INTERNATIONAL JOURNAL OF ENERGY RESEARCH

Int. J. Energy Res. 2008; 32:698–721

Published online 17 December 2007 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/er.1384

A parametric study of multi-phase and multi-species transport

in the cathode of PEM fuel cells

Nada Zamel and Xianguo Li*,y

Department of Mechanical Engineering, University of Waterloo, Waterloo, Ont., Canada

SUMMARY

In this study, a mathematical model is developed for the cathode of PEM fuel cells, including multi-phase and
multi-species transport and electrochemical reaction under the isothermal and steady-state conditions.
The conservation equations for mass, momentum, species and charge are solved using the commercial software
COMSOL Multiphysics. The catalyst layer is modeled as a finite domain and assumed to be composed of a uniform
distribution of supported catalyst, liquid water, electrolyte and void space. The Stefan–Maxwell equation is
used to model the multi-species diffusion in the gas diffusion and catalyst layers. Owing to the low relative
species’ velocity, Darcy’s law is used to describe the transport of gas and liquid phases in the gas diffusion and catalyst
layers. A serpentine flow field is considered to distribute the oxidant over the active cathode electrode surface,
with pressure loss in the flow direction along the channel. The dependency of the capillary pressure on the saturation is
modeled using the Leverette function and the Brooks and Corey relation. A parametric study is carried
out to investigate the effects of pressure drop in the flow channel, permeability, inlet relative humidity and shoulder/
channel width ratio on the performance of the cell and the transport of liquid water. An inlet relative humidity
of 90 and 80% leads to the highest performance in the cathode. Owing to liquid water evaporation, the relative
humidity in the catalyst layer reaches 100% with an inlet relative humidity of 90 and 80%, resulting in a high electrolyte
conductivity. The electrolyte conductivity plays a significant role in determining the overall performance up to a
point. Further, the catalyst layer is found to be important in controlling the water concentration in the cell. The
cross-flow phenomenon is shown to enhance the removal of liquid water from the cell. Moreover, a shoulder/channel
width ratio of 1:2 is found to be an optimal ratio. A decrease in the shoulder/channel ratio results in an increase in
performance and an increase in cross flow. Finally, the Leverette function leads to lower liquid water saturations in the
backing and catalyst layers than the Brooks and Corey relation. The overall trend, however, is similar for both
functions. Copyright # 2007 John Wiley & Sons, Ltd.

KEY WORDS: PEM fuel cell; liquid water removal; cross flow; multi-phase transport

*Correspondence to: Xianguo Li, Department of Mechanical Engineering, University of Waterloo, Waterloo, Ont., Canada.
y
E-mail: x6li@uwaterloo.ca

Contract/grant sponsor: Natural Sciences and Engineering Research Council (NSERC) of Canada

Received 14 September 2007

Revised 15 October 2007
Copyright # 2007 John Wiley & Sons, Ltd. Accepted 15 October 2007
 THE CATHODE OF PEM FUEL CELLS
1. INTRODUCTION
The loss in cell voltage due to the slow rate of
oxygen transport and weak oxygen reduction
reaction (ORR) at the cathode poses a major
challenge in the commercialization of PEM fuel
cells. This is further augmented by liquid water
flooding in the backing (or gas diffusion) and
catalyst layers. The liquid water hinders the
transport of species in the layers to reach the
reactive surfaces.
Numerical modeling is widely used to investi-
gate the performance of polymer electrolyte
membrane fuel cells. Many multi-phase models
are available in literature and can be classified on
the basis of half cell and full cell models. The half
cell models describe the transport of liquid water
in the cathode side due to the nature of the
electrochemical reaction at this half of the cell. He
et al. [1] developed a two-dimensional, multi-
phase, multicomponent transport model that uses
an interdigitated gas distributor to take into
account the forced flow of the gas through the
porous electrode. In this model, the catalyst layer
is considered as a boundary. Therefore, water
production is considered to occur at the gas
diffusion layer (GDL)/catalyst layer interface. In
addition, they assumed that the capillary pressure
(the capillary diffusion) has a constant slope across
the GDL. Natarajan and Nguyen [2] developed a
two-dimensional, multi-phase, multicomponent,
transient model of the cathode of the PEM fuel
cell. In this model, conventional gas distributors
are used and therefore, diffusion is the dominant
transport process. The catalyst layer was assumed
to be very thin and is considered as a boundary.
The performance of the cathode was studied under
different conditions. Wang et al. [3] modeled the
multi-phase flow and transport of reactants and
products in the cathode. They assumed the GDL
to be a hydrophilic material. The capillary pressure
was described by the Leverette function [4] for
hydrophilic media. In the backing layer, a multi-
phase region co-exists with a single-phase region.
This model, like many in literature, assumes that
the catalyst layer is a thin interface (boundary).
The same group [5] generated a mathematical
model to investigate the flooding dynamics in
Copyright # 2007 John Wiley & Sons, Ltd.
699
PEM fuel cells. The model is three dimensional
and describes the multi-phase phenomena in the
cathode backing layer. The catalyst coverage
model in the catalyst layer was also accounted
for. The capillary pressure was described by the
Leverette function for a hydrophobic media (the
backing layer was considered to be hydrophobic).
Senn and Poulikakos [6] put forward a model of
the mass and heat transport in the backing layer of
a PEM fuel cell. They assumed that the capillary
pressure (driving force of liquid water in the layer)
is governed by the expression developed by
Leverette. The backing layer was taken as a
hydrophobic medium. The catalyst layer was
taken as a thin layer. The effects of the GDL
thickness, number of channels and the shoulder
width on the performance of the cell were
investigated.
The full cell models take into account the anode
and cathode half cells as well as the membrane.
You and Liu [7] are among the many who have
developed a full cell model to investigate multi-
phase flow in PEM fuel cell. They simulated the
transport of species and the electrochemical
reaction and compared the resultant polarization
curve with experimental data. Hu et al. [8, 9]
developed a three-dimensional, multi-phase flow
model. The model investigates a conventional and
an interdigitated flow field. The model results
show that the saturation of liquid water in the
conventional flow field is less than that in the
interdigitated flow field; therefore, higher ohmic
losses are encountered with the interdigitated flow
field. Siegel et al. [10] developed a two-dimen-
sional, multi-phase model. It is shown that
consideration of liquid water is essential to
accurately simulate the fuel cell operation. The
results of the model suggest that the water transfer
from the membrane to the cathode through
electro-osmotic drag is very important. Baschuk
and Li [11] developed a general formulation for a
polymer electrolyte membrane fuel cell. The model
formulation covers the electrochemical reactions,
proton migration and mass transport of the
gaseous reactants and liquid water. The formula-
tion can be applied to the gas channels, bipolar
plate, gas diffusion and catalyst layers on the
anode and the cathode sides of the fuel cell.
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
700 N. ZAMEL AND X. LI

The conservation of mass, momentum, species and Li [12] shows that due to the pressure drop along
energy are applied to each phase, with the the channel flow field with a serpentine shape
technique of volume averaging being used to (Figure 1), the reactants can leak or cross to
incorporate the interactions between the phases neighboring channels via the porous GDL next
as interfacial source terms. In order to avoid to the bipolar plate. The cross flow is found to lead
problems arising from phase discontinuities, the to larger effective flow area altering reactant flow
gas and liquid phases are considered as a mixture. in the flow channel and therefore changing the
The literature review revealed that many of the flow distribution. Many other studies have shown
numerical models available treat the gas diffusion that the cross-flow phenomenon achieved by using
and catalyst layers as hydrophilic media, assume a serpentine flow field leads to significant enhance-
that the catalyst layer is a thin interface and ment of the overall performance as well as to
therefore is just a boundary and the effects of proper liquid water removal (see [13–15]).
cross-flow phenomenon are lacking. Hence, the To analyze the effects of the cross flow on the
main objectives of this study are to treat the transport phenomena, a cross section of the
catalyst layer as a finite modeling domain and to cathode with three channels has been modeled as
take into account the cross-flow phenomenon shown in Figure 2. The x- and y- axes have been
between adjacent channels. Then, a parametric normalized. The x-axis has been normalized by the
study is carried out to investigate the effects of height of the channel, which is 1 mm: The bipolar
inlet relative humidity (RH) and shoulder to plate thickness ð04y42Þ has been normalized by
channel width ratio. the channel height, which is 1 mm: The section ð2
4y43Þ has been normalized by the thickness of
the GDL which is 150 mm: Finally, the thickness of
the catalyst layer, 5 mm; was used for the normal-
2. MODEL FORMULATION
ization of section ð34y44Þ: The non-uniform
normalization of the y-axis was necessary for the
2.1. Physical problem
The parameteric study focuses on the multi-phase
and multi-species transport phenomena and elec-
trochemical reaction in the cathode of a PEM fuel
cell. The electrochemical reaction at the cathode is
known as the ORR. Electrons, protons and
oxygen molecules combine to produce water as
follows:
1
2 O2 þ 2Hþ þ 2e
! H2 O ð1Þ
Water is formed at the reaction sites as a result
of the oxygen reduction. Liquid water build-up in
the void region of the porous gas diffusion and
catalyst layers will block oxygen diffusion. On the
other hand, liquid water is needed for the
hydration of the membrane for proton transport.
Therefore, it is crucial to investigate the transport
of liquid water in the cathode.
As mentioned earlier, the effects of cross flow on
the transport of liquid water will be taken into
account in this study. A comprehensive study on
the cross-flow phenomenon in a PEM fuel cell with Figure 1. Diagram of a serpentine flow field configura-
a serpentine flow channel carried out by Park and tion [13].

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS
Figure 2. Modeling domain}x- and y-axes have been
normalized}x-axis is normalized by the height of the
flow channel which is 1 mm, here x ¼ 6 denotes a
distance of 6 mm; the section ð04y42Þ is normalized by
the channel height which is 1 mm and y ¼ 2 denotes a
distance of 2 mm. The section ð24y43Þ is normalized
by the gas diffusion layer thickness which is 150 mm and
the section ð34y44Þ is normalized by the catalyst layer
thickness which is 5 mm:
clear presentation of the results due to the small
thickness of the gas diffusion and catalyst layers in
comparison with the channel dimension. The flow
in the channels is not modeled since the flow is
considered fully developed and uniform; thus, the
reactant concentration and pressure can be speci-
fied at the channel/GDL interface. The electrons
travel through an external electric circuit from the
anode side to the cathode. They are transported
via the bipolar plate to the GDL and then to the
catalyst layer. The membrane is used to transport
the protons to the catalyst layer. The oxygen
transports in the GDL to the catalyst layer for
reaction.
2.2. Assumptions
1. The process is steady and isothermal.
2. The flow is laminar and incompressible.
3. Effects of gravity are negligible.
4. The model is two dimensional}dependency in
the x- and y-directions only.
5. The macro-homogeneous approach is consid-
ered for the gas diffusion and catalyst layers.
The macro-homogeneous structure leads to the
assumption that all the ingredients of the layers
are evenly distributed over the layer.
Copyright # 2007 John Wiley & Sons, Ltd.
701
6. The gas diffusion and catalyst layers are
hydrophobic (i.e. the contact angle is greater
than 908 ) [16–20].
7. Both vapor and liquid phases co-exist in the gas
diffusion and catalyst layers.
8. The production of water in the catalyst layer is
in the liquid form.
2.3. Governing equations
2.3.1. Conservation of momentum. Darcy’s law is
used to describe the relative velocity of species
given that the flow of species in the gas diffusion
and the catalyst layers is expected to be small.
Thus, the velocity and pressure of the phases can
be solved for using the generalized Darcy term as
follows [21, 22]:
m
0 ¼
rPk
k uk ð2Þ
kkrk
where Pk is the pressure of phase k; mk is the
viscosity of phase k; k is the permeability of the
porous cathode, krk is the relative permeability of
phase k and uk is the relative velocity of phase k:
Here, the phase pressure is dependent on the
phase velocity, viscosity and permeability. The
permeability is independent of the fluid, whereas
the relative permeability of the phase is dependent
on the liquid water saturation in the layers. In
addition, the relative permeability is dependent on
the capillary pressure, Pc ; which is the difference
between gas and liquid pressures:
Pc ¼ Pg
Pl ð3Þ
Normally, a liquid water droplet is taken to be
in the shape of a cylindrical pore. For a cylindrical
pore, the capillary pressure can be expressed in the
following general form as a function of the pore
size, contact angle and surface tension [23]:
2s cos yc
Pc ¼
ð4Þ
r
where s is the surface tension, yc is the contact
angle of the wetting phase with the solid phase in
the porous media and r is the radius of the
cylindrical pore.
According to Ustohal et al. [24], there exists a
minimum (residual) saturation at which the
capillary pressure approaches infinity. At the same
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
702 N. ZAMEL AND X. LI

time, there exists a maximum saturation at which where Sr and Sm are the residual and
the capillary pressure approaches negative infinity. maximum saturation, respectively, of the wetting
One can think of the residual saturation as the fluid. Their values have been found by Ustohal
saturation at which the fluid loses its capability to et al. [24].
move [25]. A general correlation for the capillary The relative permeability of the wetting and
pressure is given below [26]: non-wetting phases are described as follows [3]:
rffiffiffi
E krw ¼ S3 ; krnw ¼ ð1
SÞ3 ð10Þ
Pc ¼ 2$syc f ðSÞ ð5Þ
k
where Pc is the capillary pressure, $ is the 2.3.2. Conservation of mass. The conservation of
correction factor to account for the change in mass is applied to solve for the saturation of liquid
contact angle due to roughness, yc is the contact water as follows:
angle of the wetting fluid to the solid, f ðSÞ is a   
mg krl
function of the effective saturation of the wetting r  rl
Dc rS þ ug ¼ Rv$l þ ðRH2 O Þl
ml krg
fluid.
For the purpose of this study, two different ð11Þ
relationships that describe the capillary pressure
where the capillary diffusion coefficient, Dc ; is
will be used and discussed. The most widely used
defined as
relationship in the fuel cell literature is that
relationship derived by Leverette and simply kl dPc
Dc ¼
ð12Þ
known as the Leverette function [4]. ml dS
 E 0:5 and
Pc ¼ s cosðyc Þ f ðSÞ ð6Þ
k kl ¼ kkrl ; kg ¼ kkrg ð13Þ
where the expression of the Leverette function is The two source terms in the mass conservation
dependent on the contact angle as follows: equation of liquid water describe the condensation
8
>
> 1:417ð1
SÞ
2:120ð1
SÞ2 þ1:263ð1
SÞ3 or evaporation of liquid water due to phase change
>
>
>
< if 084yc 5908 from vapor to liquid or vice versa in the gas
f ðSÞ ¼ ð7Þ diffusion and catalyst layers and the addition of
>
> 1:417S
2:120S 2 þ 1:263S 3 liquid water via the electrochemical reaction in the
>
>
>
: catalyst layer. The condensation, evaporation term
if 9085yc 51808
is evaluated as follows:

The other relationship that will be used in this * H2 O
M
study is known as the Brooks and Corey relation- Rv$l ¼ wE ðxH2 O Pg
Psat Þð1
SÞq
RT
ship for a porous media and is shown below [27] 
8 * H2 O
M
> Pd þEE SðxH2 O Pg
Psat Þð1
qÞ
>
> if 084yc 5908 RT
< ðSÞ1=d
Pc ¼ ð8Þ
>
> Pd ð14Þ
>
: if 9085y c 51808
ð1
SÞ1=d where w is the condensation constant, E is the
where Pd is the displacement pressure and d is the evaporation constant and q is the switch function
pore size distribution. The values of Pd and d are and is dependent on the RH inside the gas
found from [24]. diffusion and catalyst layers:
8 xH2 O Pg
The effective saturation is needed to solve the >
>
momentum equation and is defined as < 0 if P 41
sat
q¼ ð15Þ
s
Sr >
> xH2 O Pg
S¼ ð9Þ : 1 if >1
Sm
Sr Psat

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS 703

The saturation pressure in atmospheres is evaluate the vapor water that will condense or
estimated using the following function [28]: evaporate
 X
log10 Psat ¼
2:1794 þ 0:02953ðT
273Þ r  ðruYH2 O Þ ¼ r  YH2 O ð1
SÞ1:5 Deff
H2 O;j

9:1837  10
5 ðT
273Þ2  

rM *
 rYH2 O þ YH2 O
þ 1:4454  10
7 ðT
273Þ3 ð16Þ *
M
The addition of liquid water through the
Rv$l ð20Þ
electrochemical reaction is evaluated in the cata-
Bruggemann’s correction has to be used to
lyst layer only. It is equal to zero in the GDL since
obtain the effective diffusion coefficient of species i
no reaction takes place there. The production of
in the gas diffusion and catalyst layers. The
water follows the subsequent expression:
! gO correction is as follows:
* H2 O
M ref CO2
2

ðRH2 O Þl ¼ ð1
SÞAv i0 Deff 1:5

i;j ¼ EGDL Di;j
2F ref
CO
Deff 1:5
2
   i;j ¼ ECL Di;j ð21Þ
 aF
  exp ðfs
fm Þ
RT
  2.3.4. Conservation of charge. The conservation
aF
 of charge is applied to the charge due to electron

exp
fs
fm  ð17Þ
RT  and proton transports as given below
r  is ¼ r  ð
seff
s rfs Þ ¼
J ð22Þ
2.3.3. Conservation of species. The conservation where seff is the effective electrical conductivity
s
of species is applied to solve for the transport of

oxygen and water vapor. The Stefan–Maxwell r  im ¼ r 
seff

m rfm ¼ J ð23Þ
equation is used to solve for the mass fractions of where sm is the electrolyte conductivity, and the
oxygen and water vapor as follows: volumetric current density, J; is determined from
 X
r  ðruYO2 Þ ¼ r  YO2 ð1
SÞ1:5 Deff the Butler–Volmer equation for the cathodic
O2 ;j
reaction:
  ! gO  
rM * 2

 rYO2 þ YO2
RO2 ð18Þ ref C O aF
*
2
M J ¼ ð1
SÞAv i0 exp ðf
fm Þ
COref
2
RT s
The source term in the above equation is equal  
to zero in the GDL since there is no sink source. In aF

exp
fs
fm ð24Þ
the catalyst layer, the rate of oxygen reaction is as RT
follows: Since the gas diffusion and catalyst layers are
!gO
M* O2 2 porous media, the solid and electrolyte conductiv-
ref CO2
RO2 ¼ ð1
SÞAv i0 ref ity should be corrected:
4F CO (
ss ð1
EGDL Þ1:5 in the GDL
2
   eff
 aF ss ¼ ð25Þ
  exp ðfs
fm Þ ss ðes Þ1:5 in the CL
RT
  where EGDL is the void region in the GDL and es is
aF 

exp
ðfs
fm Þ  ð19Þ the solid phase fraction in the catalyst layer, and
RT
1:5
seff
m ¼ sm ðECL em Þ ð26Þ
The conservation of water vapor in the gas
diffusion and catalyst layers is presented in where ECL is the void region in the catalyst layer
Equation (20). The source term, Rv$l ; is present and em is the electrolyte fraction in the catalyst
in the gas diffusion and catalyst layers and used to layer.

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
704
2.4. Boundary conditions
The boundary conditions employed to solve
the governing equations are described as
follows, referring to the physical problem shown
in Figure 1:
* At the GDL/channel 1 inlet interface, the mass
fraction of oxygen is specified. It is assumed
that the flow in the channel is fully developed
and uniform; thus, the mass fraction can be
calculated knowing the inlet RH and that the
reactant gas is air. That is,
YO2 jat x2½0:5;1:5&y¼2 ¼ YO2;in
* At the GDL/channel 2 and 3 interfaces, the
mass fraction of oxygen is specified. There is a
loss in concentration of the reactant gas along
the serpentine flow field, such that,
YO2 jat x2½2:5;3:5x2½4:5;5:5&y¼2 ¼ YO2;in
DYO2
where the concentration drop between the
successive flow channels shown in Figure 1, D
YO2 ; can be determined [29].
n  ðrfs Þjat x¼0;x¼6;y¼4&x2½0;0:5;x2½1:5;2:5;x2½3:5;4:5;x2½5:5;6&y¼2
n  ðrfs Þjat x2½0:5;1:5;x2½2:5;3:5;x2½4:5;5:5&y¼1;x¼0:5;1:5;2:5;3:5;4:5;5:5&y2½1;2
* At all the other boundaries, the symmetry
boundary condition is used for oxygen trans-
port:
n  ðrYO2 Þjat x2½0;6&y¼4;x¼0;6;&y2½2;4
* Darcy’s law governs the gaseous pressure
distribution in the gas diffusion and catalyst
layers. At the GDL/channel 1 interface, the gas
pressure is specified as equal to the cathode inlet
pressure:
Pg jat x2½0:5;1:5&y¼2 ¼ Pcathode
* At the GDL/channel 2 and 3 interfaces, the gas
pressure is known. Because of the channel
geometry, there exists a pressure drop between
Copyright # 2007 John Wiley & Sons, Ltd.
N. ZAMEL AND X. LI
the channels:
Pg jat x2½2:5;3:5&x2½4:5;5:5&y¼2 ¼ Pcathode
DPg
where DPg is the pressure drop between the
successive flow channels shown in Figure 1 and
it can be estimated [29].
* The symmetry boundary condition is used at all
the other boundaries for the gas pressure:
n  ðrPg Þjat ¼0
y¼4;x¼0;6&y2½2;4
* The solid potential due to electron transfer from
the anode to the cathode is specified at the
bipolar plate boundary. It should be pointed
out here that the cathode potential is specified
at the bipolar plate boundary:
fs jy¼0 ¼ fcathode
* At all the other boundaries, the insulation
boundary condition is used to solve for the
solid potential. It implies that electrons cannot
travel through that boundary; the correspond-
ing current density would then equal zero:
¼0
¼0
* The electrolyte potential due to the
proton transfer from the membrane to the
cathode is specified at the catalyst layer/
membrane interface. It is assumed to be
¼0 uniform along this boundary and to be equal
to 0 V:
fm jy¼4 ¼ 0
* At all the other boundaries, the insulation
boundary condition is used to solve for the
electrolyte potential. It implies that electrons
cannot travel through that boundary condition;
the corresponding current density would equal
zero:
n  ðrfm Þjat y¼2;x¼0;6&y2½2;4 ¼0
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS 705

* The mass fraction of water vapor is specified !gO

2
CO2
at the inlet of the three channels. The inlet ijy¼4 ¼ ð1
SÞi0ref ref
mass fraction can be calculated using the CO 2

inlet RH. The RH of the oxidant is constant   

aF
in the flow stream, while the concentration  exp ðfs
fm Þjy¼4
RT
of oxygen decreases along the channel flow  
field. Thus, the mass fraction of water vapor aF

exp
ðf
fm Þjy¼4
differs at the inlet of each channel and it is RT s
specified at the inlet of the three channels as
shown below: * The symmetry boundary condition is employed
at all the other boundaries for the liquid water
At the inlet of channel 1: YH2 O jat x2½0:5;1:5&y¼2 transport:
¼ YH2 O1 :
At the inlet of channel 2: YH2 O jat x2½2:5;3:5&y¼2
n  ðrSÞjat x¼0;6;&y2½2;4 ¼0
¼ YH2 O2 :
At the inlet of channel 3: YH2 O jat x2½4:5;5:5&y¼2
¼ YH2 O3 : 2.5. Parameters
The mass fraction at each channel is calcu- The electrolyte conductivity, sm ; is dependent
lated using the RH and the concentration of on the activity of water, a; and the hydration
oxygen as follows: of the membrane, l; inside the catalyst
  layer [28]:
* H2 O
M   
YH2 O ¼ xH2 O 1 1
M* mix sm ¼
100 exp 1268

303 T
P
where M* mix ¼ xi M * i ; xH2 O ¼ RHPsat =Pg and  ½0:00326
0:005139l ð27Þ
xN2 ¼ 1
xO2
xH2 O : The RH for channel 1 is
where
the same as for the cathode inlet.
* The symmetry boundary condition is used l ¼ 0:3 þ 6a½1
tanhða
0:5Þ
at all other boundaries for the water vapor   
pffiffiffi a
0:89
transport: þ 3:9 a 1 þ tanh ð28Þ
0:23
n  ðrYH2 O Þjat x2½0;6&y¼4;x¼0;6;&y2½2;4 ¼0 where
xH2 O P
a¼ ð29Þ
Psat
* The effective saturation is specified at the GDL/
channel 1; 2 and 3 interfaces. The electro-osmotic drag coefficient, nd ; of
water molecules through the electrolyte
membrane is also dependent on the water activity
Sjat x2½0:5;1:5&y¼2;x2½2:5;3:5x2½4:5;5:5&y¼2 ¼0
[30, 31]:
(
1:0 if l59
* At the catalyst layer/membrane interface, a nd ¼ ð30Þ
0:117l
0:053 if l59
liquid water flux is specified. This flux is due to
the electro-osmotic drag effect: The reference exchange current density, i0ref ;
in A cm
2 is obtained using the following correla-
tion [32]:
* H2 O
nd ijy¼4 M 4001
n  ðDc rSÞjy¼4 ¼ log10 ði0ref Þ ¼ 3:507
ð31Þ
Frl SECL T

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
706 N. ZAMEL AND X. LI

Table I. Operating and physical conditions.

Parameter Value Unit
Cathode inlet pressure (P) 5 atm
Operating temperature (T) 353 K
Cathode potential at the boundary of the bipolar plate (fcathode ) 0.3 V
Stoichiometry (Sc ) 2
Cell active surface area (A) 100 cm2
Channel width (Wd ) 1 mm
Channel height 1 mm
Gas diffusion layer width (WGDL ) 150 mm
Catalyst layer width (WCL ) 5 mm
Gas diffusion layer porosity (EGDL ) 0.5
Catalyst layer porosity (ECL ) 0.4
Electrolyte fraction in catalyst layer (em ) 0.15
Solid fraction in catalyst layer (es ) 0.3
Permeability (k) 5:5  10
11 m2
Electronic conductivity (ss ) 115 S m
1
Oxygen reference concentration (CO ref
2
) 3.38 mol m
3
Order of reaction (gO2 ) 1.0
Minimum saturation level (Sr ) 0.102
Maximum saturation level (Sm ) 0.924
Evaporation constant (E) 100 1 s
1
Condensation constant (w) 100 1 s
1
Displacement pressure (Pd )
2266.11 Pa
Pore size distribution (d) 7.2
Surface tension (s) 0.062 ðN m
1 Þ
Contact angle (yc ) 1008

Water diffusion through the electrolyte, Dl
m ; in 3. NUMERICAL PROCEDURES

m2 s
1 is found using the correlation [33]:
 

10 l 0:15
Dl
m ¼ 4:1  10
0:25
  
l
2:5
 1 þ tanh ð32Þ
1:4
The operating and physical conditions are
summarized in Table I. These properties are used
to solve the governing equations unless otherwise
stated. In the results section, a parametric study
will be presented; thus, some of the operating and
physical parameters will be varied to investigate
their effects. It should be pointed out here that the
cathode potential is specified at the bipolar plate
boundary. Thus, for the base condition, the
current density is measured in the ohmic polariza-
tion region and the polarization region varies from
ohmic to concentration once the parametric values
are varied.
The conservation equations are solved using
COMSOL MultiphysicsTM ; a commercial software
package based on finite element methods. To
accurately describe the electrochemical phenom-
ena, the element size in the catalyst layer is refined
1
to 10 of the catalyst layer thickness and the
resulting mesh has approximately 4  105 degrees
of freedom (DOF).
3.1. Integrating the math model into COMSOL
multiphysics
In order to solve the governing equations, the
built-in modules along with the coefficient PDE
have been used.
* The Stefan–Maxwell equation module is used to
solve for the species. The species specified are
oxygen, water vapor and nitrogen.

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS
This module exits in two domains out of the
three computational domains, GDL and cata-
lyst layer. It is set to inactive in the bipolar plate
domain since no gaseous species’ transport
occurs in the plate. It is important to choose
oxygen as the main species, followed by water
vapor and then nitrogen. Since nitrogen is not
involved in any reactions, its mass fraction is
computed using the knowledge that summation
of species’ mass fractions would equal one.
* The AC/DC module is employed to solve for
the solid and electrolyte potentials.
Two modules are needed to solve for the
potentials; one represents the solid potential
while the other represents the electrolyte poten-
tial. The solid potential applies to all the three
domains, while the electrolyte potential has to
be set as inactive in the bipolar plate and GDL
since protons do not travel through these two
domains.
* The Darcy equation found under the chemical
engineering module is used to solve for the gas
velocities.
The Darcy equation is applied to the gas
diffusion and catalyst layers, while it is set to
inactive in the bipolar plate domain.
* The coefficient PDE is used to define the mass
transfer of liquid water in the gas diffusion and
catalyst layers and is set to inactive in the
bipolar plate domain.
A numerical error occurred when solving this
equation due to the capillary diffusion coefficient.
The capillary diffusion coefficient is equal to zero
when the effective liquid saturation is zero (refer to
Equations (11) and (12)). The capillary diffusion
coefficient is dependent on the relative permeabil-
ity of liquid water in the layers defined by
Equation (10). A constant was used in order to
fix this numerical problem. The relative perme-
ability of liquid water was therefore changed to the
3 PEM fuel cells using COMSOL Multiphysics.
following form, kl ¼ kðS þffl}tÞ; where t is a
|fflfflffl{zfflffl
krl
constant used to ensure convergence when the
effective saturation is equal to zero. A value of 0.01
was first chosen for t to compute the saturation in
the layers. The value was then decreased to ensure
that it did not have any effect on the final result. It
Copyright # 2007 John Wiley & Sons, Ltd.
707
was seen that the solution becomes independent of
t with a value of 1  10
6 : A similar approach was
taken by Vynnycky [34].
3.2. Validation of model
The verification of the COMSOL Multiphysics is
used to verify the numerical model since the
governing equations are all solved using the
built-in modules. Thus, it is important to verify
that the software is correctly assigning and solving
the matrices. The verification is carried out as
follows:
1. First, each term in the built-in modules is
verified. Let us take the Stefan–Maxwell equa-
tion module as an example. In this module,
there exists a diffusion term, a convection term
and a source term. First the diffusion term is
turned on while the convection and source
terms are turned off. This is done to ensure
that the diffusion term is working properly.
It is seen that the transport of species is purely
due to diffusion and there exists no sink source
since the reaction (source) term does not exist at
this moment. Then, the convection term is
turned on while keeping the source term off.
For this test, two velocity gradients are used, a
zero velocity gradient and a velocity gradient of
1 ðm s
1 Þ: This test was done to ensure that the
convection term is working properly. Finally, a
sink source is added. A positive, negative and a
zero source terms are used to ensure that the
source term is working properly. Finally,
similar tests are done on the boundary condi-
tions to ensure that a proper solution is
obtained. All the tests mentioned above were
carried out for all the built-in modules used.
2. The second stage of verification was to find
some numerical models of a full PEM fuel cell,
which use COMSOL Multiphysics. Ziegler et al.
[35] created a model for planar self-breathing
They solved the transport of mass and heat in
the full cell (anode, cathode and membrane)
using the built-in modules in the software. They
were able to solve for the three kinds of
overpotential: activation, ohmic and con-
centration overpotential. Thus, they created a
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
708 N. ZAMEL AND X. LI

polarization curve and were able to validate

their curve with that of experimental
data. They found very close agreement between
the model and the experimental data.
Consequently, it can be said that COMSOL
Multiphysics is capable of solving the governing
equations of a PEM fuel cell system.
In addition, Wu [36] used COMSOL Multi-
physics to solve a two- and three-dimensional
models of a full PEM fuel cell. He solved
the mass and heat transfer in the cell.
He was then able to validate the resultant
polarization curve to that of experimental
data and found good agreement between
the numerical and the experimental results.
Again, this shows that COMSOL Multi- Figure 3. Error in the average current density vs the
physics is capable of solving a PEM fuel cell number of meshing elements.
system.
3. To ensure the proper formulation of solution
4. RESULTS AND DISCUSSION
matrices using the specified boundaries, bound-
ary conditions were imposed on multiple
A parametric study has been carried out to
boundaries.
investigate the effect of various operating and
design parameters on the cathode processes and
performance. The results are presented for the
3.3. Meshing and grid independency effect of the inlet RH in the air stream, the pressure
The mesh type used is a non-uniform, triangular drop between the successive flow channels, which
mesh. The element size in the gas diffusion and affects the strength of the cross flow between the
1 channels, the permeability of the cathode and the
catalyst layers is refined to 10 of the layer’s
thickness. At the channel inlet boundaries, the flow channel shoulder to channel width ratio that
mesh size is further decreased. This decrease is influences the relative importance of the mass and
necessary in order to ensure the convergence of the electron transport.
solution. In the bipolar plate, the solution is
independent of the mesh size since the thickness 4.1. Effect of inlet RH
of the plate is very large compared with that of the In this section, the effect of the inlet RH on the
gas diffusion and catalyst layers. In order to ensure overall performance of the cathode will be studied.
that the solution is grid independent, 88 332 In PEM fuel cells, the inlet RH is used to control
elements were used. Analysis was carried out in the amount of vapor water entering the cell; hence,
order to study the effect of decreasing element affecting the hydration of the membrane. Vapor
number on the solution. Figure 3 is constructed to water will condense once the RH in the gas
show the relation between the error and the diffusion and catalyst layers reaches 1 and in turn
number of meshing elements. The error was liquid water will evaporate when the RH inside the
calculated as follows: layers is less than 1. In addition, as seen in
jJ2N
Jj Equation (27), the electrolyte conductivity is
error ¼  100% ð33Þ directly influenced by the RH, in other words
J2N
water activity, in the catalyst layer. The electrolyte
where J2N is the current density when the number conductivity is important in determining the PEM
elements is doubled. fuel cell performance.

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS
In order to determine the effect of the inlet RH
on the performance of the cathode, six different
values for the inlet RH will be examined (100, 90,
80, 70, 60 and 50%). In addition, the comparison
will be carried out for the Leverette function and
the Brooks and Corey relation used to determine
the capillary pressure.
The effects of inlet RH on the current density
and the electrolyte conductivity are shown in
Figure 4. It is clear that the best performance is
reached with an inlet RH of 90%. However, this
performance does not correspond to the highest
electrolyte conductivity. It is also interesting to
Figure 4. Effect of the inlet relative humidity on (a) the
current density of the cell and (b) the electrolyte
conductivity}for the Leverette function and Brooks
and Corey relation.
Copyright # 2007 John Wiley & Sons, Ltd.
709
note here, that an inlet RH of 80% leads to a
slightly higher performance (current density) than
an inlet RH of 100% even though it corresponds
to a lower electrolyte conductivity. Further exam-
ination of the results obtained indicates that the
variation in current density with the inlet RH is
not very appreciable.
In order to be able to explain the behavior of
the current density versus the inlet RH, Figures 5
and 6 should be examined. The RH in the backing
and catalyst layers for different inlet RH values is
described in Figure 5, while Figure 6 shows
the effective liquid water saturation in the backing
and catalyst layers. The RH in the backing
and catalyst layers for an inlet RH of 100% is 1
everywhere; thus, its corresponding plot is not
shown. The RH in the backing layer for all the
values of inlet RH seems to vary between the inlet
RH and 1. Meanwhile, in the catalyst layer, the
RH seems to be closer to 1. This is so since the
evaporation term is much higher in the catalyst
layer due to the presence of more liquid water that
is readily available for evaporation. This helps in
explaining the high electrolyte conductivity with
an inlet RH of 80 and 90%. Also, as expected, the
saturation levels are higher with a higher inlet RH.
Similarly, the RH and the effective liquid satura-
tion in the backing and catalyst layers have been
examined and are shown in Figures 7 and 8. The
RH in the backing and catalyst layers correspond-
ing to an inlet RH of 100% is not presented since it
has a value of 100% everywhere.
From these results, the following observations
can be made:
* With the Brooks and Corey relation, effective
liquid water saturation in the backing
and catalyst layers is higher than that obtained
with the Leverette function. The Brooks
and Corey relation leads to a higher capillary
pressure; thus, the capillary diffusion coefficient,
Dc ; will be lower affecting the liquid water
removal.
* With an inlet RH of 90, 80, 70, 60 and 50%, the
corresponding electrolyte conductivity is higher
with the Brooks and Corey relation and the
performance is slightly higher. This occurs since
more liquid water is available for evaporation;
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
710 N. ZAMEL AND X. LI

Figure 5. Relative humidity in the backing and catalyst layers for the inlet relative humidity of (a) 50%; (b) 60%;
(c) 70%; (d) 80%; and (e) 90% using the Leverette function.

thus, the RH in the catalyst layer is increased. to be better. This is so since for both relations
However, for an inlet RH of 100%, the the RH in the catalyst layer is 1, but the
performance using the Leverette function seems saturation level is higher for the Brooks and

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS 711

Figure 6. Liquid water saturation in the backing and catalyst layers for the inlet relative humidity of (a) 50%; (b) 60%;
(c) 70%; (d) 80%; (e) 90%; and (f) 100% using the Leverette function.

Corey relation. Therefore, the available sites for
reaction are reduced.
* Finally, the overall trend of performance in the
catalyst layer using both relations seems to be
similar. The electrolyte conductivity has a
significant effect on the performance, but up
to a point.
At the end of this section, it is important to note
that the catalyst layer plays a very important role

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
712 N. ZAMEL AND X. LI

Figure 7. Relative humidity in the backing and catalyst layers for the inlet relative humidity of (a) 50%; (b) 60%;
(c) 70%; (d) 80%; and (e) 90% using Brooks and Corey relation.

in controlling the water concentration in the cell catalyst layer. This is expected since the produc-
both in vapor and liquid forms. The evaporation tion of water occurs in the catalyst layer; hence,
of liquid water seems to mostly occur in the depending on the thermodynamic properties of the

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS 713

Figure 8. Liquid water saturation in the backing and catalyst layers for the inlet relative humidity of (a) 50%; (b) 60%;
(c) 70%; (d) 80%; (e) 90%; and (f) 100% using the Brooks and Corey relation.

cell at the time of water production, water
will either evaporate or condense before leaving
the catalyst layer. This piece of information is
in agreement with the conclusions made by
Eikerling [37].
4.2. Effect of pressure drop
As mentioned earlier, a comprehensive study on
the cross-flow phenomenon in a PEM fuel cell with
a serpentine flow channel has been carried out by

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
714 N. ZAMEL AND X. LI
Park and Li [12]. Owing to the cross leakage
between neighboring channels with a serpentine
flow field, liquid water is ensured to be properly
removed from the cell. This section focuses on the
effect of the pressure drop on liquid water
removal. The cathode potential is 0:3 V; the
operating conditions are 808C and the permeabil-
ity of the porous media is 5:5  10
11 m2 : The
effective saturation level in the gas diffusion and
catalyst layers is shown in Figure 9 using the
Leverette function and the Brooks and Corey
relation. Liquid water accumulates in the areas
under the channel shoulder. Due to no pressure
drop, the fluid is not allowed to pass from one
channel to another unlike with a pressure drop as
shown in Figures 6(f) and 8(f). It is interesting to
notice that the highest saturation level in both
Figure 9. Liquid water saturation in the backing and
catalyst layers with no pressure drop: (a) Leverette
function and (b) Brooks and Corey relation.
Copyright # 2007 John Wiley & Sons, Ltd.
cases is the same. This is so since the permeability
of the backing layer plays a significant role in
obtaining this value and it is the same in both
cases. However, the distribution of the saturation
is quite different with and without the pressure
drop between the flow channels.
Figure 10 is a pictorial illustration of how
diffusion is the only means of water removal from
the cell with no pressure drop along the channel
flow field. Due to no pressure drop along the
channel flow field, the corresponding velocity
gradient for the gas is zero. Revisiting the
definition of the liquid water velocity, its depen-
dency on the gas phase velocity is obvious:
mg kl
ul ¼ ug
Dc rS ð34Þ
ml kg
where ul is the liquid phase velocity and ug is the
gas phase velocity.
Thus, the presence of cross flow, or in other
words a gas phase velocity, will aid in the removal
of liquid water from the cell as shown in Figure 11.
With the presence of cross flow, liquid water is
removed from the cell via capillary diffusion and
the gas pressure difference.
The corresponding current density in the cata-
lyst layer for this case (without cross flow) is
3833.06 and 3809:40 A m
2 using the Leverette
function and the Brooks and Corey relation,
respectively; a decrease of 2% compared with the
case with cross flow.
Finally, it should be pointed out here that even
though a high permeability value is used to
Figure 10. Plot of liquid water velocity streamlines with
no pressure drop along the channel flow field.
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS
Figure 11. Streamline plot of the liquid water velocity
with pressure drop along the flow channel field.
generate the results of this section, the lack
of a pressure drop along the flow field results in
a decrease in performance. Consequently, for a
better performance, a high-pressure drop accom-
panied with a high permeability should be
beneficial.
4.3. Effect of permeability
It is vital to consider the effects of permeability
since it is a measure of the ability of the material to
transport fluids. Therefore, it will be an important
measure of how liquid water is being transported
in the layers and of the cross-flow phenomenon. In
addition, the permeability plays a significant role
in determining the relative permeability of the gas
and liquid phases; hence, affecting the velocity of
both phases. Further, the permeability will have a
significant effect on the so-called capillary diffu-
sion coefficient, Dc ; described by Equation (12).
The capillary diffusion coefficient is a measure of
how fast liquid water will be transferred in the
backing and catalyst layers. The permeability
affects the capillary diffusion coefficient not just
directly through the relative permeability of liquid
water, but also through the capillary pressure. The
capillary pressure is expressed through the Lever-
ette function and the Brooks and Corey relation.
In this section, two permeability values will be
considered for comparison. The permeability
values used to generate the results of this subsec-
tion are 5:5  10
13 and 5:5  10
10 m2 : The inlet
RH is 100%, the pressure drop along the channel
Copyright # 2007 John Wiley & Sons, Ltd.
715
Figure 12. Liquid water saturation in the backing and
catalyst layers for the low permeability value of
5:5  10
13 m2 : (a) Leverette function and (b) Brooks
and Corey relation.
flow field is 349 Pa and the cathode potential is
0.3 V. With a low permeability value, the cross-
flow phenomenon is not observed even with a
pressure drop along the channel flow field as
shown in Figure 12. It is clear that there is water
build-up in the areas under the flow channel
shoulder between the channels. The only means of
liquid water removal is through diffusion.
With a high permeability value, the cross-flow
phenomenon is greatly observed in Figure 13. The
liquid water removal is much more improved and
the highest liquid water saturation is decreased.
With the low permeability value, the current
density in the catalyst layer is 3777.37 and 3610:6
1 A m
2 with the Leverette function and Brooks
and Corey relation, respectively. With the high
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
716 N. ZAMEL AND X. LI
Figure 13. Liquid water saturation in the backing and
catalyst layers for the high permeability value of
5:5  10
10 m2 : (a) Leverette function and (b) Brooks
and Corey relation.
permeability value, the current density in the
catalyst layer is 3942.79 and 3941:14 A m
2 with
the Leverette function and the Brooks and Corey
relation, respectively.
Finally, the x-direction velocity component will
be investigated in order to better comprehend the
effect of permeability on the cross flow or leakage
phenomenon in the fuel cell. The x component of
the velocity is an indicator of the flow rate between
the adjacent channels. In Figure 14, a comparison
of the x-velocity component for the three different
permeability values, 5:5  10
13 ; 5:5  10
11 and
5:5  10
10 m2 ; is shown. The velocity is plotted
against the normalized x-direction at the normal-
ized y value of 2.5 (the mid plane in the GDL).
It is obvious that the velocity increases with
an increase in permeability. At a permeability of
Copyright # 2007 John Wiley & Sons, Ltd.
Figure 14. x-Component of the velocity in m s
1 against
the normalized x-direction for the different permeability
values plotted at a normalized y value of 2.5, that is, at
the middle plane of the gas diffusion layer.
5:5  10
13 m2 the velocity is almost negligible.
Further investigation of the x-component of the
velocity illustrates the convection effects between
the channels. Moreover, comparison of the corre-
sponding velocity magnitudes for the permeability
values of 5:5  10
11 and 5:5  10
10 m2 shows
that the magnitude increase is proportional to the
increase in permeability.
4.4. Effect of shoulder/channel width ratio
The shoulder to channel ðS=CÞ width ratio is
another important parameter to be investigated.
The channel width controls the amount of oxidant
being supplied, while the shoulder width controls
the area available for electron transfer through the
bipolar plate. The ratio is defined as
Shoulder width Ws
¼ ð35Þ
Channel width Wd
where Ws is the width of the shoulder and Wd is
the width of the channel as shown schematically in
Figure 15.
In this section, five ratios are considered. The
results are summarized in Tables II and III. When
the shoulder to channel width ratio is decreased,
the channel area is increased and therefore, the
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS 717

Figure 15. Illustration of shoulder to channel width ratio.

Table II. Effect of shoulder/channel width ratio

on the performance of the cathode using the
Leverette function.
Ratio Current density ðA m
2 Þ
2:1 3764.36
1:2 4170.63
1:1 3899.81
3:1 3551.23
1:3 4194.06

Table III. Effect of shoulder/channel width ratio

on the performance of the cathode using the Brooks
and Corey relation.
Ratio Current density ðA m
2 Þ
2:1 3749.80
1:2 4145.31
1:1 3884.39
3:1 3539.12
1:3 4166.88

amount of oxidant being supplied to the cell is

increased.
The decrease in the width ratio results in an
increase in the total current density in the catalyst
layer. The gain in current density with the decrease
of the ratio is attributed to the increase of oxidant
flow into the cathode. On the other hand, when the
ratio is increased, the overall performance of the Figure 16. Total current density in the catalyst layer
ðA m
2 Þ along the normalized x-direction at different y
cathode decreases due to a lower gas flow rate. positions: (a) 3; (b) 3.5; and (c) 4; with the shoulder to
This is also illustrated in Figure 16. The total channel width ratio of 1:2 and 2:1 and the capillary
current density along the normalized cell length for pressure is represented by the Leverette function.

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
718 N. ZAMEL AND X. LI
different parts of the catalyst layer is shown. It is
apparent that with the decrease in the width ratio,
the overall current density increases and the
difference is most apparent around the catalyst
layer/membrane interface. Thus, to obtain higher
performance, shorter shoulder width is preferred.
However, it should be noted that shoulders are
needed for current collection. Decreasing the
shoulder/channel width ratio even further might
result in higher cell ohmic resistance and therefore
higher losses. This helps in explaining the small
difference in performance gain when the width
ratio is decreased from 1:2 to 1:3 shown in
Tables II and III.
Figure 17. Profile of liquid water saturation in the
backing and catalyst layers for the shoulder to channel
width ratio of (a) 2:1 and (b) 1:2 using the Leverette
function.
Copyright # 2007 John Wiley & Sons, Ltd.
The effect of the shoulder/channel width ratio
on the cross-flow phenomenon should also be
mentioned in this section. Figures 17(a) and (b)
show the liquid water saturation profile in the
backing and catalyst layers with an S=C width
ratio of 2:1 and 1:2, respectively. The cross flow
between two adjacent channels is decreased with
an increase in S=C width ratio. This is so since the
species (liquid water, vapor water and oxygen)
need to travel a longer distance from one channel
to reach the adjacent channel. However, when the
S=C ratio is 1:2, the distance between two adjacent
channels is decreased; therefore, the cross flow is
enhanced. This is also illustrated through the
velocity in the x-direction as seen in Figure 18. The
magnitude of the velocity is halved for the S=C
width ratio of 2:1. Therefore, the mass flow rate of
liquid water is reduced significantly affecting the
liquid water removal in the end.
Finally, studying the saturation profiles in
Figure 17 demonstrates that the decrease in
shoulder/channel width ratio results in better
removal of liquid water from the cathode. The
total liquid water saturation in the cathode with an
S=C width ratio of 1:2 is lower than that with an
S=C width ratio of 2:1. This is attributed to the
increased channel area; consequently, an increased
area for liquid water removal.
Figure 18. x-Component of the velocity in m s
1 against
the normalized x-direction for the shoulder to channel
width ratios of 1:2 and 2:1 plotted at a normalized y
value of 2.5, that is, the mid-plane of the gas diffusion
layer.
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS 719

5. CONCLUSIONS C ¼ molar concentration ðmol m
3 Þ

D ¼ mass diffusivity of species ðm2 s
1 Þ
A two-dimensional, multi-phase, multi-species, E ¼ evaporation constant ð1 s
1 Þ
isothermal and steady flow and transport in the F ¼ Faraday’s constant 96 487 ðC mol
1 Þ;
cathode of a polymer electrolyte membrane fuel force (N)
cell has been investigated numerically, including i0ref ¼ reference exchange current density
liquid water transport in the gas diffusion and ðA m
3 Þ
catalyst layers. A serpentine channel flow field is J ¼ phase current density ðA m
2 Þ
used in order to investigate the cross-flow phenom- M* ¼ molecular weight ðkg mol
1 Þ
ena between the neighboring channels. Conserva- n ¼ number of electrons transferred in the
tion equations are developed for the solid, gaseous half cell reaction
and liquid phases. The commercial software, nd ¼ electro-osmotic drag coefficient ðH2 O=
COMSOL Multiphysics, was used to solve the Hþ Þ
governing equations numerically. The parametric P ¼ pressure (Pa)
study carried out revealed that an inlet RH Pd ¼ displacement pressure
between 90 and 80% led to the highest perfor- q ¼ switch function; phase
mance in the cathode. Owing to liquid water R ¼ universal gas constant 8.314 ðJ mol
1
evaporation, the RH in the catalyst layer reaches K
1 Þ
100% with an inlet RH between 90 and 80%; R ¼ source term
resulting in a high electrolyte conductivity. The RH ¼ relative humidity
electrolyte conductivity plays a significant role in s ¼ liquid water saturation
determining the overall performance, but up to a S ¼ effective saturation; entropy ðJ K
1 Þ
point. Once the electrolyte conductivity reaches 7: Sm ¼ maximum saturation
7 S m
1 ; the liquid water saturation in the layers Sr ¼ residual (minimum, immobile) saturation
becomes the determining step in the performance T ¼ temperature ðKÞ
of the cathode. The analysis of the inlet RH effects u ¼ phase velocity ðm s
1 Þ
showed that the catalyst layer plays a very W ¼ width (m)
important role in controlling the water concentra- Y ¼ mass fraction
tion in the cathode both in vapor and liquid forms.
The evaporation of liquid water seems to mostly Greek letters
occur in the catalyst layer. The cross-flow phenom-
ena resulted in an increase in water removal from a ¼ transfer coefficient; volume fraction of
the areas under the shoulder. Further, a shoulder/ a phase
channel width ratio of 1:2 was shown to be an g ¼ reaction order
optimal ratio. A decrease in the shoulder/channel d ¼ pore size distribution
ratio resulted in an increase in performance and an e ¼ volume fraction
increase in cross flow. Finally, it was shown that E ¼ porosity
the Leverette function leads to lower liquid water Z ¼ overpotential (V)
saturations in the backing and catalyst layers than yc ¼ contact angle
the Brooks and Corey relation. The overall trend, k ¼ permeability ðm2 Þ
however, is similar for both functions. l ¼ membrane hydration
m ¼ viscosity ðPa s
1 Þ
$ ¼ correction factor
NOMENCLATURE r ¼ density ðkg m
3 Þ
s ¼ conductivity ðS m
1 Þ; surface tension
a ¼ water activity ðN m
1 Þ
Av ¼ density of active area in the catalyst f ¼ phase potential (V)
layer ðm2 m
3 Þ w ¼ condensation constant ð1 s
1 Þ

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
720 N. ZAMEL AND X. LI

Subscripts 7. You L, Liu H. A two-phase flow and transport model for

the cathode of PEM fuel cells. International Journal of Heat
and Mass Transfer 2002; 45:219–230.
a ¼ anode 8. Hu M, Zhu X, Wang M, Gu A, Yu L. Three dimensional,
act ¼ activation two phase flow mathematical model for PEM fuel cell: Part
c ¼ cathode; capillary I. Model development. Energy Conversion and Management
2004; 45:1861–1882.
CL ¼ catalyst layer 9. Hu M, Zhu X, Wang M, Gu A, Yu L. Three dimensional,
g ¼ gas phase two phase flow mathematical model for PEM fuel cell: Part
GDL ¼ gas diffusion layer II. Analysis and discussion of the internal transport
mechanisms. Energy Conversion and Management 2004;
H2 O ¼ water 45:1883–1916.
in ¼ inlet 10. Siegel NP, Ellis MW, Nelson DJ, von Spakovsky MR.
j ¼ the jth species A two-dimensional computational model of a PEMFC with
liquid water transport. Journal of Power Sources 2004;
l ¼ liquid phase 128:173–184.
m ¼ electrolyte membrane 11. Baschuk JJ, Li X. A general formulation for a mathema-
mix ¼ mixture tical PEM fuel cell model. Journal of Power Sources 2004;
142:134–153.
ohm ¼ ohmic 12. Park J, Li X. An experimental and numerical investigation
O2 ¼ oxygen on the cross flow through gas diffusion layer in a PEM
s ¼ solid phase fuel cell with a serpentine flow channel. Journal of Power
Sources 2006; 163:853–863.
sat ¼ saturation 13. Feser JP, Prasad AK, Advani SG. On the relative influence
of convection in serpentine flow fields of PEM fuel cells.
Superscripts Journal of Power Sources 2006; 161:404–412.
14. Kanezaki T, Li X, Baschuk JJ. Cross-leakage flow between
adjacent flow channels in PEM fuel cells. Journal of Power
eff ¼ effective value Sources 2006; 162:415–425.
ref ¼ reference state 15. Li X, Sabir I, Park J. A flow channel design procedure for
PEM fuel cells with effective water removal. Journal of
Power Sources 2007; 163:933–942.
16. Yu HM, Ziegler C, Oszcipok M, Zobel M, Hebling C.
Hydrophilicity and hydrophobicity study of catalyst layers
in proton exchange membrane fuel cells. Electrochimica
ACKNOWLEDGEMENTS Acta 2005; 51:1199–1207.
17. Ihonen J, Mikkola M, Lindbergh G. Flooding of gas
The financial support by the Natural Sciences and diffusion backing in PEFCs: physical and electrochemical
Engineering Research Council (NSERC) of Canada is characterization. Journal of Electrochemical Society 2004;
151:1152–1161.
greatly appreciated.
18. Williams MV, Begg E, Bonville L, Kunz HR, Fenton JM.
Characterization of gas diffusion layers for PEMFC.
Journal of Electrochemical Society 2004: 151:1173–1180.
19. Pasaogullari U, Wang CY. Liquid water transport in gas
REFERENCES
diffusion layer of polymer electrolyte fuel cells. Journal of
1. He W, Yi JS, Nguyen TV. Two-phase flow model of the Electrochemical Society 2004; 151:399–406.
cathode of PEM fuel cells using interdigitated flow fields. 20. Lim C, Wang CY. Effects of hydrophobic polymer content
AICHE Journal 2000; 46:2053–2064. in GDL on power performance of a PEM fuel cell.
2. Natarajan D, Nguyen TV. A two-dimensional, two-phase, Electrochimica Acta 2004; 49:4149–4156.
multicomponent, transient model for the cathode of a 21. Wang C. Multiphase flow and heat transfer in porous
proton exchange membrane fuel cell. Journal of Electro- media. Advances in Heat Transfer 1997; 30:93–196.
chemical Society 2001; 148:A1324–A1335. 22. Vafai K, Tien C. Boundary and inertia effects on flow and
3. Wang ZH, Wang CY, Chen KS. Two phase flow and heat transfer in porous media. Journal of Heat and Mass
transport in the air cathode of proton exchange membrane Transfer 1981; 24:195–203.
fuel cells. Journal of Power Sources 2001; 94:40–50. 23. Adamson A. Physical Chemistry of Surfaces (5th edn).
4. Leverette MC. Capillary behavior in porous solids. Wiley: New York, 1990.
Transactions of AIME 1941; 142:151–169. 24. Ustohal P, Stauffer F, Dracos T. Measurement and
5. Meng H, Wang CY. Model of two-phase flow and flooding modeling of hydraulic characteristics of unsaturated porous
dynamics in polymer electrolyte fuel cells. Journal of the media with mixed wettability. Journal of Contaminant
Electrochemical Society 2005; 152:A1733–A1741. Hydrology 1998; 33:5–37.
6. Senn SM, Poulikakos D. Multiphase transport phenomena 25. Weber AZ, Darling RM, Newman J. Modeling two phase
in the diffusion zone of a PEM fuel cell. Journal of Heat behavior in PEFC’s. Journal of the Electrochemical Society
Transfer 2005; 127:1245–1259. 2004; 151:A1715–A1727.

Copyright # 2007 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er
 THE CATHODE OF PEM FUEL CELLS
26. Demond A, Roberts P. Effect of interfacial forces on two-
phase capillary pressure saturation relationships. Water
Resources Research 1991; 27:423–437.
27. Corey AT. Mechanics of Heterogenous Fluids in Porous
Media. Water Resources Publications: Fort Collins, 1977.
28. Springer TE, Zawodinski TA, Gottesfeld S. Polymer
electrolyte fuel cell model. Journal of Electrochemical
Society 1991; 138:2334–2342.
29. Zamel N, Li X. Numerical investigation of transport
phenomena and electrochemical reactions in PEM fuel cell
cathode. International Journal of Low Carbon Technology
2007; 2:126–148.
30. Zawodzinski TA, Davey J, Valerio J, Gottesfeld S. The
water content dependence of electro-osmotic drag in
proton-conducting polymer electrolytes. Electrochimica
Acta 1995; 40:297–302.
31. Büchi FN, Scherer GC. Investigation of the transversal
water profile in nafion membranes in polymer electrolyte
fuel cells. Journal of the Electrochemical Society 2001;
148:A183–A188.
32. Parthasarathy A, Srinivasan S, Appleby J. Temperature
dependence of the electrode kinetics of oxygen reduction at
Copyright # 2007 John Wiley & Sons, Ltd.
721
the platinum/nafion interface}a microelectrode investiga-
tion. Journal of Electrochemical Society 1992; 139:
2530–2537.
33. Motupally S, Becker JA, Weidner JW. Diffusion of water in
Nafion 115 membranes. Journal of the Electrochemical
Society 2000; 147:3171–3177.
34. Vynnycky M. On the modelling of two-phase flow in
the cathode gas diffusion layer of a polymer electrolyte
fuel cell. Applied Mathematics and Computation 2007;
189:1560–1575.
35. Ziegler C, Tranitz M, Schumacher JO. A model for planar
self-breathing proton exchange membrane fuel cells in
FEMLAB. Proceedings of COMSOL Multiphysics Con-
ference, Cambridge, MA, 2005.
36. Wu H. Computational analysis on proton exchange
membrane fuel cell technology. Master’s Thesis, Depart-
ment of Mechanical Engineering, Ryerson University,
2005.
37. Eikerling M. Water management in cathode catalyst
layers of PEM fuel cells: a structure-based model.
Journal of the Electrochemical Society 2006; 153:
E58–E70.
Int. J. Energy Res. 2008; 32:698–721
DOI: 10.1002/er