas diffuser, and dissolves into cathode catalyst
layer. Hydrogen protons diffuse into the cathode catalyst layer from anode catalyst
layer through the membrane. On the surface of the catalyst particles, the oxygen is
consumed along with the protons and electrons, and the product, H2 O, is produced
along with the waste heat. The overall electrochemical reactions occurring at the
reaction site may be represented as:
4H+ + 4e- + O2 → 2H2 O + heat. (2)
Therefore, the overall electrochemical reaction of the PEM fuel cell is:
2H2 + O2 → 2H2 O + heat + electric energy. (3)
Figure 1: Schematic diagram of PEMFC.

The proton, dissolved oxygen, and dissolved hydrogen concentrations depend

on the kinetic expressions of four-electron-transfer reaction for oxygen reduction
and the proton-transfer reaction for hydrogen oxidation.
2.1.2 Model assumptions

The cell is assumed to operate at steady state conditions. Since the cell thickness
is very small compared with the other dimensions of the cell, a one-dimensional
approximation is used in the model formulation. The entire system is taken to be
at constant temperature, and the gases are assumed to be ideal and well mixed in
the chambers. The steady state and isothermal assumptions are valid because these
conditions are often achieved in a small single-cell experimental investigation. The
model requires some properties inputs such as water-diffusion coefficients, electro-
osmotic drag coefficients, water sorption isotherms, and membrane conductivities,
which are all measured in single-cell experimental studies. The temperature of
fuel cell is assumed to be well controlled. The heat transfer is supposed to be
very efficient so that the heat generation, which is due to the irreversibility of the
electrochemical reaction, ohmic resistance, as well as mass transport overpotentials,
can be taken out quickly. The inlet gaseous temperature is assumed to be preheated
to the cell temperature. Fully hydrated membrane, wet gas diffuser, and saturated
chamber gases are considered.
The total gas pressure within the gas channel is assumed to be constant and same
as the pressure in gas diffuser since the gas velocity is very slow, the pressure
variation can be neglected along the gas channel. However, the pressure can vary
between anode and cathode.
With these assumptions the cell model is formulated with transport equations for
the electrodes, catalyst layers and membrane as described below.
2.1.3 Governing equations

The governing equations constituting the mathematical model of the PEM fuel
cell are derived by applying the conservation equations for an ideal gas in porous
medium, the Butler-Volmer equations, and the Stefan-Maxwell equation for gas-
phase transport.
Electrodes:
Continuity:
d
(ρv) = 0. (4)
dx
Species:
d
(ρi v + ρi Vi ) = mi . (5)
dx
Potential:
d s I
= − eff , (6)
dx σ
where ρi is the partial density of species i, ρ is average density, v is the mass-
averaged velocity in x direction, Vi is the species i diffusion velocity, mi is the mass
of species i, s is the electrical potential in the solid matrix of the electrode, I is cell
current density, and σ eff is the effective electrical conductivity. The diffusion veloc-
ity Vi can be determined from the Stefan-Maxwell equation for multi-component
gas diffusion,
xi xj
∇xi = Vj − Vi , (7)
i
Dijeff
where Dieff is the effective binary diffusion coefficient for species i in j, and xi is the
mole fraction for species i. Species i for cathode electrode can be taken as O2 , N2
and H2 O; species i for anode electrode can be taken as H2 and H2 O.
2.1.4 Catalyst layers

The catalyst layers are very thin compared to the other components in PEMFC, but
they are the heart of the fuel cell. Electrochemical reaction occurs in this region
to produce electrical energy and products. In this region, the transfer of mass and
energy is coupled with reaction kinetics when the cell is loaded and results in a
potential difference between electrodes. How this potential difference varies as
functions of mass transfer, electrode kinetics, and energy flux determines the fuel
cell performance. The mathematical description of the active-catalyst-layer region
is based on Butler-Volmer relations [13, 14] for electrochemical reaction and species
diffusion. The macro-homogeneous approach is used in developing the governing
equations. Current in the catalyst layer can transfer to electronically-conductive
solid portion of the catalyst layer (carbon and catalyst particles). The rate of elec-
trochemical reaction is given by the Butler-Volmer expression [13, 14] . The mass
transfer is modeled by using species conservation and Fick’s Law of diffusion. The
governing equations can be expressed as follows:

di CO2 αc Fηc
= −aj0 ref
exp − , (8)
dx CO2 ref RT
where i is the current density, a is the catalyst reactive surface area per unit volume,
j0ref is the reference exchange current density at the reference concentration C ref ,
CO2 is the oxygen concentration, αc is the transfer coefficients in Butler-Volmer
relation, F is the Faraday’s number, R is the universal gas constant, ηc is the over-
potential at cathode side, and T is the reaction temperature.
From the material balance based on standard porous-electrode theory, we have:
dNi si di
=− , (9)
dx nF dx
where Ni is the superficial flux of species i, and si is the stoichiometric coefficient
for species i in the cathode reaction and anode reaction. The right hand side of
eqn (9) is the source term for the species conservation equation as shown below.
For hydrogen:
d 2 CH2 dCH2 di sH 2
eff
DH =v + . (10)
2
dx2 dx dx nF
For oxygen:
d 2 CO2 dCO2 di sO2
eff
DO 2 2
=v + . (11)
dx dx dx nF
For H2 O:
d 2 CH2 O dCH2 O di sw
eff
DH 2O
= v + − . (12)
dx2 dx dx nF
Potential:
d
i = −σ eff , (13)
dx
where i is the current density in the electron-conducting solid, and σ eff is the
conductivity of the electronically conductive catalyst layer.
2.1.5 Membrane
The membrane of PEM fuel cell acts as the hydrogen proton conductor. The trans-
port processes in the membrane can be described by the conservation of species.
The flux of species in the membrane is determined by the net effect of electro-
osmotic drag, diffusion due to concentration gradient, and the convection due to a
pressure gradient. A form of the Nernst-Planck equation is used to describe the flux
of species in the membrane [15, 16].
F d m dCi
Ni = −zi Di Ci − Di + vCi , (14)
RT dx dx
where Ni is the superficial flux of species i, zi is charge number of species i, Ci is the
concentration of species i, and Di is the diffusion coefficient of species i, and m
is electrical potential in membrane. v is the velocity of H2 O, which is generated by
electric potential and pressure gradient, and can be described by a form of Schögl’s
equation:
k d m kp dp
v= zf cf F − , (15)
µ dx µ dx
where k is electro-kinetic permeability, µ is pore-fluid viscosity, zf is fixed-site
charge, cf is fixed-charge concentration, and kp is hydraulic permeability.
Current conservation:
di
= 0. (16)
dx
Mass conservation for liquid flow:
dv
= 0. (17)
dx
Electric potential:
d i F
= − + cH + v, (18)
dx σ σ
F2
σ= DH + cH + . (19)
RT
Furthermore, the potential and pressure profiles throughout the membrane region
are assumed to be linear with constant velocity. This is basically true for fully-
developed porous media case.
2.1.6 Boundary conditions

Before solving the governing equations, appropriate boundary conditions must be
specified. The temperature, pressure, relative humidity, flow rate, compositions of
the reactant gases both in anode and cathode channels are specified according to
the cell operation condition.
In the following model description, c− is the anode catalyst layer, c+ is the
cathode catalyst layer, d− is the anode gas diffuser; d+ is the cathode gas diffuser,
and m is membrane. lc− is the thickness of anode catalyst layer, lc+ is the thickness
of cathode catalyst layer, lm is the thickness of membrane, ld− is the thickness of
anode gas diffuser, ld+ is the thickness of cathode gas diffuser. At the interface
between the anode gas diffuser and catalyst layer, the solid phase potential gradient
is related to the cell operating current density by Ohm’s law:

d m c
i= σ −, (20)
dx c− eff
c
where σeff− is electronic conductivity of anode catalyst layer.
The membrane concentration of oxygen set to be zero as,
cO2 = 0. (21)
Since the hydraulic pressure distribution throughout the anode gas diffuser is linear,
we can impose the boundary condition of pressure at the interface between anode
gas diffuser and catalyst layer as:
µ d
po = p− − l v −,
d− d− s
(22)
kp
d
where p− is the pressure in anode gas chamber, kp − is anode gas diffuser perme-
d
ability, and vs − is water velocity in anode gas diffuser.
Because the total water flux is continuous, we have

d d c
Cw vs − + Nw− = Cw εm− εm
wv c , −
(23)
d
where Nw− is water vapor flux from Stefan-Maxwell equation [17], which is
sat
d I xw −
Nw− = − , (24)
2F (1 − xwsat− )
where
sat
sat− pw −
xw = . (25)
p−
The concentration of hydrogen is given by

p−
sat
cH 2
= (1 − xwsat ) . (26)
KH2
At the interface between the membrane and anode catalyst-layer, the current con-
tinuity should be satisfied:

d d
σm = σ . (27)
dz m dz c−
eff
The superficial flux of liquid water is continuous:

c
v|m = εm− v|c− (28)
and the flux of dissolved hydrogen is continuous through the internal boundary:

dcH2
eff dcH2
DH2 = DH2 . (29)
dx m dx c−
At the membrane /cathode-catalyst-layer interface, the current, superficial flux

of liquid water, and the flux of dissolved oxygen are continuous, and the dissolved
hydrogen concentration is zero.

d c+ d
σm = σ eff dz , (30)
dz m c+
c
v|m = εm+ v|c+ , (31)

dcO2
eff dcO2
DO2 = DO2 , (32)
dx m dx c+
cH2 = 0. (33)
At the cathode-catalyst-layer/gas-diffuser interface, the current in the solid phase

is continuous.
c d solid d+ d solid
σeff+ = σ (34)
dx c+ eff dx d+
and the superficial flux of water is given by

d d c
ρvs + + Nw+ = ρεm+ εm
wv c , +
(35)
where
sat
d I xw +
Nw+ = sat xN2 (36)
4F 1 − xw + − xN2 +
rw
and rw is the diffusivity ratio given by
Dw−N2
rw = . (37)
Dw−O2
The concentration of dissolved-oxygen in the catalyst layer is given by

sat p+
sat
cO 2
= (1 − xN2 − xw + ) , (38)
K O2
where KO2 is Henry’s constant, and

sat
sat+ pw +
xw = . (39)
p+
2.1.7 Results and discussion

Ticianelli et al. [11] and Srinivasan et al. [18] published their experimental results
respectively. Their experimental data have been used extensively as standard vali-
dation by many numerical studies such as Bernardi and Verbrugge [7]. Since most
of the numerical simulation parameters and properties are based on their experi-
mental results, these parameters and properties are adopted in this chapter. Table 1
shows the detailed physical parameters and properties.
Figure 2 shows the comparison of Bernardi and Verbrugge’s [7] calculated results
with Ticianelli’s experimental data [11]. Note that, in their model the exchange-
current density a+ ioref was adjusted to yield model results that are suitable mimic the
experimental results. With this one adjustable parameter, the agreements between
model and experimental results are quite good.
Figure 3 shows the components of the overall cell polarization for the base case.
At the low current densities (less than 100 mA/cm2 ), the activation overpotential of
the oxygen reduction reaction is almost entirely responsible for the potential losses
of the cell. For current densities greater than 200 mA/cm2 , the ohmic potential
loss due to the membrane and electrodes become more significant, and the cathode
activation overpotential reaches a relatively constant value.
Figure 4 shows the effect of membrane thickness on the potential. Generally,
the electronic conductivity is proportional to the thickness of the membrane. The
results show the fuel cell potential increases as the membrane thickness decreases
from 7 mil (1 mil = 1/1000 inch) to 2 mil under dry membrane state.
The modeling value for the wet membrane thickness was obtained by assuming
that the ratio of the wet to dry thickness was a constant. The comparison between
experimental results (not shown) and model predictions is generally satisfactory.
The effect of cathode pressurization on the fuel cell potential is depicted in Fig. 5
When the cathode pressure decreases from 5 to 3 atm, the open-circuit potential is
only slightly affected by the change in cathode pressure. The cell potential decrease
is primarily due to oxygen mole fraction decreased because of lower pressure; this
leads to the increase in cathode activation overpotential loss.
2.1.8 Summary
1-D model provides a good preliminary foundation for PEM fuel cell modeling.
These 1-D models base on the fundamental transport properties where the potential
losses incurred by the activation overpotential of the anode and cathode reactions,
Table 1: The values of the parameters.

Wet membrane thickness, lm 0.023 cm
Gas diffuser thickness, ld+ = ld− 0.026 cm
Catalyst layer thickness, lc+ = lc− 0.001 cm
Relative humidity of inlet air 100%
Relative humidity of inlet hydrogen 100%
Inlet fuel and air temperature, T0 105 ◦ C
Cell temperature, T 80 ◦ C
Inlet nitrogen-oxygen mole ratio 0.79/0.21
Air-side pressure, p+ 5 atm
Fuel-side pressure, p− 3 atm
Thermodynamic open-circuit potential, U 1.194 V
Estimated proton diffusion coefficient, DH+ 4.5 × 10−5 cm2 /s
Estimated ionic conductivity, σm 0.17 ohm/cm
Fixed charge site concentration, cf 1.2 × 10−3 mol/cm3
Charge of fixed site, zf −1
Dissolved oxygen diffusivity, DO2 1.2 × 10−6 cm2 /s
Electrokinetic permeability, k 1.13 × 10−15 cm2
Hydraulic permeability, kp 1.58 × 10−14 cm2
Pore-fluid (water) viscosity, µ 3.565 × 10−4 kg/m · s
Pore-fluid (water) density, ρ 0.054 mol/cm3
Saturated water vapor pressure, pw sat 0.467 atm
Henry’s law constant for oxygen in membrane, KO2 2 × 105 atm · cm3 /mol
Henry’s law constant for hydrogen in membrane, KH2 4.5 × 104 atm · cm3 /mol
Volume fraction membrane in active layer, εm,c 0.5
Volume fraction water in membrane, εw,m 0.28
d = σc
Electronic conductivity, σeff 0.53 ohm/cm
eff
Reference kinetic parameter of anode, ajoref 1.4 × 105 A/cm3
Reference kinetic parameter of cathode, ajoref 1 × 10−5 A/cm3
Cathodic transfer coefficient, αc 2
Anodic transfer coefficient, αa 1/2
H+ reference concentration, cH ref
+ 1.2 × 10−3 mol/cm3
ref
O2 reference concentration, cO 3.39 × 10−6 mol/cm3
2
ref
H2 reference concentration, cH 5.64 × 10−5 mol/cm3
2
and the ohmic losses incurred by the membrane. The membrane is assumed fully
hydrated and the cell is assumed isothermal. However, it cannot model the depletion
of the reactants and the accumulation of products in the flow direction.
2.2 General 2-D model
Two-dimensional (2-D) models are developed to improve the earlier 1-D models.
Fuller and Newman [19] modeled and solved the transport across the fuel cell
Figure 2: Model and experimental data for fuel cell at 80 ◦ C, p+ = 5 atm,

p− = 3 atm.
Figure 3: The contribution of fuel cell.
sandwich at certain location along the gas channel and then integrated in the sec-
ond direction. In this 2-D model, the gas outside the gas diffusers was assumed to
have uniform composition in the direction across the cell. Nguyen and White [20],
Amphlett et al. [21], and Yi and Nguyen [22] developed pseudo 2-D models account-
ing for composition changes along the flow path. These models are useful for small
single cells, however, when it is applied to large-scale fuel cells, particularly under
high fuel utilization conditions, the applicability is limited.
Figure 4: The effects of membrane thickness on fuel cell potential.
Figure 5: Effect of cathode gas-channel pressure.
Later, Gurau et al. [23] presented a 2-D model of transport phenomena in PEM
fuel cells. They developed a model to understand the transport phenomena such as
the oxygen and water distributions. They considered the interaction between the gas
channels and the rest of the fuel cell sandwich. Yi and Nguyen [24] also formulated a
2-D model to explore hydrodynamics and multi-component transport in the cathode
Figure 6: PEM fuel cell schematic diagrams.
of PEMFCs with an interdigitated flow field. More recently, Um et al. [25] presented
a transient 2-D model based on finite-volume CFD approach. They also explored
hydrogen dilution effects on PEMFC running on reformate gas.
The following section describes a general two-dimensional model for electro-
chemical and transport processes occurring inside a PEMFC.
2.2.1 Model description and assumptions

Figure 6 shows a typical 2-D PEM fuel cell. It includes the collector plates, fuel
and gas channels, gas-diffusers, catalyst layers and membrane. Hydrogen and water
vapor flow through anode fuel channel and humidified air is fed into the cathode
channel. Assume that hydrogen oxidation and oxygen reduction reactions occur
only within active catalyst layers.
Additional assumptions of this 2-D model are described below:
• The gas mixtures including anode channel and cathode channel are ideal gases.
The density is treated as constant.
• The gas flows are laminar and incompressible.
• The electrodes, gas diffusers, catalyst layers, and membrane are isotropic and
homogeneous.
• The heat generated under reversible condition is neglected.
• Only steady state condition is considered.
• Cell temperature is held constant.
• The contact electrical potential drop of different cell components is negligible.
2.2.2 Mathematical model

A single-domain approach developed by Um et al. [24] is used in this section
to describe fuel cell transport processes. Therefore, no boundary conditions are
required at the interface of different fuel cell components.
The governing equations can be written as:

Mass conservation:
∂u ∂v
+ = 0. (40)
∂x ∂y
Momentum conservation:
2
1 ∂u ∂u ∂p ∂ u ∂2 u
ρ u +v = −ε + µ + + Sx , (41)
ε ∂x ∂y ∂x ∂x2 ∂y2
2
1 ∂v ∂v ∂p ∂ v ∂2 v
ρ u +v = −ε + µ + + Sy . (42)
ε ∂x ∂y ∂y ∂x2 ∂y2
Species conservation:
2
1 ∂Xk ∂Xk ∂ Xk ∂ 2 Xk
u +v = εDk + + Sk . (43)
ε ∂x ∂y ∂x2 ∂y2
Charge conservation:

∂ ∂ ∂ ∂
σm + σm + S = 0, (44)
∂x ∂x ∂y ∂y
where ε is porosity (for gas channel ε equals 1), S is the source term.
Table 2 shows the values for different region of fuel cell. Xk is the mole fraction
of species k, Dk is diffusivity of species k.
Table 2: Source terms for the above governing equations.

Sx Sy Sk S
Gas channel 0 0 0 0
µ 2 µ 2
Gas diffuser − ε u − ε v 0 0
K K
ja
For H2 − 2Fctotal For anode ja
µ ∂ µ ∂ jc
Catalyst layer − εc u + E − ε c v + E For H2 O 2Fctotal For cathode jc
Kp ∂x Kp ∂y
jc
For O2 − 4Fctotal
µ ∂ µ ∂
Membrane − ε m u + E − εm v + E 0 0
Kp ∂x Kp ∂y
where E = KKp zf cf F, K is electro-kinetic permeability, Kp is hydraulic per-

meability of membrane, zf is fixed site charge, cf fixed charge concentration, and
F is Faraday constant. ctotal is the total species concentration. ja and jc are transfer
current densities defined by the Bulter-Volmer equations:

a c
XO2 αa F αa F
ja = aj0 ref
exp ηc − exp − ηC , (45)
XOref2 RT RT
1/2 a c
XH2 αa F α F
jc = aj0 ref
exp ηa − exp − a ηa , (46)
XHref RT RT
2
where σm is the proton conductivity in the membrane, which was correlated by

Springer et al. [8] as:

1 1
σm = exp 1268 − (0.005139λ − 0.00326). (47)
303 T
The water content λ, in eqn (47) can be expressed as follows [25]:
λ = 0.043 + 17.81A − 39.85A2 + 36A3 for 0 < A < 1,

(48)
λ = 14 + 1.4(A − 1) 1 ≤ A ≤ 3.
Note that A is defined as the vapor activity at the cathode gas diffuser catalyst layers
interface assuming thermodynamic equilibrium, which is given by
Xw p
A= . (49)
psat
Here Xw is the water vapor molar fraction. The saturated water partial pressure,
psat is expressed by the following empirical equation.
log10 psat = −2.1794 + 0.2953T − 9.1837 × 10−5 T 2 + 1.4454 × 10−7 T 3 .

(50)
Once the electrical potential is determined in the membrane, the local current
density can be calculated by:
∂e
i(y) = −σm |interface . (51)
∂x
Then the averaged current density is determined as follows:
!
1
iavg = i(y)dy. (52)
L mem
The cell voltage can be calculated as:
E = E0 − |ηa | − |ηc | − iavg /σm , (53)
where E0 is the reference open-circuit potential of the fuel cell, which can be
expressed as a function of temperature [25]:
E0 = 0.0025T + 0.2329. (54)


The boundary conditions for u, v, p, XH2 , XH2 O , XO2 , e at inlet of the channel
are fixed.
uin,anode = uo− , vin,anode = 0, uin,cathode = uo+ , vin,cathode = 0,

−
pin,anode = po− , pin,cathode = po+ , XH2 O,anode = XHo2 O ,
+
XH2 O,cathode = XHo2 O ,
where superscript or subscript o− and o+ represent inlet condition at anode and

cathode, respectively.
All velocities at solid walls are set to be zero due to no-slip conditions. The bound-
ary condition for the electric potential is no-flux everywhere along the boundaries
of computational domain. At the outlet, both channels are assumed sufficiently long
that the velocity and species concentration are fully developed.
2.2.4 Numerical procedures

Gurau et al. [23] solved these governing equations in three different domains:
the cathode gas channel-gas diffuser-catalyst layer for air mixture, the cathode gas
diffuser-catalyst layer-membrane-anode catalyst layer-gas diffuser for liquid water,
and the anode gas channel-gas diffuser-catalyst layer for hydrogen. The continuity
equations and momentum equations for the gas mixture were solved first in the
coupled gas channel and gas diffuser domains. Then the species concentration
equations together with the Butler-Volmer equations were solved iteratively. After
convergence is achieved, the transport equations related to water vapor flow as
well as the equations for cell potential and current density can be solved. Since
the source term and boundary conditions for the two domains have to be matched
at the interface, new level iterations have to be introduced. When they solved the
electrochemical cell efficiency, they assumed the total overpotential first and then
calculated the transfer current density together with other unknowns. Once the
correct current density is found, the ohmic losses can be calculated, and the cell
potential is set to be the open potential minus the sum of the total over-potential,
and the total ohmic losses.
Um et al. [25] solved the governing equations in a single domain. Although
some species are practically non-existing in certain regions of a fuel cell, the species
transport equations can still be applied throughout the entire computational domain
by using the large source term technique, which is often assigned a sufficiently large
source term in this sub-region, that effectively freezes the non-exsiting species mole
fraction to zero.

Figure 7 shows the basic profile of fuel cell potential-current density characteris-
tic for a 2-D model solution [26]. The potential loss generally has three sources:
(1) activation polarization (act), (2) ohmic polarization (ohm), and (3) concentra-
tion polarization (conc). The activation polarization loss is dominant at low current
Figure 7: Fuel cell voltage-current density.
Figure 8: Effect of the gas diffuser porosity on the voltage-current density charac-
teristic.
density. At this point, electronic barriers have to be overcome prior to current and
ion flow. Activation losses increase slightly as current increases. Ohmic polariza-
tion (loss) varies directly with current, increasing over the whole range of current
because cell resistance remains essentially constant when temperature does not
change too much. Gas transport losses occur over the entire range of current den-
sity, but these losses become prominent at high limiting currents where it becomes
difficult to provide enough reactant flow to the cell reaction sites.
Figure 8 demonstrated the effect of gas diffuser porosity on fuel cell performance
presented by Gurau et al. [23]. In their work, an isothermal condition at T = 353 K
and an inlet air velocity U0 = 0.35 m/s with 100% humidity were assumed. For
lower values of porosity, lower values of the limiting current density were found.
The cell performance in the region where the concentration over-potentials are
predominant was also predicted.
Figure 9: Effect of the air inlet velocity on the performance.
Figure 9 presents the current density for different air inlet velocity at the cathode
gas channel, assuming the gas diffuser porosity ε = 0.4 and a constant temperature
T = 353 K with 100% humidity. For higher inlet air velocity, more oxygen is fed;
therefore, more oxygen is likely to arrive at the catalyst layer, with the result of
a higher limiting current density. This is explained by the fact that for the same
pressure field, the axial momentum transfer across the gas channel-gas diffuser
interface becomes more important for higher velocities. This is a consequence when
more “fresh” air is able to arrive at the catalyst layer. For the inlet air velocity higher
than approximately 2 m/s, the limiting current becomes constant, which shows that
there is a limiting effect of the momentum transfer across the gas channel-gas
diffuser interface.
Figure 10 shows the oxygen mole fraction field in the cathode gas channel-
gas diffuser coupled domain for a case of current density Iavg = 2.89 × 103 A/m2 ,
Ui = 0.35 m/s, and T = 353 K (iso-thermal case) with humidity = 100%. The
oxygen is consumed in the catalyst layer. The mole fraction is decreased along the
flow direction. The oxygen consumption depends on the operating current density.
The higher current density, the more oxygen is consumed, thus, the faster mole frac-
tion decreases along the flow direction. A limiting current density may occur when
the oxygen or hydrogen is completely depleted at the reaction surface. The hydro-
gen mole fraction field in the anode gas channel-gas diffuser coupled domain has a
similar distribution as the oxygen distribution in cathode side. In the present model,
it is assumed that water only exists in the vapor state. However, if the electrochem-
ical reaction rate is sufficiently high, the amount of water produced is condensed
into liquid phase. Under this situation, two-phase flow model has to be considered.
2.2.6 Concluding remarks

The 2-D model is a significant improvement over the 1-D models and could provide
simulations that are more realistic. It can predict the transport phenomena in the
Figure 10: Oxygen distribution in the cathode gas channel and gas diffuser.
entire fuel cell sandwich, including the gas channels. No assumptions are necessary
for the distribution of the species concentrations or current density. The input data
are only those parameters that can be controlled in real fuel cell applications.
2.3 Three-dimensional (3-D) model
Recently, a few research groups extended 2-D model to 3-D model such as Shim-
palee and Dutta [27], Zhou and Liu [28], Um and Wang [29], and Jen et al. [30].
One of the major improvements of the 3-D model over the 2-D model is its ability to
study the blocking effect of the collector plates and the effectiveness of the interdig-
itated flow field [31]. Here we introduce a generalized 3-D model as well as some
numerical approaches and results. To solve the 3-D model, many different numer-
ical method have been used such as CD-Star [24], FLUENT [27], Semi-Implicit
Method [28], and Vorticity-Velocity Method [30]. In this section, 3-D formulations
are developed in such a way that they allow using the same code for solving the
Navier-Stokes equations of the gas channel, gas diffuser and catalyst layers in a
coupled domain. The potential equation and species concentration equations are
solved in a single domain of the whole fuel cell. The solutions of the hydrodynam-
ics of the flow and polarization curves are analyzed and presented in details. The
results of this study may be beneficial for further and more complete analyses of
the performance of fuel cells.
2.3.1 Model development

A typical PEM fuel cell configuration is shown in Fig. 11. The physical fuel cell
model consists of anode gas channel, anode gas diffuser that formed by porous
Figure 11: The schematic of PEMFC.
media, anode catalyst layer, membrane, cathode catalyst layer, cathode gas diffuser,
and cathode gas channel. In reality, when the fuel cell works, fuel and oxidant can
be viewed as a steady, laminar, developing forced convection flow in an isothermal
rectangular channel, and penetrating through the gas diffuser to catalyst layers.
In the model presented here, the flow is assumed to be steady, constant proper-
ties, and incompressible. The viscous dissipation, compression work and buoyancy
are assumed negligible. The gas mixtures are considered as perfect gases, and the
species concentrations are considered as constant at the inlet of the channel. The
concentrations along the gas channel and the gas-diffuser will vary due to diffusion-
convection transport and electron kinetics in catalyst layers, and the distributions
will depend on the gas properties and reaction rate. Water transport in and out of
the electrodes is assumed to be in the form of vapor only. This assumption may be
questionable in this model, in particular when the reactants flow into the gas chan-
nel under saturated conditions. The water generation rate is very likely to exceed
its removal rate and thus condensation formed in the cathode. As a result, two-
phase flow forms in the cathode channel. This complex case is, however, neglected
in this chapter. The gas-diffuser, catalyst layers, and the membrane materials are
considered as isotropic porous media. Contraction of the porous media is also
neglected.
2.3.2 Mathematical model

The governing equations base on conservation of mass, momentum, energy and
species.
The model governing equations can be written as:
Mass conservation:
∂u ∂v ∂w
+ + = 0. (55)
∂x ∂y ∂z
Momentum conservation:
2
ρ ∂u ∂u ∂u ∂p ∂ u ∂2 u ∂2 u
u +v +w = −ε + µ + 2 + 2 + Sx , (56)
ε ∂x ∂y ∂z ∂x ∂x2 ∂y ∂z
2
ρ ∂v ∂v ∂v ∂p ∂ v ∂2 v ∂2 v
u +v +w = −ε + µ + + + Sy , (57)
ε ∂x ∂y ∂z ∂y ∂x2 ∂y2 ∂z 2
2
ρ ∂w ∂w ∂w ∂p ∂ w ∂2 w ∂2 w
u +v +w = −ε + µ + + + Sz . (58)
ε ∂x ∂y ∂z ∂z ∂x2 ∂y 2 ∂z2
Species conservation:
2
1 ∂Xk ∂Xk ∂Xk ∂ Xk ∂ 2 Xk ∂2 Xk
u +v +w = εDk + + + Sk . (59)
ε ∂x ∂y ∂z ∂x2 ∂y2 ∂z2
Charge conservation:

∂ ∂ ∂ ∂ ∂ ∂
σm + σm + σm + S = 0, (60)
∂x ∂x ∂y ∂y ∂z ∂z
where ε is porosity, for gas channel ε equals 1, for gas diffuser ε = εd , for catalyst
layer ε = εc , and for membrane ε = εm , and S is the source term. Table 3 shows
the source terms for different region of fuel cell.
The parameters of above governing equations are the same as the 2-D model.
The boundary conditions and numerical procedures are similar to the 2-D case and
will not repeat here.
Here we introduce a new method, which was used by Jen et al. [30], to sim-
plify the 3-D model by making the usual parabolic assumption. With the parabolic
2 2 2
assumption, the diffusion term in axial direction such as ∂∂zu2 , ∂∂z v2 , ∂∂zw2 , as well as
∂ 2 Xk
∂z 2
can be neglected. Furthermore, the modified pressure P may be defined as
P(x, y, z) = p(z) + p∗ (x, y), (61)
where p(z) is the pressure over the cross section at each axial location, and p∗ (x, y)
is the pressure variation in the x, y direction, which drives the secondary flow.
Pressure gradient for axial direction:
∂P ∂p ∂p∗
= + , (62)
∂z ∂z ∂z
∗
where ∂p ∂p
∂z ∂z is due to the useful parabolic assumption. So the axial pressure
gradient can be written as:
∂P ∂p
= = f (z). (63)
∂z ∂z
Table 3: Source terms for the above governing equations.
Sx Sy Sz Sk Sφ
Gas channel 0 0 0 0 0
µ 2 µ 2 µ
Gas diffuser − ε u − ε v − ε2 w 0 0
K K K
ja
For H2 − For anode ja
2Fctotal
µ ∂e µ ∂e µ ∂e jc
Catalyst layer − εc u + E − εc v + E − εc w + E For H2 O For cathode jc
Kp ∂x Kp ∂y Kp ∂z 2Fctotal
jc
For O2 −
4Fctotal
µ ∂m µ ∂m µ ∂m
Membrane − εm u + E − εm v + E − εm w + E 0 0
Kp ∂x Kp ∂y Kp ∂z
And we also have:

∂P ∂p∗
= , (64)
∂x ∂x
∂P ∂p∗
= . (65)
∂y ∂y
With the above parabolic flow assumptions, it is now possible to simply march
through the computational domains in the main flow direction without worrying
about the downstream conditions as those in elliptic flow cases. A novel vorticity-
velocity method was used here to solve this problem. The axial vorticity function
can be defined as:
∂u ∂v
ζ= − . (66)
∂y ∂x
Applying it to continuity eqn (55) for u and v, respectively.
∂ζ ∂2 w
∇ 2u = − , (67)
∂y ∂x∂z
∂ζ ∂2 w
∇ 2v = − − . (68)
∂x ∂y∂z
Cross differentiation of x and y momentum equations of gas channel to eliminate

pressure terms yield:

∂ζ ∂ζ ∂ζ ∂u ∂v ∂w ∂u ∂w ∂v
u +v +w +ζ + + −
∂x ∂y ∂z ∂x ∂y ∂y ∂z ∂x ∂z
2
µ ∂ ζ ∂2 ζ
= 2
+ 2 . (69)
ρ ∂x ∂y
Using the same method we can get similar equation for gas diffuser

1 ∂ζ ∂ζ ∂ζ ∂u ∂v ∂w ∂u ∂w ∂v ε2 µ
u +v +w +ζ + + − + ζ
ε ∂x ∂y ∂z ∂x ∂y ∂y ∂z ∂x ∂z κ

µ ∂2 ζ ∂2 ζ
= + 2 . (70)
ρ ∂x2 ∂y
An additional constraint, which will be used to determine f (z), is that global mass
conservation must be satisfied as follows:
!!
(1 + r)2
wdxdy = Uin De2 , (71)
4r
where r = ab is the aspect ratio of the gas channel, Uin is inlet velocity, De is
hydraulic diameter (4A/S).

Boundary conditions at the gas channel entries, such as gas mixture velocities,
pressure, and component concentrations, are specified. No slip boundary conditions
are specified at the gas channel walls. At the interface between the gas channel
and electron collectors, the boundary conditions of components concentration are
assumed as:
∂Xi
= 0. (72)
∂n
For the membrane potential equations, the boundary conditions are:
∂
= 0. (73)
∂y
This means that no proton current leaves top and bottom boundary. In x-direction,

∂

anode = 0, = 0. (74)
catalyst ∂x cathode
surface catalyst
surface
No boundary conditions are needed at the interface between gas channel and gas
diffuser because they are coupled in a single domain.
2.3.4 Discretization strategies

In order to solve the above equations, some terms need to be discretized furthermore.
The strategies of discretization is as follows:
The values of ∂w ∂w ∂u ∂v
∂x , ∂y , ∂x , and ∂y are discretized using central differencing at
∂u ∂v ∂
each grid point. The values of ∂z and∂z , ∂z are computed by two points backward
∂2 w ∂2 w∂ζ ∂ζ
differencing. The values of ∂x∂z , ∂y∂z , ∂x and ∂y are calculated by using backward
differences axially and central differences in the transverse directions.
The values of vorticity on the boundary walls can be evaluated by following
equations.
1 1 v2, j
ζ1 1 ,j = (ζ1, j + ζ2, j ) = (u1 1 , j+1 − u1 1 , j−1 ) − , (75)
2 2 2y 2 2 x
ζ1, j = −2v2, j /x + (u2, j+1 − u2, j−1 )/(2y) − ζ2, j , (76)
where u1 1 , j+1 = 12 u2, j+1 ; u1 1 , j−1 = 12 u2, j−1

2 2
2.3.5 Solution algorithms

1. Solve velocity and pressure field for anode domain first. The initial values of
the unknowns u, v, and ζ are assigned to be zero at the entrance, z = 0. Uniform
inlet axial velocity (i.e., w = 1) is used. Note that ζ = 0 at z = 0 results from
the vorticity definition.
2. Discretizing eqns (69), (70) with power-law scheme [33] and solving them in
the coupled domain. Vorticity ζ can be calculated for gas channel, gas diffuser
and catalyst layer.
3. The elliptic-type eqns (67) and (68) are solved for u and v iteratively. During
the iteration process, the values of vorticity on the boundaries are evaluated
using eqn (76).
4. Using control volume method with power-law scheme to discretize momentum
equation in z direction, plug in new values u and v solved in step 2, one can solve
axial velocity w in coupled domain with constraint (71) to meet the requirement
of the constraint flow rate.
5. Steps 3 – 4 are repeated at a cross section until the following convergence
criterion is satisfied for the velocity components u and v:

n
max ui,n+1
j − ui, j
< 10−5 , (77)
n+1
max ui, j
where n is the nth iteration of steps 3 – 4.

6. Repeat steps 1– 5 for cathode domain. One can calculate velocity distributions
for gas channels, gas diffusers and catalyst layers in cathode domain.
7. With obtained solutions u, v, and w for both anode and cathode channels,
species concentration equations and potential equations for a single domain
from anode electrolyte to cathode electrolyte can be solved iteratively. These
steps are repeated until the following convergence criterion is satisfied:


max Xkm+1 − Xkm m+1 − m
max  ,  < 10−6 , (78)
m+1 m+1
max Xk
where m is the mth iteration of step 6, and k is the kth species.

8. Steps 3 –7 are repeated at the next axial location until the final z location is
reached.
9. Once the electrolyte potential is obtained, the local current density can be
calculated along the axial direction using following equation:
∂
I ( y, z) = −σm . (79)
∂x
The average current density is then determined by
! b ! L
11
Iavg = dy I ( y, z)dz. (80)
bL 0 0

3-D models are generally more accurate and more detailed in transport phenomena
than 2-D models. Jen et al. [30] investigated the secondary flow patterns in fuel cell
channels. Zhou and Liu [28] analyzed detailed temperature distribution in fuel cell.
Um and Wang [29] studied the effectiveness of the interdigitated flow field on fuel
cell performance. Figure 12 shows the comparison of Zhou and Liu’s [28] 3-D
model numerical results with experimental data [32]. The operating conditions are:
cathode flow rate is 1200 cm3 /s; anode flow rate is 1200 cm3 /s; cathode temperature
Figure 12: 3-D Comparison of computational result with experimental data.
is 60 ◦ C; anode temperature is 60 ◦ C; cathode pressure is 3 atm; anode pressure is

1 atm. It can be found that the 3-D model numerical results for the polarization
curve agree well with the experimental data except in the region of the concentration
polarization loss dominate. Even if some discrepancies can be seen in this region,
the basic trend is still good. Note that the polarization of concentration region was
not given in Ticianelli et al. [11, 12] experimental investigations. Most early 1-D
or 2-D models did not investigate this region because no experimental data are
available for this region.
2.3.6.1 Secondary flow pattern In Jen et al’s 3-D model [30], a general sec-
ondary flow pattern in fuel cell channel were also investigated as shown in Fig. 13.
At the location z̄ = 0.0005, as shown in Fig. 13(a), the secondary flow moves out
from porous media in the core region. This is simply due to the acceleration of
the core gas channel flow and the strong porous media resistance to push the fluid
out of the gas diffuser in the early entrance region. Figure 13(b) shows the vector
secondary flow pattern at z̄ = 0.008, which has two pairs of vortices, with one
small pair counter rotating cells near the corner of the gas channel. It is observed
that, near the core region, a fairly uniform secondary flow running from the left
side to right side into the gas diffuser, and there are outflows from the gas dif-
fuser to gas channel near the top and bottom wall. It is also interesting to see two
counter rotating cells at the top and bottom left corner, which are generated due to
the corner effect of the gas channel. In nearly fully developed region at z̄ = 0.2
(Fig. 13(c)), the secondary flow strength has decreased significantly, and the effect
of cross-sectional convection is negligible.
2.3.6.2 Temperature distributions Zhu and Liu [28] analyzed the temperature
distribution in PEMFC. The temperature distributions across the whole fuel cell
Figure 13: Secondary flow vector plots.
sandwich depend on the heat generation of the chemical reaction, Joule heating of
the current, and cooling from channel walls. Due to very effective cooling, constant
cell wall temperature is assumed; the typical temperature distribution is shown in
Fig. 14 as an illustrative case. The inlet temperature of the air and fuel are both
assumed to be 82 ◦ C, inlet pressure at anode/cathode is 1 atm/3 atm. The major heat
source is the chemical reaction and Joule heating in the cathode side catalyst layer.
The maximum temperatures are located within the cathode side catalyst layer near
air entrance region, as shown in Fig. 14. This is due to the high chemical reaction
rate. The outer walls of the gas channels are kept at the constant temperature of
82 ◦ C, heat is transferred out of the fuel cell effectively by the cooling along the
outer walls of the gas channel so the inside of the fuel cell is prevented from
overheating. The temperature distribution profiles can be used to improve fuel cell
design to avoid membrane burning and drying out.
2.3.6.3 Water concentration variation Water management is very important for

PEM fuel cell performance. During fuel cell operation, water within membrane is
driven from the anode side to the cathode side by electro-osmosis, and at the same
time it is driven in the opposite direction by diffusion [28]. If the water generation
is more than water transport from the cathode, it causes cathode side flooding. On
the other hand, if the water loss from anode is more than water supplies, it causes
membrane dehydration which leads to high ohmic losses. The conductivity of the
polymer electrolyte is a strong function of the degree of hydration, and flooding of
Figure 14: Temperature distributions across the fuel.
the cathode. It is common practice to saturate inlet fuel and air streams. However,
due to the water generation along the cathode and electrical-osmosis drag from
the anode to cathode, the water vapor concentration increases along the cathode
channel. As the inlet cathode stream is already fully saturated, the added water
from the chemical reaction and osmosis drag causes the cathode side to become
over-saturated. In other words, liquid water exists along the cathode side and this
could lead to flooding. To avoid this phenomenon, it is then suggested that the
water content in the cathode inlet stream should be reduced. Figures 15 and 16
show the water vapor distributions along the cathode side and anode side when
the hydrogen stream at the anode is fully saturated and the inlet air stream at the
cathode is completely dry. Water vapor concentration along the cathode is increased
due to the combined effect of water generation from the chemical reaction and
electrical-osmosis drag from the anode to cathode as shown in Fig. 15. Under such
operation conditions, the water vapor concentration along the anode is reduced due
to the electrical-osmosis drag as shown in Fig. 16. Under this operation condition,
membrane flooding can be avoided. However, the water vapor concentration near
the inlet of the cathode is very low, the membrane dehydration may occur. Therefore,
a proper water vapor added to the cathode stream and fully saturated anode stream
are beneficial for fuel cell performance.
2.3.6.4 The effect of the flow fields Um and Wang [29] investigated the effects of
conventional flow fields and interdigitated flow fields on fuel cell performance. Con-
ventional flow is usually straight streamtraces due to them all having open inlet and outlet.
In contrast to conventional flow fields, interdigitated flow channels indicate that air
comes in along the lower channel, penetrates through the porous backing layer and
Figure 15: Water vapor mole fraction varies along the flow direction in cathode
channel.
Figure 16: Water vapor mole fraction varies along the flow direction in the anode
channel.
exits through the upper channel. During the process, more oxygen is brought to the
cathode catalyst reaction site by the forced convection, which results in better cell
performance. Figure 17 shows the effect of the interdigitated flow field on the cell
performance. There is a little difference in the current densities until cell potential
reaching 0.55 V because the cell potential loss for current densities below 1 A/cm2 is
primarily dominated by the ohmic resistance. However, for Iavg > 1 A/cm2 , the cell
polarization curve begins to be limited by mass transport. In this region the positive
Figure 17: The polarization curve for conventional and interdigtitated flow fields
at 353 K.
role of the interdigitated flow field becomes apparent. It is seen that the mass trans-
port limiting current density is much improved by the use of interdigitated flow field.
2.4 Summary and conclusion
The numerical simulation presented here enables prediction of phenomena in the

entire fuel cell sandwich, including the two gas flow channels, two gas diffusers, two
catalyst layers and membrane. It is able to predict detailed distributions of velocity
fields, species concentration, current density, temperature, and polarization curves.
It can be used to understand the interacting, complex electrochemical and transport
phenomena that cannot be visualized experimentally. It also provides the ways to
increase fuel cell power output for future fuel cell design, such as increase cell
temperature, pressure, decrease the membrane thickness, choosing proper water
vapor contents, and flow fields etc.
References
[1] Crowe, B.J. Fuel Cells: A Survey, NASA Rep. (SP-5115), Washington, DC,
1973.
[2] Savinell, R.F. & Fritts, S.D., Theoretical performance of a hydrogen-bromine
rechargeable SPE fuel cell. J. Power Sources, 22, p. 423, 1988.
[3] Ridge, S.J., White, R.E., Tsau, Y., Beaver, R.N. & Eisman, G.A., Oxygen
reduction in a proton exchange membrane test cell. J. Electrochem. Soc.,
p. 136, 1902, 1989.
[4] Fritts, S.D. & Savinell, R.F., Simulation studies on the performance of the
hydrogen electrode bonded to proton exchange membranes in the H2 -Br2
fuel cell. J. Power Source, 28, pp. 301–315, 1990.
[5] Yang, S.C., Cutlip, M. B. & Stonehart, P., Further development of an approxi-
mate model for mass transfer with reaction in porous gas-diffusion
electrodes to include substrate effects. Electrochem. Acta, 34, p. 703, 1989.
[6] Verbrugge, M.W. & Hill, R. F., Ion and solvent transport in ion-exchange
membranes. J. Electrochem. Soc., 137, pp. 886, 1990a.
[7] Bernardi, D.M. & Verbrugge, W.M., Mathematical model of a gas diffusion
electrode bonded to a polymer electrolyte. AIChE, 37, pp. 1151–1163, 1991.
[8] Springer, T.E., Zawodinski, T.A. & Gottesfeld, S., Polymer electrolyte fuel
cell model. Electrochem. Soc., 136, pp. 2334–2341, 1991.
[9] False, J.L., Vanderborgh, N.E. & Stroeve, P., The influence of channel geome-
try on ionic transport, Diagrams. Separators, and Ion-Exchange Membranes,
86, pp. 13, 1986.
[10] Fuller, T.F. & Newman, J., Water and heat management in solid-polymer-
electrolyte fuel cell. J. Electrochem. Soc., 140, pp. 1218–1225, 1993.
[11] Ticianelli, E.A., Derouin, C. R., Redondo, A. & Srinivasan, S., Methods
to advance technology of proton exchange membrane fuel cells.
J. Electrochem. Soc., 135, pp. 2209–2214, 1988a.
[12] Ticianelli, E.A., Derouin, C. R. & Srinivasan, S., Localization of platinum
in low catalyst loading electrodes to attain high power densities in SRE fuel
cells. J. Electroanal. Chem., 251, pp. 275–295, 1988b.
[13] Newman, J., Electrochemical Systems, Prentice-Hall: Englewood Cliffs, NJ,
1973.
[14] Bard, A.J. & Faulkner, L.R., Electrochemical Methods, Wiley: New York,
1980.
[15] Nernst, W., Die elektromotorische wirksamkeit der Jonen: I. Theorie der
diffusion. Z. Physic Chem., 4, pp. 129, 1889.
[16] Planck, M., Ueber die erregung vol electricitat und warme in electrolyten.
Ann, d. Phys. U. Chem., 39, p. 161, 1890.
[17] Bird, R.B., Stewart, W.E. & Lightfoot, E.N., Transport Phenomena, Wiley:
New York, 1960.
[18] Srinivasan, S., Manko, D.J., Koch, H., Enayetullab, M.A. & Appleby, J.A.,
J. Power Sources, 29, p. 367, 1990.
[19] Fuller, T.F. & Newman, J., Water and thermal management in solid-polymer-
electrolyte fuel cells. J. Electrochem. Soc., 5, pp. 1218, 1993.
[20] Nguyen, T.V. & White, R.E., A water and heat management model for
proton-exchange membrane fuel cells. J. Electrochem. Soc., 140, pp.
2178–2186, 1993.
[21] Amphlett, J.C., Baumert, R.M., Mann, R.F., Peppley, B.A. & Roberge, P.R.,
Performance modeling of ballard mark IV solid polymer electrolyte fuel cell.
J. Electrochemical Society, 142, pp. 1, 1992.
[22] Yi, J.S. & Nguyen, T.V., An along-the-channel model of PEM fuel cells,
J. Electrochem. Soc., 145, pp. 1149–1159, 1998.
[23] Gurau, V., Liu, H.T. & Kakac, S., Two dimensional model for proton
exchange membrane fuel cells. AIChE Journal, 44, pp. 2410–2422, 1998.
[24] Yi, J.S. & Nguyen, T.V., Multi-component transport in porous electrodes of
proton exchange membrane fuel cells using inerdigitated gas distributors. J.
Electrochem. Soc., 146, pp. 38–45, 1999.
[25] Um, S. & Wang, C.Y., Computational fluid dynamics modeling of proton
exchange membrane fuel cells. Journal of The Electrochem. Soc., 147, pp.
4485–4493, 2000.
[26] Fuel Cell Hand Book, Fifth edition, EG&G Services Parsons, Inc. Science
Applications International Corporation, 2000.
[27] Shimpalee, S. & Dutta, S., Effect of humidity on PEM fuel cell performance.
Part-II Numerical Simulation, HTD, 364,1. Heat Transfer Division. ASME
1999.
[28] Zhou, T. & Liu, H., A general three-dimensional model for proton
exchange membrane fuel cells. I.J. Trans. Phenomena, 3(3), pp. 177–198,
2001.
[29] Um, S. & Wang, C.Y., Three dimensional analysis of transport and reaction
in proton exchange membrane fuel cells. Proceedings of the ASME Fuel
Cell Division – 2000: The 2000 ASME 5, 10, Walt Disney World Dolphin,
Orlando, FL.
[30] Jen, T. C., Yan, T.Z. & Chan, S.H., Chemical reaction transport phenomena
in a PEM fuel cell. International Journal of Heat and Mass Transfer, 46, pp.
4157–4168, 2003.
[31] Wood, D.L., Yi, J.S. & Nguyen, T.V., Effect of direct liquid water injec-
tion and interdigitated flow field on the performance of proton exchange
membrane fuel cells. Electrochem. Acta, 43, pp. 3795–3809, 1998.
[32] Huang, Z., Experimental and mathematical studies for PEM fuel cell per-
formances, M.S. thesis, University of Miami: Coral Gables, Florida.
[33] Patankar, S. V., Numerical Heat Transfer and Fluid Flow, Hemisphere:
Washington, DC, 1980.
CHAPTER 7
Mathematical modeling of fuel cells: from

analysis to numerics
M. Vynnycky & E. Birgersson
FaxénLaboratoriet, Royal Institute of Technology, Stockholm, Sweden.
Abstract
The mathematical modeling of transport phenomena in fuel cells is rendered dif-
ficult by the vast array of effects that have to be considered: flows are often
highly three-dimensional, non-isothermal, multiphase, multicomponent and time-
dependent, and occur over several media: flow channels, porous electrode, catalytic
layer and electrolyte. Contrary to the most recent trends in fuel cell modeling,
which typically involve extensive 3D computational fluid dynamics, the approach
we advocate here involves the use of scaling arguments, nondimensionalization
and asymptotic techniques to identify the main governing parameters in a model
and, subsequently, to derive a reduced model. The benefit of this is a model that
is considerably cheaper to compute, but which does not sacrifice any significant
physical features. These ideas are illustrated initially for the case of 2D momentum
and multicomponent mass transfer in the performance-limiting electrode of a poly-
mer electrolyte fuel cell (PEFC) and a direct methanol fuel cell (DMFC). For the
second case, quantitative comparison is provided of the computed results from, and
the computation times for, the full and reduced models; all in all, good agreement
is achieved, but with two orders of magnitude less computing time being required
for the reduced formulation.
1 Introduction
The recent explosive increase in interest in fuel cells has been accompanied by an
increase in the use of mathematical modeling as a tool, both to interpret experimental
results and to provide insight on how to improve fuel cell design and performance.
The task is exacerbated by the vast array of physical phenomena that can occur
simultaneously: for example, in the operation of a typical seven-layer polymer
electrolyte fuel cell (PEFC), consisting of two sets of flow channels, an anode and
Figure 1: Heart of a PEFC.

cathode porous electrode (often referred to as porous backings or gas diffusion
layers), anode and cathode catalytic layers (also referred to as active layers) and a
polymer electrolyte as membrane shown in Fig. 1, some, or all, of the following
phenomena are thought to be of importance:
• water transport;
• phase change in the porous backings and flow channels;
• conjugate heat transfer between bipolar plates/cooling channels and bipolar
plates/flow channels;
• two-phase (gas and liquid) flow in the cathode porous backing and flow channels;
• electron transport in the cathode porous backing;
• ohmic heating in the membrane and the catalytic layers;
• water production at the cathode catalytic layer;
• convective heat/mass transfer in the flow channels;
• electro-osmotic drag in the membrane;
• electrochemical reactions at the catalytic layers;
• diffusion of H2 /O2 through the membrane;
• proton transport in the membrane;
• heat generation in the catalytic layers;
• Knudsen diffusion in the porous backing.
In addition, flows can often be highly three-dimensional, as is evident from the
figure, and fuel cell operation for a variety of applications can be time dependent.
The state of the art in the modeling of transport phenomena in fuel cells is
widely recognized to involve the use of 3D computational fluid dynamics (CFD)
to describe flow in the channels and porous backings, coupled to models for the
electrochemical reactions in the anode and cathode catalytic layers, as well as water
transport through the membrane; most recent examples of this for PEFCs are [1–6]
with, in several cases, the implementation of computational models in commercially
available software [7–10].
There can be little doubt that 3D CFD, provided it is carried out correctly, is able
to provide, geometrically at least, the most detailed model description of transport
phenomena in a fuel cell. It does, however, have its drawbacks, mostly associated
with lengthy computing times:
1. To solve the model equations numerically for one combination of operat-
ing parameters can be prohibitively expensive; furthermore, there are often
many possible such combinations (humidification, temperature, anode/cathode
pressure, stoichiometry), and there may well be numerous model constants
also, particularly for multi-phase flows where empirical relations have to be
used.
2. Often, it is stacks of cells, rather than individual cells, that are of interest; this
increases the number of mesh points required for a numerical solution.
3. It is difficult to incorporate this approach into system studies, the model equa-
tions for which can be solved several orders of magnitude more quickly.
4. The simulation of cells operating in transient mode is also often of most interest.
5. There is no real possibility to use the models further, e.g. as the basis for fuel
cell control strategies or for optimisation studies, again because of the lengthy
computing times that this would entail.
6. It would be difficult, if not impossible, to identify model simplifications that
could be exploited to reduce computing times.
There is, therefore, a need for modeling that describes the physics that 3D CFD
describes, but without the prohibitively large computational requirements. Prior to
the use of CFD, this was first done for PEFCs through one-dimensional models
[11–14], and later through pseudo-two-dimensional ‘along-the-channel’ models
[15–18]. The respective disadvantages of these approaches are:
1. One-dimensional approaches, whilst they are able to address some aspects of
the three issues related to fuel cell performance mentioned above, are not able
to address these questions at a local level: that is to say, where oxygen depletion
occurs, where there is flooding or inadequate heat removal.
2. ‘Along the channel’ models result in ordinary differential equations with the
coordinate along the fuel cell as the independent variable and do not satisfy
mass, momentum and energy balance locally, although they may do so globally;
consequently, they are formally valid only in a global sense.
In both cases, almost all studies have been carried out in dimensional variables;
it is therefore difficult to establish whether model simplifications are possible, or
what the relative importance of competing physical effects might be.
In this article, we fill the niche that exists between the types of modeling indi-
cated above by highlighting the power of a reduced model approach. How this
is carried out will be shown in later Sections 2 and 3, but the general idea will
be to take the full model equations, e.g. as would be implemented in a CFD
model, nondimensionalize, identify key dimensionless parameters and with luck (if
there exist parameters which are much larger or smaller than one) reduce the
model’s complexity, but without reducing its practical usefulness. It should be
emphasized that, in the present context, the reduced model approach is not the
same as a simple reversion to the non-CFD fuel cell models mentioned earlier
[11–18]; rather, it starts with an analysis of the full equations, and reduces these to
a formulation whose solution, if not obtainable analytically, is then almost invari-
ably less computationally-demanding than the original one. The emphasis then
is on making the behavior of a complex model more transparent by means of
mathematical analysis; through this approach, time spent doing ‘pen-and-paper’
analysis at an early stage of model development is recouped later through a much
clearer understanding of the model’s structure, as well as a much faster solution
time.
The development presented here reflects our own efforts in establishing reduced
nondimensional models for steady state multicomponent flow in slender fuel cell
geometries, with a particular focus on the hydrodynamics in the performance-
limiting electrode of a particular type of cell: cathode for the PEFC and anode
for the DMFC. The essence of the mathematical approach advocated can, how-
ever, be used for more general problems: the whole cell, multiphase flow, transient
behavior, etc. We begin with a formulation for gaseous flow in the cathode of a
PEFC, from which we derive a reduced model. Subsequently, we show that that
model itself can be reduced further to describe flow in the anode of a DMFC.
In both cases, we find algebraic relations between certain variables that would
not have been apparent had a dimensional formulation been used. Furthermore,
the reduced model equations are parabolic and are solved numerically using a
marching scheme; this is found to be considerably faster than a solution to the
full elliptic formulation of the original equations, and not significantly less accu-
rate. The work is based on already published work [19, 20], and on work still in
progress [21].
2 PEFC
A schematic diagram of the operation of a PEFC is given in Fig. 2. Essentially,
this entails a polymer electrolyte membrane sandwiched between two gas-diffusion
electrodes, which are each adjacent to flow channels contained within bipolar plates.
The oxidant, usually oxygen from air which is either dry or humidified to some
extent, is fed in at the inlet of the channel on the cathode side, and is transported to
the electrolyte/cathode interface; the fuel on the other hand, normally hydrogen, is
fed at the anode channel inlet and is transported to the electrolyte/anode interface.
Figure 2: Schematic of a PEFC.
Both interfaces contain catalyst, often platinum, to accelerate the reactions
2H2 → 4H+ + 4e− at the anode, (1)

+ −
O2 + 4H + 4e → 2H2 O at the cathode, (2)
in the course of which an electric current is produced to drive a given load. In

particular, the reaction at the cathode also produces both heat and water as by-
products, the latter of which may be present throughout the system as either vapor
or liquid, or both; the production of the former can lead to temperatures at the
catalytic layer in the order of 80–90 ◦ C. Optimal fuel cell performance is achieved
at typical voltages of around 0.5 V at current densities of about 1 Acm−2 .
2.1 Mathematical formulation for flow in the cathode
In what follows, we begin by considering a steady 2D flow of a gaseous air/water

mixture in the slender channel and porous backing of a fuel cell.
2.1.1 Channel
Consider flow in a channel of height hf , adjacent to a porous medium of length L
and height hp (see Fig. 3). The equations of continuity of mass and momentum for
the mixture are taken as
∇ · (ρv) = 0, (3)

2µ
∇ · (ρv ⊗ v) = −∇ p + ∇ · v + µ∇ 2 v − ρgj, (4)
3
where ρ, µ, v are the density, viscosity, and mass-averaged velocity of the mixture,
respectively, g is the acceleration due to gravity and j is the unit vector in the
positive y-direction; for later use, it is also convenient to define p , the modified
Figure 3: Cathode of a PEFC.
pressure, given in terms of the pressure p by

2
p = p + µ∇ · v.
3
The continuity equation for each of the species is given in terms of the mass
flux, n, by
∇ · ni = 0, i = O2 , H2 O, N2 , (5)
with
ρ
n
ρMi xi
ni = v+ 2 Mi Mj Dij ∇xj , i = O2 , H2 O, N2 ,
M M
j=1
where xi is the mole fraction of species i, and the second term on the right-hand
side is the mass diffusive flux for an ideal gas mixture [22], due to concentration
diffusion alone, relative to the mass-averaged velocity v. Also, (Mi )i=1,..,O2 ,H2 O,N2
are the molecular weights,
M = MO2 xO2 + MH2 O xH2 O + MN2 xN2 ,
and (Dij )i, j=1,..,O2 ,H2 O,N2 are the multicomponent diffusion coefficients, given by
( ) *+
xk Mk /Mj Dik − Dij
Dij = Dij 1 + , i, j, k = O2 , H2 O, N2 ;
xi Djk + xj Dik + xk Dij

here, the Stefan-Maxwell diffusion coefficients, Dij i, j=1,..,n , are independent of
composition, and can in principle be measured experimentally.
Equation (5) can then be recast in the form of two transport equations,

ρv xO2 ρ ∇xO2
∇· =∇· M , (6)
M xH2 O M2 ∇xH2 O
where

DO2 ,N2 DO2 ,N2 0 MH2 O DO2 ,H2 O
M = M N2 − .
DH2 O,N2 DH2 O,N2 MO2 DH2 O,O2 0
Here, use has been made of the relation xO2 + xH2 O + xN2 = 1 to eliminate xN2 .
For the mixture density, we use the constitutive relation for an ideal gas,
pM
ρ= , (7)
RT
where T is the temperature and R is the universal gas constant (8.314 kg m2 s−2
mol−1 K−1 ). We note, in addition, the possibility that the mixture viscosity, µ, will
not necessarily be constant either, although we have treated it as so here.
2.1.2 Porous backing

For the porous backing, volume-averaging techniques are required, as is care in
distinguishing between intrinsic and superficial quantities. First, let B be a quantity
(either scalar, vector, or tensor) associated with the gas phase, and let the quantity
B be the local volume (or superficial) average of B, defined by
!
1
B ≡ Bd V, (8)
V V (g)
and let B(g) be the intrinsic volume average of B in the gas phase, where
!
(g) 1
B ≡ Bd V. (9)
V (g) V (g)
Also, let γ be the porosity, given by γ = V (g) /V. A comparison of eqns (8)
and (9) shows that the local and intrinsic volume average for the gas phase is
given by
B = γ B(g) . (10)
With these definitions, the governing equations in the porous backing become

∇ · ρ(g) v = 0, (11)
κ , -(g) κ
v = − ∇ p + ρ(g) gj + ∇ 2 v , (12)
µ γ
 
, - 3
, -
( ) (g) ( ) (g )
ρ v xO2  γ 2 ρ
g g
(g) ∇, xO2 - 
∇·
( ) , -(g) = ∇ ·  2 M ( g)  ,
M g
xH 2 O M ( g) ∇ x H2 O
(13)
where
, -(g) , -(g) 
DO2 ,N2 DO2 ,N2
M(g) = MN2 , -(g) , -(g) 
DH2 O,N2 DH2 O,N2
, -(g)
0 MH2 O DO2 ,H2 O
− , -(g) ,
MO2 DH2 O,O2 0
with
 ) * 
, -(g)  xk (g)
Mk /Mj Dik − Dij 
Dij = Dij 1 + , -g ,
 xi (g) Djk + xj ( ) Dik + xk (g) Dij 
i, j, k = O2 , H2 O, N2 .
Here, (12) is Darcy’s law with Brinkman extension, and the Darcy law permeability
tensor is assumed to be isotropic and constant, leading simply to a constant value
of the permeability, κ. Formal details of how (11)–(13) can be arrived at are given
in [19].

2.1.3.1 Inlet, outlet, upper wall, vertical walls For boundary conditions in the
channel, we prescribe inlet velocity and composition at x = 0, 0 ≤ y ≤ hf , so that
u = U in , v = 0, xO2 = xO
in
2
, xH2 O = xH
in
2O
, (14)
where v = (u, v). At the upper channel wall (0 ≤ x ≤ L, y = hf ), there is no slip,

no normal flow and no componental flux, so that
∂xO2 ∂xH2 O
u=v= = = 0. (15)
∂y ∂y
At the outlet at x = L, 0 ≤ y ≤ hf , we have constant pressure and no diffusive
componental flux, so that
∂v ∂xO2 ∂xH2 O
p = pout , = = = 0. (16)
∂x ∂x ∂x
At the vertical walls of the porous electrode (x = 0, L, −hp ≤ y ≤ 0), we prescribe
no normal flow, no tangential shear and no mass flux for the gas components, so
that , -(g) , -(g)
∂ v ∂ xO2 ∂ xH2 O
u = = = = 0, (17)
∂x ∂x ∂x
where v = (u, v).
2.1.3.2 Channel/porous backing interface In addition, matching conditions are

required for the fluid-porous interface at y = 0, 0 ≤ x ≤ L. The conditions for
continuity of normal velocity and normal stress are given respectively as
v = v , (18)
∂v ∂ v
p−µ = p(g) − µeff , (19)
∂y ∂y
where µeff (=µ/γ) is termed the effective viscosity of the porous medium. The
remaining two conditions that are required have been the subject of longstanding
debate ever since the work of Beavers and Joseph [23], and a summary of possible
options for the momentum equation is given by Alazmi and Vafai [24]; of these,
the most relevant for this application is one due to Ochoa-Tapia and Whitaker [25]
when inertial effects are important:
u = u , (20)
µ ∂ u ∂u β1 µ
−µ = 1 u + β2 ρu2 , (21)
γ ∂y ∂y κ2
respectively. Here, β1 and β2 are O (1) constants which would need to be determined
experimentally, although it turns out here that the leading order problem is dictated
more by (20) than by (21).
Finally, analogous volume-averaging techniques at the interface to those used
for heat transfer by [26] are required for the mole fraction transport equations. We
do not pursue the details, but simply assume the point values for the mole fractions
of O2 and H2 O in the channel to be equal to their intrinsic values in the porous
backing, so that
, -(g) , -(g)
xO2 = xO2 , xH2 O = xH2 O , at y = 0, (22)
and in addition that the point values for the mole fraction fluxes of O2 and H2 O are
equal to their superficial values in the porous medium, so that
, - , -
nO2 · N = nO2 · N , nH2 O · N = nH2 O · N ,
where N is the outward unit normal; using (18) and (22), we arrive at
, -
3 ∂
( g) ∂ xO2
xO2
γ2 , - = , (23)
∂y xH2 O (g) ∂y xH2 O
respectively.
2.1.3.3 Catalyst/porous backing interface At y = −hp , we would expect u,

v, xO2 (g) and xH2 O (g) to match to their counterparts in the catalytic layer,
although naturally this approach would require us to model this layer, and then
by extension the membrane and the corresponding regions on the other electrode.
An alternative approach, often adopted when the flow field in the porous backing
and gas channels rather than the electrochemistry in the catalyst and the membrane
is of interest, is to prescribe a functional form for the current density, I , at this
interface. Using Faraday’s Law, the superficial mass flux of the reactant is given as
a function of current density, so that
, - MO2 I
nO2 · N = − , (24)
4F
where F is the Faraday constant. The corresponding expression for water is then
taken to be
, - MH2 O (1 + 2α)I
nH2 O · N = , (25)
2F
where α is a parameter accounting for the total water transport by all mechanisms,
e.g. electro-osmosis, in the membrane; typical values encountered in the literature
are α = 0.3 [28] and 0.5 ≤ α ≤ 1.7 [18, 29]. Furthermore, since nitrogen does not
participate in the reaction at the catalytic layer,
, -
nN2 · N = 0. (26)
, -(g) , -(g)
This leads to the following boundary conditions for v, xO2 and xH2 O :
I
ρ(g) v = (2 + 4α)MH2 O − MO2 , (27)
4F
and
, - 3
, -
ρ(g) v xO2 (g) γ 2 ρ(g) ∂ xO2

(g) I −1
, - − M , - = .
M (g) xH2 O
(g) 2 ∂y xH2 O (g) 4F 2(1 + 2α)
M (g)
(28)
2.2 Nondimensionalization
Writing
x y v , - v ρ ρ(g)
x̃ = , ỹ = , ṽ = in , ṽ = in , ρ̃ = , ρ̃(g) = ,
L L U U [ρ] [ρ]

, -(g)
(g) = p − p , p̃ = p − p , ,p̃ -(g) = p
p − pout (g) out out − pout
p̃ = 2
, p̃ 2 2 2
,
[ρ] U in [ρ] U in [ρ] U in [ρ] U in
I M M (g) c
Ĩ = , M= , M(g) = , c̃ = ,
[I ] [M ] [M ] [ρ] / [M ]
Mi Dij Dijeff
Mi = , D̃ij = , D̃ijeff = , i, j = O2 , H2 O, N2
[M ] [D] [D]
[ρ] U in L µ κ U2
Re = , Sc = , Da = , Fr = ,
µ [ρ] [D] L2 gL
M , -(g) M(g)
M̃ = , M̃ = ,
[M ] [D] [M ] [D]
where [ρ] is a density scale, [D] is a diffusion scale, [I ] is a current density scale and
[M ] is a molecular weight scale (all to be either determined or specified shortly),
and Re, Sc, Da and Fr are the Reynolds, Schmidt, Darcy and Froude numbers,
respectively, we drop the tildes and arrive at the following nondimensionalized
forms. For the channel (0 ≤ x ≤ 1, 0 ≤ y ≤ hf /L),
∇ · (ρv) = 0, (29)

2δ2
∇ · (ρv ⊗ v) = −∇ p + ∇ · v + δ2 ∇ 2 v−Fr −1 ρj, (30)
3

ρv xO2 δ2 ρ ∇xO2
∇· = ∇· M , (31)
M xH2 O Sc M2 ∇xH2 O
where δ2 = Re−1 , and for the porous medium (0 ≤ x ≤ 1, −hp /L ≤ y ≤ 0),

∇ · ρ(g) v = 0, (32)

δ2 , -(g) 2 2 v
v = −∇ p +δ ∇ − Fr −1 ρ(g) j, (33)
ε2 γ
 
, - , -
( ) (g) ( ) (g)
ρ v xO2  3 ρ
g 2 g
δ (g) ∇, xO2 - 
∇·
( g)
, -(g) = ∇ · γ 2 2 M ( g)  ,
M xH 2 O Sc
M(g) ∇ xH 2 O
(34)
where ε2 = Da. The boundary conditions are now
u = 1, xO2 = xO
v = 0, in
2
, xH2 O = xH
in
2O
, at x = 0, 0 ≤ y ≤ hf /L; (35)
∂xO2 ∂xH2 O
u=v= = = 0, at 0 ≤ x ≤ 1, y = hf /L; (36)
∂y ∂y
∂v ∂xO2 ∂xH2 O
p = 0, = = = 0, at x = 1, 0 ≤ y ≤ hf /L; (37)
∂x ∂x ∂x
, -(g) , -(g)
∂ v ∂ xO2 ∂ xH2 O
u = = = = 0, at x = 0, 1, −hp /L ≤ y ≤ 0.
∂x ∂x ∂x
(38)
The boundary conditions for 0 ≤ x ≤ 1, y = −hp /L are now:

I
u = 0, ρ(g) v = 2(1 + 2α)MH2 O − MO2 , (39)
4
, - 3
, -
ρ(g) v xO2 (g) δ2 γ 2 ρ(g) (g) ∂ xO2
(g)
, - (g) − M , -
M(g) xH2 O 2 ∂y xH2 O (g)
Sc M(g)

I −1
= , (40)
4 2(1 + 2α)
where = [I ] [M ] /FU in [ρ]. Finally, the boundary conditions along the fluid-
porous interface on y = 0 reduce to
v = v , (41)
∂v ∂ v
p − δ2 = p(g) − δ2 , (42)
∂y ∂y
u = u , (43)

1 ∂ u ∂u β1 β2
− = u+ ρu2 , (44)
γ ∂y ∂y ε δ2
and
, -(g) , -(g)
xO2 = xO2 , xH2 O = xH2 O , (45)
, -
∂ (g) ∂ xO2
3 xO2
γ2 , - = . (46)
∂y xH2 O (g) ∂y xH2 O
2.3 Parameters
Typically, U in ∼ 1 ms−1 , hf ∼ 10−3 m, hp ∼ 3 × 10−4 m, L ≥ 10−2 m, [I ] ∼

104 Am−2 , pout ∼ 1 atm ∼ 105 kg m−1 s−2 , T ∼ 300–350 K, 0.1 ≤ γ ≤ 0.5,
0.3 ≤ α ≤ 1.7, µ ∼ O(10−5 ) kg m−1 s−1 . In addition, MO2 = 0.032 kg mol−1 ,
MH2 O = 0.018 kg mol−1 , MN2 = 0.028 kg mol−1 , F = 96487 Asmol−1 , from
which we note that Mmin ≤ M ≤ Mmax , where
Mmin = M |xH2 O =1, xO2 =0 = 0.018 kg mol−1 ,

Mmax = M |xH2 O =0, xO2 =1 = 0.032 kg mol−1 .
Further, we use the constitutive relation for an ideal gas in order to obtain the density
scale [ρ]; with p ∼ pout , we have ρ ∼ 1 kg m−3 , so that [ρ] ∼ 1 kg m−3 seems
appropriate. For [D], we take O(10−5 ) m2 s−1 from available literature, e.g. [11, 12].
Thence, for the nondimensional parameters Re, Sc, Da, Fr, , we arrive at
Re ∼ 104 , Sc ∼ 1, Da ≤ 10−6 , Fr ∼ 1, ≤ 10−2 , σ ∼ 10−2 ,
so that δ ∼ 10−2 and ε ≤ 10−3 .
2.4 Narrow-gap approximation
Typically, hf /L, hp /L 1, which leads us to further rescaling as follows. Writing

y v v
X = x, Y = , U = u, V = , U = u, V = ,
σ σ σ
, - , -
P = p, P = p , P = p , P = p ,
where σ = hf /L, we simplify further by neglecting terms in O (σ) or lower,

although we retain for the time being terms which contain multiples of σ and
the other dimensionless parameters. We introduce the dimensionless parameters
, and , given by
= δ2 /σ 2 , = σ 2 /ε, = /σ,
−1
and note that an alternative expression for is = Reσ 2 , i.e. the reciprocal
of the reduced Reynolds number. We have now, for the channel,
∂ ∂
(ρU ) + (ρV ) = 0, (47)
∂X ∂Y

∂U ∂U ∂P ∂2 U
ρ U +V =− + 2, (48)
∂X ∂Y ∂X ∂Y
∂P
0=− , (49)
∂Y

∂ ∂ 1 xO2 ∂ ρ ∂ xO2
ρ U +V = M , (50)
∂X ∂Y M xH2 O Sc ∂Y M2 ∂Y xH2 O
and for the porous medium,
∂ ( g) ∂ (g)
0= ρ U + ρ V , (51)
∂X ∂Y
, -(g)
ε ∂ P ε ∂2 U
U = − + , (52)
∂X γ ∂Y 2
, -(g)
1 ∂ P ε ∂2 V
V = − + , (53)
2 ∂Y γ ∂Y 2
, -(g)
(g ) ∂ ∂ 1 xO2
ρ U + V , -(g)
∂X ∂Y M(g) xH2 O
 
3
, -(g)
∂  γ ρ(g) 2
(g) ∂ xO 
=  2 M , 2 -(g)  . (54)
Sc ∂Y ∂Y xH2 O
M(g)
Note also that

, -(g)
P = P + O δ2 , P = P(g) + O δ2 ,
and since δ2 1, henceforth, we use the actual pressure rather than the modified
pressure. In addition, the gravitational terms in (49) and (53) are O(Fr −1 σ) and
have therefore been dropped. The boundary conditions are: for 0 ≤ X ≤ 1, Y = 1,

∂xO2 ∂xH2 O
U =V = = = 0; (55)
∂Y ∂Y
for 0 ≤ X ≤ 1, Y = 0,
V = V , (56)
P = P(g) , (57)
U = U , (58)

1 ∂ U ∂U β1 σ β2 σ
− = U+ ρU 2 , (59)
γ ∂Y ∂Y ε δ2
, -(g) , -(g)
xO 2 = xO2 , xH2 O = xH2 O , (60)
, -(g)
∂3 xO2 ∂ xO2
γ 2 , -(g) = ; (61)
∂Y xH 2 O ∂Y xH2 O

for 0 ≤ X ≤ 1, Y = −H =hp /hf ,

I
U = 0, ρ(g) V = 2(1 + 2α)MH2 O − MO2 , (62)
4
, - 3
, -
ρ(g) V xO2 (g) γ 2 ρ(g) ( g) ∂ xO2
(g)
, - ( g) − 2 M , -
M(g) xH 2 O ∂Y xH2 O (g)
Sc M(g)

I −1
= . (63)
4 2(1 + 2α)
The neglect of streamwise diffusion terms will of course imply that not all of
the original boundary conditions at X = 0 and 1 in this reduced formulation can be
satisfied and those terms would need to be reinstated for X ∼ O(σ) and 1−X ∼ O(σ).
This is beyond the scope of interest here, and for a consistent formulation we
simply retain
U = 1, xO2 = xO
in
2
, xH2 O = xH in
2O
, at X = 0, 0 ≤ Y ≤ 1; (64)
, -(g) , -(g)
∂ xO2 ∂ xH 2 O
U = = = 0, at X = 0, −H ≤ Y ≤ 0. (65)
∂X ∂X
For the initial discussion, we proceed under the assumption that , , ∼
O(1); later, we will consider 1 also. Further simplification is now possible
by noting from (52) that U = 0 to leading order, which reduces (51) and (54)
still further. In addition, Vynnycky and Birgersson [19] identify a porous boundary
1
layer of thickness ε 2 which has to be accounted for, although this proves to be
of no consequence for (54). Furthermore, the leading order equations prove to be
independent of β1 and β2 ; these, and a variety of other details can be found in [19].
2.5 Further simplifications and observations
Invoking the constitutive relation for an ideal gas in dimensionless variables, with
2 in 2
[ρ] U in [ρ] U
ρ = M+ P, ρ(g) = M(g) + P(g) (66)
pout pout
for the channel and porous medium respectively, which can be reduced to just
ρ = M, ρ(g) = M(g) , respectively, for the pressures and velocities being con-
sidered here. The reduced system of equations is now, for 0 ≤ X ≤ 1, 0 ≤ Y ≤ 1,
∂ ∂
(ρU ) + (ρV ) = 0, (67)
∂X ∂Y

∂U ∂U dP ∂2 U
ρ U +V =− + 2, (68)
∂X ∂Y dX ∂Y

∂ xO2 ∂ xO2 ∂ M ∂ xO2
U + V = , (69)
∂X xH2 O ∂Y xH2 O Sc ∂Y M ∂Y xH2 O
and for 0 ≤ X ≤ 1, −H ≤ Y ≤ 0,

I
ρ(g) V = 2(1 + 2α)MH2 O − MO2 , (70)
4
1 ∂ P(g)
V = − , (71)
2 ∂Y
, - 3
, -
(g) (g)
xO2 γ 2 ( g) ∂ xO2 I −1
V , -(g) − M , - = .
xH 2 O Sc M(g) ∂Y xH2 O (g) 4 2(1 + 2α)
(72)
The boundary conditions are: for 0 ≤ X ≤ 1, Y = 1,

∂xO2 ∂xH2 O
U =V = = = 0; (73)
∂Y ∂Y
and for X = 0, 0 ≤ Y ≤ 1,
U = 1, xO2 = xO
in
2
, xH2 O = xH
in
2O
, at X = 0, 0 ≤ Y ≤ 1; (74)
no boundary conditions as such prove to be necessary for X = 0, −H ≤ Y ≤ 0,

since only ordinary differential equations are solved for −H ≤ Y ≤ 0. At Y = 0
for 0 ≤ X ≤ 1, porous and fluid quantities are matched through
U = 0, V = V , P = P(g) , (75)
, -(g) , -(g)
x O2 = xO2 , xH2 O = xH2 O , (76)
, -(g)
∂3 xO2 ∂ xO2
γ 2 , -(g) = . (77)
∂Y xH 2 O ∂Y xH2 O
In general, I will not be constant; even more generally, it cannot be described

a priori, but is determined by considering the transport of species in the catalyst,
membrane and the anode side also. However, a common practice in studies which
emphasize the investigation of flow in the porous backing and the gas channel is
simply to prescribe a current density as a function of mole fraction. For example.
if we use the dimensional form of the Tafel law given by He, Yi and Nguyen [27],

aρ αc Fη
I= exp ,
M RT
where αc (= 2) is the transfer coefficient of the oxygen reduction reaction, eqn (2), η
is the overpotential for the oxygen reaction and a ( = 10−6 Am mol−1 ) is a constant
related to the exchange current density and oxygen reference concentration for the
oxygen reaction, we obtain the appropriate scale for [I ] as

a [ρ] αc Fη
[I ] = exp ; (78)
[M ] RT
consequently, in dimensionless form,
, -g
,
-(g) , -(g) ρ(g) xO2 ( ) , -(g)
I ρ , xO2 , xH2 O
(g)
= = xO2 . (79)
( g)
M
A dimensional quantity of importance for the determination of polarization curves

is the average current density, Iav , which is then given by
! 1
Iav = [I ] I dX .
0
This completes the formulation and necessary definitions. The data given in Sec-
tion 2.3 for the base case physical parameters indicates that ∼ O (1). Obviously,
taking channels with smaller aspect ratio, or operating the fuel cell at lower inlet
gas velocity would increase , motivating consideration of the lubrication theory
limit ( 1), since it provides qualitatively useful analytical solutions, as well as
a quantitative comparison with our numerical method; this is done in [19].
An interesting and instructive simplification occurs when 1 [21]. We intro-

duce the asymptotic series

χ = χ0 + −1 χ1 + O −2 , where χ = (U , V , P, ρ) ,

χ = χ( 0) + −1 χ( 1) + O −2 , where χ = xO2 , xH2 O , M, M
into eqns (67)–(77). These, taken at leading order in −1 , are for the most part
unchanged, except that eqn (70) and the two equations contained within (72) appear
to collapse onto just one equation:
(0)
xO 2
= 0 at 0 ≤ X ≤ 1, Y = −H. (80)
Examining (70) and (72) at the next order in gives, respectively,

(1)
ρ0 V0 = 1
4 2(1 + 2α)MH2 O − MO2 xO2 at 0 ≤ X ≤ 1, Y = −H, (81)
and
(0) 3
(0)
xO2 γ2 ∂ xO 1 −MO2 (1)
V0 − M(0) 2
= x . (82)
(0)
xH2 O Sc M (0) ∂Y xH2 O
(0) 4 2(1 + 2α)MH2 O O2
At first sight, this suggests difficulties since the leading order problem appears to
(1)
require information from higher orders; however, we note that xO 2
can be eliminated
via (81) and the necessary boundary conditions are (80) and
3
(0)

0 γ2 ∂ xO2
V0 (0) − M(0)
xH2 O Sc M (0) ∂Y (0)
xH2 O

ρ0 V0 −1
= . (83)
2(1 + 2α)MH2 O − MO2 2(1 + 2α)
This reveals that the limiting current for a particular cell can be found, regard-
less of how the electrochemistry in the active layer has been modelled: once ρ0
and V0 have been computed, the (dimensionless) limiting current density profile
would be
4ρ0 (X ) V0 (X )
Ilim (X ) = .
2(1 + 2α)MH2 O − MO2
Further, these observations suggest that the use of (80) and (83) as boundary con-
ditions should assist numerical convergence for higher current densities.
2.6 Numerics and results
The simplified parabolized equations were solved numerically using the Keller-
Box discretization scheme and Newton iteration (see, for example, Cebeci and
Bradshaw [30]). The system of partial differential equations to be solved in the
channel, (67)–(69), is of 8th order, and this is coupled to a 6th order system of
ordinary differential eqns (70)–(72), in the porous region. As is well-known, the
scheme is second-order accurate in both time-like and space-like variables, and we
omit any further details here. As an indication of the speed of the computations,
we note that a typical run with 500 points across, and 200 points along, the chan-
nel required around 100 CPU seconds on a 500 MHz Compaq Alphaserver with
3 GB RAM.
Results are presented for the Tafel law given in dimensionless form by (79), and
used previously for PEFC studies by [18, 27, 29]. Throughout, we keep γ = 0.3,
T = 353 K, and concentrate more on the effect of changes in channel height
and length, porous backing thickness and permeability, pressure, inlet speed and
composition. Physically realistic and implementable changes in any of these will
result in, at most, an order of magnitude change in the relevant dimensionless
parameter. The most sensitive parameter is , which varies over several orders of
magnitude as the cell voltage Ecell decreases; note here that we revert to using the
cell voltage rather than the overpotential, η, with the two being related by
Ecell = E0 − η,
where E0 (=1.1 V) is termed the open circuit voltage of the fuel cell.
Analytical solutions are given in [19] for when 1. Here, however, we
concentrate on solutions for physically more realistic operating parameters when
∼ O (1).
2.6.1 Effect of and

We show first results for Ecell = 0.75 V , corresponding to = 10.2, ranging over
several orders of magnitude in , and compare these with the analytical results in
the lubrication theory limit. Figures 4 and 5 are for intrinsic oxygen and water mole
fraction at Y = −H, respectively, and demonstrate that the lubrication solution
works well for as low as O 102 ; this is, however, considerably higher than
the base case physical values given in Section 2.3, which correspond to = 1.89.
Figure 6 shows the streamwise velocity U at Y = 12 , and illustrates the extent of
deviation from the classical value 32 .
An interesting limit occurs as Ecell is decreased. In this case, increases although
the quantity xO2 (g) at Y = −H remains O(1); this corresponds to the attainment
of the limiting current and the corresponding plots are given in Figs 7–9; observe
that in Figs 7 and 8 the limiting values for intrinsic oxygen and water mole fraction
for the analytical solution are reached very rapidly, so that in Fig. 7 the curve for
xO2 (g) effectively lies on the X -axis.
As regards the numerics, it was found that considerably more outer loop iterations
for the density were required as was increased. For instance, whereas 4 iterations
, -(g)
Figure 4: Comparison of analytical solution for xO2 at Y = −H with numerical
solutions for = 1.89, 1.89 × 101 , 1.89 × 102 (Ecell = 0.75 V).
, -(g)
Figure 5: Comparison of analytical solution for xH2 O at Y = −H with numeri-
cal solutions for = 1.89, 1.89 × 101 , 1.89 × 102 (Ecell = 0.75 V).
Figure 6: Comparison of analytical solution for U at Y = 12 with numerical solu-

tions for = 1.89, 1.89 × 101 , 1.89 × 102 (Ecell = 0.75 V).
, -(g)
Figure 7: Comparison of analytical solution for xO2 at Y = −H with numerical
solutions for = 1.89, 1.89 × 101 , 1.89 × 102 (Ecell = 0.65 V).
, -(g)
Figure 8: Comparison of analytical solution for xH2 O at Y = −H with numeri-
cal solutions for = 1.89, 1.89 × 101 , 1.89 × 102 (Ecell = 0.65 V).

tions for = 1.89, 1.89 × 101 , 1.89 × 102 (Ecell = 0.65 V).
sufficed for Ecell = 0.75 V, it was common for 20–30 to be necessary for Ecell =
0.65 V. In addition, there were difficulties in initiating the marching scheme at
X = 0 for higher values of ; we surmise this to be due to the phenomena illustrated
at the end of the previous section. Whilst setting the channel inlet values as an initial
guess for the first step along the channel was adequate for lower values of , this
was found not be sufficient for Ecell lower than 0.71 V; for those cases, the first-
step solution for Ecell = 0.71 V had to be used instead, then enabling numerical
solutions to be obtained for higher and higher values of until the limiting current
was reached.
2.6.2 ‘Polarization surfaces’

It is customary for fuel cell performance to be given in terms of a polarization curve
where the cell potential, Ecell , is given as function of the average current density, Iav .
Generally speaking, if the analysis is done dimensionally, this leads to a vast number
of graphs for each alteration made in one of the physical parameters. However, a
major benefit of the nondimensional analysis carried out here is that the results
can be expressed considerably more compactly by plotting polarization ‘surfaces’;
individual polarization curves will therefore be curves lying on those surfaces. We
explain this as follows. From the nondimensionalization given above, the emergent
non-dimensional parameters were , , and Sc. In addition, there is γ, which we
hold fixed in this study, and xO in , x in and H, whose effect on fuel cell performance
2 H2 O
one would like to explore. First, we observe that, in the parameter range of interest,
has no effect on Iav , since the dimensionless density is independent of pressure
and the pressure in the channel serves as a boundary condition for the pressure in the
porous medium. In addition, a change in Sc can only be effected by changes in [ρ],
which only occurs if the cathode is run at a different pressure. Consequently, a tidy
representation of Iav is to plot it as a function of and , for fixed Sc, xO in , xin and
2 O2
H, the benefit of this being that the effect of four parameters, hf , L, U in and Ecell , are
displayed on one graph; since can vary over several orders of magnitude, it proves
more convenient to use log () as a variable. Examples of this are given below.
Figure 10 gives polarization surfaces for H = 0.3, with the pressure at 1 and 3
atmospheres. The limiting current phenomenon is observed as increases, and
its value is observed to increase moderately with increasing , but strongly with
increasing pressure. Figure 11 shows a similar plot, except with computations now
for pout = 1 atm, for H = 0.15 and H = 0.6. Average current densities are
found to be higher for the thinner porous backing, and in both cases a limiting
value is evident as is increased. Figure 12 compares the base case for pout =
1 atm and xO in = 0.21 with two other cases at 1 atm which have differing inlet
2
compositions: dry oxygen (xO in = 1) and partially humidified air, for which x in =
2 O2
0.13 and xH in = 0.36 (corresponding to 76% relative humidity) [1, 2, 16]. As is
2O
evident, increased oxygen content at the inlet raises the average current density; for
xOin = 1, convergence difficulties were experienced for quite low values of , which
2
explains the rather narrow range of values presented for this case, but nonetheless
the average current density is much higher than that for the other two cases.
Figure 10: Polarization surfaces for pout = 1, 3 atm (H = 0.3).
Figure 11: Polarization surfaces for H = 0.3, 0.6 ( pout = 1 atm).


tions for = 1.89, 1.89 × 101 , 1.89 × 102 (Ecell = 0.75 V).
Figure 13: Schematic of a DMFC.
3 DMFC
A schematic diagram of a DMFC is given in Fig. 13. Essentially, this entails a
polymer membrane sandwiched between two gas-diffusion electrodes, which are
each adjacent to flow channels contained within bipolar plates. The oxidant, usually
humidified oxygen from air, is fed in at the inlet of the channel on the cathode side,
and is transported to the electrolyte/cathode interface; the fuel on the other hand,
normally dilute liquid methanol, is fed at the anode channel inlet and is transported
to the electrolyte/anode interface. Both interfaces contain catalyst to accelerate the
reactions
CH3 OH + H2 O → CO2 + 6H+ + 6e− at the anode, (84)
O2 + 4H+ + 4e− → 2H2 O at the cathode, (85)
in the course of which an electric current is produced to drive a given load. Optimal
fuel cell performance is achieved at typical voltages of around 0.4 V at current
densities of about 0.1 Acm−2 .
3.1 Mathematical formulation for flow in the anode
In what follows, we begin by considering a steady 2D flow of a liquid water/

methanol/carbon dioxide mixture in the slender channel and porous backing of
a DMFC. For brevity, we highlight here the similarities and differences with the
PEFC cathode model already derived in Section 2.3. Throughout, we make the
following associations from the three species in the cathode of a PEFC to those in
the anode of a DMFC:
O2 ⇔ CH3 OH, H2 O ⇔ CO2 , N2 ⇔ H2 O.
3.1.1 Channel
Equations (3)–(6) are unchanged. However, the diffusion tensor in the channel and
the porous backing can be simplified to a diagonal tensor, since the system can be
treated as a dilute solution, rendering the cross terms redundant, i.e

DCH3 OH,H2 O 0
M = MH2 O ,
0 DCO2 ,H2 O
, -(l)
DCH3 OH,H2 O 0
M(l) = MH2 O , -(l) .
0 DCO2 ,H2 O
For the mixture density, the constitutive relation for an ideal gas is now of course
entirely inappropriate. By extrapolating literature data to 70 ◦ C, the influence of
methanol on the density is estimated to lower the density by 0.4% at 2 wt.% methanol
in the mixture [31]; this contribution can safely be neglected, whence we take the
density of the mixture, ρ, as that of pure water, and hence constant.
3.1.2 Porous backing

Superscripts (g) are replaced by (l) to denote liquid phase.

3.1.3.1 Inlet, outlet, upper wall vertical walls These remain, for the most part,
unchanged, although we assume a developed flow profile at the inlet, so that
2
y y
u = 4U in
− , v = 0. (86)
hf hf
3.1.3.2 Channel/porous backing interface These remain unchanged.
3.1.3.3 Catalyst/porous backing interface The only changes are the replace-
ment of (24)–(26) by
, - I (1 + 6αCH3 OH )MCH3 OH
nCH3 OH · N = − , (87)
6F
, - IMCO2
nCO2 · N = , (88)
6F
, - I (1 + 6αH2 O )MH2 O
nH2 O · N = − , (89)
6F
where αH2 O and αCH3 OH are parameters accounting for the water and methanol
transport across the membrane. In the following, we have neglected the contribution
from the methanol crossover, given by αCH3 OH . This is a reasonable assumption for
any practical DMFC application, since a high crossover leads to an unfavourable
mixed potential at the cathode as well as loss of fuel at the anode. If the crossover
cannot be neglected, terms accounting for the transport of methanol in the membrane
can easily be added; see for example Scott et al.[32]. Equations (27) and (28) give
, -(l) , -(l)
then instead the following boundary conditions for v, xCH3 OH and xCO2 :
I
ρ(l) v = − (1 + 6αH2 O )MH2 O + MCH3 OH − MCO2 , (90)
6F
and
, -(l) 3
, -(l)
ρ(l) v xCH3 OH γ 2 ρ(l) ∂ xCH3 OH I −1

, -(l) − M̄ , - = . (91)
M (l) xCO2 M (l)
2 ∂y xCO2
(l) 6F 1
3.2 Nondimensionalization
This is unchanged except that (39) and (40) are now

I
u = 0, ρ v = −
(l)
(1 + 6αH2 O )MH2 O + MCH3 OH − MCO2 ,
6
(92)
, -(l) 3
, -(l)
ρ(l) vxCH3 OH δ2 γ 2 ρ(l) (l) ∂ xCH3 OH
, -(l) − 2 M , -(l)
M(l) xCO2 Sc M(l)
∂y xCO2

I −1
= . (93)
6 1
3.3 Parameters
Typically, U in ∼ 10−2 ms−1 , xCH3 OH ∼ O(10−2 ), hf ∼ 10−3 m, hp ∼ 1.8 ×

10−4 m, L ≥ 10−2 m, κ ∼ 10−12 m2 , [I ] ∼ 4 × 103 Am−2 , pout ∼ 1 atm ∼
105 kg m −1 s−2 , T ∼ 300–350 K, 0.1 ≤ γ ≤ 0.9, αH2 O = 2.5 (see [33]), αCH3 OH =
xCH3 OH αH2 O (see [32]), µ ∼ O(10−4 ) kg m−1 s−1 , ρ ∼ 978 kg m−3 . In addi-
tion, MCH3 OH = 0.032 kg mol−1 , MCO2 = 0.044 kg mol−1 , MH2 O = 0.018
kg mol−1 , F = 96487 Asmol−1 , R = 8.314 Jmol−1 K−1 , A = 157, B = 0.61,
E0 = 0.504 V, EA = 0.5 V, αA = 7.9. For [D], we take O(10−9 ) m2 s−1 , which is
the typical scale for liquid diffusion.
Thence, for the nondimensional parameters Re, Sc, Da, , we arrive at
Re ∼ 103 , Sc ∼ 50, Da ≤ 10−6 , Fr ∼ 10−2 , ≤ 4×10−4 , σ ∼ 10−2 ,
so that δ ∼ 3 × 10−2 and ≤ 10−3 .
3.4 Narrow-gap approximation
The details for this are unchanged except for

U = 0, (94)

I
ρ(l) V = (1 + 6αH2 O )MH2 O + MCH3 OH − MCO2 , (95)
6
, -(l) 3
, -(l)
ρ(l) V xCH3 OH γ 2 ρ(l) ∂ x CH OH
-(l) − 2 M
(l) 3
, , -(l)
M(l) xCO2 Sc M (l) ∂Y xCO2

I −1
= . (96)
6 1
Significantly, ∼ 10−2 , which will enable us to make simplifications that were

not possible for the PEFC cathode.
3.5 Further simplifications and observations
Since the system is dilute, xH2 O xCH3 OH , xCO2 , implying that we will be able to
write
xCH3 OH + xCO2 = 1 − xH in
2O
;
thus, we only need to solve the mass transfer equation for methanol and this rela-
tionship will provide the mass fraction of carbon dioxide. The mean molecular mass
M(l) reduces to MH2 O , since the contribution of carbon dioxide and methanol is
negligible. The fact that the density is constant enables us to write simply ρ(l) = 1.
This also simplifies the diffusion coefficient DCH3 OH,H2 O , which we can treat as
constant for the dilute system; thus, an appropriate choice for the characteristic
diffusion coefficient scale, [D], is [D] = DCH3 OH,H2 O . With the simplifications
outlined above, the dimensionless form of the equations for the channel is
∂U ∂V
+ = 0, (97)
∂X ∂Y
∂U ∂U ∂P ∂2 U
U +V =− + 2, (98)
∂X ∂Y ∂X ∂Y
∂P
= 0, (99)
∂Y
∂xCH3 OH ∂xCH3 OH ∂2 xCH3 OH
U +V = ; (100)
∂X ∂Y Sc ∂Y 2
these are the same as eqns (67)–(69) for constant density ρ, and with P = P. For
the porous backing, we have
∂ V
= 0, (101)
∂Y
U = 0, (102)
∂ P(l)
= −2 V , (103)
∂Y
, -(l) 3 , -l
∂ xCH3 OH γ 2 ∂2 xCH3 OH
V = . (104)
∂Y Sc ∂Y 2
However, V is still not correctly scaled, since it is of order (1); thus,
we set
V
V̂ = , (105)

so that (101), (103) and (104) become, respectively,
∂V̂
= 0, (106)
∂Y
∂P(l)
= −2 V̂ , (107)
∂Y
, -(l) , -(l)
∂ xCH3 OH 3 ∂2 xCH3 OH
V̂ = γ2 . (108)
∂Y Sc ∂Y 2
The interface condition for V at Y = 0, the second equality in eqn (75), implies
that V will also be of order . Setting V̂ = V /, the governing equations in the
channel, (97)–(100), are then given by
∂U ∂V̂
+ = 0, (109)
∂X ∂Y
∂U ∂U ∂P ∂2 U
U + V̂ =− + 2, (110)
∂X ∂Y ∂X ∂Y
∂P
= 0, (111)
∂Y
∂xCH3 OH ∂xCH3 OH ∂2 xCH3 OH
U + V̂ = . (112)
∂X ∂Y Sc ∂Y 2
Now, since ∼ O(10−2 ) 1, we can safely neglect terms of O(), which corre-
sponds to V = 0; furthermore, eqns (102) and (58) imply that U = 0 at Y = 0.
Consequently, we find that the inlet condition for a fully developed laminar velocity
profile satisfies the momentum equation downstream, and thence that, at leading
order, the velocity decouples from the mass transfer in the channel; all that remains
to be solved there is
∂xCH3 OH ∂2 xCH3 OH
4(Y − Y 2 ) = . (113)
∂X Sc ∂Y 2
Now, eqns (95) and (106) combine to give, for the entire porous backing,
I
V̂ = − ; (114)
6
where = (1+6αH2 O )MH2 O +MCH3 OH −MCO2 . Then, since the current density
is a function of X alone, we have V̂ = V̂ (X ). We can integrate (108) and, using
(114) and (96), we arrive at
, -(l)
I , -(l) 3 ∂ xCH3 OH I
− xCH3 OH − γ 2 = − MH2 O . (115)
6 Sc ∂Y 6
Furthermore, choosing [M ] = MH2 O , whence MH2 O = 1, eqn (115) becomes
, -(l)
, -(l) 6 3 ∂ xCH3 OH
xCH3 OH + γ2 = 1. (116)
ScI ∂Y
Integrating eqn (116) gives

, -(l) 1 ScI
xCH3 OH (X , Y ) = + C exp − 3
Y , (117)
6γ 2
where C is an integration constant to be determined shortly. At Y = −H, we have

, -(l) 1 ScI
xCH3 OH (X , −H) = + C exp 3
H , (118)
6γ 2
and at Y = 0,
, -(l)
1
xCH3 OH (X , 0) =
+ C, (119)

whence the integration constant C is given by the methanol mass fraction at the
interface between the porous backing and the channel. Recalling the boundary
conditions for the plain/porous interface, eqns (60) and (61), for the methanol
transport equation, we combine these two to obtain just one boundary condition for
the channel at Y = 0 :

∂xCH3 OH (X , 0) ScI 1
+ xCH3 OH (X , 0) − = 0. (120)
∂Y 6
This is valid for any profile I . As for the PEFC, I will not in general be constant.
In order to take the porous effects of the actual active layer into account, a previous
model [34] is used to generate an expression for the local current density that can
be used for the boundary conditions at the active layer/porous backing. This current
density implicitly accounts for the same effects as were considered in that paper,
i.e. pore diffusion in the active layer, finite ionic conductivity and the complex
methanol oxidation kinetics: in dimensional form,
,
-(l) B αA F
ρ xCH3 OH exp RT (EA − E 0 )
I =A , (121)
M (l) 1 + exp αA F (EA − E0 ) RT
which gives the appropriate scale for [I ] as

B
αA F
in
ρxCH exp RT (EA − E0 )
3 OH
[I ] = A ,
M (l) 1 + exp αA F
(EA − E0 )
RT
and consequently, in dimensionless form,

, -(l) B
, -(l) xCH3 OH
I xCH3 OH = in
. (122)
xCH 3 OH
, -(l)
This is valid for any profile I , which is a function of the value of xCH3 OH at
Y = −H; for the profile considered in this paper, I can be determined in terms of
xCH3 OH (X , 0) through the transcendental equation

1 1 1 I
in
xCH I B = + xCH OH (X , 0) − exp , (123)
3 OH 3
6
3
where = ScH/(γ 2 ).
In summary, the adaption of the reduced model for the cathode of a PEFC to
the anode of a DMFC is based on the fact that the anode operates at a large water
fraction and that the magnitude of the dimensionless parameter is much smaller
than 1.
3.6 Numerics and results
As for the PEFC, we resort to a numerical scheme to solve the transport equation
for methanol in the channel, since no further analytical simplifications are possible.
This entails solving eqn (113), subject to the boundary conditions (55) and (120)
and the inlet condition for methanol (64).
The governing equations are once again parabolic, for which a Modified Box
discretization scheme is suitable [35]. The scheme leads to a block tridiagonal
matrix, allowing fast computations. The resulting system of non-linear equations
is solved with a Newton-Raphson-based algorithm in MATLAB 6.
To confirm the validity of the reduced model, its predictions were compared with
numerical results obtained using two other softwares [7, 8], wherein the full ellip-
tic governing equations and boundary conditions are implemented. Comparison,
shown in Fig. 14, was carried out in terms of the local superficial current density
obtained along the anode for a variety of values for the nondimensional parame-
ters and . These were obtained by varying the inlet velocity, U in , and keeping
Figure 14: Verification of the reduced model. ( · · · ) corresponds to the CFX-4.4

solution with 104 number of nodes,(–) is the Femlab solution for ∼103
adapted nodes and markers are for the reduced model, with 104 cells.
(✩): = 0.932, (): = 2.79, (×): = 9.32, (+): = 27.9, ():
= 93.2, (): = 279.
all other physical parameters constant; in particular, the base case corresponds to
= 9.32 and = 3.76 × 10−2 . Several features are apparent. First, for the higher
(, )-combinations, the reduced model starts to deviate from the full solution,
since the normal velocity due to the electrochemical reaction now is of the same
order of magnitude as the streamwise inlet velocity, i.e. is no longer much smaller
than unity and the normal velocity in the channel is no longer negligible. Neverthe-
less, the local current densities from the reduced model remain close to the current
densities from the full elliptic equations, even for these combinations. Second, we
note discrepancies between the values obtained at the inlet and outlet by the two
commercial solvers. In particular, convergence difficulties were encountered with
FEMLAB 2.2 as regards the resolution of the corners, and the solutions presented
are actually for a channel that is extended at the inlet and outlet; such difficulties
were not encountered with CFX-4.4. Since our principal interest was only to verify
the reduced model, these differences were not investigated further.
A final comment here which illustrates the benefit of the reduced model approach
concerns a comparison of the computing times for the three methods. On a 1 Ghz
AMD PC, with 512 MB SDRAM the reduced model with 104 cells took ∼5 CPU
seconds to converge, whereas Femlab 2.2 required 1 − 2 CPU minutes. The
CFX-4.4 code required 1 − 2 CPU minutes on a 500 MHz Compaq Alphaserver
with 3 GB RAM. Mesh independent solutions were found for the reduced model
for 103 cells, allowing for computational times of less than 1 second.
4 Conclusions
In this chapter, we have considered the use of reduced models for describing
transport phenomena in fuel cells as an alternative to computational fluid dynam-
ics. Although the approach appears, at first sight, to be mathematically intensive,
requiring nondimensionalization, rescaling and analysis, as well as a certain level
of intuition and experience to ensure that the reduced model is indeed consistent,
it does have several advantages:
1. reduced models are, computationally speaking, orders of magnitude less inten-
sive than, yet still as practically useful as, CFD models;
2. physical trends become much more apparent, e.g. it is not necessary to perform
computations in order to carry out a sensitivity analysis, since the dependence
of the model on a particular parameter will become apparent as the model is
being analyzed;
3. a nondimensional analysis opens the possibilities for model modification, so
that a model used in one context (here, the cathode of a PEFC) can be used in
another (the anode of a DMFC).
Although the problems treated here address only the interaction of kinetics and
transport in the performance-limiting electrodes of the PEFC and DMFC, there is
little doubt that the approach can be used successfully when dealing with extensions
of these models for multiphase flow, transient behavior, whole cell models, etc.,
even for other types of fuel cells. Furthermore, the reduced model approach is the
one most likely, in practice, to integrate successfully the modeling of transport
phenomena at the heart of a fuel cell with system studies, stack modeling and fuel
cell optimization.
References
[1] Dutta, S., Shimpalee, S. & van Zee, J.W., Three-dimensional numerical sim-
ulation of straight channel PEM cells. Journal of Applied Electrochemistry,
30, pp. 135–146, 2000.
[2] Shimpalee, S. & Dutta, S., Numerical prediction of temperature distribution
in PEM fuel cells. Numerical Heat Transfer, Part A, 38, pp. 111–128, 2000.
[3] Dutta, S., Shimpalee, S. & van Zee, J.W., Numerical prediction of mass
exchange between cathode and anode channels in a PEM fuel cell. Int. J.
Heat and Mass Transfer, 44, pp. 2029–2042, 2001.
[4] Berning, T., Lu, D. & Djilali, N., Three-dimensional computational anal-
ysis of transport processes in a PEM fuel cell. J. Power Sources, 106(1),
pp. 284–294, 2002.
[5] Jen, T.-C., Tan, T. & Chan, S.-H., Chemical reacting transport phenomena in
a PEM fuel cell. Int. J. Heat and Mass Transfer, 46, pp. 4157–4168, 2003.
[6] Lee, W.-K., Shimpalee, S. & van Zee, J.W., Verifying predictions of water and
current distributions in a serpentine flow field polymer electrolyte membrane
fuel cell. J. Electrochem. Soc., 150(3), pp. A341–A348, 2003.
[7] CFX, www.waterloo.ansys.com/cfx/.
[8] Femlab, www.comsol.se/femlab/version23.php.
[9] Fluent, www.fluent.com.
[10] Star-CD, www.cd-adapco.com/apps/STAR-CDfcell.htm.
[11] Bernardi, D.M. & Verbrugge, M.W., Mathematical model of a gas diffusion
electrode bonded to a polymer electrolyte. AIChE Journal, 37, pp. 1151–
1163, 1991.
[12] Bernardi, D.M. & Verbrugge, M.W., A mathematical model of the solid-
polymer-electrolyte fuel cell. J. Electrochem. Soc., 139, pp. 2477–2491,
1992.
[13] Springer, T.E., Zawodzinski, T.A. & Gottesfeld, S., Polymer electrolyte fuel
cell model. J. Electrochem. Soc., 138, pp. 2334–2342, 1991.
[14] Gurau, V., Barbir, F. & Liu, H., An analytical solution of a half-cell model
for PEM fuel cells. J. Electrochem. Soc., 147, pp. 2468–2477, 2000.
[15] Dannenberg, K., Ekdunge, P. & Lindbergh, G., Mathematical model of the
PEMFC. Journal of Applied Electrochemistry, 30, pp. 1377–1387, 2000.
[16] Fuller, T.F. & Newman, J., Water and thermal management in solid-polymer-
electrolyte fuel cells. J. Electrochem. Soc., 140, pp. 1218–1225, 1993.
[17] Nguyen, T.V. & White, R.E., A water and heat management model for proton-
exchange-membrane fuel cells. J. Electrochem. Soc., 140, pp. 2178–2186,
1993.
[18] Yi, J.S. & Nguyen, T.V., An along-the-channel model for proton exchange
membrane fuel cells. J. Electrochem. Soc., 145, pp. 1149–1159, 1998.
[19] Vynnycky, M. & Birgersson, K.E., Analysis of a model for multicomponent
mass transfer in the cathode of a polymer electrolyte fuel cell. SIAM Journal
on Applied Mathematics, 63(4), pp. 1392–1423, 2003.
[20] Birgersson, K.E., Nordlund, J., Ekström, H., Vynnycky, M. & Lindbergh, G.,
A reduced two-dimensional one-phase model for analysis of the anode of a
DMFC. J. Electrochem. Soc., 150(10), pp. A1368–A1376, 2003.
[21] Vynnycky, M. & Birgersson, K.E., On the role of heat transfer in the cathode
of a polymer electrolyte fuel cell, in preparation.
[22] Bird, R.B., Stewart, W.E. & Lightfoot, E.N., Transport Phenomena, Wiley
and Sons, Inc.: USA, pp. 767–770, 2002.
[23] Beavers, G.S. & Joseph, D.D., Boundary conditions at a naturally permeable
wall. J. Fluid Mech., 30, pp. 197–207, 1967.
[24] Alazmi, B. & Vafai, K., Analysis of fluid flow and heat transfer interfacial
conditions between a porous layer and a fluid layer. Int. J. Heat and Mass
Transfer, 44, pp. 1735–1749, 2001.
[25] Ochoa-Tapia, J.A. & Whitaker, S., Momentum jump condition at the bound-
ary between a porous medium and a homogeneous fluid: inertial effects.
J. Porous Media, 1, pp. 201–217, 1998.
[26] Ochoa-Tapia, J.A. & Whitaker, S., Heat transfer at the boundary between a
porous medium and a homogeneous fluid: the one-equation model. J. Porous
Media, 1, pp. 31–46, 1998.
[27] He, W., Yi, J.S. & Nguyen, T.V., Two-phase flow model of the cathode of
PEM fuel cells using interdigitated flow fields. AIChE Journal, 46, pp. 2053–
2064, 2000.
pp. 40–50, 2001.
[29] Nguyen, T.V., Modeling two-phase flow in the porous electrodes of proton
exchange membrane fuel cells using the interdigitated flow fields. Electro-
chemical Society Proceedings, 99–14 pp. 222–241, 2000.
[30] Cebeci, T. & Bradshaw, P., Momentum Transfer in Boundary Layers,
Hemisphere Publishing: Washington, 1997.
[31] Landolt-Börnstein, New Series IV/1, Springer: Germany, p. 117, 1977.
[32] Scott, K., Argyropoulos, P. & Sundmacher, K., A model for the liquid
feed direct methanol fuel cell. J. Electroanal. Chem., 477, pp. 97–110,
1999.
[33] Baxter, S.F., Battaglia, V.S. & White, R.E., Methanol fuel cell model: anode.
J. Electrochem. Soc., 146(2), pp. 437–447, 1999.
[34] Nordlund, J. & Lindbergh, G., A model for the porous direct methanol fuel
cells anode. J. Electrochem. Soc., 149(9), pp. A1107–A1113, 2002.
[35] Tannehill, J.C., Anderson, D.A. & Pletcher, R.H., Computational Fluid
Mechanics and Heat Transfer, Taylor & Francis: USA, 1997.
Nomenclature
a constant for current density (PEFC), Am mol−1
A experimentally fitted parameter
B experimentally fitted parameter (DMFC)
B scalar, vector or tensor quantity (PEFC)
C integration constant
Dij diffusion coefficients, m2 s−1
Di, j binary diffusion coefficients for a pair (i, j), m2 s−1
Da Darcy number
E potential, V
F Faraday’s constant, A s mol−1
Fr Froude number
g gravity, ms−2
h height, m
H = hp /hf dimensionless height of porous backing
I current density, Am−2
L length, m
M mean molecular mass, kg mol−1
Mi molar mass of species i, kg mol−1
M(k) , M(k) diffusion tensor of phase k, kg mol−1 m2 s−1
M, Mi dimensionless molar mass
ni total mass flux of species i, mol m−2 s−1
N outward unit vector
p, p pressure, Pa
P, P dimensionless pressure
R gas constant, J mol−1 K−1
Re = [ρ]U in L/µ Reynolds number
Sc = µ/([ρ][D]) Schmidt number
T temperature, K
u, v, v, U in velocities, m s−1
U , V , V̂ dimensionless velocities
V volume, m3
xi mole fraction of species i
x, y coordinates, m
x̃, ỹ, X , Y dimensionless coordinates
Greek symbols
α coefficient for water transport in membrane

αA Tafel slope (DMFC)
αc transfer coefficient for oxygen reduction (PEFC)
β1 , β2 constants for channel/porous backing interface
γ porosity
3
= ScH/(γ 2 ) dimensionless number
δ2 = Re−1 dimensionless number

= 1/(Reσ 2 ) dimensionless number
ε = Da2 parameters for relative permeabilities
η overpotential (PEFC), V
κ permeability, m2
= [i][M ]/(ρU in F), dimensionless number
µ dynamic viscosity, kg m−1 s−1
ρ density, kg m−3
σ = hf /L dimensionless number
= σ 2 /ε dimensionless number
= ((1 + 6αH2 O )MH2 O
+MMeOH − MCO2 ) dimensionless number
χ dummy variable for asymptotic series
= /σ dimensionless number
Subscripts
0 (E0 ) open circuit voltage (PEFC), V

0 (E0 ) experimentally fitted parameter (DMFC), V
0,1 index for series expansion
A anode
av average
cell cell
CH3 OH methanol
CO2 carbon dioxide
eff effective
f flow channel
H2 O water
i, j, k species i, j, k
lim limiting
min, max min, max values
N2 nitrogen
O2 oxygen
p porous backing
Superscripts
(0), (1) index in asymptotic series

(g) gas phase
in inlet
(l) liquid phase
out outlet
Miscellaneous
[ ] characteristic value of a variable

CHAPTER 8
Modeling of PEM fuel cell stacks with

hydraulic network approach
J.J. Baschuk & X. Li
Department of Mechanical Engineering, University of
Waterloo, Canada.
Abstract
Polymer electrolyte membrane (PEM) fuel cells convert the chemical energy of
hydrogen and oxygen directly into electrical energy. Waste heat and water are the
reaction by-products, making PEM fuel cells a promising zero-emission power
source for transportation and stationary co-generation applications. In this study, a
mathematical model of a PEM fuel cell stack is formulated. The distributions of the
pressure and mass flow rate for the fuel and oxidant streams in the stack are deter-
mined with a hydraulic network analysis. Using these distributions as operating
conditions, the performance of each cell in the stack is determined with a mathe-
matical, single cell model that has been developed previously. The stack model has
been applied to PEM fuel cell stacks with two common stack configurations: the
U and Z stack design. The former is designed such that the reactant streams enter
and exit the stack on the same end, while the latter has reactant streams entering
and exiting on opposite ends. The stack analyzed consists of 50 individual active
cells with fully humidified H2 or reformate as fuel and humidified O2 or air as the
oxidant. It is found that the average voltage of the cells in the stack is lower than
the voltage of the cell operating individually, and this difference in the cell per-
formance is significantly larger for reformate/air reactants when compared to the
H2 /O2 reactants. It is observed that the performance degradation for cells operating
within a stack results from the unequal distribution of reactant mass flow among the
cells in the stack. It is shown that strategies for performance improvement rely on
obtaining a uniform reactant distribution within the stack, and include increasing
stack manifold size, decreasing the number of gas flow channels per bipolar plate,
and judicially varying the resistance to mass flow in the gas flow channels from
cell to cell.
1 Introduction
Polymer electrolyte membrane (PEM) fuel cells convert the chemical energy of
hydrogen and oxygen directly and efficiently into electrical energy with by-products
of heat and liquid water. PEM fuel cells also have a high power density, quick
start-up and load following characteristics, making them attractive zero emission
power sources [1]. Before PEM fuel cells can be successfully commercialized,
the production cost must be reduced from the current estimate of approximately
$200/kW to $30/kW [2]. Increasing the energy conversion efficiency and power
output of the PEM fuel cells could decrease the cost per kW, and thus several
empirical and mathematical modeling studies have been undertaken for the purpose
of understanding and predicting PEM fuel cell performance.
In order to satisfy the power demand of most applications, several PEM fuel cells
must be connected in series to form a PEM fuel cell stack. However, heat and water
management strategies, which are successful for single PEM fuel cells, are difficult
to implement in a stack environment; the efficiency and power output of a PEM
fuel cell operating within a stack are lower than the performance of a PEM fuel
cell operating independently [3]. Thus, single cell PEM fuel cell models cannot be
directly applied for PEM fuel cell stack optimization. This is because the operating
conditions for each cell in a stack are typically not the same as the conditions at the
stack inlet, and are different among the cells themselves due to the non-uniform
reactant flow distribution among the cells, influenced by the pressure loss associated
with each flow passage.
Several modeling studies of PEM fuel cell stacks exist in the published litera-
ture. Empirical models, originally developed for a single PEM fuel cell, have been
extended to model PEM fuel cell stacks. The empirical, single cell model of Kim
et al. [4] was applied to a stack by Chu et al. [5]. The stack voltage was char-
acterized as a function of current density using terms that represented activation
overpotential, ohmic overpotential, and mass transport limitations. The general-
ized steady state electro-chemical model (GSSEM) [6] has been applied to both
single PEM fuel cells and stacks. Stack voltage has a functional dependence on
the partial pressure of the reactants, current density and temperature through terms
accounting for activation, concentration and ohmic overpotential, CO poisoning,
and performance degradation due to aging [7]. Mathematical PEM fuel cell models
have also been extended to simulate stack performance. The single cell model of
Nguyen and White [8] was isothermal, two-dimensional, steady state, and incorpo-
rated mass transport in the electrode backing, the electro-chemical reaction of the
cathode, and proton migration in the polymer electrolyte. By modeling the reactant
flow in the gas flow channels and stack manifold as a pipe network, Thirumulai and
White [9] extended the single cell model to simulate stack performance.
Due to the exothermic nature of the electro-chemical reactions occurring within
a PEM fuel cell, thermal management within a stack is a significant considera-
tion for stack design. Maggio et al. [10] investigated the temperature and current
density distribution in a PEM fuel cell stack using a three-dimensional model.
The temperature distribution in the cooling plate, cooling water and membrane
electrode assembly was found through application of conservation of energy, while

the electro-chemical performance of the PEM fuel cell was determined with the
empirical relationship of Patel et al. [11]. The model of Maggio et al. [10] was
developed for a stack operating in steady state, but the model of Lee and Lalk [12]
allowed for a non-steady state simulation. As with the model of Maggio et al. [10],
the model of Lee and Lalk [12] determined the temperature distribution within the
stack using conservation of energy and the voltage of each cell in the stack was
found with the empirical model of Kim et al. [4].
When used as a power source for stationary or transportation applications, a PEM
fuel cell stack requires auxiliary equipment for providing fuel, oxidant and heat
removal; the stack operates as a component in a larger energy conversion system.
Barbir et al. [13] developed a model of a PEM fuel cell stack system consisting of
a stack, air compressor subsystem for the stack oxidant supply, gasoline reformer
subsystem for the stack fuel supply, and cooling subsystem. The electro-chemical
performance of the stack was modeled using an empirical, linear current-voltage
relationship and the system water balance and efficiency was investigated at various
operating pressures and temperatures. A PEM fuel cell stack system consisting of
a stack, air compression subsystem, compressed hydrogen supply subsystem, and
cooling subsystem was modeled by Cownden et al. [14]. The stack voltage and
power output were determined with the GSSEM of [6] and the efficiencies of both
the stack and system were examined.
This study formulates a PEM fuel cell stack model. The reactant distribution
within the stack is modeled by treating the stack manifold and gas flow channels
as a pipe network, and the voltages of the cells in the stack are determined with
the single cell, steady state, isothermal model developed previously by the present
authors [15]. U and Z configuration stacks operating with humidified hydrogen or
reformate as the fuel and humidified oxygen or air as the oxidant are simulated,
strategies for reducing the unequal distribution of reactants within the stack are
examined, and methods for the improvement of stack performance are described
based on the results of the present study.
2 Model formulation
In general, a PEM fuel cell stack consists of several PEM fuel cells connected in
series, as illustrated in Fig. 1. The cathode side of the PEM fuel cell is exposed to
the oxidant, while fuel is introduced to the anode side of the cell. Each cell in the
stack consists of several components. The bipolar plates conduct electrons and have
grooves, referred to as gas flow channels, that supply fuel or oxidant to the PEM fuel
cell. The anode and cathode electrode backings also conduct electrons and allow
reactants to access the catalyst layer. In the catalyst layer, electro-chemical reactions
convert the chemical energy of the fuel and oxidant to electrical energy. Reduction
of oxygen occurs in the cathode catalyst layer and the oxidation of hydrogen occurs
in the anode catalyst layer. The polymer electrolyte membrane conducts the proton
produced by hydrogen oxidation to the cathode for participation in the reduction
of oxygen.
Figure 1: Schematic of a polymer electrolyte membrane fuel cell stack.
The fuel and oxidant for each PEM fuel cell are supplied by the stack mani-
fold, with the anode manifold supplying fuel and the cathode manifold supplying
oxidant. The major component of the fuel for a PEM fuel cell is hydrogen, with
carbon dioxide and carbon monoxide being present if reformate fuel is utilized. The
presence of carbon monoxide severely degrades the performance of a PEM fuel
cell through the mechanism of CO poisoning [16]. Mitigation of CO poisoning is
possible with oxygen or air bleeding, whereby 1 to 4% oxygen is added to the fuel;
thus oxygen and nitrogen can also be present in the fuel stream. The oxidant used in
a PEM fuel cell is oxygen, with nitrogen being present if air is used as the oxygen
supply. The gas flow channels remove the water produced by the electro-chemical
reactions within the MEA and supply the humidity required to avoid polymer elec-
trolyte membrane dehydration; thus liquid and vapor phase water are present in
both the oxidant and fuel streams. In addition to fuel and oxidant, water is circu-
lated through cooling plates in order to remove the heat produced by the PEM fuel
cells and maintain a constant stack temperature.
The stack performance, often measured in terms of the stack voltage, can be
determined by:

N cell
N cell
Estack = Ecell − ηcp , (1)
1 1
where Ncell is the total number of fuel cells in the stack, Ecell is the voltage of each
cell (from bipolar plate to bipolar plate), and ηcp is the ohmic loss due to a cooling
plate. In this study, the voltage of each cell is found with the single cell model of
Baschuck and Li [15]. The single cell model is one-dimensional and assumes that
the cell is isothermal and operating in steady-state with fully humidified reactants.
Cell voltage is calculated with:
Ecell = Erev − ηa − |ηc | − 2ηbp − 2ηe − ηm , (2)
where Erev is the reversible cell voltage, ηa and ηc are the overpotentials attributed
to the anode and cathode catalyst layers, respectively. The voltage losses caused by
the bipolar plate, electrode backing and polymer electrolyte membrane are denoted
by ηbp , ηe , and ηm , respectively.
The reversible cell voltage is the cell potential obtained at thermodynamic equi-
librium. It is a function of temperature and reactant concentration through a modified
version of the Nernst equation. The cell voltage is reduced from the reversible cell
voltage by the overpotentials associated with the various components of the PEM
fuel cell. The voltage losses attributed to the bipolar plate and electrode backing
are the result of electron migration; the overpotential is calculated by considering
the electrode backing and bipolar plate as electrical resistances. Proton migration
is responsible for the voltage loss in the polymer electrolyte membrane and thus
the voltage loss is determined by the Nernst-Planck equation. The conductivity of
the polymer electrolyte is a function of hydration, but the single cell model [15]
assumes that the reactants and the polymer electrolyte are fully humidified; thus the
conductivity is constant. Therefore, the polymer electrolyte membrane overpoten-
tial is a function of the membrane properties, such as conductivity and thickness,
and current density.
For a PEM fuel cell operating with CO-free fuel, the cathode catalyst layer over-
potential is the major voltage loss. The anode and cathode catalyst layer overpoten-
tials are found by considering species conservation, proton and electron migration
within the catalyst layers. Proton and electron migration within the catalyst layers
are related to the protonic and ionic current through Ohm’s law.
Species conservation requires modeling of reaction kinetics and mass transport.
Oxygen reduction is modeled with the Butler-Volmer equation in the cathode cata-
lyst layer, while in the anode catalyst layer the adsorption and desorption of H2 , CO
and O2 , the electro-oxidation of the adsorbed hydrogen and carbon monoxide, and
the heterogeneous oxidation of H2 and CO by O2 are included in the reaction
kinetics. The reaction rates in the catalyst layers are functions of overpotential and
reactant concentrations; the functional dependency on concentration necessitates
consideration of mass transport. The concentrations within the catalyst layers are
influenced by resistance to mass transport from the gas flow channels to the electrode
backing, within the electrode backing, and within the catalyst layer. The mass trans-
fer from the gas flow channels to the gas flow channel/electrode backing interface
is calculated using a logarithmic mean concentration relationship. Mass transport
within the electrode backing and catalyst layers is assumed to be through diffusion
only and the diffusion coefficients are formulated such that a variable amount of liq-
uid water can exist within the pore space of the electrode backing and catalyst layers;
thus the PEM fuel cell can be simulated with a variable degree of water flooding.
Through consideration of species conservation, proton and electron migration,

ordinary differential equations are generated which can be solved for the catalyst
layer overpotential. Therefore, the catalyst layer overpotential depends on reactant
concentration, current density, and the properties of the gas flow channels, electrode
backing and catalyst layer, such as porosity, thickness, and conductivity.
Determination of the reversible cell potential and overpotentials requires several
input parameters, which can be classified as operating or design parameters. Design
parameters depend on the manufacture of the PEM fuel cell and include properties,
such as conductivity and porosity, and geometric dimensions. Design parameters
can be further classified according to the components of a PEM fuel cell; thus
there are bipolar plate, electrode backing, catalyst layer, and polymer electrolyte
membrane design parameters. The operating parameters include current density,
temperature, pressure, reactant composition and stoichiometry.
The PEM fuel cells in a stack will have the same design parameters. Due to the
series connection, the current density in each cell will be equal. As well, the circu-
lation of cooling water allows each cell to have the same temperature. However,
the pressure, reactant composition and stoichiometry can vary from cell to cell if
the mass flow rate and pressure distributions within the stack are unequal.
Therefore, the stack model presented here consists of two parts: the single cell
model and the stack flow model. The single cell model, as described above, deter-
mines the voltage of each cell in the stack based on the cell inlet pressure, tem-
perature, stoichiometry, and reactant composition in the gas flow channels, as well
as the current density and design parameters. In order to find the cell inlet pres-
sure, temperature, stoichiometry and reactant composition, the mass flow rate and
pressure distributions among each cell within the stack must be determined; this
constitutes the stack flow model. The voltage loss attributed to the cooling plate in
eqn (1) is determined by assuming that the cooling plate has the same overpotential
as the bipolar plate in the single cell model. The single cell model is described in
detail elsewhere [15] and only the stack flow model will be presented here.
2.1 Stack flow model
The mass flow rate and pressure distributions within the stack are coupled; thus
they must be solved simultaneously. The fuel or oxidant flow within the stack is
modeled as a pipe network, and is illustrated in Fig. 2. Two stack configurations
are considered in this study: the U and Z configurations. Other stack configurations
can be treated similarly. The U configuration is illustrated in Fig. 2(a) and is char-
acterized by the stack inlet and outlet being on the same end of the stack. The Z
configuration, which is shown in Fig. 2(b), has the flow inlet and outlet on opposite
ends of the stack. For both the U and Z configuration, the section of stack manifold
that supplies the reactants to the gas flow channels of the PEM fuel cells is referred
to as the top section, while the gas flow channels exit into the bottom section of
stack manifold.
If two PEM fuel cells and the connecting manifold sections are considered, as
illustrated in Fig. 2, then the pressure losses in the gas flow channels and manifold
Figure 2: Schematic of (a) U and (b) Z stack manifold configurations.
sections are related:

i i i+1 i+1 i
i
Pcell ṁcell,in − Ptop
i
ṁtop − Pcell ṁcell,in + αPbot
i
ṁbot = 0,
(3)
where α equals −1 for the U configuration, 1 for the Z configuration, and i = 1 at

the stack inlet. Each pressure loss is a function of the mass flow rate in the manifold
section or gas flow channel; thus the pressure and mass flow rate distributions must
be solved simultaneously. The mass flow rate in the top sections of the manifold
can be related to the mass flow rate entering the gas flow channels of each PEM
fuel cell:
i
j
ṁitop = ṁstack
in − ṁcell,in , (4)
j=1
where ṁstack
in is the mass flow rate at the inlet of the stack. The mass flow rate in the
bottom sections of the stack manifold depends on the manifold configuration:
 i

 j

 ṁcell,out Z configuration


j=1
ṁibot = (5)


Ncell

 j

 ṁcell,out U configuration,
j=i+1
where ṁicell,out is the mass flow rate exiting the gas flow channels. Because of the
reactions occurring within the anode and cathode catalyst layers, the mass flow rate
at the inlet of the gas flow channels is not equal to the value at the outlet. The inlet
and outlet mass flow rates are related:
ṁicell,out = ṁicell,in − ṁr , (6)
where ṁr is the mass consumed in the catalyst layers. Hence, the mass generated
will be written with a minus sign. The anode and cathode values of ṁr differ and
are functions of current density.
Equation (3), in conjunction with eqns (4) and (5), can be applied to generate
Ncell −1 equations with Ncell unknown ṁicell,in values. One final equation is required
to solve for the unknown ṁicell,in ’s, and this equation is the mass conservation for
the stack as a whole:

Ncell
ṁstack
in = ṁicell,in . (7)
i=1
With the addition of eqn (7), determination of the mass flow rate and pressure
distributions within the stack is possible. However, the relationship between the
pressure loss and mass flow rate must be determined first, as well as the value of
ṁr for the anode and cathode gas flow channels. These are presented in the next
sections.
2.2 Manifold pressure loss
The pressure loss in either the top or bottom section of stack manifold is found
with:
P = Pm + Pf , (8)
where Pm is the pressure loss due to the change in momentum of the fluid and
Pf is the pressure loss due to the wall friction. The losses due to the branching
of the flow, or minor losses, are not included in this formulation because appropri-
ate coefficients are not available for the flow conditions encountered. However, a
separate experimental investigation is under way to develop empirical correlation
for the minor loss coefficient associated with the branching/confluence flow, and
therefore, minor loss will be integrated in future studies. The pressure loss due to
a change in momentum is given by:
ṁ
Pm = (Vout − Vin ), (9)
Am
where ṁ is the mass flow rate in the manifold section, Am is the manifold cross-
sectional area and V is the velocity. The velocity within the manifold section can
be found with:
ṁ
V = , (10)
ρAm
where ρ is the density of the fluid. The fluid in both the manifold and gas flow
channels consists of a mixture of a multi-component gaseous phase and liquid
water. The density of this mixture depends on pressure, temperature and the mass
fraction of the gaseous phase; Appendix A describes the calculation of density.
The pressure loss due to friction in the manifold sections is given by:
2Cf Ls ρave (Vave )2
Pf = , (11)
dhstack
where Cf is the friction coefficient, Ls is the length of the manifold section between
the gas flow channels of the PEM fuel cells, and dhstack is the hydraulic diameter
of the stack manifold. The subscript “ave” refers to the arithmetic average of the
inlet and outlet values; because density is a function of pressure and mass fraction
of the gas phase, the inlet and outlet values will differ. The distance Ls depends on
the thickness of the PEM fuel cell and is equal to the thickness of two bipolar plates,
one MEA and one cooling plate. In this study, the cooling plate is assumed to have
the same dimensions as the bipolar plate. The friction coefficient is a function of
the Reynolds number [17]:
(
16(Redh )−1 Redh ≤ 2000
Cf = −1/4
(12)
0.079(Redh ) Redh ≥ 4000,
where Redh is the Reynolds number of the flow based on the hydraulic diameter:
ρave Vave dhstack

Redh = . (13)
µave
The viscosity of the fluid is denoted by µ and is calculated with the equations
described in Appendix A. As in eqn (11), the subscript “ave” denotes the arithmetic
average of the inlet and outlet values. For Reynolds numbers between 2000 and
4000, a linear relationship is used for the friction coefficient:

Redh − 2000
Cf = Cf |Redh = 2000 + Cf |Redh = 4000 − Cf |Redh = 2000 , (14)
2000
where Cf |Redh = 2000 and Cf |Redh = 4000 are the friction coefficient values at Reynolds
numbers of 2000 and 4000, respectively.
2.3 Cell pressure loss
The pressure loss calculation for the PEM fuel cell gas flow channels is similar to
the manifold sections, with the total pressure loss being calculated with eqn (8).
However, due to the change in mass flow rate between the inlet and outlet of the gas
flow channels, the formulation of Pm and Pf differ from the manifold section
formulation. The pressure loss due to momentum change is:
1
Pm = (ṁout Vout − ṁin Vin ) , (15)
Afc
where Afc is the cross-sectional area of a gas flow channel. The bipolar plate, with
the manifold and gas flow channels, is illustrated in Fig. 3. Two flow channel
configurations are shown: serpentine and parallel. Both configurations allow for
several gas flow channels to exist on a single bipolar plate. In this study, the gas
flow channels on a bipolar plate are assumed to have the same resistance to mass
flow; hence all of the gas flow channels on a single bipolar plate have the same
mass flow rate. Therefore, the pressure losses of all the gas flow channels on the
bipolar plate are found by considering only the flow in one gas flow channel. The
mass entering and exiting one gas flow channel can be found with:
ṁcell,in
ṁin = , (16)
nc
ṁcell,out
ṁout = , (17)
nc
where nc is the number of gas flow channels per bipolar plate.
Figure 3: Illustration of the bipolar plate with serpentine and parallel gas flow
channel configurations.
The pressure loss due to friction is determined with:
2Cf Lfc ρave (Vave )2

Pf = , (18)
dhfc
where Lfc is the length of a gas flow channel and dhfc is the gas flow channel
hydraulic diameter. The determination of the friction coefficient, average density
and average velocity in eqn (18) is the same as for the stack manifold sections.
This implies that the effect on the friction coefficient of wall suction due to the
reactants flowing into the catalyst layers for electrochemical reactions and the wall
blowing due to the reaction products coming out of the cathode catalyst layer are not
accounted for. The wall suction/blowing might in reality have significant impact on
the friction coefficient, their impact on the transport of momentum, heat and mass
is being investigated numerically and will be incorporated later once the relevant
information is available.
Further, in the above approach for the pressure loss calculation associated with
reactant stream in the flow channels built on bipolar plates, the minor loss associated
with the bend or turn of the flow direction, as mandated by the serpentine flow
channel design, has not been included due to the lack of relevant information. An
experimental study is currently under way to measure the associated pressure loss in
the serpentine flow channels, and correlations for the minor loss coefficient will be
developed and included to improve the model formulated here. However, it might
be pointed out that the minor loss for the branching/confluence flow associated with
the flow in the manifolds and in the serpentine flow channels will affect the total
pressure loss within the stack, but is relatively small in its impact on the reactant
mass distribution among the cells in the stack. Therefore, their effect on the final
results shown in this study is small as well, and can be neglected.
2.4 Mass consumed in the catalyst layers
The amount of mass consumed in the anode and cathode catalyst layers differ, but
in general can be written as:

ṁr = ṁir , (19)
i=species
where the summation is over all of the species present in the catalyst layers. For the
anode catalyst layer, the species present are H2 , CO, CO2 , O2 , N2 and H2 O. The
amount of H2 , CO and O2 consumed in the anode catalyst layer can be calculated
using Faraday’s law:
ṁH
r
2
ṁCO
r ṁr 2
O
Iδ Acell
+ −2 = , (20)
M̂ H2 M̂ CO M̂ O2 2F
where Iδ is the current density, Acell is the active area of a PEM fuel cell, F is the
Faraday constant, and M̂ is the molecular weight. The concentration of CO is
typically at the ppm level, while the concentration of O2 is 1 to 4 percent; as a

result the amount of O2 and CO removed from an anode gas flow channel will be
much less than the amount of H2 removed. Thus, in order to simplify the solution
O2
procedure, the values of ṁCO
r and ṁr are neglected. Since the amount of CO
consumed is negligible, the mass flow rate of CO2 will not change between the
inlet and outlet of a gas flow channel. Nitrogen, if present, does not react in the
anode catalyst layer. The water produced in the cathode catalyst layer is removed
by the cathode gas stream in practical PEM fuel cell stack operation, and thus this
study assumes that no water is added or removed from the anode gas flow channels.
Therefore, the only species that is consumed in the anode catalyst layer is H2 , and
the mass consumed for the anode is:
ṁr = ṁH 2
r . (21)
In the cathode catalyst layer, the species present are O2 , N2 , and H2 O. Nitrogen
does not react in the cathode catalyst layer and the amount of oxygen consumed
can be calculated with Faraday’s law:
Iδ Acell
r =
ṁO2 M̂ O2 . (22)
4F
All of the water produced by the PEM fuel cell is assumed to enter the cathode gas
flow channels; thus the amount of water consumed (actually generated, hence the
negative sign in the equation below) in the cathode catalyst layer becomes:
2O = −
Iδ Acell
ṁH
r M̂ H2 O . (23)
2F
The negative sign denotes that the water produced by the cathode catalyst layer is
added, not removed from, the cathode gas flow channels. Therefore, the amount of
mass consumed by the cathode catalyst layer becomes:
ṁr = ṁO
r + ṁr
2 H2 O
. (24)
2.5 Boundary conditions
The boundary conditions for the above model formulation are specified at the stack
inlet, that include the temperature, pressure, reactant composition. The reactant
flow rate is determined based on the stack current density specified and the stoi-
chiometry of the reactant desired. Then the stack performance and the reactant at
the stack outlet are determined by the model presented earlier. The following sec-
tion on numerical procedure provides the details on how the boundary conditions
are implemented for the model formulated.
3 Numerical procedure
As with the single cell model, the input parameters for the stack flow model are
classified as operating and design parameters. The design parameters are the stack
manifold dimensions and stack configuration. Operating parameters include the

stack current density, temperature, and the pressure, stoichiometry and reactant
composition at the stack inlet. The stack stoichiometry is defined as:
2FNHstack
Sastack = 2
, (25)
Ncell Iδ Acell
4FNHstack
Scstack = 2
, (26)
Ncell Iδ Acell
where the inlet molar flow rates of hydrogen and oxygen to the anode and cathode
sides of the stack, respectively, are denoted by NHstack
2
and NOstack
2
. Using stoichiome-
try and current density, the inlet molar flow rates of hydrogen in the anode manifold
and oxygen in the cathode manifold can be calculated. The mass flow rate at the
stack inlet can be determined with the molar flow rates and reactant composition.
The relationship between pressure and mass flow rate in the manifold sections
and gas flow channels is non-linear; therefore the mass flow rate and pressure dis-
tributions must be calculated using an iterative procedure. The numerical solution
begins with assumed values of ṁicell,in and two levels of iteration are then required.
The outer level of iteration solves for the pressure and mass flow rate distributions
within the stack. For a given mass flow rate, the procedure for determining the
pressure loss in the gas flow channels or stack manifold is iterative; thus an inner
i , P i and P i for the
level of iteration is required to find the values of Pcell top bot
estimate of mass flow rate.
3.1 Outer iteration
Using estimated values of ṁicell,in , values for ṁitop and ṁibot can be calculated using
eqns (4) and (5). Applying these values of mass flow rate to eqn (3) results in a
pressure residual:
i i i+1 i+1 i
rpi = Pcell
i
ṁcell,in − Ptop
i
ṁtop − Pcell ṁcell,in + αPbot
i
ṁbot . (27)
The residual can be set to zero by changing the assumed mass flow rates:
i i
0 = Pcell
i
ṁcell,in + ṁicell,in − Ptopi
ṁtop + ṁitop
i+1 i+1 i
− Pcell ṁcell,in + ṁi+1
cell,in + αPbot ṁbot + ṁbot .
i i
(28)
Combining eqns (27), (28) and using a Taylor series expansion yields a relationship
between the ṁicell,in , ṁitop and ṁibot :
i i
d Ptop
d Pcell
−rpi = ṁicell,in − ṁitop
d ṁicell,in d ṁitop
i+1 i
d Pcell d Pbot
− ṁi+1
cell,in + α ṁibot , (29)
d ṁi+1
cell,in
d ṁibot
where the derivatives are calculated numerically [18]. From eqns (4) and (5), ṁitop
and ṁibot can be rewritten in terms of ṁicell,in :

i
ṁitop = − ṁicell,in , (30)
j=1
 i




j
ṁcell,in Z configuration


j=1
ṁibot = (31)


N cell

 j

 ṁcell,in U configuration.
j=i+1
Thus, eqn (29) can be used to make Ncell − 1 equations for the Ncell values of
ṁicell,in ; it becomes for the Z configuration stack:
i i
dPtop i
dPcell j
−rpi = ṁicell,in + ṁcell,in
d ṁicell,in d ṁitop j=1
i+1
dPcell i
dPbot
i
j
− ṁi+1
cell,in + ṁcell,in . (32)
d ṁi+1
cell,in
d ṁibot j=1
and for the U configuration stack:

i i
dPtop i
dPcell j
−rpi = ṁicell,in + ṁcell,in
d ṁicell,in d ṁitop j=1
i+1
dPcell i
dPbot
i
j
− ṁi+1
cell,in − ṁcell,in . (33)
i+1
d ṁcell,in d ṁibot j=1
The final equation needed to solve for the ṁicell,in comes from the overall mass
conservation eqn (7):

N cell
ṁicell,in = 0. (34)
i=1
Equations (34) and either (32) or (33) can be solved for the corrections to the
assumed mass flow rate distribution with a linear equation solver, such as LU
decomposition [18]. The mass flow rate distribution estimation is then updated:
ṁicell,in |new = ṁicell,in |old + ṁicell,in . (35)
Using the updated values of ṁicell,in , new values of ṁicell,in are found until con-
vergence, which is achieved when:

N cell
i
rp ≤ 1 × 10−3 Pa. (36)
i=1
3.2 Inner iteration
The pressure loss in the gas flow channels and stack manifold are functions of the
inlet and outlet mass flow rate, density, and viscosity. Density and viscosity are
functions of the mass fraction of the gas phase (χ), the species mole fractions
in the gas phase (xi ), and the pressure. However, the outlet pressure is not known
until the pressure loss is calculated; hence an iterative procedure is required to
determine the pressure loss in the gas flow channels or stack manifold.
From the estimated mass flow rates in the outer iteration, the pressure, mass
fraction of gas phase and species mole fractions in the gas phase are known at the
inlet of the gas flow channel or stack manifold. In order to obtain values of χ and
xi at the outlet, the mass flow rate of each species at the outlet must be known. For
the species other than water, calculation of the mass flow rate is straightforward.
The mass flow rates of the species other than water at the inlet and outlet of the
stack manifold sections are equal; the mass flow rates of H2 in the anode and O2 in
H O
the cathode gas flow channels are reduced by ṁr 2 and ṁr 2 , respectively, and
all other non-water mass flow rates are the same at the inlet and outlet of the gas
flow channels.
The amount of water exiting in liquid or vapor phase depends on the outlet
pressure of the gas flow channels or stack manifold section. The maximum mole
fraction of water in the gaseous phase at the outlet is:
H2 O
Psat
max
xH 2 O,out
= k
, (37)
Pout
H2 O k is the estimated value of
where Psat is the saturation pressure of water and Pout
outlet pressure. With this mole fraction, the maximum mass flow rate of water vapor
can be found:
max
xH2 O(g) ,out
M̂ H2 O(g) ṁi,out
ṁmax
H2 O(g) ,out = . (38)
1 − xH
max
2 O(g) ,out
M̂ i
i =H2 O(g) ;H2 O()
The total mass flow rate of water (liquid and vapor) is:
H2 O,out = ṁH2 O(g) ,in + ṁH2 O() ,in − ṁr

ṁtotal H2 O
, (39)
where ṁrH2 O is only non-zero for the cathode gas flow channels. Using eqns (38)
and (39), the mass flow rates of the liquid and vapor water can be found:

ṁtotal
H2 O,out H2 O(g) ,out ≥ ṁH2 O,out
ṁmax total
ṁH2 O(g) ,out = (40)
ṁmax ṁmax total
H2 O(g) ,out H2 O(g) ,out < ṁH2 O,out ,

0 H2 O(g) ,out ≥ ṁH2 O,out
ṁmax total
ṁH2 O() ,out = (41)
ṁtotal
H2 O,out − ṁH2 O(g) ,out ṁH2 O(g) ,out < ṁH2 O,out .
max max total
Knowledge of the outlet mass flow rates allows for the calculation of the outlet
mole fractions in the gas phase and the mass fraction of the gas phase:
ṁi,out /M̂ i
xi,out = , (42)
j =H2 O() ṁj,out /M̂ j
ṁH2 O() ,out
χout = 1 − . (43)
ṁi,out
With the values of pressure, mole fraction, mass fraction and mass flow rate at
the inlet and outlet, the pressure loss in either the gas flow channel or stack manifold
can be determined with eqn (8). This value of P can then be used to generate a
new estimate for the outlet pressure:
k+1
Pout = Pin − P. (44)
Iteration continues until:

P k+1 − P k
out out
≤ 1 × 10−6 . (45)
k
Pout
3.3 Numerical procedure summary
Using the inner and outer iterations, the mass flow rate and pressure distributions
in the stack are solved with the following procedure, for a given current density
output from the stack:
1. An initial estimate for ṁicell,in is made.
2. Using the inner iteration and the estimated value of ṁicell,in , Pcell
i , P i , and
top
Pboti are calculated.
3. The values of ṁicell,in are calculated and used to generate new values for
ṁicell,in .
4. Steps 2 to 3 (the outer iterations) are repeated until convergence.
5. The current density output can be varied, and the above procedures repeated
in order to determine the stack performance for a range of loading conditions.
The solution of the stack flow model provides the mass flow rate, composition and
pressure at the inlet of each PEM fuel cell in the stack. These values are used by the
single cell model of [15] to calculate values of Ecell for each cell. The cell voltages,
along with the value of ηcp also calculated by the model of [15], are used in eqn (1)
to determine the stack voltage.
4 Results and discussion

Figure 4 compares the performance of a single PEM fuel cell operating indepen-
dently with U and Z configuration stacks consisting of 50 cells. This comparison
Figure 4: Polarization curves comparing the performance of a single PEM fuel

cell with 50 cell, U and Z configuration stacks operating with (a) H2 /air,
(b) reformate/air and (c) H2 /O2 reactants. The stack polarization curves
are plotted by dividing the stack voltage by the number of cells in the
stack.
is accomplished by comparing the single cell voltage to the stack voltage divided
by the number of cells in the stack (average cell voltage). Both the single cell and
the cells of the stacks have an active area of 240 cm2 and a serpentine gas flow
channel configuration with three gas flow channels per bipolar plate. The relevant
design parameters for various cell and stack components, such as the bipolar plate,
electrode backing, catalyst layer and polymer electrolyte membrane are given in
Table 1. The anode and cathode sides of the cells and stacks are assumed to be
the same. The cross-sectional areas of both the anode and cathode stack manifolds
are equal to 1.54 cm2 , and the design parameters for the anode and cathode sides
of the cell or stack are considered to be equal. The single cell and stacks operate
Table 1: Design parameters for various cell and stack components.
Parameter Value
Bipolar Plate ρbp 6 × 10−5 · m

W 0.155 m
L 0.155 m
hp 0.002 m
hc 0.002 m
ws 0.00262 m
wc 0.002 m
nc 3
ng 11
Electrode Backing ρebulk 6 × 10−5 · m
δe 2.5 × 10−4 m
φe 0.4
Catalyst Layer δc 2.0465 × 10−5 m
mPt 0.004 kg/m2
fPt 0.2
m 0.9
κs 72700 S/m
Polymer Electrolyte δm 1.64 × 10−4 m
Membrane KE 7.18 × 10−20 m2
Kp 1.8 × 10−18 m2
CH+ 1200 mole/m3
DH + 4.5 × 10−9 m2/s
Stack wm 0.0124 m
hm 0.0124 m
Ncell 50
with anode and cathode inlet pressures of 250 kPa, stoichiometries of 1.1 and 2,
respectively, and a temperature of 358 K.
The performance of a single cell and stack operating with fully humidified hydro-
gen as the fuel and fully humidified air, consisting of 21% O2 and 79% N2 , as the
oxidant (H2 /air reactants) is compared in Fig. 4(a). Figure 4(b) compares single cell
and stack performance with fully humidified reformate, consisting of 75% H2 and
25% CO2 , as the fuel and fully humidified air as the oxidant (reformate/air reac-
tants), while single cell and stack performance with fully humidified hydrogen as
the fuel and fully humidified oxygen as the oxidant (H2 /O2 reactants) is compared
in Fig. 4(c). For operation with H2 /air and reformate/air reactants, the average cell
voltage in the stack is less than the voltage of a single cell operating independently.
Figure 5: Voltage of each cell in 50 cell U and Z configuration stacks operating

with (a) H2 /air reactants and at the current density of 0.41 A/cm2 and
(b) reformate/air reactants and at the current density of 0.32 A/cm 2 .
The difference between cell and stack performance is greater when reformate/air
reactants are utilized. Operation with H2 /O2 reactants results in the average cell
voltage in the stack being almost the same as the single cell voltage; this agrees
with the experience that a stack designed for H2 /O2 reactants, in general, will not
operate properly for H2 /air or reformate/air reactants. From Fig. 4, it is also apparent
that the Z configuration stacks have a better performance than the U configuration
stacks.
The performance differences between the stacks and single cells are caused by
voltage variations among the cells in the stack. This cell-to-cell voltage variation
is illustrated for U and Z configuration stacks in Fig. 5, with Figs 5(a) and 5(b)
showing cell-to-cell variation for H2 /air and reformate/air reactants, respectively.
The average cell voltage in the stack of Fig. 5 is approximately 0.6, making the
current density used for the H2 /air simulation 0.41 A/cm2 and 0.32 A/cm2 for the
reformate/air simulation. The voltage of each cell in the stack is compared to the
voltage of a single cell operating independently, with the cells of the stack being
numbered starting at the stack inlet. Near the stack inlet and outlet, the cells of
the Z configuration stacks have a higher voltage than the single cell operating
independently, while only the cells near the stack inlet have a higher performance
than the single cell for the U configuration stacks. However, the majority of the cells
in the stacks have voltages less than the single cell, which results in the average
cell voltage of the stack being less than the single cell voltage.
The cell-to-cell voltage variation can be quantified by the voltage spread of the
stack:
E max − E min
SE = cellN cell × 100%, (46)
1 cell
Ncell 1 Ecell
max and E min are the maximum and minimum cell voltages, respectively,
where Ecell cell
within the stack. For the H2 /air reactants of Fig. 5(a), the voltage spread for the
U configuration stack is 9.8% and 5.0% for the Z configuration stack. Operation
with reformate/air reactants yields voltage spreads of 15% and 6.2% for the U and
Z configuration stacks, respectively. Although not illustrated in Fig. 5, operation
with H2 /O2 reactants and at the current density of 0.88 A/cm2 , which corresponds
to the average cell voltage of 0.6 V, results in voltage spreads of only 0.0021% and
0.0018% for the U and Z configuration stacks, respectively. These voltage spread
values show that a high voltage spread corresponds to a poor stack performance
when compared to a single cell. Z configuration stacks have a better performance
and lower voltage spread than U configuration stacks, while operation with refor-
mate/air reactants results in the lowest performance and the highest voltage spread.
Therefore, in practice, a low voltage spread and uniform cell-to-cell performance
is desirable in order to maximize cell performance.
The cell-to-cell voltage variations shown in Fig. 5 are the result of the pressure
loss distribution, leading to non-uniform distribution of mass flow rate among the
cells within the stack. This unequal distribution of mass flow rate among the cells
in the stack creates cell-to-cell variations in the anode and cathode stoichiometry.
To illustrate how the mass flow rate distribution affects the cell voltage distribution
in a stack, the anode and cathode cell stoichiometry in a Z configuration stack is
presented in Fig. 6. Operation with H2 /air reactants and at the current density of
0.41 A/cm2 results in little anode stoichiometry variation among the cells in the
stack, as illustrated in Fig. 6(a). However, the cathode stoichiometry variation is
similar to the cell voltage distribution shown in Fig. 5(a). The cathode stoichiometry
varies more than the anode stoichiometry due to the larger mass flow rate in the
cathode gas flow channels. The average Reynolds number in the anode gas flow
channels is around 32, but due to the presence of inert N2 gas, the Reynolds number
is approximately 640 in the cathode gas flow channels.
The use of reformate/air reactants and a current density of 0.32 A/cm2 results in
variation in both the anode and cathode stoichiometry, as illustrated in Fig. 6(b).
The anode stoichiometry shows more variation for reformate fuel than for hydrogen
fuel due to the higher flow rate caused by the presence of the inert CO2 ; the average
Reynolds number in the corresponding gas flow channels is about 112 for the
reformate fuel.
In a manner similar to the cell voltage variation, the cell-to-cell variation in
stoichiometry can be quantified by the stoichiometry spread of the stack:
Scell
max
− Scell
min
SS = Ncell
× 100%, (47)
1 Scell
1
Ncell
Figure 6: Anode and cathode stoichiometry of each cell in a 50 cell, Z configura-

tion stack operating with (a) H2 /air reactants and at the current density
of 0.41 A/cm2 and (b) reformate/air reactants and at the current density
of 0.31 A/cm 2 .
where Scellmax
and Scell
min
are the maximum and minimum values, respectively, of
cell stoichiometry within the stack. For the H2 /air reactants of Fig. 6(a), the anode
stoichiometry spread is only about 1%, while the cathode stoichiometry spread
is 50%. The use of reformate/air reactants, as illustrated in Fig. 6(b), results in
an anode stoichioimetry spread of 10% and a cathode stoichiometry spread of
42%. The cathode stoichiometry spread is greater for the case with H2 /air reac-
tants than for reformate/air reactants because the H2 /air case uses a larger current
density. With a larger current density, the mass flow rate increases and generates a
larger stoichiometry spread; the average Reynolds number in the cathode gas flow
channels, when reformate/air reactants and a current density of 0.32 A/cm2 are
used, is approximately 501 while, as mentioned previously, the average Reynolds
number in the cathode gas flow channels is about 640 for the H2 /air case. However,
the larger anode stoichiometry spread results in a larger cell voltage spread when
reformate/air reactants are used.
Obtaining a uniform mass flow rate distribution within a PEM fuel cell stack
reduces the voltage spread and improves stack performance. The degree of flow
uniformity through the side branches of a manifold system depends on the ratio of
cross-sectional area between the side branches and manifold, and the resistance to
mass flow in the side branches and manifold [19]. The ratio of cross-sectional area
between the side branches and manifold is defined as:
nb Ab
f¯ = , (48)
Am
Figure 7: Effect of stack manifold cross-sectional area on voltage spread for (a) Z
and (b) U configuration stack with 3 gas flow channels per bipolar plate.
The current density for H2 /air reactants is 0.41 A/cm2 , reformate/air reac-
tants is 0.32 A/cm2 , and H2 /O2 reactants is 0.88 A/cm2 .
where nb is the number of side branches, Ab is the cross-sectional area of each side
branch, and Am is the cross-sectional area of the manifold. For the PEM fuel cell
stacks considered in this study, eqn (48) becomes:
nc Ncell Afc
f¯ = . (49)
Am
In principle, a uniform mass flow rate distribution is achieved when f¯ approaches

zero. However, nearly uniform flow can be obtained in reality at a finite value of f¯.
From eqn (49), f¯ can be reduced by increasing the cross-sectional area of the stack
manifold.
The effect of stack manifold cross-sectional area, or manifold area, on volt-
age spread for stacks operating with H2 /air reactants and a current density of
0.41 A/cm2 , reformate/air reactants and a current density of 0.32 A/cm2 , and
H2 /O2 reactants and a current density of 0.88 A/cm2 is illustrated in Fig. 7.Although
the different mass flow rates in the anode and cathode manifold could warrant dif-
ferent manifold areas, the results shown in Fig. 7 are for anode and cathode manifold
areas that are equal. It is seen that the voltage spread decreases as the manifold area
is increased if H2 /air or reformate/air reactants are used. Within the range of mani-
fold areas considered in Fig. 7, the voltage spread is almost zero and independent of
Table 2: Critical stack manifold areas and f¯ values, corresponding to a voltage

spread of 1%, for stacks using three gas flow channels per bipolar plate.
Reactants Configuration Current Density Manifold Area f¯
H2 /air Z 0.41 A/cm2 3.3 cm2 1.8

reformate/air Z 0.32 A/cm2 3.8 cm2 1.6
H2 /air U 0.41 A/cm2 4.9 cm2 1.2
reformate/air U 0.32 A/cm2 5.6 cm2 1.1
manifold area if H2 /O2 reactants are utilized. In reality, the critical manifold area,
below which significant cell-to-cell voltage variations occur, is much smaller for
the H2 /O2 reactants.
The U configuration stacks exhibit greater cell-to-cell variation than the Z con-
figuration stacks, while operation with reformate/air reactants results in a larger
voltage spread than operation with H2 /air reactants. This is evident in Table 2,
which lists the critical manifold areas and f¯ values corresponding to a voltage
spread of 1%. The U configuration stacks require a larger manifold area than the
Z configuration stacks in order to obtain a voltage spread of 1%. For a given stack
configuration, reformate/air reactants require a larger manifold area than H2 /air
reactants; thus stacks designed for H2 /air reactants may not operate effectively if
switched to reformate/air reactants.
Decreasing the number of gas flow channels per bipolar plate (nc ) can also
reduce the value of f¯ in eqn (49), resulting in less cell-to-cell voltage variation.
To illustrate this, Fig. 8 shows the effect of manifold area on voltage spread for
U and Z configuration stacks employing a bipolar plate with one serpentine gas
flow channel. Table 3 tabulates the bipolar plate design parameters used for the
simulation results of Fig. 8. Other than the value of nc used (3 in Fig. 7 and 1
in Fig. 8), all other parameters used in the simulations of Fig. 8 are the same as
in Fig. 7, such as stack configuration, reactant composition and current density.
The trends illustrated in Figs 7 and 8 are also similar. Increasing the manifold area
decreases the voltage spread if H2 /air or reformate/air reactants are used, while the
manifold area does not affect the voltage spread when H2 /O2 reactants are used.
For a given manifold area, the Z configuration stack has a smaller voltage spread
than the U configuration stack.
Although reducing the number of gas flow channels does not affect the general
relationship between voltage spread and manifold area, a reduction in nc signifi-
cantly decreases the magnitude of the voltage spread. Therefore, the manifold area
corresponding to a voltage spread of 1% would be expected to be smaller if one,
rather than three, gas flow channel grooves exist on each bipolar plate. Table 4 lists
the critical manifold areas and f¯ values corresponding to a voltage spread of 1%
for stacks utilizing one gas flow channel per bipolar plate. Comparing the entries
of Tables 2 and 4, it is evident that the critical manifold areas for the nc = 1 stacks
are approximately one-third of those when nc = 3. However, less variation occurs
Figure 8: Effect of stack manifold cross-sectional area on the voltage spread for
(a) Z and (b) U configuration stack with 1 gas flow channel per bipo-
lar plate. The current density for H2 /air reactants is 0.41 A/cm2 , refor-
mate/air reactants is 0.32 A/cm2 , and H2 /O2 reactants is 0.88 A/cm2 .
Table 3: Bipolar plate design parameters utilizing one gas flow

channel per bipolar plate.
Parameter Value
ρbp 6 × 10−5 · m
W 0.155 m
L 0.155 m
hp 0.002 m
hc 0.002 m
ws 0.00213 m
wc 0.002 m
nc 1
ng 37
between the f¯ values, with the f¯ values for stacks with nc = 1 being approximately
20% larger than the values with nc = 3.
Both the manifold cross-sectional area and the number of gas flow channels can
influence the pressure loss of the stack. The pressure loss for the cathode stream in a
Z configuration stack is illustrated in Fig. 9. Air is used as the cathode reactant and
Table 4: Critical stack manifold areas and f¯ values, corresponding to a voltage

spread of 1%, for stacks using one gas flow channels per bipolar plate.
Reactants Configuration Current Density Manifold Area f¯
H2 /air Z 0.41 A/cm2 1.1 cm2 2.2

reformate/air Z 0.32 A/cm2 1.4 cm2 1.8
H2 /air U 0.41 A/cm2 1.6 cm2 1.6
reformate/air U 0.32 A/cm2 2.0 cm2 1.3
Figure 9: Cathode stream stack pressure loss for a Z configuration stack operating
with H2 /air reactants, a current density of 0.41 A/cm2 , and 1 or 3 gas
flow channels per bipolar plate (nc ).
the current density is 0.41 A/cm2 . The pressure loss increases by approximately 9
times if the number of gas flow channels per bipolar plate is reduced from three
to one, caused by the combined effects of a lengthened flow path and higher flow
velocity in the gas flow channel. However, the general effect of manifold area on
the stack pressure loss is the same regardless of the number of gas flow channels.
As the manifold area is increased, the pressure loss decreases initially, but then
becomes independent of manifold area. This plateau in the pressure loss/manifold
area plot indicates that the manifold area is sufficiently large such that the friction
loss attributed to the manifold does not contribute to the overall stack pressure
loss; the stack pressure loss depends almost solely on the gas flow channels. In
practice, nc = 1 is the minimum that can be used. However, depending on the
cell size, the flow path and the momentum loss in the gas flow channel may be
too large, resulting in an excessive pressure loss; this large pressure loss increases
the parasitic power required for gas compression and decreases the overall stack
effeciency. Thus, nc = 3 is often used for large stacks consisting of large cells.
In a pipe network, the distribution of mass flow rate is influenced by the flow
resistance in each pipe; pipes with a high resistance to mass flow have smaller mass
flow rates than pipes with a low resistance to mass flow. Therefore, one method of
reducing the cell-to-cell variation of mass flow rate and cell voltage in a PEM fuel
cell stack is to alter the resistance to mass flow in the gas flow channels. In this
study, the resistance to mass flow in the gas flow channels is altered through the
addition of a term in the pressure loss due to friction, eqn (18):
2
2Cf Lfc ρave Vave
Pf = ζ , (50)
dhfc
where ζ is the flow resistance parameter that, if greater than one, increases the
resistance to mass flow in the gas flow channel. Practically, the resistance to mass
flow can be increased by either decreasing the gas flow channel hydraulic diameter,
increasing the length of the gas flow channel, or installing a flow obstruction in the
gas flow channel.
Reducing voltage spread by varying the flow resistance parameter among the
cells in the stack can be illustrated by considering a Z configuration stack operating
with H2 /air reactants and a current density of 0.41 A/cm2 . In order to reduce cell-to-
cell cathode stoichiometry variation, the flow resistance parameter for the cathode
gas flow channels is set according to:
Sci
ζi = , (51)
Scmin
where Sci is the cathode stoichiometry of cell i in Fig. 6(a) and Scmin is the
minimum cathode stoichiometry in Fig. 6(a). Using these values of ζi , the cell-
to-cell variation of cathode stoichiometry is greatly reduced when compared to
the variation resulting from using a constant ζ = 1, as illustrated in Fig. 10. The
variable values of ζi are shown in the inset of Fig. 10. The reduction in cathode
stoichiometry variation results in a 0.5% voltage spread, which is much smaller
than the 5% voltage spread achieved by using ζ = 1 for all cells in the stack.
Significantly, the use of the variable flow resistance parameter only increases the
cathode stack pressure loss from 1.8 kPa to 2.0 kPa; this increase of 11% is much
smaller than the pressure loss increase of over 800% incurred if the voltage spread
reduction is achieved by reducing the number of gas flow channels per bipolar
plate from three to one. Thus, the energy required to overcome the stack pressure
loss would be less if a variable ζi is used, allowing for greater system efficiency.
However, varying the flow resistance from cell to cell may be difficult to implement
since it requires the customization of each bipolar plate.
Finally, it should be pointed out that the above modeling results are based on a
5 kW 50 cell PEM fuel cell stack. However, we don’t have access to the test results
Figure 10: Cathode stoichiometry for a Z configuration stack using a constant flow
resistance parameter of ζ = 1 and the variable flow resistance parameter
(dashed curve), which is given in the inset.
for the validation of the present model. Thus the above results should be treated
as qualitative for the moment. However, the single cell model used in this study
has been validated against the single cell test results [15], and the stack flow model
has been validated against experimental results for a different application [23, 24].
In this sense, the present model can be used as a useful tool for the design and
optimization of PEM fuel cell stacks.
5 Conclusions
The performance of 50 active cell, U and Z configuration stacks were simulated
using a mathematical model that consisted of two parts. The first part of the model
determined the distribution of mass flow rate and pressure in the stack manifold and
the gas flow channels of the fuel cells through a hydraulic network analysis. The
results of the hydraulic network analysis were used as an input parameter for the
second part of the model, which calculated the voltage of each cell in the stack with
a previously developed, mathematical model. Therefore, the distribution of mass
flow rate within the stack influenced the voltage of each cell in the stack; a small
mass flow rate in a cell resulted in a low cell voltage. This relationship between
mass flow rate distribution and individual cell voltage lead to the performance
of fuel cells operating within a stack being lower when compared to a fuel cell
operating independently. The magnitude of the performance difference between
single cells and cells within stacks was larger for U than for Z configuration stacks,
and greater when the anode/cathode reactant compositions were fully humidified
reformate/air, rather than H2 /air. Eliminating the performance differential could be
achieved by ensuring that each cell in the stack had the same mass flow rate, and
three methods of attaining a uniform mass flow rate distribution were examined.
The first method was increasing the cross-sectional area of the stack manifold,
while the second involved decreasing the number of gas flow channels per bipolar
plate. Finally, a uniform distribution of mass flow rate within the stack could be
achieved by varying, from cell to cell, the resistance to mass flow in the gas flow
channels.
Acknowledgments
This work is a part of a large research project on PEM fuel cells and related tech-
nologies supported by the Auto 21 NCE, NRC Institute for Fuel Cell Innovation,
Hydrogenics Corporation and PalCan Fuel Cells Ltd. Partial funding is also pro-
vided by the Natural Sciences and Engineering Research Council of Canada.
Appendix A: Property determination
The fluid in the gas flow channels and stack manifold is assumed to be composed of
a multi-component gas phase with liquid water droplets. The density and viscosity
of the fluid are required to calculate the pressure loss in the gas flow channels and
stack manifold. The density of the fluid is given by [20]:
1 χ 1−χ
= + , (52)
ρ ρg ρH2 O()
where ρg is the density of the gas mixture and ρH2 O() is the density of liquid water.
The density of the gas phase is calculated with the ideal gas relation:
xi P M̂ i
ρg = , (53)
RT
i=species
where the summation includes all gaseous species present in the gas flow channel
or stack manifold. The liquid water density can be found in [21] and is equal to
968 kg/m 3 at a temperature of 85◦ C.
The viscosity of the fluid in the gas flow channels and stack manifold can be
found with [20]:
1 χ 1−χ
= + , (54)
µ µg µH2 O()
where µg is the viscosity of the gas mixture and µH2 O() is the viscosity of liquid
water. The Wilke correlation is used to find the viscosity of the gas mixture [22]:

N
xi µi
µg = N
i=1 j=1 xj φij
−1/2   (55)
1/2 1/4 2
1 M̂ i M̂ j
φij = √ 1 + 1 + µi  ,
8 M̂ j µ j M̂ i
where the individual gas viscosities are found using a power law [17].
n
µi T
= . (56)
µi,293 K 293 K
The viscosity of liquid water can be found with the relationship [17]:

µH2 O()
ln = −1.704 − 5.306 + 7.003 2 ,
µo
273 K
= , (57)
T
µo = 1.788 × 10−3 kg/(m · s).
References
[1] Hamelin, J., Agbossou, K., Laperriere, A., Laurencelle, F. & Bose, T.,
Dynamic behavior of a PEM fuel cell stack for stationary applications. Inter-
national Journal of Hydrogen Energy, 26(6), pp. 625–629, 2001.
[2] Bar-On, I., Kirchain, R. & Roth, R., Technical cost analysis for PEM fuel
cells. Journal of Power Sources, 109(1), pp. 71–75, 2002.
[3] Costamagna, P. & Srinivasan, S., Quantum jumps in the PEMFC science and
technology from the 1960s to the year 2000 Part II. Engineering, technology
development and application aspects. Journal of Power Sources, 102(1–2),
pp. 253–269, 2001.
[4] Kim, J., Lee, S.M., Srinivasan, S. & Chamberlin, C., Modeling of proton
exchange membrane fuel cell performance with an empirical equation. Jour-
nal of the Electrochemical Society, 142(8), pp. 2670–2674, 1995.
[5] Chu, D., Jiang, R. & Walker, C., Analysis of PEM fuel cell stacks using
an empirical current-voltage equation. Journal of Applied Electrochemistry,
30(3), pp. 365–370, 2000.
[6] Amphlett, J., Baumert, R., Mann, R., Peppley, B., Roberge, P. & Harris, T.,
Performance modeling of the Ballard Mark IV solid polymer electrolyte
fuel cell II. Empirical model development. Journal of the Electrochemical
Society, 142(1), pp. 9–15, 1995.
[7] Fowler, M., Mann, R., Amphlett, J., Peppley, B. & Roberge, P., Incorporation
of voltage degradation into a generalised steady state electrochemical model
for a PEM fuel cell. Journal of Power Sources, 106(1–2), pp. 274–283, 2002.
[8] Nguyen, T. & White, R., A water and heat management model for proton-
exchange-membrane fuel cells. Journal of the Electrochemical Society,
140(8), pp. 2178–2186, 1993.
[9] Thirumulai, D. & White, R., Mathematical modeling of proton-exchange-
membrane-fuel-cell stacks. Journal of the Electrochemical Society, 144(5),
pp. 1717–1723, 1997.
[10] Maggio, G., Recupero, V. & Mantegazza, C., Modelling of temperature dis-
tribution in a solid polymer electrolyte fuel cell stack. Journal of Power
Sources, 62(2), pp. 167–174, 1996.
[11] Patel, D., Maru, H., Farooque, M. & Ware, C., Methodology for predic-
tive testing of fuel cells. Journal of the Electrochemical Society, 131(12),
pp. 2750–2756, 1992.
[12] Lee, J. & Lalk, T., Modeling fuel cell stack systems. Journal of Power
Sources, 73(2), pp. 229–241, 1998.
[13] Barbir, F., Balasubrumanian, B. & Neutzler, J., Trade-off design analysis of
operating pressure and temperature in PEM fuel cell systems. Proceedings
of the ASME Advanced Energy Systems Division, pp. 305–315, New York,
1999. ASME Advanced Energy Systems Division, The American Society of
Mechanical Engineers.
[14] Cownden, R., Nahon, M. & Rosen, M., Modelling and analysis of a solid
polymer fuel cell system for transporation applications. International Jour-
nal of Hydrogen Energy, 26(6), pp. 615–623, 2001.
[15] Baschuk, J. & Li, X., Mathematical model of a PEM fuel cell incorporating
CO poisoning and O2 (air) bleeding. International Journal of Global Energy
Issues, 20(3), pp. 245–276, 2003.
[16] Oetjen, H.F., Schmidt, V., Stimming, U. & Trila, F., Performance data of a
proton exchange membrane fuel cell using H2 /CO as fuel gas. Journal of
the Electrochemical Society, 143(12), pp. 3838–3842, 1996.
[17] White, F., Fluid Mechanics, Third Edition, McGraw-Hill: New York, 1994.
[18] Press, W., Teukolsky, S., Vetterling, W. & Flannery, B., Numerical Recipes
in C: The Art of Scientific Computing, Second Edition, Cambridge University
Press: Cambridge, 1992.
[19] Idelchik, I., Handbook of Hydraulic Resistance, Third Edition, CRC Press:
Boca Raton, 1994.
[20] Wallis, G., One-dimensional Two-phase Flow, McGraw-Hill: New York, 1969.
[21] Reynolds, W. & Perkins, H., Engineering Thermodynamics, Second Edition,
McGraw-Hill: New York, 1977.
[22] Reid, R., Prausnitz, J. & Sherwood, T., The Properties of Gases and Liquids,
McGraw-Hill: New York, 1977.
[23] Zhang, J. & Li, X., Coolant flow distribution and pressure loss in ONAN
transformer windings - Part I: Theory and model development. IEEE Trans-
actions on Power Delivery, 19(1), pp. 186–193, 2004.
[24] Zhang, J. & Li, X., Coolant flow distribution and pressure loss in ONAN
transformer windings, Part II: Optimization of design parameters. IEEE
Transactions on Power Delivery, 19(1), pp. 194–199, 2004.
Nomenclature
Acell Active area of fuel cell (m2 )
Am Cross-sectional area of stack manifold (m2 )
Afc Cross-sectional area of gas flow channel (m2 )
Cf Wall friction coefficient
CH+ Fixed charge concentration in the polymer electrolyte (mole/m3 )
DH+ ,ref Diffusion coefficient of H+ in the polymer electrolyte (m2 /s)
dh Hydraulic diameter (m)
E Cell or stack voltage (V)
fpt Mass ratio of platinum to carbon support
fw Volume fraction of electrode backing void flooded by liquid water
F Faraday constant (96495 C/mole)
hc Depth of gas flow channel (m)
hm Height of stack manifold cross-section (m)
hp Thickness of solid portion of bipolar plate (m)
Iδ Cell current density (A/m2 )
KE Electrokinetic permeability of polymer electrolyte membrane (m2 )
Kp Hydraulic permeability of polymer electrolyte membrane (m2 )
m Fraction of catalyst layer void space occupied by polymer electrolyte
H2 O() Fraction of catalyst layer void space occupied by liquid water
L Length of fuel cell active area (m)
Lfc Length of gas flow channel (m)
Ls Length of stack manifold between cells (m)
ṁ Mass flow rate (kg/s)
mPt Platinum mass loading per unit electrode area (kg/m2 )
ṁr Amount of mass consumed in the catalyst layer (kg/s)
M̂ i Molecular weight of species i (kg/mole)
Ncell Number of cells in a stack
nc Number of bipolar plate gas flow channels
ng Number of times a serpentine gas flow channel traverses the
bipolar plate
Nistack Molar flow rate of species i at the stack inlet (mole/s)
P Pressure loss (Pa)
Pf Pressure loss due to friction (Pa)
Pm Pressure loss due to momentum change (Pa)
P Pressure (Pa)
R Universal gas constant (8.314 J/mole · K)
Redh Reynolds number
rpi Pressure residual (Pa)
S Spread (%)
S Stoichiometry
T Fuel cell or stack temperature (K)
V Velocity (m/s)
W Width of fuel cell active area (m)
wc Width of gas flow channel (m)
wm Width of stack manifold cross-section (m)
ws Width of gas flow channel support (m)
xi Mole fraction of species i
Greek symbols
α Equals −1 for U and 1 for Z configuration stacks

δc Thickness of catalyst layer (m)
δe Thickness of electrode backing (m)
δm Thickness of polymer electrolyte membrane (m)
ζ Flow resistance parameter
η Overpotential (V)
ks Conductivity of catalyst layer solid phase (S/m)
µ Viscosity (N · s/m2 )
ρ Density (kg/m3 )
ρebulk Resistivity of electrode backing ( · m)
ρbp Resistivity of bipolar plate ( · m)
φe Porosity of electrode backing
φc Porosity of catalyst layer
χ Mass fraction of the gaseous phase
Subscripts
a Anode
ave Average value
bot Stack manifold bottom section
bp Bipolar plate
c Catalyst layer; cathode
cell Fuel cell
e Electrode backing
fc Gas flow channel
g Gas
in Inlet value
Liquid
m Polymer electrolyte membrane
out Outlet value
rev Reversible
top Stack manifold top section
Superscripts
cell Fuel cell

k Iteration number
max Maximum value
min Minimum value
new Current value of an iterative parameter
old Previous value of an iterative parameter
stack Stack
CHAPTER 9
Two-phase microfluidics, heat and mass

transport in direct methanol fuel cells
G. Lu & C.-Y. Wang
Department of Mechanical Engineering and Electrochemical Engine
Center (ECEC), The Pennsylvania State University, USA.
Abstract
This chapter provides an overview of the latest developments in direct methanol fuel
cell (DMFC) technology. We begin by describing major technological challenges
that DMFCs presently face for portable power, and demonstrate that the fundamen-
tal transport processes of methanol, water and heat, along with methanol oxidation
kinetics, hold the key to successfully address these challenges. We then describe
complementary experimental and modeling work to elucidate the critical trans-
port phenomena, including two-phase microfluidics, heat and mass transport. We
explain how the better understanding of these basic transport phenomena leads to a
paradigm shift in the design of portable DMFCs, and show experimental evidence
of surprisingly low methanol and water crossover through a very thin membrane,
Nafion® 112. Fuel efficiency resulting from low methanol crossover reaches 78%
and net water transport coefficient through the membrane is found to be less than
unity. These salient characteristics will enable highly concentrated methanol to be
used directly and hence lead to much higher energy density for next-generation
portable DMFCs. Finally, the latest research on micro-DMFCs is reviewed.
1 Introduction
A direct methanol fuel cell (DMFC) is an electrochemical cell that generates elec-
tricity based on the oxidation of methanol and reduction of oxygen. Figure 1
illustrates the cell construction and operating principles of a DMFC. An aque-
ous methanol solution of low molarity acts as the reducing agent that traverses
the anode flow field. Once inside the flow channel, the aqueous solution diffuses
through the backing layer, comprised of carbon cloth or carbon paper. The backing
Figure 1: Operating schematic of a DMFC [1].
layer collects the current generated by the oxidation of aqueous methanol and trans-
ports it laterally to ribs in the current collector plate. The global oxidation reaction
occurring at the platinum-ruthenium catalyst of the anode is given by:
CH3 OH + H2 O → CO2 + 6H+ + 6e− . (1)
The carbon dioxide generated from the oxidation reaction emerges from the
anode backing layer as bubbles and is removed via the flowing aqueous methanol
solution.
Air is fed to the flow field on the cathode side. The oxygen in the air combines
with the electrons and protons at the platinum catalyst sites to form water. The
reduction reaction taking place on the cathode is given by:
3
2 O2 + 6H+ + 6e− → 3H2 O. (2)
These two electrochemical reactions are combined to form an overall cell reac-
tion as:
CH3 OH + 32 O2 → CO2 + 2H2 O. (3)
Extensive work on DMFCs has been conducted by many groups, notably Halpert
et al. [2] of Jet Propulsion Laboratory (JPL) and Giner, Inc, Baldauf and Preidel [3]
of Siemens, Ren et al. [4] of Los Alamos National Laboratory (LANL), Scott and
co-workers [5–7] of University of Newcastle upon Tyne, and Wang and co-workers
[8–11] of the Pennsylvania State University. A comparative study of DMFC with
H2 /air polymer electrolyte fuel cells (PEFC) was presented by the LANL group
[4–12], demonstrating that a DMFC requires platinum-ruthenium and platinum
loadings roughly five times higher to achieve power densities of 0.05 to 0.30 W/cm2 .
A number of extensive reviews have been published in recent years as world-
wide DMFC research activities grew exponentially. Gottesfeld and Zawodzinski
[13] briefly summarized research at Los Alamos intended for transportation appli-
cation, and pointed out areas for improvement if DMFC technology is to become a
serious power plant candidate for transportation. Among these challenges, reducing
catalyst loading to compete with reformed/air fuel cells is perhaps the greatest, and
presents a difficult task for the foreseeable future. In a later book chapter, Gottesfeld
and Wilson [12] discussed perspectives on DMFC for portable applications. Lamy
et al. [14] provided an in-depth review of DMFC fundamentals, including the reac-
tion mechanisms of methanol oxidation, use of various binary and ternary electro-
catalysts, effects of electrode structure and composition on the activity of methanol
oxidation, and development of proton conducting membranes with low methanol
crossover. It was projected that DMFCs will be commercialized as portable power
sources before the year 2010 and that a quantum jump in technology will occur,
making it possible to drive DMFC-powered vehicles ten years thereafter. Arico
et al. [15] reviewed recent advances in DMFC from both fundamental and techno-
logical aspects. The fundamental aspects concerned electrocatalysis of methanol
oxidation and oxygen reduction in the presence of methanol crossover, and the
technological aspects focused upon the proton conducting membranes, as well as
MEA fabrication techniques. Neergat et al. [16] provided an excellent review of
new materials for DMFC, including novel proton conducting membranes and elec-
trocatalysts. Narayanan et al. [17] and Muller et al. [18] discussed, in detail, the
paramount importance of water balance to the portable DMFC system.
As expected, a DMFC exhibits lower power densities than that of a H2 /air
PEFC, which at present requires anode and cathode platinum loadings of less than
1 mg/cm2 to achieve power densities of 0.6 to 0.7 W/cm2 . However, the DMFC
has the advantages of easier fuel storage, no need for humidification, and simpler
design. Thus, DMFC is presently considered a leading contender for portable power
application. To compete with lithium-ion batteries, the first and foremost property of
a portable DMFC system must be higher energy density in Wh/L. This requirement
entails overcoming four key technical challenges: (1) low rate of methanol oxida-
tion kinetics on the anode, (2) methanol crossover through the polymer membrane,
(3) water management, and (4) heat management.
The present chapter deals with the fundamental transport processes of methanol,
water and heat underlying DMFCs. The basic transport phenomena, along with
electrochemical kinetics, are critical to addressing the four technical challenges
outlined above. Section 2 summarizes the thermodynamics and electrochemical
kinetics of DMFCs. Section 3 discusses the two-phase micofluidic phenomena in

the DMFC anode and cathode, respectively, based on experimental observations.
Section 4 describes mass transport phenomena in DMFC with focus on methanol
crossover and water management issues. Section 5 treats heat transfer in DMFC
and its coupling with water transport. Section 6 presents a review of mathematical
modeling and experimental diagnostic techniques presently under active research.
Finally, in Section 7 we present an exciting application of DMFC technology to
power microsystems.
2 Fundamentals of DMFC
2.1 Cell components and polarization curve
The heart of a DMFC is a membrane-electrode assembly (MEA) formed by sand-

wiching a perfluorosulfonic acid (PFSA) membrane between an anode and a cathode.
Upon hydration, the polymer electrolyte exhibits good proton conductivity. On
either side of this membrane are anode and cathode, also called catalyst layers, typ-
ically containing Pt-Ru on the anode side and Pt supported on carbon on the cathode
side. Here the half-cell reactions described in eqns (1) and (2) are catalyzed. On the
outside of the MEA, backing layers made of non-woven carbon paper or woven
carbon cloth, shown in Fig. 2, are placed to fulfill several functions. The primary
purpose of a backing layer is to provide lateral current collection from the catalyst
layer to the ribs as well as optimized gas distribution to the catalyst layer through
diffusion. It must also facilitate the transport of water out of the catalyst layer.
This latter function is usually accomplished by adding a coating of hydrophobic
polymer, polytetrafluoroethylene (PTFE), to the backing layer. The hydrophobic
character of the polymer allows the excess water in the cathode catalyst layer to be
expelled from the cell by the gas flowing inside the channels, thereby alleviating
flooding.
The microstructure of the catalyst layer is of paramount importance for the kinet-
ics of an electrochemical reaction and species diffusion. Figure 3 shows scanning
Figure 2: SEM micrographs of (a) carbon paper and (b) carbon cloth.
electron microscopy (SEM) images of such microstructures of the DMFC anode

and cathode, respectively, where high surface areas for electrochemical reactions
are clearly visible.
A cross-sectional SEM of a MEA segment consisting of a backing layer, a micro-
porous layer (MPL) and a catalyst layer, is displayed in Fig. 4 [9]. The MPL, with
an average thickness of 30 µm, overlays a carbon paper backing layer. The anode
catalyst layer of about 20 µm in thickness covers the MPL. In the anode, this MPL
provides much resistance to methanol transport from the feed to the catalyst sites,
thus reducing the amount of methanol crossover. In the cathode, the MPL helps
alleviate cathode flooding by liquid water [19].
Figure 5 displays a voltage vs. current density polarization curve of a typical
DMFC. The thermodynamic equilibrium cell potential for a DMFC, as calculated
in Section 2.2, is approximately equal to 1.21 V. However, the actual open circuit
voltage in DMFCs is much lower than this thermodynamic value, largely due to
Figure 3: SEM images of electrodes.
Figure 4: Cross-sectional SEM micrograph of backing layer, microporous layer,

and catalyst layer.
Thermodynamic reversible cell Potential

1.21
Voltage drop due to fuel crossover

Voltage (V)
Activation overpotential
Mass transport loss
Ohmic loss
0
Current density (mA/cm 2)
Figure 5: Schematic of DMFC polarization curve.
fuel crossover. Methanol crossover is an important topic in DMFCs and thus will
be fully elaborated in Section 4.1. On closed circuit, the polarization curve can be
categorized into three distinctive regions: kinetic control, ohmic control, and mass
transport control. The kinetic control region of DMFC is dictated by slow methanol
oxidation kinetics at the anode as well as oxygen reduction kinetics at the cathode.
In this region a DMFC suffers the voltage loss second only to the low open circuit
voltage caused by methanol crossover. More detailed discussion of this aspect is
provided in Section 2.3. The area where cell voltage decreases nearly linearly in the
polarization curve is recognized as the ohmic control region. For a DMFC where
the polymer electrolyte is usually well hydrated, the voltage loss in this section
is minimal. The last portion is referred to as the mass transport control region,
whereby either methanol transport on the anode side results in a mass transport
limiting current, or the oxygen supply due to depletion and/or cathode flooding
becomes a limiting step. The cell voltage drops drastically in the mass transport
control region.
2.2 Thermodynamics
The thermodynamic equilibrium potential of a fuel cell can be calculated from:
g h − T s
E = − =− . (4)
nF nF
Table 1: Thermodynamic data of fuel cell reactions (per mole of fuel) [20].
Reaction T (K) g (kJ/kg) h (kJ/kg) s (kJ/kg K) n E (V) ηrev
PEFC 298 −237 −285 −162 2 1.23 0.83

DMFC 298 −704 −727 −77 6 1.21 0.97
PEFC: H2 + 12 O2 → H2 O; DMFC: CH3 OH + 32 O2 → CO2 + 2H2 O.
Table 1 lists thermodynamic data of common fuel cell reactions at 25 ◦ C and 1 atm.
For the liquid-feed DMFC, n = 6 and the thermodynamic cell potential is 1.21 V,
similar to that of the H2 /air PEFC.
The thermodynamic efficiency of a fuel cell is defined as the ratio of maximum
possible electrical work to the total chemical energy, i.e.:
g nFE
ηrev = = . (5)
h −h
As shown in Table 1, the theoretical thermodynamic efficiency of DMFC reaches
97% at 25 ◦ C.
The practical energy efficiency, however, is much lower after accounting for
voltage and fuel losses. The voltaic efficiency is defined as the ratio of the actual
electric work to the maximum possible work, with the former given by:
Wact = −nFVcell , (6)
where Vcell is the cell voltage at a current of I . Hence the voltaic efficiency can be
written as:
Wact −nFVcell −nFVcell Vcell
ηvoltaic = = = = . (7)
Wmax g −nFE E
For example, if the cell is running at 0.4 V, then the voltaic efficiency is only 33%.
This low efficiency is caused by substantial overpotentials existed in both the anode
and cathode of a DMFC.
In a DMFC, there is also fuel efficiency due to methanol crossover defined as:
I
ηfuel = , (8)
I + Ix over
where Ix over is an equivalent current density caused by methanol crossover under
the operating current density of I .
The total energy efficiency of DMFC is therefore given by:
η = ηrev ηvoltaic ηfuel . (9)
Suppose that the fuel efficiency, ηfuel , in a DMFC is 80%, the total energy effi-
ciency becomes η = 97% × 33% × 80% = 25.6% with cell voltage of 0.4 V.
In comparison, for a PEFC η = 83% × 0.7/1.23 = 40.5% with the cell voltage of
0.7 V. The energy efficiency of the PEFC is relatively higher owing largely to its neg-
ligibly small fuel crossover and overpotential for hydrogen oxidation on the anode.
It is evident from eqn (6) that in order to achieve higher energy-conversion effi-
ciency, one must control methanol crossover so as to maintain high fuel efficiency
(e.g. >80%). In addition, it is desirable to operate DMFCs at higher voltages. Thus,
high-voltage performance is a high priority for portable DMFC development.
Waste heat produced in the DMFC can thus be expressed as:
IVcell
Q= − IVcell = IVcell (1/η − 1). (10)
η
By substituting the definition of the total energy efficiency, another expression of

heat generation results:
I + Ix over
Q = (−h) − IVcell , (11)
nF
where the first term on the right hand side represents the chemical energy of
methanol consumed for power generation and by crossover, while the second term
stands for the electric energy generated.
2.3 Methanol oxidation and oxygen reduction kinetics
Combined with methanol crossover, slow anode kinetics lead to power density in a
DMFC that is three to four times lower than that of a hydrogen fuel cell. Much work
has been focused on the anodic oxidation of methanol [21]. A multi-step mechanism
of electrocatalytic oxidation of methanol at the anode was postulated [22, 23].
Different anode catalyst structures of Pt-Ru were developed [24] and several anode
catalysts other than Pt-Ru were explored [25–27]. Additionally, the effects of the
anode electrochemical reaction on cell performance were experimentally studied
[28–30]. Lamy et al. [14] and Arico et al. [15] provided extensive reviews of the
most recent work on electro-catalysis. More active catalysts for methanol oxidation
would enable a certain power density to be realized at higher cell voltage, and hence
directly impact the energy efficiency of the cell, which translates to the energy
density if the amount of fuel carried by a DMFC system is fixed.
The activation overpotential of methanol oxidation reaction (MOR) can be des-
cribed by Tafel kinetics of the following form:

I
ηa = ba log , (12)
I0,a
where ba and Io,a are the Tafel slope and exchange current density of MOR, respec-
tively. A convenient method to characterize anode activation polarization uses a
MeOH anode vs. H2 cathode cell as proposed by Ren et al. [28], to be discussed in
more detail in Section 7.
The oxygen reduction reaction (ORR) on the DMFC cathode is similarly slow,
causing a high cathode overpotential. Thus, a Tafel expression is usually used to
describe ORR kinetics as follows:

I
ηc = bc log , (13)
Io,c
where the Tafel slope for ORR is around 70 mV/decade in the absence of methanol
oxidation. However, in DMFCs, ORR takes place simultaneously with oxidation
of crossover methanol, and consequently bc for a DMFC becomes greater than that
for a H2 /air PEFC.
3 Two-phase flow phenomena

3.1 Bubble dynamics in anode
On the anode side, carbon dioxide is produced as a result of MOR. If CO2 bubbles
cannot be removed efficiently from the surface of the backing layer, they remain,
covering the backing surface and hence decreasing the effective mass transfer area.
In addition, flow blockage results, particularly in channels of small dimensions as
required in micro or compact portable fuel cells with maximum volumetric power
and energy densities. Therefore, gas management on the anode side is an important
issue in DMFC design. Argyropoulos et al. [5, 31] was perhaps among the first to
observe the two-phase flow pattern in the anode channel under various operating
conditions. This flow visualization on the anode side yields valuable understanding
of bubble dynamics in a DMFC. This study was, however, undertaken under low cell
performance. Most recently, Lu and Wang [32] developed an improved transparent
DMFC to visualize bubble dynamics on the anode side and liquid droplet (and
flooding) dynamics on the cathode. This latest study is described in detail in the
next two subsections.
3.1.1 Flow visualization

Figure 6 displays a diagram of the experimental setup, which consists of an elec-
tronic load system to characterize polarization behaviors of the fuel cell, a peristaltic
pump to deliver the liquid fuel, an electric heater with temperature controller, pres-
sure relief valves, flow meters and pressure gauges. A Sony digital video camera
recorder was used in experiments for flow visualization, and still pictures were
captured according to the time sequence when the movie was edited offline. Also, a
Nikon N70 camera with a micro-Nikkor lens (60 mm f/2.8D) was utilized to obtain
clear pictures of small-size objects.
Figure 7 shows a picture of the transparent fuel cell. The cell was constructed of a
pair of stainless steel plates mated with a polycarbonate plate.Atotal of eight parallel
flow channels (1.92 mm width, 1.5 mm depth, 1 mm rib width) were machined
through the stainless steel plate to form an active area of approximately 5 cm2 . The
surface of the stainless steel plate contacting the MEA was coated with 30 nm Cr
and 300 nm Au to minimize contact resistance. A transparent polycarbonate plate
Figure 6: Experimental setup for flow visualization.
Figure 7: Photo of the transparent fuel cell.
covered the stainless steel plate, forming a window to allow direct observation of
flow behaviors. The polycarbonate plate was concave in design, while the stainless
steel plate had a matching convex pattern. This unique design avoided flow leakage
between neighboring parallel channels. Cell inlet and outlet manifolds were also
machined in the polycarbonate plates.
Figure 8: Images of bubble dynamics in the DMFC anode using a MEA with carbon
paper as backing layer for 2 M MeOH feed and non-humidified air at
100 mA/cm2 and 85◦ C.
Figure 8 shows a sequence of images at various times for an MEA with hydropho-
bic carbon paper as the backing layer at the feed temperature of 85 ◦ C and the
current density of 100 mA/cm2 . The images, one second apart, were captured from
the movie, with time resolution 1/30 second. In addition, the time of the first image
was chosen arbitrarily due to the fact that two-phase flow is a regularly periodic
event and the cell was operated at steady-state. As shown in Fig. 8, the CO2 bubbles
nucleate at certain locations and form large and discrete gas slugs in the channel.
The CO2 bubbles are large in size (∼2 mm) and confined by the channel dimen-
sions, elongated in shape, and distributed discretely on the backing layer along the
anode channel. This bubble flow is commonly categorized as Taylor bubbles. The
bubble motion is governed by the momentum of liquid flow, the force of buoyancy
on the bubble, and the surface tension between bubbles and substrate. It can be seen
from Fig. 8 that the bubbles are held on the carbon paper by strong surface tension
until they grow into larger slugs for detachment, clearly indicative of the dominant
Figure 9: Bubble behavior on the anode side using hydrophilic carbon cloth for 1 M
MeOH feed and non-humidified air at 100 mA/cm2 and 85 ◦ C.
effect of surface tension in bubble dynamics in DMFC. Once the bubbles grow to
a sufficient size, they detach and sweep along the backing surface in the channel.
This sweeping process clears all small bubbles pre-existing on the backing surface,
making new bubbles grow from the smallest size to full detachment diameter. As a
result, the two-phase flow becomes regularly intermittent. The flow pattern on the
MEA with carbon paper is characterized as bubble flow or slug flow, depending on
the accumulation of the bubbles.
Figure 9 shows the bubble patterns on the MEA with hydrophilic carbon cloth
also at the feed temperature of 85 ◦ C. Since the carbon cloth has a much rougher
surface, it is challenging to capture sharp still pictures due to light deflection,
although the two-phase flow could be observed clearly in the experiments and
movies. Alternatively, a Nikon N70 camera with a micro-Nikkor lens was employed
for still photos. It is seen that the CO2 bubbles are produced more uniformly and
with smaller size (∼0.5 mm) from the hydrophilic carbon cloth. Therefore, the flow
on the MEA with carbon cloth is characterized as a bubble flow.
The differences in bubble behaviors between hydrophobic carbon paper and
hydrophilic carbon cloth can be explained by considering the fundamental process
of bubble growth. It is insightful to compare the differences in the pore structure
of carbon paper and carbon cloth. Figure 2 shows SEM micrographs of these two
substrates. Clearly, carbon cloth has more regularly distributed pores, whereas car-
bon paper is more of a random porous medium. This difference in the pore size
distribution gives rise to the fact that CO2 bubbles emerge more uniformly from
the carbon cloth than carbon paper.
3.1.2 Bubble diameter and drift velocity

The bubble detachment diameter from the backing layer is strongly correlated with
surface wettability. Consider a bubble growing and detaching from a single pore
of known diameter, dp , and surface contact angle of θ. Assuming, as was indicated

in experimental observations, that the bubble detachment process is dominated by
buoyancy and surface tension effects, the force balance predicts that the diameter
of the bubble at detachment, db , is [33]:
1
4dp σ sin θ 3
db = . (14)
g(ρl − ρg )
With the typical pore size of 10 µm for both carbon paper and carbon cloth [34], eqn
(14) calculates the bubble detachment diameter of 0.68 mm for the hydrophobic
carbon paper (e.g. θ = 100◦ ) and 0.38 mm for the hydrophilic carbon cloth (e.g.
θ = 10◦ ). These theoretical estimates are consistent with experimental observations.
Once detached, bubbles stay spherical in shape due to strong surface tension. If
these bubbles are smaller than the channel dimension, the bubble drift velocity
through the liquid can be estimated from the correlation of the bubble rising velocity
through an infinite, stagnant liquid as obtained from the balance between inertia
and gravitational forces [35], i.e.:
db2 g(ρl − ρg )
ub = . (15)
12µl
3.1.3 Pressure drop

For the purpose of gross estimation, the two-phase frictional pressure drop through
the anode channel of a DMFC may be approximated by assuming a homogenous
flow. Usually, the two-phase flow in the DMFC anode is laminar as the flow rates of
both phases are quite small. For a 50 cm2 DMFC operated under typical conditions,
the anode pressure drop is of the order of a few kPa.
3.2 Liquid water transport in cathode
The importance of flooding on the cathode side in H2 /air PEFCs has been empha-
sized in the literature [36–39]. Similarly, water management on the cathode in a
DMFC was identified as a key issue [40]. A proper level of liquid water existing
in the DMFC cathode helps to hydrate the polymer membrane, thus increasing the
proton conductivity. However, severe flooding should be avoided so as to maintain
the cathode performance.
3.2.1 Flooding in the cathode

In the cathode, water is produced by the oxygen reduction reaction as well as trans-
ported from the aqueous anode due to diffusion and electro-osmotic drag. Param-
eters governing liquid water formation and distribution in the cathode include the
stoichiometry (or volumetric flow rate) of the inlet air, current density, cell tem-
perature, and membrane water transport properties such as the diffusion coefficient
and electro-osmotic drag coefficient.
Formally, the water flux arrived at the cathode by diffusion, electro-osmosis, and
hydraulic permeation across the membrane can be expressed as [40]
Cc−a I Km ρl I
jm = −D + nd − Pc−a =α . (16)
δm F µl MH 2 O F
Clearly, the three terms on the right hand side in eqn (16) represent three modes
of water transport through the membrane, respectively. The molecular diffusion is
driven by the concentration gradient. The electro-osmotic drag is proportional to the
current density, and the permeation is driven by the hydraulic pressure difference.
The net water flux through the membrane can be conveniently quantified by a net
water transport coefficient, α, as defined in eqn (16). This important parameter
dictates water management strategies in DMFC systems. It is a combined result of
electro-osmotic drag, diffusion and convection through the membrane. For thick
membranes like Nafion® 117, α approaches the electro-osmotic drag coefficient as
the other two modes of water transport are weakened with increasing membrane
thickness.
The electro-osmotic drag coefficient nd for Nafion electrolyte in contact with
liquid water depends on the temperature [41], as shown in Fig. 10. The relation can
also be fitted as:
nd = 1.6767 + 0.0155T + 8.9074 × 10−5 T2 , (17)
where T is the cell temperature in ◦ C.

The total rate of water transported to and produced at the cathode is given by:

1 I
jH2 O = α + . (18)
2 F
Figure 10: Water drag coefficient as a function of temperature [41].

At steady state, this must be balanced by the removal rate through cathode air flow.
Suppose that completely dry air is fed into the cathode channel, its molar flow rate
can be expressed as a function of air stoichiometry such that:
1 I
jair = ξ , (19)
0.21 4F
where ξ is the stoichiometry defined at the current density of I .
The depletion rate of oxygen due to the ORR is simply given by:
I
j O2 = . (20)
4F
A simple water balance thus yields the relative humidity of air at the cathode exit
as follows:

pH2 O
I 1
F 2 + α ptotal
RHexit = = ×
psat (T ) ξ 1 I
− I
+ I 1
+α psat (T )
0.21 4F 4F F 2
(21)
1 + 2α ptotal
= ξ
× .
0.42 + 0.5 + 2α p sat (T )
A critical air stoichiometry is obtained when the relative humidity RHexit =

100%, i.e.:
ptotal
ξcri = 0.21 (2 + 4α) − (1 + 4α) . (22)
psat (T )
This threshold stoichiometry represents the formation of liquid water and thus
characterizes the state of cathode flooding. If the actual stoichiometry is smaller
than that given in eqn (22), the cathode exhaust air will carry liquid water and hence
cathode flooding likely occurs. On the other hand, if the actual air stoichiometry is
higher, the cathode exhaust air is under-saturated. In this case, cathode flooding is
avoided; however, there is too much water loss through evaporation in the case of
large air flowrate. Recovery of water vapor in an external condenser proves to be a
difficult task for a compact portable system. Hence, for portable DMFC systems,
air stoichiometry ought to be designed to be smaller than the critical value given
in eqn (22), implying at the same time that a small amount of cathode flooding is
inevitable in portable systems.
3.2.2 Flooding visualization

Visualization on the cathode side is a useful diagnostic tool to understand the
nature of flooding. Figure 11 displays an image of water drop formation on carbon
paper treated with PTFE with non-humidified air preheated to 85 ◦ C and fed at a
volumetric flow rate of 68 mL/min using the transparent cell as shown in Fig. 7.
The cell current density was 100 mA/cm2 . It is shown in Fig. 11 that water droplets
are attached on the surface of the carbon paper due to its decreased hydrophilicity
Figure 11: Water droplet formation at cathode using Toray carbon paper for
2 M MeOH feed and non-humidified air (68 mL/min and 1 psig) at
100 mA/cm2 and 85 ◦ C.
Figure 12: Cathode flooding on single-sided ELAT carbon cloth for 2 M MeOH
feed and non-humidified air (161 mL/min and 1 psig) at 60 mA/cm2
and 85 ◦ C.
at elevated temperatures. It was observed that while the droplets grow slowly, the
cell voltage drops gradually when the current density is fixed.
Figure 12 shows an image of flooding on the single-sided ELAT carbon cloth
with non-humidified air preheated to 85 ◦ C. It is seen from Fig. 12 that the surface
of carbon cloth is nearly free of liquid droplets due to its higher hydrophobicity, but
liquid droplets or ‘sweating’ can be found in contact corners between the stainless
steel rib and carbon cloth. This is because the rib surface is rather hydrophilic.
Interestingly, it appears that sweating inside corners between the ribs and car-
bon cloth gas diffusion layer occurs at a rather low current density of 60 mA/cm2
and a high air flow rate of 161 mL/min. In comparison, no such sweating is seen
between the stainless steel ribs and carbon paper GDL at a higher current density of
100 mA/cm2 and a lower airflow rate of 68 mL/min (see Fig. 11).
Much remains to be learned about the fundamental process of flooding occurrence
and its relation with the backing layer material.
4 Mass transport phenomena

4.1 Methanol crossover
Methanol crossover occurs due to the inability of the commonly-used perfluoro-

sulfonic acid (PFSA) membrane to prevent methanol from permeating its
polymer structure. Diffusion and electro-osmotic drag are the prime driving forces
for methanol transport through the polymer membrane and eventual reaction with
platinum catalyst sites on the cathode, leading to a mixed potential on the cathode.
This mixed potential on the cathode causes a decrease in cell voltage. Methanol
reaching the cathode also results in decreased fuel efficiency, thus lowering the
energy density of the system.
Methanol crossover in DMFC has been extensively studied both experimentally
and theoretically. Narayanan et al. [42] and Ren et al. [43] measured the methanol
crossover flux with different membrane thickness and showed that methanol cross-
over rate is inversely proportional to membrane thickness at a given cell current
density, thus indicating that diffusion is dominant. In addition, Ren et al. [44] com-
pared diffusion with electro-osmotic drag processes and demonstrated the impor-
tance of electro-osmotic drag in methanol transport through the membrane. In their
analysis, methanol electro-osmotic drag is considered a convection effect and the
diluted methanol moves with electro-osmotically dragged water molecules. Valdez
and Narayanan [45] studied the temperature effects on methanol crossover and
showed that the methanol crossover rate increases with cell temperature. Ravikumar
and Shukla [30] operated a liquid-feed DMFC at an oxygen pressure of 4 bars and
found that cell performance is greatly affected by methanol crossover at methanol
feed concentrations greater than 2 M, and that this effect increases with increased
operating temperature. Wang et al. [46] analyzed chemical composition of the
cathode effluent of a DMFC using a mass spectrometer. They found that methanol
crossing over the membrane is completely oxidized to CO2 at the cathode in the
presence of a Pt catalyst.
Additionally, the cathode potential is influenced by the mixed potential phe-
nomenon due to simultaneous methanol oxidation and oxygen reduction as well as
poisoning of Pt catalysts by methanol oxidation intermediates. Kauranen and Skou
[47] presented a semi-empirical model to describe the methanol oxidation and oxy-
gen reduction reactions on the cathode and concluded that the oxygen reduction
current is reduced in the presence of methanol oxidation due to surface poisoning.
Kuver and Vielstich [48] studied the dependence of crossover on reaction con-
ditions, such as temperature and methanol concentration. Additionally, a cyclic
voltammetry technique was presented which allows the evaluation of methanol
crossover in a fuel cell under operating conditions. Scott et al. [49] investigated the
effect of cell temperature, air cathode pressure, fuel flow rate and methanol concen-
tration on power performance. Gurau and Smotkin [50] used gas chromatography
to measure crossover variation with temperature, fuel flow rate and concentra-
tion. Heinzel and Barragan [51] gave an extensive review of the state-of-the-art
of methanol crossover in DMFC. Influence of methanol concentration, pressure,
temperature, membrane thickness and catalyst morphology have been discussed.
Development of novel membranes with low methanol crossover would surely
increase cell performance and fuel efficiency [14, 16, 52–54]. Alternatively, Wang
and co-workers [9, 10, 55] proposed to modify the anode backing structure to miti-
gate methanol crossover. It was demonstrated that a compact microporous layer can
be added in the anode backing to create an additional barrier to methanol transport,
thereby reducing the rate of methanol crossing through the polymer membrane.
Both practices to control methanol crossover by increasing mass transport resis-
tance, either in the anode backing or in the membrane, can be mathematically
formulated by a simple relation existing between the crossover current, Ic , and
anode mass-transport limiting current, IA,lim . That is:

I
Ic = Ic,oc 1 − , (23)
IA,lim
where Ic,oc is the crossover current at open circuit, and I the operating current. In the
conventional approach using thick membranes with low methanol crossover, Ic,oc is
low and IA,lim is set to be high. In contrast, setting up a barrier in the anode backing is
essentially reducing the anode limiting current, IA,lim , and making Ic,oc a significant
fraction of IA,lim , about 50–80%. Two immediate advantages result from this latter
cell design principle. One is that more concentrated fuel can be used thus leading
to much higher energy density of the DMFC system. Lu et al. [10] successfully
demonstrated the use of 8 M methanol solution as the anode feed, and Pan [55]
most recently reported a DMFC operated with 10 M (or 30% by volume) methanol
fuel solution. Secondly, this type of DMFC permits use of thin membranes such as
Nafion 112, which greatly facilitates water back flow from the cathode to anode [10,
56], thus addressing another major challenge of portable DMFC to be discussed in
the next subsection.
As an example, Fig. 13 shows the polarization curve of a DMFC design based on
the above new concept and using a very thin membrane, Nafion 112. The cell was
operated at an anode stoichiometry of 2 and a cathode stoichiometry of 4 at a current
density of 150 mA/cm2 . It is evident from Fig. 13 that the mass transport limiting
current density, IA,lim , in this cell is equal to 205 mA/cm2 . Figure 14 displays the
polarization behavior of the cell when using humidified nitrogen in the cathode
to obtain the crossover rate at the open circuit, which gives Ic,oc = 157 mA/cm2 .
According to eqn (23), the crossover current at the operating current density of
150 mA/cm2 is Ic = 42 mA/cm2 . Therefore, the fuel efficiency defined in eqn (8)
reaches 150/(150 + 42) = 78%. This experiment provides direct evidence that it is
possible to obtain very high fuel efficiency even when using a very thin membrane,
Nafion 112, provided that the cell is well designed. Finally, this cell yields a power
Figure 13: Polarization and power density curves for a DMFC designed for portable
application.
Figure 14: Measurement of methanol crossover current density at open circuit using
humidified nitrogen in the cathode. The limiting current density signifies
the crossover rate.
density of 56 mW/cm2 at 60 ◦ C under operating conditions eminently suited for

portable systems.
4.2 Water management in portable DMFC systems
Water management emerges as a new significant challenge for portable DMFC

systems. Constrained by the methanol crossover problem, the anode fuel solution
has been very dilute, meaning that a large amount of water needs to be carried in the
system and therefore reduces the energy content of the fuel mixture. In addition,
for each mole of methanol, one mole of water is needed for methanol oxidation
at the anode and 2.5 × 6 moles of water are dragged through a thick membrane
such as Nafion 117 towards the cathode, assuming that the electro-osmotic drag
coefficient of water is equal to 2.5 water molecules per proton. This then causes
16 water molecules to be lost from the anode for every mole of methanol. Water
in the anode must therefore be replenished. On the other hand, inside the cathode,
there are 15 water molecules transported from the anode due to electro-osmosis and
3 additional water molecules produced by consuming six protons generated from
oxidation of one methanol molecule. Presence of a large amount of water floods
the cathode and reduces its performance. The difficult task of removing water from
the cathode to avoid severe flooding and supplying water to the anode to make
up water loss due to electro-osmotic drag through the membrane is referred to as
innovative water management in a portable DMFC.
Traditionally, a high cathode gas flow rate (high stoichiometry) is employed to
prevent flooding. This strategy not only increases parasitic power consumption but
also removes excessive water from the fuel cell, making external water recovery
more difficult; see the discussion in Section 3.2.1. How to minimize water removal
from the cathode and subsequent recovery externally to replenish the anode with-
out causing severe cathode flooding becomes an important engineering issue. A
greater understanding and ability to tailor water flow in the cell is of fundamental
interest for portable DMFC systems. This is an area where DMFC modeling plays
an important role.
In the open literature, Blum et al. [57] proposed a concept of a water-neutral
condition under which the anode does not need water supply and the cell maintains
perfect water balance by losing exactly 2 moles of water per mole of methanol con-
sumed. Apparently, this condition corresponds to α = −1/6. Most recently, Peled
et al. [56] reported experimental data with α being small and even negative at low
current densities by using a nonporous proton-conducting membrane and oxygen
at 3 bars as the oxidant. It was postulated that the convection effect induced by a
hydraulic pressure differential across the membrane can offset the electro-osmotic
drag, leading to α being much smaller than the electro-osmotic drag coefficient of
approximately 3 at 60 ◦ C.
Low-α DMFCs are highly desirable from the water management standpoint as
the anode does not require an excessive amount of water and, in conjunction with
the methanol transport barrier concept suggested in Section 4.1, it becomes possible
to use highly concentrated methanol fuel at the anode. In addition, there is less or
no need to recover water from the cathode exhaust, thus eliminating or reducing
the condenser in a portable system. Based on our theory of liquid water transport
in PEFCs [19, 58], we have designed a unique MEA structure which utilizes the
microporous layer, as shown in Fig. 4, to build up the hydraulic pressure on the
cathode side and which then uses a thin membrane (i.e. Nafion 112) to promote
the water back flow under this hydraulic pressure difference. Such MEAs exhibit
extraordinarily low α and hence are generally termed low-α MEA technology.
Figure 15: Evolutions in cell voltage during constant current loading to measure
the net water transport coefficient α.
Figure 15 shows the evolution of cell voltage at constant current load in a series
of experiments to measure the net water transport coefficient, α, at various tempera-
tures using a moisture trap [59]. The steady-state power density reaches 16 mW/cm2
at 23 ◦ C, 33.3 mW/cm2 at 40 ◦ C and 56 mW/cm2 at 60 ◦ C, respectively. At low cur-
rent densities, e.g. 50 mA/cm2 at 23 ◦ C and 100 mA/cm2 at 40 ◦ C, the cell voltages
remain stable for an extended period of operation. While at a high current density
such as 150 mA/cm2 at 60 ◦ C and 70 ◦ C, the cell voltage occasionally experiences
a sharp fluctuation once a slug of CO2 gas produced by the large current density
blocks the anode channels temporarily, causing a “short-lived” mass transport limi-
tation on the anode side. Paying attention to gas management in the design of anode
flowfield will likely remove this voltage oscillation.
Figure 16 displays the net water transport coefficient α measured at different
temperatures. It is seen that α is only 0.05 at room temperature and α is equal to
0.64 at 60 ◦ C for the air stoichiometry of 4 and methanol stoichiometry of 2. The
significance of this set of experiments, shown in Figs 15 and 16, is the fact that
commercially available Nafion membranes and MEA materials were used and the
cell operated with ambient air without pressurization.
5 Heat transport
Thermal management in DMFCs is intimately tied to water and methanol transport
processes. First, heat generation in DMFC is much higher than H2 /air PEFC due to
a much lower energy efficiency (only 20–25% when the cell is operated between
0.3–0.4 V). That is, for a 20 W DMFC system, 60–80 W of waste heat is produced.
The waste heat is typically removed from DMFC by liquid fuel on the anode side
Figure 16: The net water transport coefficient α at different temperatures with con-
stant flow rates of methanol and ambient air.
and by water evaporation on the cathode side. The latter also determines the amount
of water loss from a DMFC and the load of water recovery by an external condenser.
Therefore, while a higher cell temperature promotes the methanol oxidation reac-
tion, it may not be practically feasible from the standpoint of water evaporation
loss. Moreover, the higher cell temperature increases the methanol crossover rate,
thereby reducing the fuel efficiency and the system energy density.
Argyropoulos et al. [60] present a one-dimensional thermal model for direct
methanol fuel cell stacks. In this model, the variation of the temperature of the
various components in the stack as well as the heat flow inside the system has been
assessed. Dohle et al. [61] described the heat and power management of a direct
methanol fuel cell by taking the power for auxiliary equipment into consideration.
However, a heat flow analysis carried out for portable DMFC systems is absent in
the open literature.
The total waste heat produced from DMFC reactions has been derived in Sec-
tion 2.2, i.e. eqn (11). In addition, a heat sink term exists due to liquid water
evaporating into the gas phase in the cathode. This can be expressed as:
Q = −Hvapor = −je hlg , (24)
where hlg is the latent heat of evaporation. The evaporation water flux, je , can be
calculated from:
ξ psat (T ) I
je = − 0.5 , (25)
0.84 ptotal F
where the first term accounts roughly for the amount of water vapor present in the
cathode exhaust and the second term describes the water vapor produced by ORR
at the cathode. Therefore, the net heat generation from a DMFC is given by:

I + Ix over ξ psat (T ) I
Qtotal = −h · − IVcell − − 0.5 hlg . (26)
6F 0.84 ptotal F
This amount of heat must be removed either by circulating aqueous fuel solution
(fuel cooling) or by air cooling from the stack peripherals. Much work remains
to be done to analyze heat management in a DMFC and to understand how heat
flow affects water management. It is expected that heat management will become a
major technological challenge in development of portable DMFC power systems.
6 Mathematical modeling and experimental diagnostics

While attempts continue to elucidate the fundamental electrochemical reaction
mechanisms, to explore new compositions and structures of catalysts, and to develop
new membranes and methods to prevent methanol crossover, important system
issues relevant to DMFC are emerging, such as water management, gas manage-
ment, flow field design and optimization, and cell up-scaling for different applica-
tions. A number of physicochemical phenomena take place in liquid-feed DMFC,
including species, charge, and momentum transfer, multiple electrochemical reac-
tions, and gas-liquid two-phase flow in both anode and cathode. Carbon dioxide
evolution in the liquid-feed anode results in strongly two-phase flow, making the
processes of reactant supply and product removal more complicated. All these
processes are intimately coupled, resulting in a need to search for optimal cell
design and operating conditions. A good understanding of these complex, inter-
acting phenomena is thus essential and can most likely be achieved through a
combined mathematical modeling and detailed experimental approach. It is appar-
ent that three of the four technical challenges for portable DMFC systems discussed
in Section 1 require a basic understanding of methanol, water and heat transport
processes occurring in DMFC. This provides a great opportunity to exercise fun-
damental modeling.
Another good topic for modeling is the micro DMFC system. Both anode carbon
dioxide blockage and cathode flooding are especially acute in microsystems due
to the small channel length scale involved, low operating temperature, dominance
of surface tension forces, and requirement for low parasitic power losses in these
systems [62–66]. More discussion on the general characteristics of micro DMFCs
is presented in Section 7.
In addition, DMFC technology is a system requiring a high degree of opti-
mization. A multitude of operating parameters affect the performance of a DMFC.
These variables include cell temperature, molarity of aqueous methanol solution,
cathode pressure, anode and cathode stoichiometry, and flow-field design. Higher
cell temperatures improve catalytic activity, but increase water loss from the cath-
ode. Efficient removal of carbon dioxide gas bubbles and liquid water produced
on the anode and cathode, respectively, must be maintained to allow reactants to
reach catalyst sites. Removal of carbon dioxide “slugs” and prevention of cathode
“flooding” can be attained by increasing flow rates. However, increasing flow rates
requires more pumping power. Too high a flow rate on the cathode will dry out the
polymer membrane, decreasing proton conductivity and hence cell performance.
An understanding of the interdependence of these parameters plays a key role in
optimizing the performance of a DMFC.
DMFC modeling thus aims to provide a useful tool for the basic understanding
of transport and electrochemical phenomena in a DMFC and for the optimization of
cell design and operating conditions. This modeling is challenging in that it entails
the two-phase treatment for both anode and cathode, and in that both the exact role
of the surface treatment in backing layers and the physical processes which control
liquid phase transport are unknown.
6.1 Mathematical modeling
In the literature, Scott et al. [67–69] developed several simplified single-phase

models to study transport and electrochemical processes in DMFC. Baxter et al. [70]
developed a one-dimensional mathematical model for a liquid-feed DMFC anode.
A major assumption of this study was that the carbon dioxide is only dissolved in
the liquid and hence their anode model is a single-phase model. Using a macro-
homogeneous model to describe the reaction and transport in the catalyst layer of
a vapor-feed anode, Wang and Savinell [71] simulated the effects of the catalyst
layer structure on cell performance. Kulikovsky et al. [72] simulated a vapor-feed
DMFC with a two-dimensional model and compared the detailed current density
distributions in the backing, catalyst layer, and membrane of a conventional to a
novel current collector. In another paper, Kulikovsky [73] numerically studied a
liquid-feed DMFC considering methanol transport through the liquid phase and
in hydrophilic pores of the anode backing. In both publications of Kulikovsky,
the important phenomenon of methanol crossover was ignored. Dohle et al. [74]
presented a one-dimensional model for the vapor-feed DMFC and included the
crossover phenomenon. The effects of methanol concentration on cell performance
were studied.
In a three-part paper [75–77], Meyers and Newman developed a theoretical
framework that describes the equilibrium of multicomponent species in the mem-
brane. The transport of species in the membrane based on concentrated-solution
theory and membrane swelling were taken into consideration in their model. The
transport phenomena in the porous electrodes were also included in their mathe-
matical model. However, the effect of pressure-driven flow was not considered.
In addition, the transport of carbon dioxide out of the anode was neglected by
assuming that the carbon dioxide was dilute enough to remain fully dissolved in
liquid. Nordlund and Lindbergh [78] studied the influence of the porous structure
on the anode with mathematic modeling and experimental verification. In their
model, they also assumed that carbon dioxide does not evolve as gas within the
electrode. Recently, Wang and Wang [79] presented a two-phase, multicomponent
model. Capillary effects in both anode and cathode backings were accounted for. In
addition to the anode and cathode electrochemical reactions, the model considered
diffusion and convection of both gas and liquid phases in backing layers and flow
channels. The model fully accounted for the mixed potential effect of methanol
oxidation at the cathode as a result of methanol crossover caused by diffusion,
convection and electro-osmosis. The model of Wang and Wang was solved using a
computational fluid dynamics technique and validated against experimental polar-
ization curves. Results indicated the vital importance of gas phase transport in the
DMFC anode. Divisek et al. [80] presented a similar two-phase, two-dimensional
model of DMFC. Two-phase flow and capillary effects in backing layers were con-
sidered. In addition, detailed, multi-step reaction models for both ORR and MOR
were developed. Murgia et al. [7] described a one-dimensional, two-phase, multi-
component steady-state model based on phenomenological transport equations for
the catalyst layer, diffusion layer, and polymer membrane for a liquid-feed DMFC.
Despite the fact that much effort has been made to model the DMFC system,
considerable work remains, particularly in support of the emerging portable designs
and systems. Few studies have treated the dominating effects of two-phase flow.
No model to date has sufficient detail to provide a microfluidic theory for portable
systems including effects of channel geometry and wettability characteristics of the
GDL on fluid flow in the anode or cathode. Modeling studies are needed to fully
elucidate the intricate couplings of methanol, water and heat transport processes.
This understanding is key to successful design and operation of portable DMFC
systems. Finally, although the important role of a micro-porous layer in DMFC and
its tailoring to control the flow of methanol and water have begun to be recognized,
much remains to be done.
6.2 Experimental diagnostics
Similarly, experimental diagnostics are an important component of DMFC devel-

opment. Diagnostic techniques for DMFC have included:
• cyclic voltammetry (CV) to determine the electrochemically active area of the
cathode,
• CO stripping to determine the electrochemically active area of the anode,
• electrochemical impedance spectroscopy (EIS),
• anode polarization characterization via a CH3 OH/H2 cell as proposed by Ren
et al. [28],
• methanol crossover rate measurement by CO2 sensing in the cathode (via
FITR, GC or infrared CO2 sensors), or by measuring the limiting current in
a CH3 OH/N2 cell (Ren et al. [28]),
• current distribution measurements via a segmented cell in conjunction with a
multi-channel potentiostat (Mench and Wang [40]), and
• material balance analysis of CH3 OH and H2 O (Narayanan et al. [17] and Muller
et al. [18]).
In addition, two-phase visualization of bubble dynamics [5, 31, 32] on anode and
liquid droplet dynamics on cathode [32] as described in Sections 3.1.1 and 3.2.2,
respectively, is a useful tool for cell design and optimization.
Mench and Wang [40] described an experimental technique to measure current

distribution in a 50 cm2 instrumented DMFC based on a segmented cell and multi-
channel potentiostat. In this method, separate current collector ribs are embedded
into an insulating substrate (e.g. Lexan plate) to form a segmented flowfield plate.
The resulting flowfield plates for both anode and cathode are then assembled with
a regular MEA to form a fuel cell with independently controllable subcells. All
subcells are connected to a multi-channel potentiostat to undergo potentiostatic
experiments simultaneously. The subcell currents measured thus provide informa-
tion on the current density distribution for the full-scale fuel cell. The spatial and
temporal resolution of this method depends on the number of channels available
and capabilities of the potentiostat. Current density distribution measurements were
made for a wide range of cathode flow rates in order to elucidate the nature of cath-
ode flooding in the DMFC. Figure 17 displays the current density distributions for
a high and a low cathode air flow rates, respectively. In the case of high cathode
stoichiometry (Fig. 17(a)), it can be seen that the current is rather uniform for all
three cell voltages. As expected, the current density increases as the cell voltage
decreases. In the case of low cathode stoichiometry (still excessive for the oxygen
reduction reaction), Fig. 17(b) clearly shows that a portion of the cathode towards
the exit is fully flooded, leading to almost zero current. The information provided
in Fig. 17 can be used to identify innovative cathode flowfield designs and enables
the development of MEA structures with improved water management capabilities.
Material balance analysis proves to be a critical diagnostic tool for the develop-
ment of portable DMFC systems. In this analysis, methanol balance on the anode
side along with the methanol crossover rate typically measured by an infrared CO2
sensor is conducted. In addition, water balance on both anode and cathode sides
is performed, where cathode water is carefully collected by a moisture trap and
measured [17, 18, 59]. From such analyses, Müller et al. [18] revealed that the
Figure 17: Current density distributions in a 50 cm2 DMFC for: (a) high cathode
air flowrate (stoichiometry of 85 @ 0.1 A/cm2 ) and (b) low cathode air
flowrate (stoichiometry of 5 @ 0.1 A/cm2 ).
water balance on the DMFC anode is highly negative, thus calling for membrane
development with low water crossover in addition to low methanol crossover.
6.3 Model validation
Experimental validation of the two-phase DMFC model of Wang and Wang [79]
has been carried out for a 5 cm2 graphite cell. A brief description of the cell geom-
etry, MEA compositions and operating conditions is given in Fig. 18. Figure 18(a)
illustrates the capability of the model to predict the polarization curves at two cell
temperatures. Excellent agreement is achieved not only in the kinetic- and ohmic-
controlled regions of the polarization curves but also in the mass transport controlled
region, where the methanol oxidation kinetics is modeled as a zero-order reaction
for molar concentrations above 0.1 M, but a first-order reaction for a molarity below
0.1 M. This shift in the reaction order and the molarity of transition is consistent
with direct kinetics measurements. A lower mass transport limiting current density
at 50 ◦ C, seen from Fig. 18(a), is caused by the lower diffusion coefficients in both
liquid and gas phases and the lower saturation methanol vapor concentration in the
gas phase at lower temperatures. Using the same model and physical property data,
Fig. 18(b) shows equally satisfactory agreement in the polarization curves between
numerical and experimental results for different methanol feed concentrations. In
accordance with these experiments, the model prediction for the 2 M case shows
a slightly lower performance (due primarily to higher methanol crossover) and an
extended limiting current density. Similar success in validating global I-V curves
was also reported by Murgia et al. [7], among others.
While the model validation against cell overall performance data has been
satisfactory and encouraging, as evident from Fig. 18, the ultimate test of these
highly sophisticated two-phase models is comparison with detailed distribution
Figure 18: Comparisons of 2-D model predictions with experimental data for a
DMFC with: (a) temperature effect and (b) concentration effect.
0.7
(a) Exp. Data x/L = 0.16 0.8 (b) Numerical Simulation
0.6 x/L = 0.37
x/L = 0.58
0.5 x/L = 0.68
Voltage (V)
Voltage (V)
x/L = 1.0 0.6
Average 1
0.4
0.7
0.4
0.3
0.2
0.2
0.6 0.4 average x/L= 0.15
0.1
0.00 0.05 0.10 0.15 0.20 0.25 0 0.05 0.1 0.15 0.2 0.25
Current Density (A/cm2) Current Density (A/cm2)
Figure 19: Comparison of localized polarization curves between experiments

(a) and model predictions (b) for a 50 cm2 DMFC with the anode flow
stoichiometry of 27 and cathode air stoichiometry of 5 @ 0.1 A/cm2 .
measurements. Figure 19 presents such an attempt toward developing high-fidelity,

first-principles models for DMFC. Figure 19(a) shows a set of localized polarization
curves measured using the current distribution measurement technique of Mench
and Wang [40], and Fig. 19(b) displays the same set of polarization curves predicted
from the DMFC two-phase model of Wang and Wang [79]. A low air stochiometry
of 5 (although not low for the electrochemical reaction requirement) was deliber-
ately chosen so that cathode GDL flooding may occur and a non-uniform current
density distribution results.
The two graphs in Fig. 19 share a qualitative similarity. For example, both exper-
iment and model results indicate that the local polarization curves near the dry air
inlet exhibit a monotonic function between the voltage and current. Also, the shape
of the polarization curves near the exit, from both experiment and simulation, is
clearly evidence of flooding in the cathode GDL. Another interesting observation
is that the average cell polarization curves, measured and predicted, do not exhibit
any sign of cathode flooding, indicating that detailed distribution measurements
are absolutely required in order to discern complex physicochemical phenomena
occurring inside the cell. Finally, it can be seen from Fig. 19 that a satisfactory quan-
titative comparison between experiment and model is lacking on the detailed level.
Difficulties in obtaining good quantitative agreement between predicted and
measured distribution results are indicative that model refinements as well as an
improved property data base will be needed before accurate quantitative predic-
tions of not only overall polarization curve but also detailed distributions within a
DMFC may be obtained.
7 Application: micro DMFC

Micro-power sources are a key technology in future integrated micro-systems that
enable sensing, computing, actuation, control, and communication on a single chip.
Due to such advantages as easy storage of liquid fuel, ambient temperature oper-
ation, and simple construction, the direct methanol fuel cell has received much
attention as a leading candidate for micro-power sources of the future [17].
Thanks to the integrated-circuit (IC) fabrication technology, micro-channel pat-
terns of DMFC bipolar plates into which reactants are fed can be featured on the
silicon wafer with high resolution and good repeatability. Kelley et al. [81] reported
a 0.25 cm2 micro DMFC using silicon (Si) wafer as the substrate. The anode cat-
alyst in their micro DMFC was prepared by coelectrodepositing a Pt-Ru alloy
onto a carbon coated Si wafer. Using 0.5 M methanol solution, the micro DMFC
was tested and yielded an output current density very close to that of large-scale
DMFC. In a subsequent paper [62], they reported that a prototype cell (12 mm3
in volume) micro-fabricated on Si substrates and featuring electrodeposition of
Pt-Ru as the anode catalyst successfully demonstrated a lowering of catalyst load-
ing to 0.25 mg/cm2 without loss of performance. Pavio et al. instead explored
low-temperature co-fired ceramic (LTCC) material as an alternative for the bipolar
plates of micro fuel cell system, and a DMFC prototype, packaged using LTCC,
was reported in their paper [64].
In addition, micro PEM fuel cells based on a Si wafer using microelectro-
mechanical system (MEMS) technology have been under extensive development
[66, 82–85]. Lee et al. [66] reported a micro fuel cell design in which a planar
array of cells are connected in series in a “flip-flop” configuration. Maynard and
Meyers [82] proposed a conceptual design for a miniaturized DMFC for power-
ing 0.5–20 W portable telecommunication and computing devices. Heinzel et al.
[83] demonstrated a prototype miniature fuel cell stack to power a laptop computer
using hydrogen and air as reactants running at ambient pressure and temperature.
Most recently, Yu et al. [84, 85] fabricated a miniature twin-fuel-cell connected in
series by sandwiching two membrane-electrode-assemblies between two Si micro-
machined plates.
In the recent work of Lu et al. [10], the fabrication process of the silicon wafer is
illustrated in Fig. 20. Figure 21 shows a picture of the silicon wafer with fabricated
flow channel pattern. The fluid channels in the Si wafer have a depth of 400 µm.
Both the flow channel and the rib separating two neighboring channels were 750 µm
wide, with the channel length of 12.75 mm. There were a total of nine channels with
serpentine flowfield, forming a cell effective area of approximately 1.625 cm2 . In
order to collect current and to minimize contact resistance between the MEA and
the Si wafer, Ti/Cu/Au (with thickness of 0.01/3/0.5 µm) was deposited on the
front-side of each wafer by electron beam evaporation.
Figure 22 shows a series of cell polarization curves operated at different tem-
peratures using 1 M methanol solution under ambient pressure in the Si-based
micro DMFC [10]. The flow rate of non-preheated air was 88 mL/min and the
methanol feed rate was 2.83 mL/min. The maximum power density of the cell
reached 14.27 mW/cm2 at a voltage of 0.196 V at room temperature (i.e. 23 ◦ C).
The maximum power density was 24.75 mW/cm2 at a voltage of 0.214 V when the
temperature increased to 40 ◦ C. This is because the kinetics of electrodes, particu-
larly methanol oxidation at the anode, is enhanced at elevated temperatures. For the
Figure 20: Fabrication process flow of the µDMFC.
Figure 21: A silicon wafer with flow channels.
same reason, maximum power density was 47.18 mW/cm2 at a voltage of 0.258 V
and temperature of 60 ◦ C.
A problem emerging in Si-based micro DMFCs is that the silicon substrate is
too fragile and it becomes difficult to compress the cell tightly for sealing and to
reduce the contact resistance between the MEA and flow plates. In addition, a thick
gold layer has to be coated on Si substrate to improve the conductivity as a cur-
rent collector. Alternatively, one can use photochemical etching of stainless steel
to fabricate the flow plate/current corrector instead of silicon wafer [86]. Stain-
less steel provides much higher conductivity than silicon, thus avoiding thicker
metal coating on the surface. Also, photochemical etching is a simple, high-quality,
fast-turnaround, low-cost process for micro-fabrication of flow channels on stain-
less steel.
Figure 22: Polarization and power density curves at different temperatures using
1 M methanol solution with the air flow rate of 88 mL/min and methanol
flow rate of 2.83 mL/min at atmospheric pressure.
Figure 23: Polarization curve using 2 M methanol solution at 22 ◦ C, air flow rate of
161 mL/min, methanol flow rate of 2.2 mL/min, and ambient pressure.
Figure 23 shows the polarization and power density curves using 2 M methanol
solution at 22 ◦ C and atmospheric pressure. The flow plates in this DMFC were
made of stainless steel using the photochemical etching method. The flowfield
and cell size were identical to the Si-based micro DMFC reported before [10].
The methanol flow rate was 2.2 mL/min, and the air flow rate was 161 mL/min. At
room temperature, the current density reach 90 mA/cm2 at 0.3 V and the maximum
power density was 34 mW/cm2 at 0.23 V. Figure 24 displays the polarization and
power density curves of the same stainless steel cell using 2 M methanol solution
Figure 24: Polarization curve using 2 M methanol solution at 40 ◦ C, air flow rate
161 mL/min, methanol flow rate 2.2 mL/min, and ambient pressure.
Figure 25: Polarization curve using 2 M methanol solution at 60 ◦ C at different air

flow rates, methanol flow rate 2.2 mL/min, and ambient pressure.
at 40 ◦ C and atmospheric pressure. The cell performance reach about 200 mA/cm2
at 0.3 V and the maximum power density was 62.5 mW/cm2 at 0.26 V.
Figure 25 shows the polarization and power density curves using different air flow
rates. The methanol flow rate was fixed at 2.2 mL/min. At low current densities,
the cell voltages are almost identical under different air flow rates. The overall
cell performance improves with the air flow rate increasing. At the air flow rate
of 375 mL/min, the cell performance reach 330 mA/cm2 at 0.3 V and the maximum
power density was 100 mW/cm2 at 0.28 V.
Figure 26 displays the anode polarization behaviors at different temperatures
when hydrogen was used in the cathode. As expected, the anode overpotential
Figure 26: Anode overpotential with hydrogen cathode, 2 M methanol with flow
rate of 2.2 mL/min, hydrogen flow rate 161 mL/min, and ambient
pressure.
Figure 27: Methanol crossover rate at open circuit voltage, 2 M Methanol with
flow rate of 2.2 mL/min, nitrogen flow rate 161 mL/min, and ambient
pressure.
decreased as the temperature increased. Using this H2 -evolving counter electrode,

the anode overpotential may deviate somewhat from the real kinetic results using
dynamic hydrogen electrode [28]. However, using the H2 -evolving cathode is still
a simple and useful method to evaluate and screen the anode characteristics in an
assembled DMFC.
Figure 27 shows the polarization behavior when air was replaced by humidified
nitrogen in the cathode. Such cells, consisting of MeOH anode and N2 cathode,
allow measurement of the methanol crossover rate. As methanol diffuses from

the anode side to the cathode side, the electrochemical reaction at the N2 cathode
becomes:
CH3 OH + H2 O → CO2 + 6H+ + 6e− .
Protons produced from the above reaction then migrate back to the anode side
of the membrane where they are combined to evolve H2 . That is:
6H+ + 6e− → 3H2 .
Thus, the methanol crossover rate at the open circuit voltage can be determined
from the limiting current density observed from the polarization behavior of such a
cell [28]. It is seen from Fig. 27 that the methanol crossover rate is relatively large
with the small MEA area due to edge leakage [86].
To summarize, micro DMFCs fabricated by photochemical etching of stainless
steel have attained impressive performance (i.e. 100 mW/cm2 at 60 ◦ C) for high-
power applications. For future development, it is necessary to further lower the air
and fuel feeding rates so as to reduce the pumping power. In this regard, a self-
activated micro DMFC holds much promise where the cathode is air breathing and
the anode features a pumpless delivery of liquid fuel. Further development in micro-
DMFCs is to decrease the system volume and increase the methanol concentration
in the fuel tank and hence the energy density.
8 Summary and outlook

The fundamental transport processes of methanol, water and heat occurring in
DMFCs for micro and portable applications have been reviewed, along with a
summary of recent DMFC models and diagnostic techniques. Significant challenges
still exist before a DMFC can compete with the latest Li-ion battery technology.
We have stressed in this chapter that a better understanding of the basic transport
phenomena achieved through combined flow visualization studies and transport
simulations is essential to overcome these challenges and to inspire new design
concepts. We demonstrated that, contrary to conventional wisdom, a DMFC based
on thin Nafion 112 membrane can reach a fuel efficiency of ∼80% and a water
crossover coefficient lower than unity while still maintaining a power density of
56 mW/cm2 at 60 ◦ C and ambient air.
Two-phase modeling capabilities for DMFC have emerged, which unravel the
importance of gas phase transport of methanol as compared to the liquid phase
transport. In addition, much effort is being directed towards developing a coupled
model for methanol, water and heat transport processes simultaneously in a DMFC.
Such models are extremely useful for the discovery of unique design and operation
regimes of the DMFC system for portable application, where the high energy density
entails using highly concentrated methanol (preferably pure methanol), maintain-
ing low water and methanol crossover, and improving high-voltage performance.
The latter two factors will result in high efficiency DMFCs. It is expected that the
DMFC model development will be directed less towards refining model accuracy
or improving computational speed, but more towards applying the models to invent
new cell designs and pinpoint areas of improvement.
Other important research issues such as an accurate materials data base were
discussed in Wang [87].
References
[1] Boslet, S.W., Experimental study of a direct methanol fuel cell, M.S. thesis,
The Pennsylvania State University, 2001.
[2] Halpert, G., Narayanan, S.R., Valdez, T., Chun, W., Frank, H., Kindler, A.,
Surampudi, S., Kosek, J., Cropley C. & LaConti, A., Progress with the direct
methanol liquid-feed fuel cell system. Proc. of the 32nd Intersociety Energy
Conversion Engineering Conference, 2, AIChE: New York, pp. 774–778,
1997.
[3] Baldauf, M. & Preidel, W., Status of the development of a direct methanol
fuel cell. Journal of Power Sources, 84, pp. 161–166, 1999.
[4] Ren, X., Zelenay, P., Thomas, S., Davey, J. & Gottesfeld, S., Recent advances
in direct methanol fuel cells at Los Alamos National Laboratory. Journal of
Power Sources, 86, pp. 111–116, 2000.
[5] Scott, K., Taama, W.M. & Argyropoulos, P., Carbon dioxide evolution pat-
terns in direct methanol fuel cells. Electrochimica Acta, 44, pp. 3575–3584,
1999.
[6] Scott, K., Mass transfer in flow fields. Handbook of Fuel Cells, eds
W. Vielstich, H.A. Gasteiger & A. Lamm, John Wiley and Sons Ltd.:
England, 1, p. 70, 2003.
[7] Murgia, G., Pisani, L., Shukla, A.K. & Scott, K., A numerical model of a
liquid-feed Solid polymer electrolyte DMFC and its experimental validation.
Journal of the Electrochemical Society, 150, pp. A1231–A1245, 2003.
[8] Mench, M., Boslet, S., Thynell, S., Scott, J. & Wang, C.Y., Experimental
study of a direct methanol fuel cell. Direct Methanol Fuel Cells, The Elec-
trochemical Society Proceedings Series: Pennington, NJ, 2001.
[9] Lim, C. & Wang, C.Y., Development of high power electrodes for a liquid-
feed direct methanol fuel cell. Journal of Power Sources, 113, pp. 145–150,
2003.
[10] Lu, G.Q., Wang, C.Y., Yen, T.J. & Zhang, X., Development and characteriza-
tion of a silicon-based micro direct methanol fuel cell. Electrochimica Acta,
49, pp. 821–828, 2004.
[11] Yen, T.J., Fang, N., Zhang, X., Lu, G.Q. & Wang, C.Y., A micro direct
methanol fuel cell operating at near room temperature. Applied Physics
Letters, 83, pp. 4056–4058, 2003.
[12] Gottesfeld, S. & Wilson, M.S., Polymer electrolyte fuel cells as potential
power sources for portable electronic devices. Energy Storage Systems for
Electronics Devices, eds T. Osaka & M. Datta, Gordon and Breach Science
Publishers: Singapore, p. 487, 2000.
[13] Gottesfeld, S. & Zawodzinski, T.A., Adv. in Electrochemical Science and

Engineering, ed. C. Tobias, Vol. 5, Wiley and Sons: New York, 1997.
[14] Lamy, C., Leger, J.-M. & Srinivasan, S., Modern Aspects of Electrochem-
istry, eds J. Bockris & O’M et al., Kluwer Academic/Plenum Publishers:
New York, p. 53, 2001.
[15] Arico, A.S., Srinivasan, S. & Antonucci, V., DMFCs: From fundamental
aspects to technology development. Fuel Cells, 1, pp. 133–161, 2001.
[16] Neergat, M., Friedrich, K.A. & Stimming, U., New material for DMFC
MEAs. Handbook of Fuel Cells, eds W. Vielstich, H.A. Gasteiger &
A. Lamm, John Wiley and Sons Ltd.: England, Chap 63, pp. 856–877,
2003.
[17] Narayanan, S.R., Valdez, T.I. & Rohatgi, N., DMFC system design for
portable applications. Handbook of Fuel Cells, eds W. Vielstich,
H.A. Gasteiger & A. Lamm, John Wiley and Sons Ltd.: England, Chap-
ter 65, pp. 894–904, 2003.
[18] Müller, J., Frank, G., Colbow, K. & Wilkinson, D., Transport/kinetic lim-
itations and efficiency losses. Handbook of Fuel Cells, eds W. Vielstich,
H.A. Gasteiger & A. Lamm, John Wiley and Sons Ltd.: England, Chap 62,
pp. 847–855, 2003
[19] Pasaogullari, U. & Wang, C.Y., Two-phase transport and the role of micro-
porous layer in polymer electrolyte fuel cells. Electrochimica Acta, submitted
for publication: 2004.
[20] Barendrecht, E., Electrochemistry of fuel cells. Fuel cell systems, eds
L.J.M.J. Blomen & M.N. Mugerwa, New York, p. 75, 1993.
[21] Wasmus, S. & Kuver, A., Methanol oxidation and direct methanol fuel
cells: a selective review. Journal of Electrochemical Society, 461, pp. 14–31,
1999.
[22] Lin, W.F., Wang, J.T. & Savinell, R.F., On-line FTIR Spectroscopic investi-
gations of methanol oxidation in a direct methanol fuel cell. Journal of the
Electrochemical Society, 144, pp. 1917–1922, 1997.
[23] Hamnett, A., Mechanism and electrocatalysis in the direct methanol fuel cell.
Catalysis Today, 38, pp. 445–457, 1997.
[24] Thomas, S.C., Ren, X. & Gottesfeld, S., Influence of ionomer content in
catalyst layers on direct methanol fuel cell performance. Journal of the
[25] Liu, L., Pu, C., Viswanathan, R., Fan, Q., Liu, R. & Smotkin, E.S., Carbon
supported and unsupported Pt-Ru anodes for liquid feed direct methanol fuel
cells. Electrochimica Acta, 43, pp. 3657–3663, 1998.
[26] Hayden, E., The promotion of CO electro-oxidation on platinum-bismuth as
a model for surface mediated oxygen transfer. Catalysis Today, 38, pp. 473–
481, 1997.
[27] Page, T., Johnson, R., Hormes, J., Noding, S. & Rambabu, B., A study of
methanol electro-oxidation reactions in carbon membrane electrodes and
structural properties of Pt alloy electro-catalysts by EXAFS. Journal of
Electroanalytical Chemistry, 485, pp. 34–41, 2000.
[28] Ren, X., Springer, T.E. & Gottesfeld, S., Water and methanol uptakes in
nafion membranes and membrane effects on direct methanol cell perfor-
mance. Journal of the Electrochemical Society, 147, pp. 92–98, 2000.
[29] Arico, S., Creti, P., Modica, E., Monforte, G., Baglio, V. & Antonucci, V.,
Investigation of direct methanol fuel cells based on unsupported Pt–Ru
anode catalysts with different chemical properties. Electrochimica Acta, 45,
pp. 4319–4328, 2000.
[30] Ravikumar, M.K. & Shukla, A.K., Effect of methanol crossover in a liquid-
feed Polymer-electrolyte direct methanol fuel cell. Journal of the Electro-
chemical Society, 143, pp. 2601–2605, 1996.
[31] Apgyropoulos, R., Scott, K. & Taama, W.M., Gas evolution and power per-
formance in direct methanol fuel cells. Journal of Applied Electrochemistry,
29, pp. 663–671, 1999.
[32] Lu, G.Q. & Wang, C.Y., Electrochemical and flow characterization of a direct
methanol fuel cell. Journal of Power Sources, 134, pp. 33–40, 2004.
[33] Wallis, G.B., One Dimensional Two-Phase Flow, McGraw-Hill: New York,
1969.
[34] Mathias, M.F., Roth, J., Fleming, J. & Lehnert, W., Diffusion media material
and characterisation. Handbook of Fuel Cells – Fundamental, Technology
and Application, eds W. Vielstich, A. Lamm & H.A. Gasteiger, Volume 3,
John Wiley & Sons, pp. 517–537, 2003.
[35] Whalley, P.B., Boiling, Condensation and Gas-Liquid Flow, Clarendon press:
Oxford, p. 16, 1987.
[36] Springer, T.E., Zawodzinski, T.A. & Gottesfeld, S., Polymer electrolyte fuel
cell model. Journal of the Electrochemical Society, 138, pp. 2334–2341,
1991.
[37] Wang, Z.H., Wang, C.Y. & Chen, K.S., Two-phase flow and transport in
the air cathode of proton exchange membrane fuel cells. Journal of Power
Sources, 94, pp. 40–50, 2001.
[38] Wang, C.Y., Two-phase flow and transport. Handbook of Fuel Cells,
eds W. Vielstich, H.A. Gasteiger & A. Lamm, John Wiley and Sons Ltd.:
England, 3, pp. 337–348, 2003.
[39] Yang, X.G., Zhang, F.Y., Lubawy, A. & Wang, C.Y., Visualization of liquid
water transport in a polymer electrolyte fuel cell. Electrochemical & Solid-
State Letters, in press, 2004.
[40] Mench, M.M. & Wang, C.Y., An in situ method for determination of current
distribution in PEM fuel cells applied to a direct methanol fuel cell. Journal
of the Electrochemical Society, 150, pp. A79–A85, 2003.
[41] Ren, X. & Gottesfeld, S., Electro-osmotic drag of water in poly(perfluoro-
sulfonic acid) membranes. Journal of the Electrochemical Society, 148,
pp. A87–A93, 2001.
[42] Narayanan, S.R., Frank, H., Jeffries-Nakamura, B., Smart, M., Chun, W.,
Halpert, G., Kosek, J. & Cropley, C., Proton Conducting Membrane Fuel
Cells I, eds S. Gottesfeld, G. Halpert & A. Landgrebe, The Electrochemical
Society Proceedings Series: Pennington, NJ, PV 95-23, p. 278, 1995.
[43] Ren, X., Zawodzinski, T.A., Uribe, F., Dai, H. & Gottesfeld, S., Proton Con-
ducting Membrane Fuel Cells I, eds S. Gottesfeld, G. Halpert &
A. Landgrebe, The Electrochemical Society Proceedings Series: Pennington,
NJ, PV 95-23, p. 284, 1995.
[44] Ren, X., Springer, T.E., Zawodzinski, T.A. & Gottesfeld, S., Methanol trans-
port through nation membranes, electro-osmotic drag effects on potential
step measurements. Journal of the Electrochemical Society, 147,
pp. 466–474, 2000.
[45] Valdez, T.I. & Narayanan, S.R., Proton Conducting Membrane Fuel Cells
II, eds S. Gottesfeld, T.F. Fuller & G. Halpert, The Electrochemical Society
Proceedings Series: Pennington, NJ, PV 98-27, p. 380, 1998.
[46] Wang, J.T., Wasmus, S. & Savinell, R.F., Real-Time Mass spectrometric
study of the methanol crossover in a direct methanol fuel cell. Journal of the
[47] Kauranen, P.S. & Skou, E., Mixed methanol oxidation/oxygen reduction cur-
rents on a carbon supported Pt catalyst. Journal of Electroanalytical Chem-
istry, 408, pp. 189–198, 1996.
[48] Kuver, A. & Vielstich, W., Investigation of methanol crossover and single
electrode performance during PEMDMFC operation – A study using a solid
polymer electrolyte membrane fuel cell system. Journal of Power Sources,
74 (2), pp. 211–218, 1998.
[49] Scott, K., Taama, W.M., Argyropoulos, P. & Sundmacher, K., The impact
of mass transport and methanol crossover on the direct methanol fuel cell.
Journal of Power Sources, 83(1–2), pp. 204–216, 1999.
[50] Garau, B. & Smotkin, E.S., Methanol crossover in direct methanol fuel cells:
a link between power and energy density. Journal of Power Sources, 112,
pp. 339–352, 2002.
[51] Heinzel, A. & Barragan, V.M., A review of the state-of-the-art of the
methanol crossover in direct methanol fuel cells. Journal of Power sources,
84, pp. 70–74, 1999.
[52] Peled, E., Duvdevani, T., Aharon, A. & Melman, A., A direct methanol fuel
cell based on a novel low-cost nanoporous proton-conducting membrane.
Electrochemical and Solid State Letters, 3(12), pp. 525–528, 2000.
[53] Nunes, S.P., Ruffmann, B., Rikowski, E., Vetter, S. & Richau, K., Inorganic
modification of proton conductive polymer membranes for direct methanol
fuel cells. Journal of Membrane Science, 203(1–2), pp. 215–225, 2002.
[54] Yamaguchi, T., Ibe, M., Nair, B.N. & Nakao, S., A pore-filling electrolyte
membrane-electrode integrated system for a direct methanol fuel cell appli-
cation. Journal of the Electrochemical Society, 149(11), pp. A1448–A1453,
2002.
[55] Pan, Y., Ph.D. Thesis, The Pennsylvania State University, 2004.
[56] Peled, E., Blum, A., Aharon, A., Philosoph, M. & Lavi, Y., Novel approach
to recycling water and reducing water loss in DMFCs. Electrochemical and
Solid-State Letters, 6, pp. A268–A271, 2003.
[57] Blum, A., Duvdevani, T., Philosoph, M., Rudoy, N. & Peled, E., Water-
neutral micro direct-methanol fuel cell (DMFC) for portable applications.
Journal of Power Sources, 117, pp. 22–25, 2003.
[58] Pasaogullari, U. & Wang, C.Y., Liquid water transport in gas diffusion layer
of polymer electrolyte fuel cells. Journal of the Electrochemical Society,
151, pp. A399–A406, 2004.
[59] Lu, G.Q., Liu, F.Q. & Wang, C.Y., Water transport through Nafion 112 mem-
brane in direct methanol fuel cells. Electrochemical & Solid-State Letters,
8(1), pp. A1–A4, 2005
[60] Argyropoulos, P., Scott, K. & Taama, W.M., One-dimensional thermal model
for direct methanol fuel cell stacks Part I. Model development. Journal of
Power Sources, 79, pp. 169–183, 1999.
[61] Dohle, H., Mergel, J. & Stolten, D., Heat and power management of a
direct-methanol-fuel-cell (DMFC) system. Journal of Power Sources, 111,
pp. 268–282, 2002.
[62] Kelley, S.C., Deluga, G.A. & Smyrl, W.H., Miniature fuel cells fabricated
on silicon substrates. AICHE J, 48, pp. 1071–1082, 2002.
[63] Mench, M.M., Wang, Z.H., Bhatia, K. & Wang, C.Y., Design of a micro direct
methanol fuel cell. Proceedings of the International Mechanical Engineering
Congress and Exposition (IMECE), November 11–16, 2001.
[64] Pavio, J., Bostaph, J., Fisher, A., Hallmark, J., Mylan, B.J. & Xie, C.G.,
LTCC fuel cell system for portable wireless electronics. Advancing Micro-
electronics, 29, pp. 1–8, 2002.
[65] Dyer, C.K., Fuel cells for portable applications. Journal of Power Sources,
106, pp. 31–34, 2002.
[66] Lee, S.J., Chang-Chien, A., Cha, S.W., O’Hayre, R., Park, Y.I., Saito, Y. &
Prinz, F.B., Design and fabrication of a micro fuel cell array with ‘flip-flop’
interconnection. Journal of Power Sources, 112, pp. 410–418, 2002.
[67] Scott, K., Argyropoulos, P. & Sundmacher, K., A model for the liquid feed
direct methanol fuel cell. Journal of Electroanalytical Chemistry, 477,
pp. 97–110, 1999.
[68] Sundmacher, K. & Scott, K., Direct methanol polymer electrolyte fuel cell:
Analysis of charge and mass transfer in the vapour–liquid–solid system.
Chemical Engineering Science, 54, pp. 2927–2936, 1999.
[69] Argyropoulos, P., Scott, K. & Taama, W. M., Hydrodynamic modelling of
direct methanol liquid feed fuel cell stacks. Journal of Applied Electrochem-
istry, 30, pp. 899–913, 2000.
[70] Baxter, S.F., Battaglia, V.S. & White, R.E., Methanol fuel cell model: anode.
Journal of the Electrochemical Society, 146, pp. 437–447, 2000.
[71] Wang, J. & Savinell, R.F., Electrode Materials and Processes for
Energy Conversion and Storage, eds S. Srinivasan, D.D. Macdonald &
A.C. Khandkar, The Electrochemical Society Proceedings Series:
Pennington, NJ, PV 94–23, p. 326, 1994.
[72] Kulikovsky, A.A., Divisek, J. & Kornyshev, A.A., Two-dimensional simula-

tion of direct methanol fuel cell a new (embedded) type of current collector.
Journal of the Electrochemical Society, 147, pp. 953–959, 2000.
[73] Kulikovsky, A.A., Two-dimensional numerical modelling of a direct
methanol fuel cell. Journal of Applied Electrochemistry, 30(9), pp. 1005–
1014, 2000.
[74] Dohle, H., Divisek, J. & Jung, R., Process engineering of the direct methanol
fuel cell. Journal of Power Sources, 86, pp. 469–477, 2000.
[75] Meyers, J.P. & Newman, J., Simulation of the direct methanol fuel cell I.
Thermodynamic framework for a multicomponent membrane. Journal of
the Electrochemical Society, 149, pp. A710–A717, 2002.
[76] Meyers, J.P. & Newman, J., Simulation of the direct methanol fuel cell II.
Modeling and data analysis of transport and kinetic phenomena. Journal of
the Electrochemical Society, 149, pp. A718–A728, 2002.
[77] Meyers, J.P. & Newman, J., Simulation of the direct methanol fuel cell III.
Design and optimization. Journal of the Electrochemical Society, 149,
pp. A729–A735, 2002.
[78] Nordlund, J. & Lindbergh, G., A model for the porous direct methanol fuel
cells anode. Journal of the Electrochemical Society, 149, pp. A1107–A1113,
2002.
[79] Wang, Z.H. & Wang, C.Y., Mathematical modeling of liquid-feed direct
methanol fuel cells. Journal of the Electrochemical Society, 150, pp. A508–
A519, 2003.
[80] Divisek, J., Fuhrmann, J., Gartner, K. & Jung, R., Performance modeling
of a direct methanol fuel cell. Journal of the Electrochemical Society, 150,
pp. A811–A825, 2003.
[81] Kelley, S.C., Deluga, G.A. & Smyrl, W.H., A miniature methanol/air polymer
electrolyte fuel cell. Electrochemical and Solid-State Letters, 3, pp. 407–409,
2000.
[82] Maynard, H.L. & Meyers, J.P., Miniature fuel cells for portable power: design
considerations and challenges. The Journal of Vacuum Science and Technol-
ogy B, 20, pp. 1287–1297, 2002.
[83] Heinzel, A., Hebling, C., Muller, C., Zedda, M.M. & Muller, C., Fuel cells
for low power applications. Journal of Power Sources, 105, pp. 250–255,
2002.
[84] Yu, J.R., Cheng, P., Ma, Z.Q. & Yi, B.L., Fabrication of a miniature
twin-fuel-cell on silicon wafer. Electrochimica Acta, 48, pp. 1537–1541,
2003.
[85] Yu, J.R., Cheng, P., Ma, Z.Q. & Yi, B.L., Fabrication of miniature silicon
wafer fuel cells with improved performance. Journal of Power Sources,
124(1), pp. 40–46, 2003.
[86] Lu, G.Q. & Wang, C.Y., Development of micro direct methanol fuel cells for
high power applications. Journal of Power Sources, in press.
[87] Wang, C.Y., Fundamental models for fuel cell engineering. Chemical Reviews,
104, pp. 4727–4766, 2004.
Nomenclature
ba Tafel slope of anode methanol oxidation reaction
bc Tafel slope of cathode oxygen reduction reaction
D diffusion coefficient
db bubble diameter
dp backing layer pore size
E thermodynamic equilibrium potential
F Faraday constant
g gravitational acceleration
g Gibbs free energy change per mole of fuel
h enthalpy change per mole of fuel
hlg latent heat of evaporation
I operating current density
IA,lim anode mass-transport limiting current density
Ic,oc crossover current at open circuit
Io,a anode exchange current density
Io,c cathode exchange current density
Ix over methanol crossover current density
jair air molar flow rate at the inlet cathode
je water evaporation flux
jH2 O water flux
jO2 consumption rate of oxygen
Km membrane hydraulic permeability
MH2 O water molecular weight
n number of electrons transferred for each molecule of fuel
nd electro-osmotic drag coefficient of water
pH2 O water partial pressure
psat (T ) saturation pressure
ptotal operating pressure
Q heat generation rate
RH relative humidity
s entropy change per mole of fuel
T temperature
ub bubble drift velocity through the liquid
Vcell cell voltage
Wact electric work
Wmax maximum possible work
Greek symbols
α net water transport coefficient through the membrane

δm membrane thickness
η total energy efficiency
ηa anode activation overpotential
ηfuel fuel efficiency

ηrev thermodynamic efficiency
ηvoltaic voltaic efficiency
µ dynamic viscosity
µl liquid water viscosity
θ surface contact angle
ρl liquid density
ρg gas density
σ liquid/gas interfacial tension
ξ the stoichiometry defined at the current density of I
ξcri critical air stoichiometry
Exergy Method Computational Methods in
Technical and Ecological Multiphase Flow III
Applications Editors: C.A. BREBBIA, Wessex Institute
J. SZARGUT, Silesian University of of Technology, UK and A.A. MAMMOLI,
Technology, Poland University of New Mexico, USA
The exergy method makes it possible to New advanced numerical methods and
detect and quantify the possibilities of computer architectures have greatly
improving thermal and chemical processes improved our ability to solve complex
and systems. The introduction of the multiphase flow problems. In this volume
concept “thermo-ecological cost” modellers, computational scientists and
(cumulative consumption of non-renewable experimentalists share their experiences,
natural exergy resources) generated large successes, and new methodologies in order
application possibilities of exergy in to further progress this important area of
ecology. fluid mechanics.
This book contains a short presentation on Originally presented at the Third
the basic principles of exergy analysis and International Conference on Computational
discusses new achievements in the field over Methods in Multiphase Flow, the papers
the last 15 years. One of the most important included cover topics such as: BASIC
issues considered by the distinguished author SCIENCE - Creeping Flow; Turbulent Flow;
is the economy of non-renewable natural Linear and Nonlinear Fluids;
exergy. Cavitation. DNS AND OTHER
Previously discussed only in scientific SIMULATION TOOLS - Finite Elements;
journals, other important new problems Boundary Elements; Finite Volumes;
highlighted include: calculation of the Stokesian Dynamics; Large Eddy
chemical exergy of all the stable chemical Simulation; Interface Tracking Methods.
elements, global natural and anthropogenic MEASUREMENT AND EXPERIMENTS -
exergy losses, practical guidelines for Laser Doppler Velocimetry; Nuclear
improvement of the thermodynamic Magnetic Resonance; X-Ray; Ultrasound.
imperfection of thermal processes and APPLICATIONS – Combustors; Injectors,
systems, development of the determination Nozzles; Injection Moulding; Casting.
methods of partial exergy losses in thermal Series: Advances in Fluid Mechanics,
systems, evaluation of the natural mineral Vol 44
capital of the Earth, and the application of ISBN: 1-84564-030-6 2005 apx 400pp
exergy for the determination of a pro- apx £140.00/US$224.00/€210.00
ecological tax.
A basic knowledge of thermodynamics is
assumed, and the book is therefore most
appropriate for graduate students and
engineers working in the field of energy WITPress
and ecological management. Ashurst Lodge, Ashurst, Southampton,
Series: Developments in Heat Transfer, SO40 7AA, UK.
Vol 18 Tel: 44 (0) 238 029 3223
ISBN: 1-85312-753-1 2005 192pp Fax: 44 (0) 238 029 2853
£77.00/US$123.00/€115.50 E-Mail: witpress@witpress.com
Condensation Heat in industry and consultancy, this book
focuses on the modeling and simulation of
Transfer Enhancement fluid flow and thermal transport phenomena
in turbulent convective flows. Its overall
V.G. RIFERT, Thermodistillation, Kiev, objective is to present state-of-the-art
Ukraine and H.F. SMIRNOV, State knowledge in order to predict turbulent heat
Academy of Refrigeration, Odessa, transfer processes in fundamental and
Ukraine idealized flows as well as in engineering
applications.
Professors V.G. Rifert and H.F. Smirnov The chapters, which are invited
have carried out research on heat transfer contributions from some of the most
enhancement by condensation and boiling prominent scientists in this field, cover a
for over 30 years and have published more wide range of topics and follow a unified
than 200 papers on the topic. In this book outline and presentation to aid accessibility.
they provide research results from the Contents: An Overview of Turbulence
former USSR to which there has previously Modeling; Unstructured Large Eddy and
been little access and also describe different Conjugate Heat Transfer Simulations of
theoretical models of the condensation Wall-Bounded Flows; Numerical Simulation
process. of Turbulence-Radiation Interactions in
Partial Contents: Theoretical Principles Turbulent Reacting Flows; Improved
of Heat Transfer at Film Condensation; Turbulence Modeling of Film Cooling Flow
Condensation on Horizontal Low-Finned and Heat Transfer; Prediction of Turbulent
Tubes – Theoretical Models; Experimental Heat Transfer in Impinging Jet Geometries;
Study of Condensation Heat Transfer on On RANS-Based Models for Prediction of
Finned Tubes; Condensation on Vertical Turbulent Flow and Heat Transfer in Ribbed
Profiled Surfaces and Tubes; Heat Transfer Ducts; Prediction of Transitional
Enhancement at Film Condensation Inside Characteristics of Flow and Heat Transfer
Tubes; Condensation in the Electric Field; in Periodic Fully Developed Ducts;
Hydrodynamics and Heat Transfer at Film Turbulent and Conjugate Heat Transfer
Condensation of Rotating Surfaces. Simulation for Gas Turbine Application;
Series: Developments in Heat Transfer, Simulation of Turbulent Flow in a Duct With
Vol 10 and Without Rotation-Cooling Passage of
ISBN: 1-85312-538-5 2004 392pp Gas-Turbine Blades.
£144.00/US$230.00/€216.00 Series: Developments in Heat Transfer,
Vol 16
Modelling and Simulation

ISBN: 1-85312-956-9 2005 360pp
£124.00/US$198.00/€186.00
of Turbulent Heat
Transfer We are now able to supply you with details of new
WIT Press titles via
Editors: B. SUNDÉN, Lund Institute of E-Mail. To subscribe to this free service, or for
Technology, Sweden and M. FAGHRI, information on any of our titles, please contact
University of Rhode Island, USA the Marketing Department, WIT Press, Ashurst
Lodge, Ashurst, Southampton, SO40 7AA, UK
Tel: +44 (0) 238 029 3223
Providing invaluable information for both Fax: +44 (0) 238 029 2853
graduate researchers and R & D engineers E-mail: marketing@witpress.com
Heat and Fluid Flow in Thermal Analysis of
Microscale and Nanoscale Welds
Structures N.T. NGUYEN, ETRS Pty Ltd., HRL
Services, Mulgrave, Australia
Editors: M. FAGHRI, University of Rhode
Island, USA and B. SUNDÉN, Lund This book is written for postgraduate
Institute of Technology, Sweden students, and welding, mechanical design
and research engineers who deal with
Presenting state-of-the-art knowledge in heat problems such as residual stresses and
transfer and fluid flow in micro- and nanoscale distortions of welds, design of welded
structures, this book provides invaluable structures, micro-structure modelling of
information for both graduate researchers and welds and optimisation of welding sequences
R&D engineers in industry and consultancy. and procedures.
All of the chapters are invited contributions The subject is approached in a simplified
from some of the most prominent scientists in way, focusing on heat conduction and step-
this active area of interdisciplinary research by-step derivation of the analytical solutions
and follow a unified outline and presentation using fundamental calculus. Solutions are
to aid accessibility. given in closed form and are ready for use
Contents: Miniature and Microscale Energy by means of simple integrals, while
Systems; Nanostructures for Thermoelectric applications are demonstrated through easily
Energy; Heat Transport in Superlattices and accessible case studies.
Nanowires; Thermomechanical Formation Special features include new analytical
and Thermal Detection of Polymer solutions developed by the author for vaious
Nanostructures; Two-Phase Flow 3D heat sources and a CD-ROM containing
Microstructures in Thin Geometries – Multi- Visual Basic programs combined into the
Field Modeling; Radiative Energy Transport WHEATSIM (weld heat simulation)
at the Spatial and Temporal Micro/ package.
Nanoscales; Direct Simulation Monte Carlo Contents: Introduction to Welding
of Gaseous Flow and Heat Transfer in a Processes and Heat Sources; Methods of
Microchannel; DSMC Modeling of Near- Analysis; Analytical Solutions for Basic Heat
Interface Transport in Liquid-Vapor Phase- Sources; Analytical Solutions for 2D
Change Processes with Multiple Microscale Gaussian-Distributed Heat Sources;
Effects; Molecular Dynamics Simulation of Analytical Solutions for Spherical Heat
Nanoscale Heat and Fluid Flow. Sources; Analytical Solutions for Single-
Series: Developments in Heat Transfer, Ellipsoidal Density Heat Source; Analytical
Vol 13 Solutions for Double-Ellipsoidal Density
ISBN: 1-85312-893-7 2004 392pp Heat Source; Application in Weld-Pool
£117.00/US$187.00/€175.50 Shape Simulation; Thermal Stresses and
Distortions; Modelling of Residual Stresses
in Welded Joints; Microstructure Modelling
of Fusion Welds; Appendices.
Series: Developments in Heat Transfer,
Find us at
http://www.witpress.com Vol 14
ISBN: 1-85312-951-8 2004
Save 10% when you order from our encrypted 352pp+CD-ROM
ordering service on the web using your credit card. £124.00/US$198.00/€186.00
Heat Transfer in Gas Thermal Conversion of
Turbines Solid Fuels
Editors: B. SUNDÉN, Lund Institute of B. PETERS, Research Centre Karlsruhe,
Technology, Sweden and M. FAGHRI, Karlsruhe, Germany
University of Rhode Island, USA
“This is a very good text that is worthy of
Containing invited contributions from some purchase.”
of the most prominent specialists working ENERGY SOURCES
in this field, this unique title reflects recent
active research and covers a broad spectrum “...an invaluable text for students examining
of heat transfer phenomena in gas turbines. solid fuel combustion, the detailed
All of the chapters follow a unified outline exploration of different aspects of packed
and presentation to aid accessibility and the bed combustion also makes [it] an ideal
book provides invaluable information for reference for experienced engineers and
both graduate researchers and R&D postgraduates.”
engineers in industry and consultancy. PETROLEUM SCIENCE AND
Partial Contents: Heat Transfer Issues in TECHNOLOGY
Gas Turbine Systems; Combustion
Chamber Wall Cooling - The Example of This book deals with the complex process,
Multihole Devices; Conjugate Heat and various aspects, of the thermal
Transfer - An Advanced Computational conversion of solid fuel beds. All three major
Method for the Cooling Design of Modern sub-processes, computational fluid
Gas Turbine Blades and Vanes; Enhanced dynamics, motion of granular media and
Internal Cooling of Gas Turbine Airfoils; combustion, are introduced thoroughly and
Computations of Internal and Film Cooling; their contribution to the process described.
Heat Transfer Predictions of Stator/Rotor The text differs from other approaches in
Blades; Recuperators and Regenerators in that the packed bed is considered as an
Gas Turbine Systems. ensemble of fuel particles. The entire
Series: Developments in Heat Transfer, conversion process is composed of the sum
Vol 8 of the particle processes and this represents
ISBN: 1-85312-666-7 2002 536pp its structure more realistically than traditional
£159.00/US$247.00/€238.50 approaches.
Series: Developments in Heat Transfer,
Vol 15
ISBN: 1-85312-953-4 2003 220pp
WIT Press is a major publisher of engineering
£79.00/US$122.00/€118.50
research. The company prides itself on producing
books by leading researchers and scientists at the
cutting edge of their specialities, thus enabling readers
to remain at the forefront of scientific developments.
Our list presently includes monographs, edited
volumes, books on disk, and software in areas such
as: Acoustics, Advanced Computing, Architecture
and Structures, Biomedicine, Boundary Elements,
Earthquake Engineering, Environmental All prices correct at time of going to press but
Engineering, Fluid Mechanics, Fracture Mechanics, subject to change.
Heat Transfer, Marine and Offshore Engineering and WIT Press books are available through your
Transport Engineering. bookseller or direct from the publisher.

Transport Phenomena in Fuel Cells

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Transport Phenomena in Fuel Cells

Uploaded by

Copyright:

Available Formats

TRANSPORT PHENOMENA

WIT Press publishes leading books in Science and Technology.

The Developments in Heat Transfer book Series publishes state-of-the-art

C.I. Adderley S. del Guidice

C.A. Brebbia P.J. Heggs

R.M. Cotta D.B. Ingham

G. de Vahl Davies P.S. Larsen

A.J. Nowak D.B. Spalding

P.H. Oosthuizen E. Van den Bulck

Editors: B. Sundén and M. Faghri

For USA, Canada and Mexico

British Library Cataloguing-in-Publication Data

A Catalogue record for this book is available

Library of Congress Catalog Card Number: 2004116358

© WIT Press 2005.

Printed in Great Britain by Athenaeum Press Ltd.

B. Sundén and M. Faghri

Multiple transport processes in solid

Figure 1: Principle of operation of a SOFC.

2 Thermodynamic and electrochemical fundamentals for

H = (xhX + yhY + · · ·) − (ahA + bhB + · · ·), (3)

G = (xgX + ygY + · · ·) − (agA + bgB + · · ·)

−G −G0 RT (pA /p0 )a (pB /p0 )b · · ·

2.1 Operation with hydrogen fuel

298.15 −110530 −169474 −393510 −457254 0 −38968

The Nernst potential from this electrochemical reaction will be:

Formula O2 H2 O (Gas) CH4

298.15 0 −61151 −241814 −298105 – –

298.15 −241.814 −228.561 1.184 – – – –

As previously mentioned, it is necessary to have a high operating temperature in a

Reforming : CH 4 + H2 O ↔ CO + 3H2 . (15)

Shift : CO + H2 O ↔ CO2 + H2 . (16)

The dominant electrochemical reaction has been reported to be the oxidation

Rearranging this equation gives:

E(H2 +1/2O2 =H2 O) = E(CO+1/2O2 =CO2 ) , (22)

CH4 out = CH4 in − x̄, (24)

Mfout = Mfin + 2x̄. (29)

From the electrochemical reaction, the molar consumption of oxygen on the

QReforming = H Reforming · x̄, (33)

3 Electrical potential losses

Figure 2: Over-potential in the operation of a fuel cell.

interfaces, which are directly proportional to their mole fractions. It is important

3.1 Activation polarization

The Butler-Volmer equation is a well-known expression for the activation polar-

from which the activation polarization can be expressed as:

3.2 Ohmic loss

Table 2: Data and equations for resistivity of SOFC components.

Bessette 0.008114e500/T 0.00294e10350/T 0.00298e−1392/T –

∗At temperature of 1000 ◦ C.

Figure 3: Schematic of a planar type SOFC.

Figure 4: Schematic of a tubular SOFC.

Figure 5: Ion/electron conduction network in a tubular SOFC.

In the same way for a grid point P in the cathode:

QPR = (H − G) · IP /(2F). (51)

3.3 Mass transport and concentration polarization

Figure 6: The interrelation amongst concentration and other parameters.

4 Computer modeling of a tubular SOFC

4.1 Outline of a computation domain

Figure 7: Schematic of a tubular SOFC in a cell stack.

Figure 8: Computation domain for a tubular SOFC.

4.2 Governing equations and boundary conditions

Table 3: Properties of SOFC materials.

Thermal conductivity Cp Density

Cathode d 11; c 2.0; b 2.0 b 623 a 4930

ṁ = gB, (37)

be adopted, where is a generic scalar (continuity, mass fraction, enthalpy, etc.)

aW (W − P ) + aE (E − P ) + aS (S − P ) + aN (N − P )