Você está na página 1de 27

A Reviewof Mathematical Models for

Hydrogen and Direct Methanol Polymer


Electrolyte Membrane Fuel Cells
K. Z. Yao
1
, K. Karan
1*
, K. B. McAuley
1
, P. Oosthuizen
2
, B. Peppley
3
, and T. Xie
4
1
Department of Chemical Engineering, Queen's University, Kingston, Ontario, CanadaK7L 3N6
2
Department of Mechanical Engineering, Queen's University, Kingston, Ontario, CanadaK7L 3N6
3
Department of Chemistry and Chemical Engineering, Royal Military College of Canada, Kingston, Ontario, Canada K7K7B4
4
DuPont Canada Inc., Research and Business Development Center, Kingston, Ontario, CanadaK7L 5A5
Received14.04.03, in revisedform15.12.03, accepted 16.12.03
1 Introduction
The hydrogen fuel cell and direct methanol fuel cell
(DMFC) are among the most favorable candidates for repla-
cing the internal combustion engine in automobiles and bat-
teries in portable and micro applications. Validated mathe-
matical models provide a powerful tool for the development
of improved fuel cell designs. Mathematical models can be
used to describe the fundamental phenomena taking place in
the fuel cell system, predict fuel cell performance under dif-
ferent operating conditions and optimize the design of fuel
cell systems. In this paper, empirical and fundamental mod-
els for both the polymer electrolyte hydrogen fuel cell and
DMFC are reviewed:
Single cell models, which describe the electrochemical and
transport processes in each fuel cell component, and the
pressure drop, flow distribution, and temperature profile
in the flow field;
Manifold stack model for the entire fuel cell stack, which
evaluates the pressure loss and temperature variation of
the entire fuel cell stack, and flow distribution between
each individual fuel cell.
Mathematical modeling for the analysis of the integrated
fuel cell power systems, which includes the reformer, fuel
purification, exhaust fuel processing, and waste heat recov-
ery, is beyond the scope of this paper. The electrochemical
kinetics models for different catalysts have been reviewed
by various authors [17] and will not be reviewed here.
2 Single Fuel Cell Models
As shown in Figure 1 [8], a typical fuel cell consists of
seven components: anode flow channel (AF), anode diffusion
layer (AD), anode catalyst layer (AC), membrane (M), cath-
ode flow channel (CF), cathode diffusion layer (CD), and
cathode catalyst layer (CC).
The fuel (hydrogen or methanol) flows through the anode
flow channel, diffuses through the anode diffusion layer, and
reaches the anode catalyst layer, where it is catalytically oxi-
dized and releases electrons and protons. The released elec-
trons are conducted through the catalyst metal and carbon
grains and arrive at the cathode side via the external circuit,
while the protons are transported through the membrane to

[
*
] Corresponding author, karan@chee.queensu.ca
Abstract
This paper presents a review of the mathematical modeling
of two types of polymer electrolyte membrane fuel cells: hy-
drogen fuel cells and direct methanol fuel cells. Models of
single cells are described as well as models of entire fuel cell
stacks. Methods for obtaining model parameters are briefly
summarized, as well as the numerical techniques used to
solve the model equations. Effective models have been de-
veloped to describe the fundamental electrochemical and
transport phenomena occurring in the diffusion layers, cata-
lyst layers, and membrane. More research is required to de-
velop models that are validated using experimental data,
and models that can account for complex two-phase flows of
liquids and gases.
Keywords: Polymer Electrolyte Membrane Fuel Cell, Direct
Methanol Fuel Cell, Mathematical Modeling
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3
R
E
V
I
E
W
DOI: 10.1002/fuce.200300004
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
the cathode catalyst layer. At the same time, oxygen or air is
fed into the cathode flow channel, and oxygen diffuses
through the diffusion layer to the catalyst layer, where it
reacts with protons and electrons and generates water. The
overall reaction occurring in a hydrogen fuel cell is:
2H
2
O
2
2H
2
O
The overall reaction for a DMFC is:
2CH
3
OH 3O
2
2CO
2
4H
2
O
Single fuel-cell models, which aim to quantitatively
describe interactions among the various physical and electro-
chemical phenomena can be divided into two groups: empiri-
cal performance models and mechanistic models. The empiri-
cal performance models use simple empirical equations to
predict how the fuel cell voltage changes with the current
density (polarization curves) at different operating condi-
tions. The mechanistic models account for the detailed funda-
mental phenomena such as heat, momentum, multi-compo-
nent mass transport, and electrochemical processes.
2.1 Empirical Performance Models
Most empirical performance models are focused on the
prediction of polarization curves, which are used to charac-
terize the electrical performance of fuel cells. In general, the
actual fuel cell voltage is lower than its thermodynamic volt-
age when a current is drawn. The voltage loss, usually called
overpotential, originates primarily from three sources: activa-
tion overpotential (due to the activation energies of the elec-
trochemical reactions at the catalyst layers), ohmic overpoten-
tial (due to electrical resistances of the membrane, catalyst
layers, diffusion layers and flow plates), and diffusion over-
potential (due to mass transfer resistances of reactants within
catalyst layers, diffusion layers and flow channels). The fol-
lowing empirical equation developed by Kim et al. [9] is
widely used to calculate the voltage (E) of the hydrogen fuel
cell at different current densities (J):
E = E
0
b log J RJ mexp(nJ) (2.1)
where, E
0
is the thermodynamic open-circuit potential. The
first two terms of this expression together are a form of the
Tafel equation, which accounts for the activation overpoten-
tial; R is the total Ohmic resistance of the fuel cell compo-
nents, so RJ is the Ohmic overpotential; the exponential term
is used to account for the diffusion overpotential. This model
provided a good fit to experimental data over a range of tem-
peratures, pressures, and oxygen flow rates [9]. Subsequently,
Lee et al. [10] added a new term, b ln
p
pO
2
_ _
, to Eq. (2.1) to
further account for the diffusion limitations on the cathode
side. In the Lee et al. model all empirical parameters are func-
tions of temperature and pressure.
Squadrito et al. [11] used an empirical approach to account
for the mass transport limitation. They modified Kim et al.'s
model by replacing the last term on the RHS of Eq. (2.1) with
[m J
n
ln (1-J/J
lim
)], where l and n are empirically determined
constants and J
lim
is the limmiting current density obtained
by fitting experimental data. More recently, Pisani et al. [12]
and Kulikovsky [13] have reported analytical models that
employ a semi-empirical approach to account for the limiting
current behavior, the functional form of which is derived by
solving the diffusional mass transfer equation. However,
some of the parameters in this term are obtained by fitting
experimental data.
Amphlett et al. [14] developed a model using a combina-
tion of mechanistic and empirical techniques. In the mechan-
istic development, the Stefan-Maxwell equation was used to
analyze multi-component mass transport processes in diffu-
sion layers, and to calculate the partial pressures of hydrogen
(p
+
H
2
) and oxygen (p
+
O
2
) at the interfaces between the catalyst
layers and the diffusion layers. The concentration (C
+
H
2
; C
+
O
2
)
on the catalyst layer is then obtained through Henry's law.
The thermodynamic voltage is defined through the Nernst
equation as:
E = 1:229
RT
F
ln(
p
+
H
2
p
+
H
2
O
) 0:5 ln(p
+
O
2
)
_
_
_
_
(2.2)
The activation overpotential is obtained from the empirical
equation:
g
act
= n
1
n
2
T n
3
T ln(C
+
O
2
) n
4
T ln(J) (2.3)
Although no separate diffusion overpotential term is
included in the model, it is partially accounted for by the
term containing C
+
O
2
. The Ohmic overpotential is calculated
from:
g
ohm
= JR (2.4)





z
y
x
AF AD AC M CC CD CF
Fig. 1 3-dimensional schematic diagram of a PEM fuel cell. AF: anode
flow channel, AD: anode diffusion layer, AC: anode catalyst layer, M:
membrane, CF: cathode flow Channel, CD: cathode diffusion layer, CC:
cathode catalyst layer.
R
E
V
I
E
W
4 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
where R is measured experimentally.
This model was initially applied to two particular fuel cell
stacks: the Ballard Mark IV and Mark V, respectively [1517].
The parameters were obtained by fitting the experimental
data generated using the Ballard stacks. Later, Mann et al.
[18] attempted to generalize the form of the model by includ-
ing explicit equations for estimating R and adding terms to
account for varying fuel cell geometries.
The number of empirical performance models for DMFCs
is very limited. Simoglou et al. [19, 20] developed an empirical
model using statistical methods (state space canonical var-
iants analysis). This model provides one-step-ahead predic-
tions of the dynamic voltage response from online measure-
ments of cell voltage and current density for the specific fuel
cell used to fit the model. Kulikovsky [21] recently reported a
model that describes the voltage-current relationship for the
direct methanol fuel cell. The model is based on exact solu-
tion for the catalyst layer reaction and includes the overpo-
tentials due to transport limitation in diffusion backing layer
and that due to methanol crossover.
The advantage of empirical performance models is their
simple structure, and the small computational effort required
to solve the model equations. This is a valuable characteristic
when a number of cells in a stack are to be simulated simulta-
neously [10, 18]. However, the parameters estimated from the
experimental data are usually specific to certain cells and
only valid over a limited range of operating conditions. Addi-
tionally, empirical performance models do not explicitly
account for the fundamental transport phenomena and elec-
trochemical processes. Important information, such as species
concentrations, temperature, and current density distribution,
cannot be predicted, limiting the usefulness of empirical
models for the development of improved fuel cells.
2.2 Mechanistic Models
Since the pioneering work done by Bernardi and Ver-
brugge [23, 24], and Springer et al. [25], considerable effort
has been devoted to the development of a mechanistic model
of fuel cells. Both of these models are one-dimensional. Only
the direction across the fuel cell (x direction in Figure 1) is
analyzed. More recently, various two-dimensional and three-
dimensional models have been developed. The two-dimen-
sional models can be divided into two categories depending
on which two dimensions are accounted for. One group of
models [2633] describe the plane perpendicular to the flow
channels (x-z plane in Figure 1), while the other group of
models [3440] simulate the behavior in the plane formed by
the direction across the fuel cell (x direction in Figure 1) and
the direction along the flow channel (y direction in Figure 1).
As shown in Figure 2, in the first group of models the effect
of flow channel dimension and configuration can be studied
simply by changing the boundary conditions.
However, the changes in the temperature and reactants
fraction along the direction of the flow channel cannot be
accounted for. On the other hand, the second group can pre-
dict the temperature and concentration profiles along the
direction of the flow channel, but cannot simulate the effect
of flow channel and rib size. A relatively simpler approach to
modeling the x-y plane is the so-called quasi-two-dimen-
sional model [4144]. In these models, a one-dimensional
model for the direction along the flow channel (y direction) is
coupled with another one-dimensional model for the direc-
tion across the fuel cell to simulated profiles in the x-y plane.
A similar approach is used in quasi three-dimensional mod-
els, in which a one-dimensional model for the y direction is
coupled with a two-dimensional model in the x-z plane [45],
or a one-dimensional model for x direction is coupled with a
two-dimensional model in the y-z plane [46]. Most recently,
(b)
z
2
3
4
1
x
x=h
x=0
z=w w
2
w
1 z=0















(a)
3
2
4
1
x
x=h
x=0
z=w w
2
w
1 z=0
z
Fig. 2 2-dimensional model in x-z direction. (a) Flow field with straight
channel structure. (b) Flow field with interdigital structure: 1. flow plate,
2. diffusion layer, 3. catalyst layer, 4. membrane.
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 5
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
three-dimensional models have been developed by various
research groups [8, 4749]. Berning et al. [49] described each
fuel cell component with different governing equations, and
boundary conditions were specified between each compo-
nent. A single domain approach was used by Dutta et al. [47,
48] and by Um and Wang [8]. A single set of conservation
equations was applied to all components with different
source terms and physical property parameters used for dif-
ferent regions. The single-domain model is considered super-
ior to the multiple-region approach since no boundary condi-
tions are required between each region.
For a hydrogen fuel cell, humidified hydrogen is fed to the
anode channel. On the cathode side, water is generated
through electrochemical reaction as well as transported from
the anode through electro-osmotic drag. Therefore, two-
phase flow may exist at both the anode and cathode sides.
For a DMFC, the situation on the cathode side is similar to
that of the hydrogen fuel cell. On the anode side, the physical
picture can however be substantially different, depending on
how the methanol is fed. If vaporized methanol and water
are used, the condensation of methanol and water may lead
to two-phase flow with the gas phase as primary phase. On
the other hand, when methanol is fed as water solution, the
gas phase comprises methanol, water vapor, and the carbon
dioxide produced by the oxidization of methanol. In this case,
the liquid phase dominates the two-phase flow. The presence
of two phases at both electrodes has a significant impact on
the mass transfer and heat balance. The liquid water on the
cathode side and the carbon dioxide on the anode side can
reduce the diffusion of oxygen and methanol to the catalyst
sites. Also, the enthalpy of the evaporation and condensation
will affect the temperature distribution [50]. In the early
stages of fuel cell modeling, only single-phase flow was con-
sidered. Recently, more and more work has been devoted to
the study of two-phase flow in fuel cells [28, 37, 40, 46, 5055].
A comprehensive model has been developed by researchers
at Pennsylvania State University [37, 40]
Depending on whether the temperature profile is simu-
lated, fuel cell models can also be categorized as isothermal
models or non-isothermal models. Many of the non-isother-
mal models assume that temperature is constant across the
fuel cell (over the x-z plane) [34, 35, 42, 43, 46]. Therefore, only
the temperature variation along the flow channel is calcu-
lated. The temperature profile across the fuel cell was studied
by Gurau et al. [29], Berning et al. [49], Bevers et al. [51], and
Rowe and Li [50].
While most fuel cell models only study the steady state, a
few dynamic models have been developed by different
research groups [8, 28, 36, 39, 40, 56]. Van Bussel et al. [39]
have shown the simulation results of the current density pro-
file along the flow channel at different times. However,
details of their model are not given in their paper. Sundma-
cher et al. [57] developed a simple dynamic model for DMFC.
The cell voltage response to dynamic changes of methanol
feed concentration was predicted. Dynamic models have
been developed for both hydrogen fuel cells and DMFCs by
researchers from Pennsylvania State University [8, 36, 40].
However, few dynamic simulation results have been shown
in their papers. The best dynamic model so far is the one
developed by Natarajan and Nguyen [28]. In their work, the
simulation results for transient profiles of current density,
oxygen, and water concentration were presented. However,
their model only accounts for the cathode of the hydrogen
fuel cell.
In the following sub-sections, details are presented to
show how each component is modeled in both the hydrogen
and methanol fuel cell.
2.2.1 Anode Flow Channel (AF)
Hydrogen Fuel Cell
The primary purpose of the anode flow channel is to sup-
ply fuel to the fuel cell. The fuel is transported from the chan-
nel through a diffusion layer to the catalyst layer, where it is
consumed via the appropriate electrochemical reaction. To
avoid drying of the membrane at the anode side, hydrogen
gas is usually fed into the anode flow channel humidified
with water vapor. Water is transported from the anode to the
cathode by electro-osmotic drag, as described later in section
2.2.8. Liquid water will appear in the channel when the tem-
perature drops below the saturation point. Therefore,
depending on the temperature profile, a single phase or two-
phase flow may be present in the anode flow channel. In early
work, only gas flow was considered. Springer et al. [25] trea-
ted the anode flow channel as a continuous stirred tank reac-
tor (CSTR). Accordingly, the composition inside the flow
channel was treated as being uniform and equal to the value
at the outlet of the flow channel. The mass balance equation
can be applied for water as follows [25]:
M
I
w;AF
M
O
w;AF
= N
w;A
W L (AF.1)
Here, M
I
w;AF
and M
O
w;AF
are anode inlet and outlet water
flow rate, respectively, N
w;A
is the water flux through the
interface between anode flow channel and diffusion layer, W
and L are the width and length of the flow channel, respec-
tively.
The mole fraction of water in the anode flow channel can
be obtained by solving Eq. (AF.1):
x
w;AF
=
t
H
x
I
w;AF
a(1x
I
w;AF
)
x
I
w;AF
a(1x
I
w;AF
)t
H
1
(AF.2)
where, t
H
, the stoichiometric coefficient, is the ratio between
M
I
w;AF
and N
H2;A
(W L), and a is the ratio of N
w;A
to N
H2;A
.
The calculation of a is discussed later in the section on the
membrane.
The mole fraction of hydrogen is:
x
H
2
;AF
= 1 x
w;AF
(AF.3)
R
E
V
I
E
W
6 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Although the CSTR approach results in simple model
equations, the assumption that the reactant composition is
uniform inside the flow channel is not applicable especially
when the fuel cell is operated under conditions of high fuel
utilization. In later studies [34, 4144], the anode flow channel
was treated as a plug flow reactor:
dM
w;AF
(y)
dy
= N
w;A
(y) W (AF.4)
Note that N
w;A
(y) is a function of y if current density
changes along the channel.
The same approach can be applied to describe the energy
balance [34, 42, 43]:
d

i
(M
i;AF
Mw
i
C
p
i
)T
AF
dy
= k
h
A
h
(T
P
T
AF
)i = w; H
2
(AF.5)
where Cp
i
is the heat capacity of species i, Mw
i
is the molecu-
lar weight of species i, T
P
and T
AF
are the temperature of the
flow field plate and gas flow in the channel, respectively, k
h
is
the heat transfer coefficient, and A
h
is the heat transfer area
per unit length.
The boundary conditions for Eqs. (AF.4) and (AF.5) are:
y = 0 M
w;AF
= M
I
w;AF
; T
AF
= T
I
AF
y = L
dM
w;AF
dy
= 0;
dT
AF
dy
= 0 (AF.6)
The plug flow approach only accounts for the changes
along the flow channel (y direction). To predict the profile of
species concentration and temperature in the x and z direc-
tion (refer to Figure 1), two [29, 36, 38, 46] and three-dimen-
sional [8, 4749] models have been developed to simulate the
gas phase flow in the anode flow channel. The equations of
continuity, momentum, and mass balance for each species are
applied:
equation of continuity:
@r
@t
= (\ rv) (AF.7)
equation of motion:
@rv
@t
= (\ rvv) \P \ (l\v) (AF.8)
mass balance equation for species i:
@x
i
@t
= (\ x
i
v) \ (D
i
\x
i
) i = w; H
2
(AF.9)
For non-isothermal models [29, 46, 49], the equation of
energy balance is applied as:
@rC
p
T
AF
@t
= (\ r C
p
T
AF
v) \ (k
h
\T
AF
) S
h
(AF.10)
Here, r is the average density, v is the vector of velocity, P
is the pressure, l and k
h
are the viscosity and thermal conduc-
tivity of the gas flow, respectively, D
i
is the diffusivity of spe-
cies i, S
h
is the heat source term that accounts for the heat
transfer between the gas and the flow field plates.
The left hand sides (LHS) of Eqs. (AF.7)(AF.10) are only
required for a dynamic model [8, 44], and are zero in steady-
state models. Also, these equations can be simplified by
assuming constant density [29, 38, 46], viscosity [29, 38, 46],
diffusion coefficient [29, 38], heat capacity [29, 46], and ther-
mal conductivity [29, 46, 49]. In Hsing and Futerko's model
[38], the anode channel gas flow is treated as a fluid passing
through a porous medium. Therefore, Darcy's law could be
used to replace the equation of motion:
\P =
l
k
m
v (AF.11)
Here, k
m
is the permeability of the imaginary porous medi-
um.
Condensation and evaporation of water in the anode chan-
nel were studied by Nguyen and White [34]. In their model,
Nguyen and White assumed that water transport between
the flow channel and diffusion layer occurs only in the vapor
phase. The following equations were used to describe the
mass balance for liquid and vapor water:
dM
l
w;AF
(y)
dy
= (
k
vl
WH
RT
)(x
v
w;AF
P P
sat
w;AF
) (AF.12)
dM
v
w;AF
(y)
dy
=
dM
l
w;AF
(y)
dy
N
w;A
(y) W (AF.13)
where, k
vl
is the condensation rate constant of water vapor, H
is the height of flow channel, and P
sat
w;AF
is the saturation vapor
pressure.
When condensation and evaporation are considered, the
energy balance can be described as:
d

i
(M
i;AF
Mw
i
Cp
i
)T
AF
dy
=
DH
vl
dM
l
w;AF
(y)
dy
UA
h
(T
P
T
AF
) i = w; H
2
(AF.14)
where, DH
vl
is the enthalpy of evaporation. Boundary condi-
tions similar to Eq. (AF.6) can be applied here.
Direct Methanol Fuel Cell
In the DMFC, fuel can be fed to the anode flow channel
either as methanol and water vapor or as an aqueous metha-
nol solution. Using an approach that is similar to that used
for the hydrogen fuel cell, the anode channel in a DMFC has
also been treated as a CSTR. The following equation was
applied by Dohle et al. [58] for vapor-fed methanol:
M
I
i;AF
M
O
i;AF
= N
i;A
W L i =
w; CH
3
OH; andCO
2
(AF.15)
Sundmacher et al. [57] developed a dynamic model for the
anode channel:
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 7
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
dc
i;AF
dt
=
1
s
(c
I
i
c
i;AF
)
k
LS
A
a
V
AF
(c
i
c
i;AC
) i =
CH
3
OH; andCO
2
(AF.16)
Here, c
i;AF
, c
I
i
and c
i;AC
are the molar concentrations of spe-
cies i in the anode channel, at the inlet, and in the anode cata-
lyst layer, T is the residence time in the anode channel, k
LS
is
a mass transfer coefficient between the anode flow and the
anode diffusion layer, A
a
and V
AF
are the anode cross-sec-
tional area and volume of the anode channel, respectively.
Note that the diffusion layer is not described by Sundma-
cher's model.
In both models only one phase is considered: the vapor
phase in Dohle's model and the liquid phase in Sundmacher's
model. In reality, a two-phase flow can exist for both vapor-
fed and liquid-fed DMFCs due to the evaporation and con-
densation of water and methanol as well as the release of
CO
2
. Condensation in a vapor-fed DMFC is less likely to
occur than the formation of a gas phase in a liquid-fed DMFC.
Therefore, the modeling of two-phase flow in liquid-fed
DMFC is more important. In fact, in a study of liquid-fed
DMFC [54], the onset of two-phase flow was observed at
methanol conversions as low as 2.4 610
3
. At these low con-
versions, the vapor phase consisted of approximately 47%
water and 14% methanol (mole percentage), which needs to
be recovered outside the DMFC; the methanol fraction in the
vapor phase reduces with increasing conversion. At high con-
version, the vapor phase may contain CO
2
in addition to
water and methanol vapor. The presence of a large amount of
CO
2
will reduce the contact surface area between the liquid
methanol and the diffusion layer [59]. Two-phase flow in a
liquid-fed DMFC has been modeled by Wang and Wang [40].
A plug flow assumption was applied:
continuity equation for each phase:
dr
l
v
l
y
(1a
v
)
dy
= N
l
A
(y)=H (AF.17)
dr
g
v
g
y
a
v
dy
= N
g
A
(y)=H (AF.18)
mass balance for each species:
dr
l
v
l
y
(1a
v
)c
l
i
dy
=

r
l
v
l
x
c
l
i

x=FD
k
l
LS
r
l
(c
l
i
c
l
i

x=FD
)
H
i = w; CH
3
OH; CO
2
(AF.19)
dr
g
v
g
y
(1a
v
)c
g
i
dy
=

r
g
v
g
x
c
g
i

x=FD
k
g
LS
r
g
(c
g
i
c
g
i

x=FD
)
H
i = w; CH
3
OH; CO
2
(AF.20)
Here, a
v
is the vapor volume fraction, N
l
A
(y) and N
g
A
(y) are
the mass exchange fluxes of the liquid and gas between the
flow channels and the diffusion layer, c
l
i

x=FD
and c
g
i

x=FD
are
the concentration of species i at the interface between flow
channel and diffusion layer. Note that both fluid convection
and species diffusion are accounted for within the flow chan-
nel and diffusion layer.
Boundary conditions for Eqs. (AF.17)(AF.20) are:
y = 0 v
l
y
= v
l;I
y
; v
g
y
= v
g;I
y
; c
l
i
= c
l;I
i
; c
g
i
= c
g;I
i
y = L
dv
l
y
dy
= 0;
dv
g
y
dy
= 0;
dc
l
i
dy
= 0;
dc
g
i
dy
= 0 (AF.21)
Divisek et al. [109] have recently reported a performance
model of the DMFC but did not simulate processes occurring
in the anode channel.
Comments
The anode flow channel in the hydrogen fuel cell has been
studied extensively. A possible development would be the
application of two or three-dimensional equations of continuity,
motion, mass balance, andenergy balance to the two-phase flow
condition. However, from a practical operation standpoint,
the condensation of water in the anode channels is less of a
concern compared to the drying of the membrane on the
anode side and the flooding problem on the cathode side.
On the other hand, the presence of two-phase flow in a
DMFC cannot be ignored, especially in a liquid-fed DMFC.
To date, the anode flow in a DMFC has been treated as flows
with two separated phases, and both phases are assumed to
be continuous. However, flow visualization experiments [59,
60] have shown that the gas phase exists as bubbles and slugs
in most parts of the flow channel. One would expect that the
interfacial properties of the liquid and gas phase as well as
the size and distribution of bubbles and slugs may have sig-
nificant effects on the fluid dynamics and transport phenom-
ena within the anode channel flow. To study the process of
gas removal, detailed fluid dynamic models coupled with
thermodynamic equilibrium equations are needed to describe
the release of CO
2
, evaporation and condensation of water
and methanol, and formation and evolution of gas bubbles
and slugs.
2.2.2 Cathode Flow Channel (CF)
Hydrogen Fuel Cell
The cathode flow channel provides a path for oxidant sup-
ply and for water removal from the diffusion layer. Although
pure oxygen could enhance the fuel cell performance, it is
more practical to use air as the oxidant due to safety and eco-
nomic concerns. Water is generated on the cathode catalyst
via the electrochemical reaction between protons and oxygen.
Water is also transported from the anode side to the cathode
side through the membrane by electro-osmotic drag. Accu-
mulation of liquid water within the cathode channel can
result in partial or complete blockage of pores in the diffusion
layer and limit the access of oxygen to catalyst sites.
R
E
V
I
E
W
8 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Therefore, the air-flow rate needs to be high enough to
remove the liquid water from the cathode.
The same approach used for modeling the anode flow chan-
nel has also been applied to the cathode flow channel. The
cathode flowchannel has been described as a CSTR[2325] or a
plug flow reactor [34, 4144]. Two-dimensional [29, 36, 38, 46]
and three-dimensional [8, 4749] models have been devel-
oped based on the equations of continuity (Eq. (AF.7)),
motion (Eq. (AF.8)), mass (Eq. (AF.9)), and energy balances
(Eq. (AF.10)). Comparedwith two species (hydrogenandwater)
in the anode flow channel, there are three species (oxygen,
nitrogen, and water) in the cathode flow channel. Therefore,
the mass balance equation needs to be applied to two species.
In some work [23, 24, 61], the gas phase is assumed to be
always in equilibrium with the liquid water phase; hence the
mole fraction of water in the vapor is known if the tempera-
ture and pressure are known. This assumption is applicable
when the cathode flow is fed saturated with water vapor.
In a few studies, the evaporation and condensation of
water were also investigated. In Nguyen and White's work
[34], the same mass and energy balance equations (Eqs.
(AF.12)(AF.14)) were used for both anode and cathode flow
channels. Kimble and Vanderborgh [62] developed a model
for the two-phase flow in the cathode flow channel. In their
model, both gas and liquid phases are assumed to be continu-
ous and separate from each other (Figure 3).
The cross-sectional area occupied by each phase changes
along the channel due to the accumulation of water and may
be accounted for by the following two equations:
d(c
i
v
g
A
g
)
dy
= R
g
i
W i = w; O
2
; andN
2
(CF.1)
d(c
l
w
v
l
A
l
)
dy
= R
l
i
W (CF.2)
Here, R
g
i
and R
l
i
are the generation or consumption rate of
species in gas and liquid phase, c is the mole concentration.
Since water is the only species in the liquid phase, c
l
w
can be
calculated from density, once the temperature is known. A
g
and A
l
are the cross-sectional area for the gas and liquid
phases, respectively, W is the width of the flow channel. If the
total cross-sectional area of the flow channel is A
T
, then:
A
T
= A
g
A
l
(CF.3)
Kimble and Vanderborgh [62] assumed that water enters
the flow channel in the vapor phase and a single phase is
present until the saturation point is reached. Once the gas is
saturated, any water entering the channel from the diffusion
layer will contribute to the liquid phase.
The equation of motion is applied to each phase:
d(r
g
v
g
v
g
A
g
)
dy
= (r
l
v
l
v
l
r
g
v
g
v
g
)
dA
g
dy

(CF.4)
d(r
l
v
l
v
l
A
l
)
dy
= (r
l
v
l
v
l
r
g
v
g
v
g
)
dA
l
dy

(CF.5)
The RHS of the above equations is an inertial effects term,
which accounts for the momentum transfer between two
phases due to the mass transfer between the two phases and
the change of cross-sectional area occupied by each phase.
The interfacial drag effect between two phases and the fric-
tion at the channel surface are neglected here. The pressure is
assumed constant along the flow channel, but the tempera-
ture profile along the flow channel is calculated.
A more complicated model was developed by Wang et al.
[37]. In their work, the so-called two-phase mixture model is
applied. The two-phase system is viewed as a binary chemi-
cal mixture. Hence, the two-phase flow can be described by a
mass-averaged mixture velocity and diffusion flux represent-
ing the difference between mixture velocity and an individual
phase velocity [63]. The equations of continuity (Eq. (AF.7))
and motion (Eq. (AF.8)) are incorporated with the following
mass balance equations for each species:
@(rc
i
)
@t
= \(c
c
rvc
i
)
\(rD\c
i
) \ (sr
l
D
l
\c
l
i
(1 s)r
g
D
g
\c
g
i
(rD\c
i
)
_ _
\(c
l
i
c
g
i
)j
l
i = w; O
2
; andN
2
(CF.6)
Here, r is the density of the mixture and c
i
is the average
weight fraction of species i in the mixture. i can be water, oxy-
gen, or nitrogen. r and c
i
can be calculated from the density
of the gas and liquid phase (r
l
and r
g
), and the weight frac-
tion of species i in the two phases (c
l
i
and c
g
i
):
r = sr
l
(1 s)r
g
(CF.7)
rc
i
= sr
l
c
l
i
(1 s)r
g
c
g
i
i = w; O
2
; andN
2
(CF.8)
where, s indicates the liquid saturation, which is equal to the
volumetric fraction of the liquid phase. Wang et al. also as-
sumed that liquid water appears only when the water vapor
pressure reaches the saturation value. Hence, c
g
w
is known as
long as the temperature and pressure are available. It is also
assumed that oxygen and nitrogen are insoluble in the liquid
phase and water is the only species in the liquid phase. There-
y
x
2
1
Fig. 3 Two-phase representation of the gas channel in a PEM fuel cell flow
field (Kimble and Vanderborgh [62]): 1. gas phase, 2. liquid phase.
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 9
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
fore, s can be solved from Eqs. (CF.7) and (CF.8):
s =
r
g
(c
w
c
g;sat
w
)
r
l
(c
l;sat
w
c
w
)r
g
(c
w
c
g;sat
w
)
(CF.9)
c
c
in Eq. (CF.6) is the advection correction factor:
c
c
=
r(k
l
c
l
k
k
g
c
g
k
)
sr
l
c
l
k
(1s)r
g
c
g
k
(CF.10)
k
l
and k
g
are the relative mobilities:
k
l
=
k
l
r
t
l
=l
l
k
l
r
t
l
=l
l
k
g
r
t
g
=l
g
, k
g
= 1 k
l
(CF.11)
k
l
r
and k
g
r
are the relative permeabilities for the liquid and
gas phases. They are calculated using the empirical correla-
tion:
k
l
r
= s
3
k
g
r
= (1 s)
3
(CF.12)
j
l
in Eq. (CF.6) is the flux caused by capillary pressure (P
cp
)
and gravity:
j
l
=
k
l
k
g
k
l=r
\P
cp
(r
l
r
g
)g
_ _
(CF.13)
The individual phase velocities are calculated from:
r
l
v
l
= j
l
k
l
rv (CF.14)
r
g
v
g
= j
l
k
g
rv (CF.15)
To date, Wang's model [37] is the most comprehensive
two-phase model for hydrogen fuel cells.
DMFC
The cathode flow channel in the DMFC is quite similar to
that in hydrogen fuel cell. In the DMFC, an additional species,
CO
2
, must be accounted for. CO
2
is produced by the oxidation
of cross-over methanol on the cathode catalyst layer and also
possibly by liquid transport of dissolved CO
2
. The cathode
flow channel has been modeled as either a CSTR [58] or a
plug flow reactor [40]. In the CSTR approach, equations simi-
lar to Eq. (AF.15) were used. Using a plug-flow approach, an
equation similar to Eq. (AF.20) was applied by Wang and
Wang for the mass balance in the cathode flow channel:
dr
g
v
g
y
c
g
i
dy
=
r
g
v
g
x
c
g
i

x=FD
k
g
LS
r
g
c
g
i
c
g
i

x=FD
_ _
H
i = w; O
2
; andN
2
(CF.16)
Note that only the gas phase is considered here. In the
recently reported DMFC model of Divisek et al. the cathode
channel was not included [109].
Comments
Current two-phase flow models assume that both the liq-
uid and gas phases are continuous. This would be a reason-
able assumption if the water came out from the diffusion
layer and formed a film at the surface of the diffusion layer.
However, there is no visualization experiment that supports
this picture. When liquid water exists in the form of droplets
or slugs, the two-phase flow would consist of a continuous
gas phase and a dispersed liquid phase. The size and distri-
bution of the water droplets or slugs would have a significant
effect on the removal of liquid water. Future studies are
needed to visualize the evolution process of two-phase flow
in the cathode flow channel and to develop corresponding
mathematical models. Also, interfacial friction at the solid
surface and the complex structure of the flow channel should
be incorporated in future models.
2.2.3 Flow-Field Plate Model (FF)
The performance of fuel cells strongly depends on the flow
rate of fuel and oxidant. However, in a fuel cell stack, the
reactant flow may not be equally distributed to each individ-
ual cell; also, in each individual cell, the flow rate may change
from one channel to the other. A non-uniform flow distribu-
tion could lead to localized reactant starvation and, conse-
quently, lead to operational instability and/or lower the per-
formance of the fuel cell stack. The main factors that affect the
flow distribution are hydraulic resistance, temperature distri-
bution, flow-field plate design, and stack configuration. A
flow-field plate model is essentially a model of the flow chan-
nel and the solid flow-field plate. Whereas a flow channel
model calculates only the pressure and temperature profiles
of the fluid in the channel, the flow-field plate model also cal-
culates the temperature distribution within the plates and the
flow rate distribution among the channels. On the other hand,
a manifold stack model describes the pressure loss and tem-
perature variation of the entire fuel cell stack, and flow distri-
(a)
(c)
(b)
(d)
Fig. 4 Structure of different flow fields: a. grid structure, b. straight chan-
nel, c. serpentine structure, d. interdigitated structure.
R
E
V
I
E
W
10 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
bution between each individual fuel cell. In this section, flow-
field plate models for both the hydrogen fuel cell and DMFC
are reviewed. Manifold stack models are described in the
next section.
The pressure drop and flow distribution in a flow-field
plate strongly depend on its structure. Four different flow
plate designs are shown in Figure 4 [86].
It can be seen that the flow-field plate with a grid structure
provides the maximum direct contact surface between reac-
tant flow and the diffusion layer. However, the water gener-
ated in the cathode is not easily removed by gas flow in the
grid structure. On the other hand, the conventional channel
structure allows higher flow velocity, which aids the removal
of water. In the serpentine structure, reactant flows across the
whole surface of the diffusion layer while maintaining a high
flow velocity. Although only one flow channel is shown for
the flow plate with serpentine channels in Figure 4, multiple
channels are more common in practical applications. In the
interdigitated structure, the inlet and outlet channels are sep-
arated by the diffusion layer. The reactants are forced
through the diffusion layer instead of allowing a flow over
the diffusion layer surface. The transport of reactants is con-
verted from diffusion to a forced convection. Currently, the
serpentine configuration is most commonly used.
Hydrogen Fuel Cells
A simple model was developed by Thirumalai and White
[87] to describe the flow field in a hydrogen fuel cell with a
grid structure. In this model, reactant flow was assumed to
be laminar. Pressure changes caused by expansion and con-
traction were considered negligible compared to the surface
friction losses. The flow-field pattern used in their analysis is
shown in Figure 5.
From this figure, it can be seen that the mass conservation
equation can be applied to each node (i, j) as:
q1
i;j
q2
i;j
q1
i1;j
q2
i;j1
Q
i;j
= 0 (FF.1)
Here, Q
i,j
is the consumption rate of all reactants at the
node, q is the volume flow rate as described in Figure 5.
For laminar flow, the pressure drop (DP) can be linearly
related to the flow rate as:
q1
i;j
= k1
i;j
(P
i1;j
P
i;j
)
q2
i;j
= k2
i;j
(P
i;j1
P
i;j
)
(FF.2)
where, k1
i,j
and k2
i,j
are constants dependent on the geometry
of the flow channels.
Eqs. (FF.1) and (FF.2), together with the inlet flow rate and
pressure, form a set of algebraic equations, which can be
easily solved to predict the pressure and flow distribution in
the flow field.
For the flow field with interdigitated structure, the pres-
sure drop from the inlet channel to the outlet channel has
been calculated using Darcy's law [26, 88]:
DP = k
m
v (FF.3)
Here, k
m
is a constant related to the geometry of the flow
field and the permeability of the diffusion layer. Dutta et al.
[48] and Kulikovsky [45] have modeled the fuel cell with ser-
pentine flow channels. However, the effects of the bends in
the flow channel on the flow dynamics were not studied in
their work.
Thirumalai and White's model only considers the gas phase
flow. However, as discussed in the flow channel sections (sec-
tions 2.2.1 and 2.2.2), the condensation of water vapor and the
production of water on the cathode side may lead to two-
phase flow in the flow field. A model has been developed by
Kimble and Vanderborgh [62] to study two-phase flow in the
cathode flow channel. This model was described in section
2.2.2 on the cathode flow channel. However, Kimble and Van-
derborgh's model only considered the inertial effects, which
are caused by the condensation and evaporation of water,
and the change in cross-sectional area. The interfacial drag
effect between the two phases was not accounted for. The
friction loss at the channel surface was also neglected. There-
fore, Kimble and Vanderborgh's model can describe the flow
rate change of each phase along the flow channel, but it can-
not predict the pressure drop in the flow field.
The temperature profile can be simulated by considering the
heat generatedinthe fuel cell, the heat removedthrough natural
convection occurred at the end of the plates or by conduction to
the cooling plates, andheat effects relatedto the mass transfer of
species with different temperature [34, 42, 43, 62, 89, 90]. How-
ever, so far, only the temperature change along the flowchannel
;has been studied. The temperature profile has been assumed to
be identical for each channel. Most of these studies assumed
that the temperature of the flow-field plate is uniform.
DMFC
Pressure drop in the anode and cathode flow fields of a
DMFC has been investigated by Argyropoulos et al. [9193].











q2
i,j+1
q1
i+1,j
q2
i,j
q1
i,j
i,j
Fig. 5 Geometry of the flow field [83].
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 11
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
In their work, a complex equation was used to calculate the
pressure drop of the two-phase flow incorporating the fol-
lowing factors:
(i) friction loss on the surface of the flow channel,
(ii) pressure change due to flow splitting at the inlet and
flow combining at the outlet,
(iii) acceleration of gas and liquid phases due to the change
in density,
(iv) gravitational head,
(v) vaporization of water and methanol,
(vi) limited solubility of CO
2
in the liquid phase.
In Argyropoulos's work, the temperature of the reactants
flow was assumed to increase linearly from inlet to outlet.
Comments
So far, most of the flow-field models have focused on
either simple fluid dynamic analysis or thermal balance anal-
ysis. There is no reported study on a single model that
accounts for both flow distribution and temperature distribu-
tion in the flow field. Additionally, only the flow-field plates
with simple structure, like conventional straight flow and
grid flow channels, have been studied. A model needs to be
developed for the more commonly used serpentine flow field.
This future model would be able to simulate the fluid
dynamics in a single serpentine flow channel and predict the
flow distribution among channels. The complex geometry of
the flow-field plates should be included in the governing
equations. Special attention should be paid to the region near
the bends in the channels. The thermal balance analysis
should be carried out simultaneously to predict the tempera-
ture distribution of both fluid and flow plate. Two-phase flow
also needs to be considered. The evolution of an additional
phase and its effects on the fluid dynamics and the tempera-
ture profile should be part of the model. Also, the flow-field
plate model needs to be coupled with models for other fuel
cell components to account for the inter-component transport
phenomena.
2.2.4 Anode Diffusion Layer (AD)
Hydrogen Fuel Cell
Direct contact between the flow plates and the catalyst
layer can have negative effects including erosion of the cata-
lyst layer, corrosion of flow plates, uneven fuel distribution
on the catalyst layer, and Ohmic overpotential due to the
irregular surfaces of the flow plates and catalyst layer. An
anode diffusion layer made of porous carbon cloth or paper
is therefore applied to connect the flow plates and catalyst
layer. This layer prevents the flow plates from coming in
direct contact with the catalyst layer, gets rid of the effects of
irregular surfaces, and allows an even fuel distribution to the
catalyst layer. However, the diffusion layer does create resis-
tance of species transport. Hence, the focus of mathematical
modeling is on the calculation of the effect of the diffusion
layer on the species concentration distribution.
The transport phenomena in the anode diffusion layer are
usually modeled as multi-component diffusion in a porous
medium. The Stefan-Maxwell equations can be applied:
dx
i
dx
= RT

j
x
i
N
j
x
j
N
i
PD
eff
i;j
(AD.1)
For the hydrogen fuel cell, there are only two species in
the anode diffusion layer: water and hydrogen. Therefore, the
Stefan-Maxwell equations become [25]:
dx
w;AD
dx
=
RTI
PD
eff
w;H
2
x
w;AD
(1 a) a
_ _
(AD.2)
Here I = J=2F = N
H
2
;A
, J is the current density, and F is the
Faraday constant. The calculation of a will be discussed later
in the section on the membrane.
If pressure is assumed to be constant, Eq. (AD.2) can be
integrated across the diffusion layer to obtain:
x
w;ADC
=
(x
x;AFD

a
1a
) exp (1 a)
RTIh
AD
PD
eff
w;H
2
_
_
_
_

a
1a
(AD.3)
Here, h
AD
is the thickness of the anode diffusion layer.
x
w;ADC
is the water mole fraction at the interface between
the anode diffusion layer and the catalyst layer, while
x
w;AFD
is the mole fraction at the interface between the
anode flow channel and the diffusion layer. In the work of
Springer et al., water is assumed to be in vapor form in the
diffusion layer, and x
w;AFD
is equal to the water mole frac-
tion in the anode flow channel, x
w;AF
, which is obtained
from Eq. (AF.2). A similar one-dimensional approach has
been used in other studies [4144, 46, 64].
A Knudsen diffusion term is sometimes added to
Eq. (AD.1) to account for the interaction between the reac-
tants and the pore-wall [31, 51]:
dx
i
dx
= RT

j
x
i
N
j
x
j
N
i
PD
eff
i;j

RTN
i
PD
k
i
(AD.4)
where, D
k
i
, the Knudsen diffusion coefficient is proportional
to the product of the average pore radius (r
p
) and the mean
thermal velocity of the fluid molecules:
D
k
i
= kr
p

8RT
pMw
i

(AD.5)
In more comprehensive models, two-dimensional [29, 36,
38] and three-dimensional [8, 4749] equations of continuity
(AF.7), motion (AF.8), and species mass balance (AF.9) have
been applied with modification for the porous medium:
equation of continuity:
@er
@t
= (\ erv) (AD.6)
equation of motion:
@erv
@t
= (\ ervv) e\P \ (el\v) e
l
k
m
v (AD.7)
R
E
V
I
E
W
12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
equation of mass balance for species i:
@ex
i
@t
= (\ ex
i
v) \ (D
eff
i
\x
i
) i = w; H
2
(AD.8)
In the non-isothermal model [29, 49], the equation of
energy balance is expressed as:
@erCpT
AF
@t
= (\ erCpT
AF
v) \ (k
h
\T
AF
) (AD.9)
Note that in Eq. (AD.7), a term has been added to account
for the drag force exerted by the fluid on the solid pore sur-
face. If the permeability is very small, the pore velocity
becomes so small that the inertia and viscous term in (AD.7)
can be neglected, and the equation of motion (AD.7) will be
reduced to Darcy's law under steady state conditions [38, 49]
\P =
l
k
m
v (AD.10)
Eqs. (AD.6)(AD.9) can be solved together to obtain the
distribution of species concentration and temperature. Note
that the effective diffusion coefficient, D
eff
i
, is used here.
D
eff
i
could be related to the non-porous diffusion coefficient
through the Bruggeman correlation:
D
eff
i
= D
i
e
s
(AD.11)
Here, e is the porosity, and T is the tortuosity, which is
often assumed to be 1.5.
The solution domain of Eqs. (AD.6)(AD-9) is shown in
Figure 2(a). z = 0 and w are the center points of the flow chan-
nels, hence, the symmetry conditions are used here. At the
interface between the anode diffusion layer and the catalyst
layer (x = h), the flux of hydrogen is related to the current
density through Faraday's law, while the water flux is gov-
erned by the electro-osmotic drag force. At the boundary
where the reactant contacts the diffusion layer, the pressure
and species concentration are assumed to be the same value
as that in the flow channel. At x = 0 and xw
1
_ z _ w
2
, where
the diffusion layer contacts the rib of the flow plate, all fluxes
in the x direction are assumed to be zero. It can be seen that
the effect of the rib size can be incorporated into the model
through the boundary conditions [32].
Theoretically, the condensation of water vapor can lead to
a two-phase flow in the anode diffusion layer. However, the
two-phase flow in the anode diffusion layer is not considered
as important as the two-phase flow on the cathode side. The
detail of two-phase flow in the diffusion layer will be pro-
vided in the section on the cathode diffusion layer.
For a 1-D model, the Ohmic overpotential in the anode dif-
fusion layer can be calculated by Ohm's law:
g
r
=
_
J
r
eff
d
dx (AD.12)
Here r
eff
d
is the effective conductivity of the diffusion layer,
which is related to the conductivity of a non-porous medium
as [23]:
r
eff
d
= e
1:5
r
d
(AD.13)
For 2-D and 3-D models, Eq. (AD.12) can be modified
appropriately to reflect the vector nature of current density J.
DMFC
The Stefan-Maxwell equations used for the hydrogen fuel
cell have also been applied to simulate the anode diffusion
layer in the DMFC. Dohle et al. [58] developed a one-dimen-
sional model for a vapor fed DMFC. Kulikovsky et al. [33, 55]
extended their hydrogen fuel cell model to both the vapor fed
and liquid fed DMFC. In all these models, only one phase
was considered. However, in the DMFC, two-phase flow
exists over a wide range of operating conditions as a result of
the production of CO
2,
and evaporation and condensation of
water and methanol. A comprehensive two-phase model has
been developed by Wang and Wang [40] for the liquid fed
DMFC. In their model, the equation of continuity (AD.6) is
used, and the equation of motion (AD.7) becomes:
v =
k
m
l
(\P rg)
n
drag
Mw
r
J
F
(AD.14)
Here, fluid flow is affected by the net effect of pressure
gradient, gravity, and electro-osmotic drag. n
drag
is the fluid
drag coefficient. The equation of species conservation is:
@(erc
i
)
@t
= \(c
c
rvc
i
) \(r
l
D
l;eff
\c
l
i
r
g
D
g;eff
\c
g
i
)
\(c
l
i
c
g
i
)j
l
i = w; CH
3
OH; CO
2
(AD.15)
Note that Eq. (AD.15) is similar to Eq. (CF.6), which has
been used to simulate two-phase flow in the cathode flow
channels of hydrogen fuel cells [37]. The same equations
(from (CF.10)(CF.14)) can be used to calculate the advection
correction factor (c
c
), relative mobilities (k
l
and k
g
), relative
permeabilities (k
l
r
and k
g
r
), capillary-diffusion flux (j
l
), and in-
dividual phase velocities v
l
and v
g
. The effective coefficients
can be calculated as:
D
l;eff
= D
l
(es)
s
(AD.16)
D
g;eff
= D
g
[e(1 s)[
s
(AD.17)
The tortuosity (T) was assumed to be unity here.
s in Eq. (AD.17) indicates the degree of liquid saturation:
s =
r
g
(c
co
2
c
g
co
2
)
r
l
(c
l;sat
co
2
c
co
2
)r
g
(c
co
2
c
g
co
2
)
(AD.18)
In Wang and Wang's work, the gas phase at the anode was
assumed to be saturated with water and methanol, and the
liquid phase was assumed to be saturated with CO
2
. There-
fore, the gas phase concentration of all three species and the
CO
2
in the liquid phase can be calculated through thermody-
namic equilibrium relationships.
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 13
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Divisek et al.'s DMFC model considers two-phase (water
and gas) transport in the diffusion layer [109]. The continuity
and species conservation equations are essentially the same
as Eqs. (AD.14) and (AD.15). One key difference, however, is
the functional form of saturation, s, which is considered to
vary according to capillary effects. The relative permeability,
k
r
, is a function of saturation, s, which in turn is dependent on
capillary pressure, p
c
, defined by the following equation:
p
c
= p p
w
(AD.19)
where, p is the total pressure and p
w
is the water pressure.
In Divisek's work, mass transfer between the liquid and the
gas phase is modeled as the rate of condensation and evapo-
ration, which is proportional to the transfer surface area and
temperature.
Comments
A number of studies have been devoted to the anode diffu-
sion layer in both the hydrogen fuel cell and the DMFC.
Among them, the two-phase flow model developed by Wang
and Wang [50] for the DMFC, which could also be applied to
the anode diffusion layer in the hydrogen fuel cell, although
two-phase flow is not a major concern here.
2.2.5 Cathode Diffusion Layer (CD)
Hydrogen Fuel Cell
The cathode diffusion layer has functions similar to those
of the anode diffusion layer. It prevents direct contact
between the catalyst layer and flow plates, and provides even
oxygen distribution. Water is generated on the cathode side
through electrochemical reactions and is transported from
the anode side through the membrane by electro-osmotic
drag. Under certain conditions, liquid water can fill the pores
and reduce the diffusion rate of oxygen. Therefore, two-phase
flow plays a more important role in the cathode diffusion
layer than it does on the anode side.
The one-dimensional Stefan-Maxwell equations used to
describe species transport in the anode diffusion layer have
also been applied to the cathode side by Springer et al. [25,
41, 64]:
dx
o
2
;CD
dx
=
RTI
P
x
o
2
;CD
(1a)0:5x
w;CD
D
eff
w;O
2

1x
o
2
;CD
x
w;CD
D
eff
w;N
2
_
_
_
_
(CD.1)
dx
w;CD
dx
=
RTI
P
(1x
o
2
;CD
x
w;CD
)(1a)
D
eff
w;N
2

x
o
2
;CD
(1a)0:5x
w;CD
D
eff
O
2
;N
2
_
_
_
_
(CD.2)
The boundary conditions are:
x
o
2
;CDF
= x
o
2
;CF
; x
w;CDF
= x
w;CF
(CD.3)
A similar approach has been used by other researchers [23,
24, 4244, 46, 61]. Bernardi and Verbrugge [23] assumed that
the water vapor is saturated. Hence, only one equation is
needed to simulate the gas phase. This assumption is applic-
able when the cathode flow is fed with saturated water
vapor.
In most models, the diffusion layer has been considered to
be isotropic and homogeneous, and characterized by a single
porosity and tortuosity. A simple model was developed by
Gurau et al. to account for the uneven composition distribu-
tion in the diffusion layer [61]. In their model, the diffusion
layer was represented by a series of parallel sub-layers, which
were defined by different porosities and tortuosities. How-
ever, there is no evidence that this treatment is necessary.
Two-dimensional [2632, 36] and three-dimensional [8, 38,
45, 4749] models have been developed by using the equa-
tions of continuity (AD.6), motion (AD.7), mass balance
(AD.8), and energy balance (AD.9). Nguyen et al. [26, 27] have
studied the cathode diffusion layer coupled with the interdi-
gitated gas flow channels. The geometrical design layout of
the interdigitated and conventional flow channels are shown
in Figure 2 (a) and (b). It can be seen that the difference lies in
the boundary conditions. At the boundary, x = 0 and
w
2
_ z _ w, it is the inlet for the diffusion layer of the conven-
tional flow channel, but the outlet for the one with an interdi-
gitated flow channel. In the latter case, the pressure was spe-
cified as the outlet pressure, and changes in the species
concentration were assumed to be zero. Similar work has
been done by Kazim et al. [30].
The modeling of two-phase flow in the cathode diffusion
layer has received more attention than in the anode diffusion
layer. Wang's two-phase flow model for the cathode flow
channel is conveniently extended to the cathode diffusion
layer by adding porosity [37]. Eq. (CF.6) becomes:
@(erc
i
)
@t
= \(ec
c
rvc
i
) \(erD\c
i
)
\ e (sr
l
D
l
\c
l
i
(1 s)r
g
D
g
\c
g
i
_
(rD\c
i
)
_ __
\[(c
l
i
c
g
i
)j
l
[ i = w; O
2
; andN
2
(CD.4)
Eqs. (CF.7)(CF.13) can be applied in the same format,
while Eqs. (CF.14) and (CF.15) become:
er
l
v
l
= j
l
ek
l
rv (CD.5)
er
g
v
g
= j
l
ek
g
rv (CD.6)
A different approach was used by He et al. [27]. The aver-
age concentration and flow velocity of the two-phase mixture
were not used. Instead the equation of continuity was applied
to the two phases:
0 = \e
0
(1 s)c
g
v
g
R
w
(CD.7)
0 = \e
0
sc
l
v
l
R
w
(CD.8)
R
E
V
I
E
W
14 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Here, R
w
is the inter-phase mass-transfer rate of water by
condensation or evaporation,
R
w
= k
vl
e
0
(1s)
M
w
(x
w
P P
sat
w
)sw
k
lv
e
0
sr
w
M
w
(x
w
P P
sat
w
)(1 sw) (CD.9)
where, k
lv
and k
vl
are the evaporation and condensation rate
constants. sw is a switch, which equals 1 when x
w
P _ P
sat
w
,
and 0 when x
w
P < P
sat
w
.
Darcy's law and the equations of mass balance were
applied to the gas phase. For the liquid phase, the equation of
motion becomes:
v
l
= f
d
v
g
D
cp
\s (CD.10)
f
d
is a defined interfacial drag coefficient:
f
d
=
k
l
l
g
k
g
l
l
(CD.11)
D
cp
is defined as the capillary diffusion coefficient:
D
cp
=
k
l
kl
l
dP
cp
ds
(CD.12)
In He's work, f
d
and D
cp
were both assumed to be constant.
f
d
was taken as 0.005 while D
cp
had a value of 1:0 10
8
m
2
s
1
. This assumption is considered reasonable for a diffusion
layer with interdigitated flow channels where the water is
flushed out by forced convection, and liquid saturation, s, is
small (about 0.1 for He's case). In a later study, which consid-
ered a fuel cell with a conventional flow channel [28], k
l
and
dP
cp
ds
were given as functions of liquid saturation.
In relatively simple models, the liquid water flow caused
by capillary pressure was neglected. The porosity for the gas
phase was related to the liquid water content through various
correlations [51, 52]. One example is given here:
e
g
= e
g
0
1 (
x
w
x
max
w
)
m
_ _
x
max
w
=
r
l
w
r
s
1e
s
e
el
e
s
e
g
0
= 1 e
s
e
el
(CD.13)
e
s
and e
el
are the volume fractions of the solidphase andelec-
trolyte in the diffusion layer. These simple models can account
for the reduction of the diffusion rate of oxygen, but cannot
be used to calculate the velocity of an individual phase.
DMFC
In the cathode diffusion layer of the DMFC, in addition to
the expected species such as water, oxygen, and nitrogen,
CO
2
may also be produced due to the reaction of cross-over
methanol. It is usually assumed that all cross-over methanol
will be consumed completely in the catalyst layer. Therefore,
there is no methanol in the cathode diffusion layer. Dohle et
al. [58] applied the same equations as those used to describe
transport in the anode diffusion layer to the cathode diffusion
layer. Kulikovsky et al. [33, 55] expanded their model for the
hydrogen fuel cell to the DMFC. Wang and Wang's [40] model
for two-phase flowinthe anode diffusionlayer was also applied
to the cathode diffusionlayer. The degree of liquidsaturation on
the cathode side was calculatedwith Eq. (CF.9), which was used
for the modeling of two-phase flow in the cathode flow chan-
nel of the hydrogen fuel cell. In Wang and Wang's work, it
was assumed that the water vapor is saturated and oxygen
and CO
2
are insoluble on the cathode side. In Divisek et al.'s
work, the cathode diffusion layer is modeled with the same
set of equations as used for the anode diffusion layer.
Comments
It can be seen that the cathode diffusion layer has been
studied intensively. So far, all models have adopted the
macro-homogenous approach, which characterizes the diffu-
sion layer as a homogeneous porous medium with a single
porosity and tortuosity. One may expect that the oxygen dif-
fusion rate would be affected by the micro-structure of the
porous diffusion layer, such as the distribution of pore, pore
size and its distribution, and how the pores are occupied by
water at variable degrees of water flooding. In addition, these
pore parameters are a function of fuel cell design and assem-
bly. However, simulation at the microscope level would need
complicated models with a large number of unknown param-
eters. Also, it is very difficult to design experiments to investi-
gate the evolution process of water flooding in the diffusion
layer and validate the microscope models directly.
2.2.6 Anode Catalyst Layer (AC)
Hydrogen Fuel Cell
The anode catalyst layer is a porous medium consisting of
catalyst metal, carbon grains, which support the catalyst, a
polymer binder (usually Teflon), and an electrolyte, which
serves as a proton conductor. Fuel diffuses through the pores
to the catalyst surface where the electrochemical reaction
takes place. The electrons released are conducted through the
catalyst metal and carbon grains, while the protons are trans-
ported through the electrolyte to the membrane. The reaction
occurring on the anode side of the hydrogen fuel cell is:
H
2
2H

2e

(AC.1)
The transport process can be described by the same equa-
tions as used for the anode diffusion layer ((AD.6)(AD.9))
with an additional source term [8, 24, 29, 36, 45]:
equation of continuity:
@er
@t
= (\ erv)
(Mw
H
2
Mw
w
a)j
a
2F
(AC.2)
equation of motion:
@erv
@t
= (\ ervv)
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 15
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
e\P \ (el\v) e
l
k
m
v e
k
U
k
m
z
f
c
f
F\U
e
(AC.3)
equation of mass balance for species I:
@ex
H
2
@t
= (\ ex
H
2
v) \ (D
eff
H
2
\x
H
2
)
j
a
2Fc
tot
@ex
w
@t
= (\ ex
w
v) \ (D
eff
w
\x
w
)
aj
a
2Fc
tot
(AC.4)
Here, j
a
is the transfer current density at the anode. k
U
is
the electro-kinetic permeability of the membrane electrolyte.
c
tot
is the total mole concentration of the fluid. z
f
and c
f
are the
charge and concentration of the fixed charge sites in the elec-
trolyte, respectively. U
e
is the potential of the electrolyte
phase. The last term in Eq. (AC.3) accounts for the flow
caused by the electric potential gradient.
Another, simpler approach is to treat the catalyst layer as
an infinitely thin interface between the diffusion layer and
the membrane [25, 34, 4649, 65]. The source term is applied
as a boundary condition:
N
H
2
;AC
= J=2F (AC.5)
This simplification is applicable since the catalyst is much
thinner than the diffusion layer and membrane (typically, the
thickness of the catalyst layer is 1/30 of the diffusion layer).
The transfer current density can be defined by the Butler-
Volmer expression [24, 29, 45, 49, 65]:
j
a
= A
CV
J
ref
0;a
c
H
2
c
ref
H
2
_
_
_
_
1=2
exp(
a
a
F
RT
g
a
) exp(
a
c
F
RT
g
a
)
_ _
(AC.6)
Here, A
CV
is the effective catalyst area per unit volume, a
a
and a
c
are the anodic and cathodic transfer coefficient, g
a
is
the anode side overpotential for the electrochemical reaction.
The anode side exhibits fast electro-kinetics and a low over-
potential compared to the cathode, thus a linear kinetic rate
equation can be used [32]:
j
a
= A
CV
J
ref
0
c
H
2
c
ref
H
2
_
_
_
_
1=2
(a
a
a
c
)F
RT
g
a
_ _
(AC.7)
In many studies, the anode side overpotential has been
completely ignored [25, 34, 4648].
In Um et al.'s [36] and Bernardi and Verbrugge's [24] mod-
els, c
H
2
in Eq. (AC.6) is the concentration in the electrolyte
phase. It was related to the gas phase concentration at the
interface between the diffusion layer and the catalyst layer
through Henry's law. Also, in these two models, the equation
of mass balance (AC.4) was applied to the electrolyte phase
only. The other models [29, 45, 49] assumed that the electro-
chemical reaction occured at the interface between the gas-
phase reactants and the catalyst surface. Therefore, c
H
2
in Eq.
(AC.6) represents the concentration in the gas phase and the
concentration ratio in (AC.6) can be replaced by either a
hydrogen mole fraction or a partial pressure ratio.
In the models mentioned above, it was assumed that the
catalyst layer is a homogenous mixture consisting of four
components: the diffusion medium through which the trans-
port of reactants and products takes place; the electrolyte that
conducts protons; the electrically conductive medium that
conducts electrons; and the catalyst where the electrochemi-
cal reaction takes place. In these models, the reactant only dif-
fuses from the interface between the diffusion layer and the
catalyst layer towards catalyst sites through the electrolyte
phase [24, 36] or gas phase [29, 45, 49]. The effects of the por-
ous structure were taken into account through the effective
diffusion coefficient, which is a function of porosity. As a
result, the transport limitations were considered on the scale
of the entire catalyst layer. Agglomerate models (Figure 6)
were proposed to study diffusion limitations at a microscopic
level [6668].
The catalyst layer was assumed to contain small agglomer-
ates consisting of carbon, catalyst metal, and electrolyte,
which were separated by gas pores. The reactants from the
diffusion layer diffuse first through the gas pores, then
through the electrolyte to reach the catalyst sites. Another
microscopic model proposed by Bultel et al. [6971] assumed
that the catalyst particles are distributed following a hexagon-
ally packed three-dimensional network of spherical particles
within the electrolyte (Figure 7).
The simulation of Bultel et al. [71] showed that the conven-
tional homogeneous model remains appropriate when mass
and charge transport resistances are insignificant at the parti-
cle level. However, when the reaction kinetics are fast, the
microscopic model becomes more worthwhile because the
local transport limitations have significant effects on the elec-
trochemical reaction rate.
The Ohmic overpotential in the catalyst layer can be de-
rived from Ohm's law:














3
2 4 1
Fig. 6 Agglomerate model for the catalyst layer: 1. diffusion layer, 2. ag-
glomerate, 3. pore, 4. membrane.
R
E
V
I
E
W
16 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
\(r
eff
e
\g
r
) = j
a
(AC.8)
Here r
eff
e
is the effective conductivity of the catalyst layer.
DMFC
The electrochemical reactionat the anode side of a DMFCis:
CH
3
OH H
2
O CO
2
6H

6e

The multi-component diffusion process can also be


described by adding a source term to the mass balance equa-
tions used for the diffusion layer [33, 40, 55]:
@(erc
i
)
@t
= \(c
c
rvc
i
) \(rD
l;eff
\c
l
i
r
g
D
g;eff
\c
g
i
)
(AC.9)
Here, the S
i
is the stoichiometric coefficient of species i.
The transfer current density j
a
can be defined through:
j
a
= A
CV
J
ref
0;a
c
CH
3
OH
c
ref
CH
3
OH
_
_
_
_
a
c
w
c
ref
w
_ _
b
exp(
a
a
F
RT
g
a
) (AC.10)
The reaction-order parameters a and b have been assigned
different values by various researchers. Dohle et al. [58]
assumed that a and b were 1, while a = 1 and b = 0 in the work
of Kulikovsky et al. [33, 55]. Wang and Wang [40] assumed
that methanol oxidation is a zero-order reaction when the
concentration of methanol was greater than 0.1 M. Therefore,
both a and b were zero in their model.
A multi-step reaction mechanism was applied to describe
the oxidation of methanol at the platinum-ruthenium catalyst
by Sundmacher et al. [57]:
(R1) 3Pt CH
3
OH

Pt
3
COH 3H

3e
(R2) 3Ru H
2
O

3Ru OH 3H

3e
(R3)Pt
3
COH 2Ru OH

COOH H
2
O2Pt 2Ru
(R4) Pt COOH Ru OH

CO
2
H
2
OPt Ru
It is believed that the first step is the rate-determining step.
Sundmacher et al. assume that concentrations are uniform
within the anode catalyst layer, so that:
dc
CH
3
OH;AC
dt
=
k
LS
A
a
V
a;AC
(c
CH
3
OH;AF
c
CH
3
OH;AC
)
A
a
V
a;AC
N
c
CH
3
OH

A
a
V
a;AC
R
1
(AC.11)
dc
CO
2
;AC
dt
=
k
LS
A
a
V
a;AC
(c
CO
2
;AF
c
CO
2
;AC
)
A
a
V
a;AC
R
1
(AC.12)
The notation is similar to that used in Eq. (AF.16), which is
applied to the anode flow channel.
A three-phase model was developed by Baxter et al. for the
anode catalyst layer in the DMFC [72]. As shown in Figure 8,
the catalyst particle (matrix layer) is covered with a polymer
electrolyte (bond layer) in a porous anode catalyst layer.
The void pores are filled with a methanol and water solu-
tion (liquid layer). The reactants (methanol and water) diffuse
through the bond layer in the y
'
direction, to the surface of
the matrix layer, where reaction occurs. The protons pro-
duced are transported through the bond layer, in the x-direc-
tion, to the membrane.
In the recent paper by Divisek et al., the anode reaction is
modeled via a ten step reaction mechanism instead of the 4-
step reaction mechanism used by Wang and Wang. Another













2 3 1
Fig. 7 Bultle et al.'s discrete distribution model for the catalyst layer: 1. dif-
fusion layer, 2. catalyst layer, 3. membrane.











y

4
2
3 5 1
Fig. 8 Three-phase model for the anode catalyst layer in the DMFC: 1. dif-
fusion layer, 2. catalyst matrix layer, 3. electrolyte bond layer, 4. liquid
pore layer, 5. membrane.
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 17
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
feature of their model is the usage of occupancy number to
account for the lack of catalytic sites available for a certain
type of reaction. It is not clear how the numerous reaction
and catalyst parameters were obtained.
Comments
When modeling the anode catalyst layer, questions remain
as to the requirements of the microscopic level model. An
ideal microscopic model needs to consider the detailed mor-
phology of the porous catalyst layer, which consists of a mix-
ture of four different solid species (catalyst metal, carbon
grains, polymer binder, and electrolyte), and gas and/or liq-
uid-phase reactants. The micro-structure of the catalyst layer
is expected to have a considerable effect on the mass and
charge transfer rate. In hydrogen fuel cells, the anode exhibits
fast kinetics, hence the mass and charge transfer at the parti-
cle level cannot be ignored [71], on the other hand, the over-
potential on the anode side is much less than that on the cath-
ode side, and it can therefore be argued that the complexity
of a microscopic model is not justified. For DMFCs, the over-
potential on the anode side cannot be ignored. Detailed mod-
els such as the microscopic level model of Baxter et al. [72] or
the 10-step reaction model of Divisek et al. [109] seems worth-
while. However, such models usually need more model pa-
rameters and a longer computation time compared to the
simpler macro-homogeneous models. Also, it is difficult to
validate a detailed microscopic model directly through
experiment.
2.2.7 Cathode Catalyst Layer (CC)
Hydrogen Fuel Cell
The cathode catalyst layer has a porous structure similar to
that of the anode catalyst layer. The protons transported from
the anode and through the membrane react with the oxygen
supplied through the cathode diffusion layer:
4H

O
2
4e

2H
2
O
The methods used for simulation of transport processes in
the anode catalyst layer can be applied to the cathode catalyst
layer. The equations of continuity, motion, and the species
mass balances are identical, except that the source terms can
be used by altering the source terms final terms on the RHS
of Eqs. (AC.2) and (AC.4) [8, 23, 24, 29, 31, 36, 45, 51, 61]. The
equation of continuity becomes:
@er
@t
= (\ erv)
0:5Mw
O
2
Mw
w
(1a)
_ _
j
c
2F
(CC.1)
Species balance equations are:
@ex
O
2
@t
= (\ ex
H
2
v) \ (D
eff
H
2
\x
H
2
)
j
c
4Fc
tot
@ex
w
@t
= (\ ex
w
v) \ (D
eff
w
\x
w
)
(1a)j
c
2Fc
tot
(CC.2)
Although it may appear that the mass transport of species
in the diffusion and catalyst layers may occur only by diffu-
sion, crossover from one channel to the other underneath the
land areas may occur in the serpentine flow-field plate
design. Accordingly, a convective term is included in the spe-
cies balance equation (CC.2).
The cathode catalyst layer can also be treated as an infi-
nitely thin layer, where the source terms serve as boundary
conditions for the reactant flux [2528, 30, 32, 34, 37, 46, 47,
49].
The transfer current density, j
c
, can be defined by the But-
ler-Volmer expression [23, 24, 29, 45, 49, 51]:
j
c
= A
CV
J
ref
0;c
c
O
2
c
ref
O
2
exp(
a
a
F
RT
g
c
) exp(
a
c
F
RT
g
c
)
_ _
(CC.3)
where, g
c
is the cathode side overpotential. The cathode reac-
tion has relatively slow kinetics, and the first term on the
RHS of Eq. (CC.3) can be neglected. Hence Eq. (CC.3)
becomes the Tafel equation [2528, 3032, 34, 36, 4648, 61]:
j
c
= A
CV
J
ref
0;c
c
O
2
c
ref
O
2
exp(
a
c
F
RT
g
c
) (CC.4)
Water generated by the electrochemical reaction at the
anode, migrates from the anode through the membrane by
electro-osmotic drag, resulting in two-phase flow on the cath-
ode side. The presence of a large amount of water may fill the
pores in the diffusion layer and catalyst layer, and reduce the
rate of oxygen diffusion to the catalyst particle surface. This
phenomenon is called flooding. The two-phase model devel-
oped by Wang et al. [37] for the cathode flow channel and dif-
fusion layer has been extended to the catalyst layer by You
and Liu [73]. Details of this model were described in section
2.2.5 on the cathode diffusion layer. Li and coworkers [53, 74]
have also modeled the flooding problem. In their early work
[53], the cathode catalyst layer was assumed to be fully
flooded, with void space completely filled with a mixture of
liquid water and electrolyte. To reach the cathode, oxygen
must first diffuse through a liquid water layer, and then
through a thin catalyst membrane layer. The overall oxygen
diffusion process is approximated with a series resistance
model. Therefore, the apparent oxygen diffusion coefficient,
D
O
2
, can be obtained from:
L
w
L
m
D
O
2
=
L
w
D
w;O
2

L
m
D
O
2
;m
(CC.5)
Here, L
w
and L
m
are the thickness of the liquid water layer
and the membrane, respectively. In later work [74], Baschuk
and Li studied the effect of variable degrees of water flood-
ing. The pores in the catalyst layer are filled with the gas mix-
ture, liquid water, and membrane. An apparent oxygen diffu-
sion coefficient, D
O
2
, was calculated from:
1
D
O
2
=
RT
H
O
2
(1l
w
l
m
)
D
g;O
2

l
w
D
w;O
2

l
m
D
m;O
2
(CC.6)
where, l
w
and l
m
are the volume fraction of liquid water and
membrane. H
O
2
is the Henry's law constant for oxygen in
R
E
V
I
E
W
18 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
water. l
w
, which changes with the degree of flooding, was
used as an adjustable parameter to fit the experimental data.
The microscopic models described in section 2.2.6, describ-
ing the anode catalyst layer [6671], can also be applied to the
cathode side. However, the study of Bultel et al. [71] showed
that this approach is not required because the macro-homoge-
neous model remains applicable due to slow kinetics on the
cathode side.
DMFC
In addition to the reduction of oxygen, the oxidization of
methanol also occurs at the cathode side of the DMFC due to
methanol cross-over:
4H

O
2
4e

2H
2
O (R.1)
CH
3
OH H
2
O CO
2
6H

6e

(R.2)
The source term in Eq. (AC.9) becomes:
@(erc
i
)
@t
= \(c
c
rvc
i
) \(rD
l;eff
\c
l
i
r
g
D
g;eff
\c
g
i
)
(CC.7)
S
R
1
i
and S
R
2
i
are the stoichiometric coefficients, and j
p
is
the transfer current density for methanol cross-over.
Note, that both the protons transported from the anode
side and the protons generated at the cathode side (R.2) parti-
cipate in the reduction reaction (R.1), accordingly [40]:
j
c
= j
a
j
p
(CC.8)
Methanol crossover was neglected in the work of Dohle et
al. [58] and Kulikovsky et al. [33, 55], and the cathode transfer
current density, j
c
, was calculated through the Tafel equation
(CC.4). Wang and Wang [40] assumed that the transfer cur-
rent density for methanol cross-over, j
p
, is dictated by the
crossover flux, N
p
CH
3
OH
, because of the complete oxidation of
methanol at the cathode:
j
p
= A
CV
N
p
CH
3
OH
6F
(CC.9)
N
P
CH
3
OH
is calculated from:
N
p
CH
3
OH
= r
l
v
l
c
l
CH
3
OH
r
l
eD
l
CH
3
OH
@c
l
CH
3
OH
@y
_
_
_
_

x=X
MCC
(CC.10)
X
MCC
indicates the position of the interface between the
membrane and the cathode catalyst layer.
The Tafel equation was also used in Wang and Wang's
work:
j
c
= j
a
j
p
= A
CV
J
ref
0;c
c
O
2
c
ref
O
2
exp(
a
c
F
RT
g
c
) (CC.11)
The cross-over overpotential, g
p
, is not considered expli-
citly in Wang and Wang's work [40]. An empirical correlation
was used by Scott et al. [7577] to calculate g
p
:
g
p
= K
cross
N
p
CH
3
OH
(CC.12)
In the model of Divisek et al., the oxidation reaction is rep-
resented by a 4-step process, each involving an adsorbed spe-
cies, whose concentration depends on the occupation number
[109]. The crossover effect is modeled by incorporating the
10-step methanol oxidation reaction mechanism, however,
only a fixed fraction of catalyst sites are made available for
this reaction.
Comments
The cathode catalyst layer has been well studied. One may
argue that the effects of flooding phenomena need to be
described using microscopic models that consider the micro-
structure of the catalyst layer. However, using a detailed
microscopic approach may increase the number of model pa-
rameters and computational time considerably. Also, it is
very difficult to obtain direct experimental evidence to vali-
date these microscopic models.
2.2.8 Membrane (M)
Hydrogen Fuel Cell
The proton exchange membrane is used to separate the
hydrogen on the anode side from the oxygen on the cathode
side. It also conducts protons produced at the anode catalyst
layer to the cathode catalyst layer, and prevents electrons
from crossing through. The most commonly used membrane
material is Nafion

, which is a copolymer of tetrafluoroethy-


lene and a perfluorinated vinyl with sulfonate endgroups. In
the presence of water, the formation of interconnected ionic
clusters provides pathways for the transport of proton and
associated water molecules. Membrane properties like the
water diffusion coefficient and proton conductivity strongly
depend on the water content of the membrane [7885].
Water transport through the membrane has been thor-
oughly studied by Springer et al. [25]. In their work, the water
flux though the membrane was expressed as:
N
w;M
= n
drag
J
F
D
w;M
r
dry
Mw
m
dk
dx
(M.1)
Here, r
dry
is the density of the dry membrane. k is the
water content, defined as the ratio of the number of water
molecules to the number of charge sites (SO

3
H

). When the
membrane is immersed in boiling water, k can be as high as
22 [25]. The correlation used between k with a
w
, the water
vapor activity, is as follows:
k = 0:043 17:81a
w
39:85a
2
w
36:0a
3
w
0 < a
w
_
k = 14 1:4(a
w
1) 1 < a
w
_ 3 (M.2)
where, a
w
can be calculated from x
w
P=P
sat
w
.
The first term on the RHS of Eq. (M.1) accounts for the
water flux caused by the electro-osmotic drag. In the mem-
brane, protons travel in a complex form, i.e., H

(H
2
O)
n
. The
electro-osmotic coefficient, n
drag
, defined as the average num-
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 19
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
ber of water molecules dragged per proton, depends on the
water content of the membrane. Springer et al. [25] found that
in a fully hydrated membrane, the average number of water
molecules carried by a proton (n
drag
) is approximately 2.5. A
linear correlation was used in Springer's model to calculate
n
drag
at different water contents:
n
drag
= 2:5k=22 (M.3)
A quadratic expression was used by Dutta et al. [47, 48]:
n
drag
= 0:0029k
2
0:05k 3:4 10
19
(M.4)
The second term on the RHS of Eq. (M.1) accounts for the
diffusion of water, which is not associated with migrating
proton. D
w;M
, the diffusion coefficient of water in the mem-
brane, is also a function of the water content. Correlations for
D
w;M
are described later, in section 4.
Costamagna added a third term to Eq. (M.1) to account for
the water flux caused by the pressure gradient [46]. This term
is derived from Darcy's law:
k
m
l
kr
m
Mw
m
\P (M.5)
kr
m
Mwm
is the molar water concentration in the membrane. In
Costamagna's model, the viscosity, l, was related to the dif-
fusion coefficient, D
w;M
, through the Stokes-Einstein equation.
In the section on the anode flow channel, a parameter, a, is
defined as the ratio between the water flux and the hydrogen
flux. From Eq. (M.1):
a = N
w;M
=N
H
2
= 2n
drag
2D
w;M
F
J
r
dry
Mw
m
dk
dx
(M.6)
Note that N
H
2
=
J
2F
. In Springer's work [25], a is obtained
through an iterative procedure. A relatively simple approach
is to replace
r
dry
Mw
m
dk
dx
with a difference approximation over the
width of the membrane [34, 47, 48]. In this case, Eq. (M.6) be-
comes:
a = 2n
drag
2D
w;M
F
J
C
w;CCM
C
w;ACM
h
M
(M.7)
Proton transport through the membrane was studied by
Bernardi and Verbrugge [23, 24]. A form of the Nernst-Planck
equation that includes convection was used to describe the
flux of both water and protons:
N
i
= z
i
F
RT
D
i
c
i
\UD
i
\c
i
c
i
v i = w; H

(M.8)
z
i
is the charge on the species. The three terms on the RHS of
the above equation account for migration due to the electro-
potential gradient, diffusion owing to the concentration gra-
dient, and convection caused by fluid flow. The velocity of
the bulk flow through the membrane is governed by Schgl's
equation:
v =
k
U
l
z
f
c
f
F\U
k
m
l
\P (M.9)
k
U
is the electrokinetic permeability; z
f
and c
f
are the
charge and concentration of the fixed charge site in the mem-
brane, respectively.
For the membrane, the following electro-neutrality expres-
sion applies:
c
H

= z
f
c
f
(M.10)
This implies that the proton concentration in the mem-
brane can be considered spatially uniform since the mem-
brane is assumed to be homogeneous. Therefore, diffusion
due to a concentration gradient is not a mode of proton trans-
port, and Eq. (M.8) becomes:
N
H

=
F
RT
D
H

c
H

\Uc
H

v (M.11)
In Bernardi and Verbrugge's model, the membrane was
considered fully hydrated and the gas phase was assumed to
be in equilibrium with the liquid water phase in the diffusion
layer and flow channel. Therefore, water transport was not a
concern. However, Eqs. (M.8) and (M.9) can also be applied
to water:
N
w;M
= D
w
\c
w
c
w
k
U
l
z
f
c
f
F\Uc
w
k
m
l
\P (M.12)
Comparing Eqs. (M.12) and (M.1), it can be seen that the
electro-osmotic drag is implicitly included in Bernardi and
Verbrugge's model through the migration driven by the elec-
tro-potential gradient.
Eq. (M.8) becomes a mass balance equation at steady state,
hence, the following expression applies:
\N
i
= 0 (M.13)
Also, Eq. (M.9) is a special form of the equation of motion.
More comprehensive forms for the equations of motion and
mass balance were developed by Um et al. [8, 36] and Gurau
et al. [29]:
equation of motion:
@erv
@t
= (\ ervv) e\P \ (el\v)
e
l
k
m
v e
k
U
k
m
z
f
c
f
F\U
e
(M.14)
equation of mass balance:
@ex
i
@t
= (\ ex
i
v) \ (D
eff
i
\x
i
) i = w; H

(M.15)
Note, that the equation of motion (M.14) is similar to Eq.
(AC.3), used for the anode catalyst layer, and the equation of
mass balance (M.15) is similar to that for the anode diffusion
layer (AD.8).
R
E
V
I
E
W
20 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
The Ohmic overpotential in the membrane can be calcu-
lated by Ohm's law:
g
r
=
_
J
r
m
dx (M.16)
The calculation of membrane conductivity, r
m
, at different
water contents is described in section 4. It may be noted that
for 2-D and 3-D models, Eq. (M.16) must be modified appro-
priately to reflect the vector nature of current density, J. A
surface integral or a volume integral may be required for the
2-D and 3-D model, respectively.
DMFC
In a DMFC, in addition to water transport through the
membrane, transport of methanol across the membrane also
occurs. Springer's method for water transport through the
membrane in a hydrogen fuel cell has been applied to the
DMFC by Dohle et al. [58]. The water flux was obtained
through Springer's expressions, and the methanol flux was
calculated in a similar manner:
N
CH
3
OH;M
= n
M
drag
J
F
D
CH
3
OH;M
dk
CH
3
OH
dx
(M.17)
n
M
drag
is the electro-osmotic drag coefficient for methanol [54]:
n
M
drag
= 1:5
P
CH
3
OH
P
sat
CH
3
OH
(M.18)
Note that pressure rather than concentration was used
here since Dohle's model was devoted to a gas fed methanol
fuel cell. k
CH
3
OH
is the methanol content, defined as the ratio
of the number of methanol molecules to the number of charge
sites:
k
CH
3
OH
= 14
P
CH
3
OH
P
sat
CH
3
OH

1:5
2:5
(M.19)
It should be noted that, in this equation, 14 is the water
content (k) in a fully hydrated membrane, 2.5 and 1.5 are the
electro-osmotic coefficients for water and methanol in the
fully saturated membrane, respectively.
Bernardi and Verbrugge's method [23] for mass transport
in a membrane was used in the Kulikovsky [55] and Sund-
macher [57] models. Schgl's equation was used to relate
the fluid velocity to the electro-potential and pressure, while
the Nernst-Planck equation was applied for the flux of
species.
Wang and Wang's two-phase model [40] for diffusion in
the catalyst layers (Eq. (AD.14) and (AD.15)) also applies to
the membrane. The membrane was assumed to be fully
hydrated with liquid (s = 1). The electro-osmotic drag coeffi-
cient for methanol, n
M
drag
, was assumed to be proportional to
the methanol mole fraction, x
CH
3
OH
:
n
M
drag
= 2:5x
CH
3
OH
(M.20)
Comments
Transport phenomena in the membrane for the hydrogen
fuel cell have been studied both through experiment and
modeling simulation. Much work has also been devoted to
the study of membrane properties. A review of this work is
beyond the scope of this paper. For the DMFC, the mecha-
nism of methanol transport through membrane is still
unclear. Correlations are used based on the assumption that
methanol is dragged by the proton in a similar way to how
water is dragged by protons. More work is needed to validate
this assumption.
3 Stack Manifold Model
The configuration of the stack manifold is dictated by the
piping connections and flow arrangement inside the stack.
Figure 9 shows two commonly used manifold types.
The parallel flow arrangement is also called a U-manifold,
while the reverse flow arrangement is called a Z-manifold
[87]. The parallel and reverse manifolds are two different
combinations for dividing (inlet flow) and combining (outlet)
flow via a series of interconnected lateral branches. The flow
distribution among branches depends on the pressure drop
in each branch.
A large number of studies have been devoted to modeling
the flow distribution in manifolds [94100]; however, they
date back to pre-1980's. A comprehensive review of these
studies has been presented by Bajura and Jones [94]. How-
ever, most of these models were developed for heat exchangers









(a)










(b)
v2(n) P2(n n ) P2d( )
v1(n) P1(n) P1d(n)
v1(1)
ffo(n)
ffi(n)
Fig. 9 Schematic of gas manifolds: (a) parallel (U) manifold, (b) reverse
(Z) manifold.
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 21
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
or steam generators. They were based on mass conservation,
which is not applicable to fuel cell stacks, in which species
are consumed and produced through electrochemical reac-
tions. A limited number of models have been developed for
flow distribution in fuel cell stack manifolds. [87, 101106].
Hydrogen Fuel Cells
Thirumalai and White [87] developed a stack manifold
model for both configurations, as shown in Figure 9. This
model consists of the following equations:
the momentum balance in the manifold region,
P1d(n) P1(n 1) =
a
i
b
i
k
i
v1(n)
P2d(n) P2(n 1) = sw
a
o
b
o
k
o
v2(n)
(3.1)
mechanical energy balance at the branch point,
P1d(n) P1(n) = k
E
r v1(n) [ [
2
v1(n 1) [ [
2
_ _
P2d(n) P2(n) = k
D
r v2(n) [ [
2
v2(n 1) [ [
2
_ _ (3.2)
mass balance at the branch point,
a
i
b
i
v1(n) v1(n 1) [ [ = ffi(n)
a
o
b
o
v2(n) v2(n 1) [ [ = sw ffo(n)
(3.3)
where, v is the flow velocity, a
i
, b
i
, a
o
, and b
o
are the dimen-
sions for the inlet and outlet flow channels, k
i
and k
o
are con-
stants similar to the k1
i,j
and k2
i,j
in Eq. (FF.2), and k
E
and k
D
in Eq. (3.2) are used to account for the losses associated with
dividing and combining flow. Most symbols are described in
Figure 9. The parameter sw in Eqs. (3.1) and (3.3) takes the va-
lue +1 for the U-manifold and 1 for the Z-manifold. ffi(n)
and ffo(n) are the inlet and outlet flow rate of the nth fuel cell.
Eqs. (3.1)(3.3) form a set of nonlinear algebraic equations,
which can be solved with the aid of a single cell model. The
single cell model may be used to calculate ffo(n) from ffi(n),
and to relate p1(n) and p2(n) through the pressure drop across
an individual flow field.
A similar model was developed by Costamagna et al.
[101]. In their work, instead of using k
E
and k
D
(Eq. (3.2)), the
following empirical equations were used to account for the
losses of flow dividing and combining:
for the dividing manifold,
P1d(n) P1(n) =
1
2
r v1(n) [ [
2
v1(n 1) [ [
2
_ _

0:4 1
v1(n1)
v1(n)
_ _
2
1
2
r v1(n) [ [
2
(3.4)
for the combining manifold,
P2d(n) P2(n) =
1
2
r v2(n) [ [
2
v2(n 1) [ [
2
_ _
(3.5)
The variation in fluid and/or plate temperature along the
stack would also affect the reactant flow distributions. In an
early study, Amphlett et al. [102] predicted the dynamic
change of temperature by accounting for the electrical energy
output, heat loss from the stack surface, and a reasonable heat
for all flow species. However, only the average temperature
of the entire stack was considered in this work. Argyropoulos
et al. [103, 104] developed a one-dimensional model to predict
the temperature profile along a DMFC stack. The energy bal-
ance equation was applied to each component of each single
fuel cell. The heat losses due to radiation and convection were
only considered for the bipolar plates and the end plates. The
simulation results showed that the number of fuel cells, cur-
rent density, and the anode side inlet temperature are critical
parameters that determine the stack temperature profile.
However, the assumption that the temperature is uniform in
the y-z plane (the plane normal to the direction along the
stack) is not always true. For example, the temperature is
usually higher near the inlet of the flow-field plate, owing to
high reaction rate and current density. A three-dimensional
model was developed by Maggio et al. [105] for the PEM fuel
cell stack. In this work, a cooling plate with cooling tubes was
placed on each side of every individual cell, and the tempera-
ture profile of the cooling plates was predicted. However, this
model considered each fuel cell as a working plate with uni-
form thermal conductivity. It has been found that a uniform
temperature distribution in the stack could be achieved by
choosing appropriate operating conditions. Similar results
were also presented in Lee's work [37, 106].
DMFC
The manifold flow distribution in a liquid feed DMFC was
studied by Argyropoulos et al. [107, 108]. Their work used an
analogy to plate and frame heat exchangers, with modifica-
tions to account for the electrochemical phenomena that occur
in fuel cells. A manifold was described as inlet and outlet
flow channels having a number of discrete branches in the
side through which the flow enters or leaves the header
(Refer to Figure 9). To predict the flow distribution in the
stack manifold, equal flows into each cell was used as an
initial conjecture. The pressure drop in the inlet flow channel
was calculated as:
DP =
1
2
rv
2
l fk ( )
d
h
_ _
(3.6)
Here, d
h
is the hydraulic diameter; v is the flow velocity; l is
the length of the flow path, which varies for different
branches, f is the surface friction coefficient, and k is the
hydraulic resistance due to flow splitting.
The pressure drop in the flow field of each individual fuel
cell was calculated using the flow field model [91, 92], which
also predicts the flow rate and composition of the flow leav-
ing the flow plates. Then the pressure loss in the outlet flow
channel was obtained using Eq. (3.6). In this equation, k is the
hydraulic resistance due to flow combination.
R
E
V
I
E
W
22 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Theoretically, the pressure drop in each branch should be
equal to that in the adjacent branches. However, the initial
assumption of equal flow into each cell may not result in
equal pressure drop across each branch. Accordingly, an
iterative procedure was used to obtain the correct flow distri-
bution, i.e., when all the branch pressure drops differed by
only a few Pa. It was found that an increase in the number of
fuel cells and the reactant flow rate can lead to serious flow
distribution problems.
Comments
Similar to the situation for flow-field models, a compre-
hensive model is needed to predict the temperature profile
and flow distribution within a fuel cell stack, simultaneously.
This model could be based on a fundamental analysis of the
fluid flow in the stack by using the equations of continuity,
motion, and energy balance, and solved together with a
detailed flow field model and a single cell model for each cell.
However, for a stack with a large number of fuel cells, this
model would be very intensive computationally. A more
practical way to simulate the fuel cell stack would be to sim-
ply combine the current flow distribution model with the
temperature distribution model.
4 Model parameters
To model the transport and electrochemical processes in
fuel cells, one needs several input parameters. These include
the inlet flow conditions such as the flow rate of each species,
temperature, pressure; dimensions of the flow channel (e.g.
width, length, and height); the thickness of each diffusion
and catalyst layer and the membrane; porosities and tortuos-
ities of each diffusion and catalyst layer and the membrane;
geometry of the manifold and the flow field plates, and the
number of fuel cells in the stack.
Other model parameters include the thermodynamic
open-circuit potential, physical properties of the chemical
species present in the reactant and product streams, and the
electrochemical reaction kinetics. The typical correlations for
the calculation of these parameters are summarized in Tabs.
111. Among these parameters, the exchange current density
and the conductivity are usually used as adjustable parame-
ters to fit the experimental data [23, 24, 40, 51].
5 Numerical methods and model validation
One-dimensional models can be easily solved through
numerical integration [22, 23, 25]. In some special cases, even
analytical solutions are available [61, 65]. However, solving
two and three-dimensional models is more complicated and
computer intensive. The model equations can be solved
through a direct finite difference method [2628, 30, 31,
33, 46, 54]. Commercially available, computational fluid
dynamics (CFD), software packages have also been used [8,
36, 40, 4749].
To validate the mathematical models, model results have
been compared with limited experimental results. So far only
polarization curves have been used for model validation [23,
24, 2729, 36, 40, 46, 49, 51, 57]. No one has compared the
computed current density distributions, pressure, and tem-
perature profiles or reactant concentrations with experimen-
tal measurements.
6 Summary
Mathematical models for both hydrogen fuel cells and the
DMFC have been reviewed. Empirical models and mechanis-
tic models of individual fuel cells and fuel cell stacks are
examined. The fundamental electrochemical processes and
transport phenomena occurring in the diffusion layers, cata-
lyst layers, and membrane have been studied extensively.
However, more studies are needed to visualize the evolution
process of two-phase flow in the flow channels, and to
develop corresponding mathematical models to calculate the
flow dynamics and temperature profiles in the complex flow
field geometries. These studies should help to enhance the
understanding of water and reactant management in hydro-
gen fuel cells and DMFCs. Improved and validated mechan-
istic models are required to enable better design of fuel cells
and fuel cell components.
Acknowledgements
The authors would like to acknowledge the financial sup-
port of the Natural Sciences and Engineering Research Coun-
cil (NSERC) of Canada and the Center for Automotive Mate-
rials and Manufacturing (CAMM), Kingston, Canada for this
work.
List of Tables
Table 1 Diffusion coefficient (m
2
s
1
).
Correlations References
D
i
=
1r
i
=r

j,=i
c=(c
i
D
i;j
)
[25]
D
g
H
2
= 2:63 10
6
at 80 C [32, 8]
D
g
O
2
= 5:2197 10
6
at 80 C
D
g
O
2
= 01775 10
4
at 0 C [22, 23]
D
g
w
= 0:256 10
4
at 34 C
D
g
O
2
= 1:77(
T
273
)
1:823
(
1:01310
5
P
) [36]
D
g
CO
2
= 3 10
5
, D
l
CO
2
= 1 10
10
D
g
CH
3
OH
= 3 10
6
4:5986 10
8
T 9:4979 10
11
T
2
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 23
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Correlations References
D
l
CH
3
OH
= 10
5:4163999:778=T
D
g
w
= 2:56 10
5
T
305
_ _
2:334
1:01310
5
P
_ _
, D
l
w
= 0
Proton diffusion coefficient in membrane (m
2
s
1
)
D
H
= 4:5 10
9
at 80 C
[19, 45]
Table 2 Binary diffusion coefficient (m
2
s
1
).
Correlations References
Binary diffusion coefficient (cm
2
s
1
)
PD
i;j
= a
T

Tc
i
Tc
j
_
_ _
b
Pc
i
Pc
j
_ _
1=3
Tc
i
Tc
j
_ _5=12
1
Mw
i
Mw
j
_ _
0:5
[21]
P in bar and T in K
The value of a and b[57]
D
w;H
2
, D
w;O
2
and D
w;N
2
: a = 0.00364 and b = 2.334 ;
D
O
2
;N
2
a = 0.002745 and b = 1.823 ;
Tc and Pc are critical temperature and pressure. [25,57]
Water: Tc = 647.3 K and Pc = 221.2 atm
Oxygen: Tc = 154.4 K and Pc = 49.7 atm
Nitrogen: Tc = 126.2 K and Pc = 33.5 atm
D
i;j
= D
i;j
T
ref
; P
ref
_ _
P
P
ref
T
T
ref
_ _
1:5
[8, 32, 45]
D
w;H
2
= 0:915 10
4
T
ref
= 307:1 K P
ref
= 1 atm
D
w;CO
2
= 0:202 10
4
T
ref
= 307:5 K P
ref
= 1 atm
D
H
2
;CO
2
= 0:646 10
4
T
ref
= 298:0 K P
ref
= 1 atm
D
w;O
2
= 0:282 10
4
T
ref
= 308:1 K P
ref
= 1 atm
D
w;N
2
= 0:256 10
4
T
ref
= 307:5 K P
ref
= 1 atm
D
O
2
;N
2
= 0:220 10
4
T
ref
= 293:2 K P
ref
= 1 atm
cD
w;H
2
= 3:68 10
3
(T=307)
1:10
,
cD
w;O
2
= 1:01 10
3
(T=293)
1:10
,
[34]
cD
w;N
2
= 1:01 10
3
(T=293)
1:10
,
cD
O
2
;N
2
= 0:84 10
3
(T=293)
1:10
c is the mole concentration (mol m
3
), T in K
PD
O
2
;N
2
= 0:279 10
4
,
PD
w;N
2
= 0:387 10
4
,
PD
w;O
2
= 0:370 10
4
, at 80 C
[19]
P is the pressure in atm
D
CH
3
OH;w
= 1:58 10
9
exp[0:02623(T 298)[, T in K [51]
D
CH
3
OH;CO
2
= 0:0174 10
4
, D
CH
3
OH;w
= 0:02357 10
4
, [29]
D
w;CO
2
= 0:0318 10
4
, D
w;O
2
= 0:0368 10
4
,
D
w;N
2
= 0:0386 10
4
, D
O
2
;N
2
= 0:02669 10
4
at 110 C
Table 3 Water diffusion coefficient in membrane (m
2
s
1
).
Correlations References
D
w;M
= 10
10
exp 2416
1
303

1
273T
_ _ _ _
(2:563 0:33k 0:026k
2
0:000671k
3
)
[21, 30, 42, 54]
D
w;M
= exp 2416
1
303

1
273T
_ _ _ _
D
30
0
w;M
[43, 44]
D
30
0
w;M
= 10
10
k < 2
D
30
0
w;M
= 10
10
(2k 3) 2 _ k _ 3
D
30
0
w;M
= 10
10
(1:67k 2:01) 3 < k < 4:5
D
30
0
w;M
= 1:25 10
10
4:5 _ k
D
w;M
= 3:5 10
6

k
14
exp
2436
273T
_ _
[34, 39]
Table 4 Knudsen diffusion coefficient (m
2
s
1
).
Correlations References
Diffusion layer D
CH
3
OH
= 0:0785 10
4
,
D
CO
2
= 0:0670 10
4
,
D
N
2
= 0:08395 10
4
[29]
D
O
2
= 0:07853 10
4
,
D
w
= 0:1047 10
4
Catalyst layer D
CH
3
OH
= 0:00785 10
4
,
D
CO
2
= 0:00670 10
4
,
D
N
2
= 0:008395 10
4
[29]
D
O
2
= 0:007853 10
4
,
D
w
= 0:01047 10
4
Table 5 Viscosities (kg m
1
s
1
).
Correlations References
Water viscosity (kg m
1
s
1
) [25]
l
w
= 8:91 10
4
Water viscosity (kg m
1
s
1
) [23]
l
w
= 5:0 10
4
at 60 C
Viscosity of anode flow in hydrogen fueled fuel cell (atm s) [34]
l = 1:08 10
10
(T=298)
1:05
Viscosity of cathode flow in hydrogen fueled fuel cell
(kg m
1
s
1
)
[23]
l
g
= 2:03 10
5
, at 60 C
Viscosity of cathode flow in hydrogen fueled fuel cell (atm s) [34]
l = 1:36 10
10
(T=298)
1:05
Viscosity of liquid phase in DMFC (kg m
1
s
1
)
l
l
= 0:458509 5:304741 10
3
T 2:31231 10
5
T
2
4:49161 10
8
T
3
3:27681 10
11
T
4
[36]
Viscosity of gas phase in DMFC(kg m
1
s
1
)
l
g
= 2:03 10
5
[36]
R
E
V
I
E
W
24 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Table 6 Hydraulic permeability (m
2
, used in Darcy's law).
Flow channel k
m
= 5 10
10
[34]
Diffusion layer k
m
= 1 10
13
[34]
k
m
= 1 10
12
[22, 26]
k
m
= 1:2 10
12
[23]
k
m
= 1:76 10
11
[8, 25, 32]
k
m
= 1 10
11
[36]
k
m
= 1:8 10
16
[51]
Membrane k
m
= 1:58 10
18
[19, 25]
k
m
= 1:8 10
18
[8, 20, 32, 45]
k
m
= 1 10
21
[36]
k
m
= 7:18 10
18
[51]
Table 7 Electrokinetic permeability of membrane (m
2
).
Correlations References
k
U
= 1:13 10
19
[19,25]
k
U
= 7:18 10
20
[8, 20, 32]
k
U
= 2:0 10
20
[45]
Table 8 Electrochemical reaction kinetics parameters.
Parameters Correlations Reference
Anodic and cathodic transfer 0.5 (a
c
) [2124, 30, 43]
coefficient for hydrogen fueled Athode side:0.5, 0.5 [20]
fuel cell: a
a
and a
c
Cathode side: 2, 2
2, 2 [8, 19, 25, 32]
0.5, 1 [45]
Anodic and cathodic transfer 0.5, 2 [29, 51]
coefficient for DMFC: a
a
and a
c
0.875, 0.239 [36]
Ref concentration of Hydroden: 1200 [20]
c
ref
H
2
(mol m
3
)
Ref concentration of Oxygen: c
ref
O
2
3.39 [20]
(mol m
3
) 4.62 [19]
3.57 [26]
Ref concentration of Oxygen for 31.8 [29, 51]
Parameters Correlations Reference
DMFC: c
ref
O
2
(mol m
-3)
0.23 [36]
Ref pressure of methanol in
vapor fed DMFC:p
ref
M
(atm)
0.97 [29, 51]
Reference current density for Cathode : 0.01 [26]
hydrogen fueled fuel cell: J
ref
0
(A cm
2
)
Cathode : 0.9 [42]
Anode: 0.6 [45]
Cathode: 4:4 10
7
Reference current density for Anode: [36]
DMFC: J
ref
0
(A m
2
) 94:25 exp
35570
R
1
353

1
T
_ _ _ _
Cathode: 0.0422
Reference kinetic parameter for Anode : 1:4 10
5
[20]
hydrogen fueled fuel cell:
A
CV
J
ref
0
Cathode: 1:0 10
5
(A cm
3
) Cathode: 5:0 10
4
[19, 25]
Anode : 5 10
2
[8, 32]
Cathode: 1:0 10
4
Reference kinetic parameter for Anode : 1.0 [29, 51]
DMFC: A
CV
J
ref
0
(A cm
3
) Cathode: 1:0 10
5
Table 9 Membrane conductivity (X
1
cm
1
).
Correlations References
r
m
= exp[1268(
1
303

1
273T
)[(0:005139k 0:00326) [21, 30,
4244, 54]
r
m
= 0:0051638 0:00020217k 0:0022154k
2
0:0002772k
3
[39]
1:4657 10
5
k
4
2:7746 10
7
k
5
r
m
= 0:06 [45]
r
m
= 0:334 [29, 51]
Table 10 Thermodynamic open-circuit potential (V).
Correlations References
Hydrogen
fueled fuel
cell
E
0
= 1:23 0:9 10
3
(T 298)
2:3
RT
4F
log(p
2
H
2
p
O
2
)
[19, 20, 25, 45]
E
0
= 0:2329 0:0025T, T in K [8, 32]
E
0
=1.1 [21, 22, 26, 30]
DMFC E
0
=1.21 [36, 53]
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 25
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
Table 11 Other parameters.
Para-
meters
Correlations References
c
f
Fixed charge concentration (mol m
3
) [19, 20, 25, 45]
c
f
= 1:2 10
3
Cp
i
Heat capacity, (J mol
1
K
1
) [30]
Cp
l
w
= 75:38
Cp
v
w
= 33:46 6:88 10
3
T 7:60 10
6
T
2
3:59 10
9
T
3
Cp
H
2
= 28:84 7:65 10
5
T 3:29 10
6
T
2
8:70 10
10
T
3
DH
vl
Heat of evaporation (J mol
1
) [30]
DH
vl
= 45070 4:19T 3:44 10
3
T
2
2:54 10
6
T
3
8:98 10
10
T
4
, T in C
k
lv
Rate constant of liquid water evaporation (atm s
1
) [23]
k
lv
=100
k
vl
Rate constant of water vapor Condensation r (s
1
) [23]
k
vl
=100
P
sat
w
Saturation vapor pressure (atm) [21, 30, 32]
log
10
P
sat
w
= 2:1794 0:02953 T 9:1837
10
5
T
2
1:4454 10
7
T
3
, T in C
Saturation vapor pressure (atm)
P
sat
w
= 218:3 exp(7:76451 q 1:45838 q
1:5
2:77580 q
3
1:23303 q
6
,
[34]
q = 1 T=(647:3), T in K
z
f
Charge of fixed charge in membrane [19, 20, 25]
z
f
= 1
k
h
Thermal conductivity of water (W m
1
K
1
)
k
h
= 0:6
[25]
List of Symbols
A
a
Anode cross-sectional area (m
2
)
a
i
and b
i
Dimensions for the stack manifold inlet flow duct
(m)
a
o
and b
o
Dimensions for the stack manifold outlet flow duct
(m)
A
CV
Effective catalyst area per unit volume (m
1
)
A
g
Cross-sectional area available for gas phase in flow
channel (m
2
)
A
h
Heat transfer area per unit length (m)
A
l
Cross-sectional area available for liquid phase in
flow channel (m
2
)
A
T
Total cross-sectional area of the flow channel (m
2
)
a
w
Water vapor activity
c
f
Concentration of the fixed charge site in the electro-
lyte (mol m
3
)
c
I
i
Mole concentration at the inlet (mol m
3
)
c
i;AF
Mole concentration in anode channel (mol m
3
)
Cp
i
Heat capacity (J kg
1
K
1
)
c
i;AC
Mole concentration in the anode catalyst layer
(mol m
3
)
c
tot
Total mole concentration of the fluid (mol m
3
)
D
cp
Capillary diffusion coefficient (m s
2
)
D
i
Diffusion coefficient (m s
2
)
D
i,j
Binary Diffusion coefficient (m s
2
)
D
k
i
Knudsen diffusion coefficient (m s
2
)
E Cell voltage (V)
E
0
Thermodynamic open-circuit potential (V)
F Faraday constant, 96,484 (C mol
1
)
f
d
Interfacial drag coefficient
ffi(n)
and ffo(n) Inlet and outlet flow rate of the nth fuel cell (m
3
s
1
)
H Height of flow channel (m)
h
AD
Thickness of the anode diffusion layer (m)
DH
vl
Heat of evaporation (J mol
1
)
I Hydrogen flux produced at anode,
I = J=2F = N
H
2
;A
(mol m
2
s
1
)
J Current density (A m
2
)
j
a
, j
c
Transfer current densityat anode andcathode (Am
3
)
k
h
Heat transfer coefficient (w m
2
K
1
)
k
LS
Mass transfer coefficient between anode flow and
anode diffusion layer (m s
1
)
k
lv
Evaporation rate constant of liquid water (atm s
1
)
k
vl
Condensation rate constant of water vapor (s
1
)
L Length of the flow channel, (m)
M
I
w;AF
, Anode inlet water flow rate, (mol S
1
)
M
O
w;AF
Anode outlet water flow rate, (mol S
1
)
Mw
i
Molar weight (kg mol
1
)
n
drag
Electro-osmotic coefficient, the average number of
water molecules dragged per proton
n
M
drag
Electro-osmotic drag coefficient for methanol
N
w;A
Water flux through the interface between the anode
flow channel and the diffusion layer (mol m
2
S
1
)
P Pressure (N m
2
)
P
cp
Capillary pressure (N m
2
)
P
sat
w
Saturation vapor pressure (N m
2
)
Q The consumption rate of all reactants (m
3
s
1
)
Q Volume flow rate (m
3
s
1
)
R
g
i
Generation or consumption rate of species in the
gas phase (mol m
3
s
1
)
R
l
i
Generation or consumption rate of species in the li-
quid phase (mol m
3
s
1
)
R
w
Inter-phase mass-transfer rate of water through
condensation or evaporation (mol m
2
s
1
)
S Degree of liquid saturation
S
h
Heat source term, which accounts for the heat
transfer between gas and flow plates (J m
3
s
1
)
T
AF
Temperature of gas flow in the channel (K)
T
P
Temperature of flow plate (K)
V Vector of velocity (m s
1
)
V
AC
Volume of anode catalyst layer (m
3
)
V
AF
Volume of anode flow channel (m
3
)
W Width of the flow channel (m)
x
H
2
;AF
; x
w;AF
Mole fraction of hydrogen and water at anode flow
channel
R
E
V
I
E
W
26 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
x
I
H
2
;AF
; x
I
w;AF
Mole fraction of hydrogen and water at inlet of an-
ode flow channel
z
f
Charge of the fixed charge site in the electrolyte
z
i
Charge on species i
A Ratio between N
w;A
and N
H
2
;A
a
a
and a
c
Anodic and cathodic transfer coefficient
c
c
Advection correction factor
E Porosity
g
a
Anode side overpotential.
g
a
Cathode side overpotential.
g
r
Ohmic overpotential.
k
h
Thermal conductivity (W m
1
s
1
)
k
m
Permeability of porous media (m
2
)
k
g
r
Relative permeability for gas phase (m
2
)
k
l
r
Relative permeability for liquid phase (m
2
)
k
U
Electrokinetic permeability (m
2
)
k Water content, the ratio of the number of water mo-
lecules to the number of charge sites
k
CH
3
OH
Methanol content, the ratio of the number of metha-
nol molecules to the number of charge sites
k
g
Relative mobility of gas phase
k
l
Relative mobility of liquid phase
M Viscosity (kg m
1
s
1
)
r Density (kg m
3
)
r
eff
e
Effective conductivity of catalyst layer (X
1
m
1
)
r
m
Membrane conductivity (X
1
m
1
)
T Tortuosity
T
a
Residence time in anode channel (s
1
)
t
H
The stoichiometric coefficient, the ratio between
M
I
w;A
and N
H
2
;A
(W L)
U
e
Potential of electrolyte phase (V)
References
[1] N.M. Markovic, T.J. Schmidt, V. Stamenkovic, P.N.
Ross, Fuel Cells 2001, 1(2), 105116.
[2] P.M. Urban, A. Funke, J.T. Muller, M. Himmen, A. Doc-
ter, Applied Catalysis A: General 2001, 221(12), 459470.
[3] J.M. Leger, J. Appl. Electrochem. 2001, 31(7), 767771.
[4] P. Costamagna, S. Srinivasan, J. Power Sources 2001,
102(12), 242252.
[5] T. Schultz, S. Zhou, K. Sundmacher, Chemical Engineer-
ing & Technology 2001, 24(12), 12231233.
[6] S. Gamburzev, A.J. Appleby, Fuel Cells 2001, 1(2),
105116.
[7] S. Srinivasan, O.A. Velev, A. Parthasarathy, D.J. Manko,
A.J. Appleby, J. Power Sources 1991, 36(3), 299320.
[8] S. Um, C.Y. Wang, HTD (Am. Soc. Mech. Eng.) Proceed-
ings of the ASME Heat Transfer Division 2000 2000, 366
(1), 1925.
[9] J. Kim, S.M. Lee, S. Srinivasan, C.E. Chamberlin, J. Elec-
trochem. Soc. 1995, 142(8), 2670-4.
[10] J. H. Lee, T.T. Lalk, A.J. Appleby, J. Power Sources 1998,
70(2), 258268.
[11] G. Squadrito, G. Maggio, E. Passalacqua, F. Lufrano, A.
Patti, J. Applied Electrochemistry 1999, 29, 14491455.
[12] L. Pisani, G. Murgia, M. Valentini, B. D'Aguanno,
J. Power Sources 2002, 108, 192203.
[13] A. A. Kulikovsky, Electrochemistry Communications 2002,
4, 845852.
[14] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley,
P.R. Roberge, T.J Harris, J. Electrochem. Soc. 1995, 142(1),
18.
[15] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley,
P.R. Roberge, T.J Harris, J. Electrochem. Soc. 1995, 142(1),
915.
[16] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley,
P.R. Roberge, T.J Harris, Proc. Intersoc. Energy Convers.
Eng. Conf. 1993, 28(1), 1.12151.1220.
[17] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley,
P.R. Roberge, A. Rodrigues, J. Power Sources 1994,
49(13), 34956.
[18] R.F. Mann, J.C. Amphlett, M.A. Hooper, H.M. Jensen,
B.A. Peppley, P.R. Roberge, J. Power Sources 2000,
86(12), 173180.
[19] A. Simoglou, P. Argyropoulos, E.B. Martin, K. Scott,
A.J. Morris, W.M. Taama, Chem. Eng. Sci. 2001, 56(23),
67616772.
[20] A. Simoglou, P. Argyropoulos, E.B. Martin, K. Scott,
A.J. Morris, W.M. Taama, Chem. Eng. Sci. 2001, 56(23),
67736779.
[21] A.A. Kulikovsky, Electrochemistry Communications 2002,
4, 939946.
[22] D. Chu, R. Jiang, J. Power Sources 1999, 83(12), 128133.
[23] D.M. Bernardi, M.W. Verbrugge, AIChE J. 1991, 37(8),
115163.
[24] D.M. Bernardi, M.W. Verbrugge, J. Electrochem. Soc.
1992, 139(9), 247791.
[25] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, J. Electro-
chem. Soc. 1991, 138(8), 233442.
[26] J.S. Yi, T.V.Nguyen, J. Electrochem. Soc. 1999, 146(1),
3845.
[27] W. He, J.S. Yi, T. Van Nguyen, AIChE J. 2000, 46(10),
20532064.
[28] D. Natarajan, T. Van Nguyen, J. Electrochem. Soc. 2001,
148(12), A1324A1335.
[29] V. Gurau, H. Liu, S. Kakac, AIChE J. 1998, 44(11),
24102422.
[30] A. Kazim, H.T. Liu, P. Forges, J. Appl. Electrochem. 1999,
29(12), 14091416.
[31] A.A. Kulikovsky, J. Divisek, A.A. Kornyshev, J. Electro-
chem. Soc. 1999, 146(11), 39813991.
[32] A.C. West, T.F. Fuller, J. Appl. Electrochem. 1996, 26(6),
557565.
[33] A.A. Kulikovsky, J. Divisek, A.A. Kornyshev, J. Electro-
chem. Soc. 2000, 147(3), 953959.
[34] T.V. Nguyen, R.E. White, J. Electrochem. Soc. 1993,
140(8), 217886.
[35] G.E. Suares, K.A. Hoo, Chem. Eng. Sci. 2000, 55(12),
22372247.
[36] S. Um, C.Y. Wang, K.S. Chen, J. Electrochem. Soc. 2000,
147(12), 44854493.
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 27
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
[37] Z. H. Wang, C.Y. Wang, K.S. Chen, J. Power Sources
2001, 94(1), 4050.
[38] I.M. Hsing, P.M. Futerko, Chem. Eng. Sci. 2000, 55(19),
42094218.
[39] H. van Bussel, F. Koene, R. Mallant, J. Power Sources
1998, 71(12), 218222.
[40] Z. H. Wang, C.Y. Wang, Proc. Electrochem. Soc. (Direct
Methanol Fuel Cells) 2001, 20014, 286319.
[41] T.E.Springer, T. Rockward, T.A. Zawodzinski, S. Gottes-
feld, J. Electrochem. Soc. 2001, 8(1), A11A23.
[42] T.F. Fuller, J. Newman, J. Electrochem. Soc. 1993, 140(5),
121825.
[43] K. Dannenberg, P. Ekdunge, G. Lindbergh, J. Appl. Elec-
trochem. 2000, 30(12), 13771387.
[44] G. J. M. Janssen, J. Electrochem. Soc. 2001, 148(12),
A1313A1323.
[45] A.A. Kulikovsky, Fuel Cells 2001, 1(2), 162169.
[46] P. Costamagna, Chem. Eng. Sci. 2001, 56(2), 323332.
[47] S. Dutta, S. Shimpalee, J.W. Van Zee, J. Appl. Electro-
chem. 2000, 30(2), 135146.
[48] S. Dutta, S. Shimpalee, J.W. Van Zee, Int. J. Heat Mass
Transfer 2001, 44(11), 20292042.
[49] T. Berning, D.M. Lu, N. Djilali, J. Power Sources 2002,
106(12), 284294.
[50] A. Rowe, X. Li, J. Power Sources 2001, 102(12), 8296.
[51] D. Bevers, M. Wohr, K. Yasuda, K. Oguro, J. Appl. Elec-
trochem. 1997, 27(11), 12541264.
[52] G. Maggio, V. Recupero, L. Pino, L. J. Power Sources
2001, 101(2), 275286.
[53] C. Marr, X. Li, Xianguo. J. Power Sources 1999, 77(1),
1727.
[54] K. Sundmacher, K. Scott, Chem. Eng. Sci. 1999,
54(1314), 29272936.
[55] A.A. Kulikovsky, J. Appl. Electrochem. 2000, 30(9),
10051014.
[56] M. Eikerling, Y.I. Kharkats, A.A. Kornyshev, Y.M. Volf-
kovich, J. Electrochem. Soc. 1998, 145(8), 26842699.
[57] K. Sundmacher, T. Schultz, S. Zhou, K. Scott, M. Ginkel,
E.D. Gilles, Chem. Eng. Sci. 2001, 56(2), 333341.
[58] H. Dohle, J. Divisek, R. Jung, J. Power Sources 2000,
86(12), 469477.
[59] P. Argyropoulos, K. Scott, W.M. Taama, J. Appl. Electro-
chem. 1999, 29(6), 661669.
[60] M.M. Mench, S. Boslet, S. Thynell, J. Scott, C.Y. Wang,
Proc. Electrochem. Soc. (Direct Methanol Fuel Cells) 2001,
20014, 241253.
[61] V. Gurau, F. Barbir, H. Liu, J. Electrochem. Soc. 2000,
147(7), 24682477.
[62] M.C. Kimble, N.E. Vanderborgh, Proc. Intersoc. Energy
Convers. Eng. Conf. 1992, 27( 3), 3.4133.417.
[63] C.Y. Wang, P. Cheng, Int. J. Heat & Mass Transfer 1996,
39(17), 36073618.
[64] T.E. Springer, M.S. Wilson, S. Gottesfeld, J. Electrochem.
Soc. 1993, 140(12), 351326.
[65] G. Murgia, L. Pisani, M. Valentini, B. D'Aguanno, J.
Electrochem. Soc. 2002, 149(1), A31A38.
[66] F. Gloaguen, R. Durand, J. Appl. Electrochem. 1997, 27(9),
10291035.
[67] K. Broka, P. Ekdunge, J. Appl. Electrochem. 1997, 27(3),
281289.
[68] C. Boyer, S. Gamburzev, A.J. Appleby, J. Appl. Electro-
chem. 1999, 29(9), 10951102.
[69] Y. Bultel, P. Ozil, R. Durand, J. Appl. Electrochem. 1998,
28(3), 269276.
[70] Y. Bultel, P. Ozil, R. Durand, J. Appl. Electrochem. 1999,
29(9), 10251033.
[71] Y. Bultel, P. Ozil, R. Durand, J. Appl. Electrochem. 2000,
30(12), 13691376.
[72] S.F. Baxter, V.S. Battaglia, R.E. White, J. Electrochem. Soc.
1999, 146(2), 437447.
[73] C.Y. Wang, P. Cheng, Int. J. Heat Mass Transfer 1996,
39(17), 36073618.
[74] J.J. Baschuk, X. Li, J. Power Source 2000, 86(12), 181196.
[75] J. Cruickshank, K. Scott, J. Power Sources 1998, 70(1), 4047.
[76] K. Scott, W. Taama, J. Cruickshank, J. Appl. Electrochem.
1998, 28(3), 289297.
[77] K. Scott, W. Taama, J. Cruickshank, J. Power Sources
1997, 65(12), 159171.
[78] X.D. Din, E.E. Michaelides, AIChE J. 1998, 44(1), 3547.
[79] T. Thampan, S. Malhotra, H. Tang, R. Datta, J. Electro-
chem. Soc. 2000, 147(9), 32423250.
[80] T.A. Zawodzinski, T.E. Springer, F. Uribe, S, Gottesfeld,
Solid State Ionics 1993, 60(13), 199211.
[81] M.W. Verbrugge, R.F. Hill, E.W. Schneider, AIChE J.
1992, 38(1), 93100.
[82] M.W. Verbrugge, E.W. Schneider, R.S. Conell, R.F. Hill,
J. Electrochem. Soc. 1992, 139(12), 34218.
[83] M.W. Verbrugge, R.F. Hill, J. Electrochem. Soc. 1990,
137(3), 8939.
[84] M. Eikerling, A.A. Kornyshev, U. Stimming, J. Phys.
Chem. B 1997, 101(50), 1080710820.
[85] F.N. Buchi, G.G. Scherer, J. Electrochem. Soc. 2001,
148(3), A183A188.
[86] T. Bewer, T. Beckmann, H. Dohle, J. Mergel, R. Neitzel,
Proceedings Electrochemical Society (Direct Methanol Fuel
Cells) 2001, 20014, 320330.
[87] D. Thirumalai, R.E. White, J. Electrochem. Soc. 1997,
144(5), 17171723.
[88] E. Hontanon, M.J. Escudero, C. Bautista, P.L. Garcia
Ybarra, L. Daza, J. Power Sources 2000, 86(12), 363368.
[89] N.E. Vanderborgh, J.R. Huff, J. Hedstrom, Proc. Intersoc.
Energy Convers. Eng. Conf. 1989, 24(3), 163740.
[90] J.S. Yi, T.V. Nguyen, J. Electrochem. Soc. 1998, 145(4),
11491159.
[91] P. Argyropoulos, K. Scott, W.M. Taama, Chemical Engi-
neering Journal 1999, 73(3), 217227.
[92] P. Argyropoulos, K. Scott, W.M. Taama, Chemical Engi-
neering Journal 1999, 73(3), 229245.
[93] P. Argyropoulos, K. Scott, W.M. Taama, Chemical Engi-
neering Journal 1999, 78(1), 2941.
[94] R.A. Bajura, E.H. Jones, J. Fluid Engineering 1976, 12,
654666.
R
E
V
I
E
W
28 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim FUEL CELLS 2004, 4, No. 12
Yao et al.: A Review of Mathematical Models for Hydrogen and Direct Methanol Polymer Electrolyte
[95] A.B. Datta, A.K. Majumdar, Int. J. Heat & Mass 1980,
2(4), 253262.
[96] J.B. Riggs, Ind. Eng. Chem. Res. 1987, 26(1), 129133.
[97] A. Acrivos, B.D. Babcock, R.L. Pigford, Chem. Eng. Sci.
1959, 10(1), 112124.
[98] J.D. Keller, P.A. Pittsburgh, J. Appl. Mechanics 1949,
19(3), 7785.
[99] A.B. Datta, A.K. Majumdar, Int. J. Heat & Mass 1983,
26(9), 13121328.
[100] E. Markland, Engineering 1959, 187(2), 150151.
[101] P. Costamagna, E. Arato, E. Achenbach, U. Reus, J.
Power Sources 1994, 52(2), 243249.
[102] J.C. Amphlett, R.F. Mann, B.A. Peppley, P.R. Roberge,
A. Rodrigues, J. Power Sources 1996, 61(12), 183188.
[103] P. Argyropoulos, K. Scott, W.M. Taama, J. Power Sources
1999, 79(2), 169183.
[104] P. Argyropoulos, K. Scott, W.M. Taama, J. Power Sources
1999, 79(2), 184198.
[105] G. Maggio, V. Recupero, C. Mantegazza, J. Power
Sources 1996, 62(2), 167174.
[106] J.H. Lee, T.R. Lalk, J. Power Sources 1998, 73(2), 229241.
[107] P. Argyropoulos, K. Scott, W.M. Taama, Chemical Engi-
neering & Technology 2000, 23(11), 985995.
[108] P. Argyropoulos, K. Scott, W.M. Taama, J. Appl. Electro-
chem. 2000, 30(8), 899913.
[109] J.Divisek, J. Fuhrmann, K. Gartner, R. Jung, J. Electro-
chem. Soc. 2003, 150(6), A811A825.
______________________
R
E
V
I
E
W
FUEL CELLS 2004, 4, No. 12 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 29

Você também pode gostar