Journal of Power Sources 182 (2008) 197–222

Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

A general model of proton exchange membrane fuel cell

Anh Dinh Le, Biao Zhou ∗
Department of Mechanical, Automotive & Materials Engineering, University of Windsor, Ont., Canada N9B 3P4

a r t i c l e i n f o a b s t r a c t

Article history: In this study, a general model of proton exchange membrane fuel cell (PEMFC) was constructed, imple-
Received 18 January 2008 mented and employed to simulate the fluid flow, heat transfer, species transport, electrochemical reaction,
Received in revised form 20 March 2008 and current density distribution, especially focusing on liquid water effects on PEMFC performance. The
Accepted 20 March 2008
model is a three-dimensional and unsteady one with detailed thermo-electrochemistry, multi-species,
Available online 27 March 2008
and two-phase interaction with explicit gas–liquid interface tracking by using the volume-of-fluid (VOF)
method. The general model was implemented into the commercial computational fluid dynamics (CFD)
Keywords:
software package FLUENT® v6.2, with its user-defined functions (UDFs). A complete PEMFC was consid-
PEM fuel cell
UDF
ered, including membrane, gas diffusion layers (GDLs), catalyst layers, gas flow channels, and current
FLUENT® collectors. The effects of liquid water on PEMFC with serpentine channels were investigated. The results
3D showed that this general model of PEMFC can be a very useful tool for the optimization of practical
Multi-phase engineering designs of PEMFC.
VOF © 2008 Elsevier B.V. All rights reserved.

1. Introduction mass transport, thus decreasing fuel cell performance. However,

due to the special chemical structure of the proton exchange mem-
In recent years, high-performance computing and advanced brane (PEM), the membrane must be well hydrated to ensure that
numerical algorithms have allowed researchers to model proton a sufficient amount of hydrogen ions could cross. Water content
exchange membrane fuel cell (PEMFC) systems as well as individual is also an important factor that affects the Ohmic resistance in
components with greater fidelity than ever before. In gen- the membrane. Therefore, keeping an appropriate amount of water
eral, PEMFC operations involve simultaneously multi-component, content in the fuel cell to avoid both membrane dehydration and
multi-phase, multi-dimensional fluid flow with heat and mass water vapor condensation has been a critical issue in improving
transfer and electrochemical reactions. Hence, a complete math- fuel cell performance. In reality, however, it is almost impossible
ematical model is necessary to characterize more fully the physical to manage water on both the anode and cathode sides without
behavior, to aid our understanding of complex phenomena occur- dehydration and condensation; this is simply because water vapor
ring in a fuel cell system and to provide powerful tools for fuel cell condensation in the gas flow channels of practical fuel cell appli-
design and optimization. cations is unavoidable. Therefore, water management, especially
Fuel cell performance is evaluated by the relation between the liquid water management, to which many engineers and scientists
cell voltage and current density. The voltage loss is mainly caused have recently paid particular attention, has been a critical challenge
by the activation loss due to electrochemical reactions, the Ohmic for a high-performance fuel cell design and optimization.
loss due to resistance and fuel crossover and the mass transport In the last decade, water management-related studies were per-
loss due to limitation of gas transport inside the cell. During fuel formed numerically and experimentally for different purposes and
cell operation, by-product water is generated in the cathode cata- in several ways. Recently, three-dimensional models based on the
lyst layer in the form of liquid, leading to a gas–liquid flow in the computational fluid dynamics (CFD) approach have been presented
porous media and the channels. On the other hand, due to the low by different commercial CFD software packages such as FLUENT® ,
operating temperature of PEMFCs (30–100 ◦ C), excessive humid- CFX® , Star-CD® , and CFDRC® . A CFD modeling of PEMFCs which
ification could result in water vapor condensation. These could simultaneously considered the electrochemical kinetics, current
subsequently block the gas flow channels resulting in a lower air- distributions, hydrodynamics, and multi-component transport was
flow rate on the cathode side, increasing the voltage loss due to conducted by Um et al. [1]. A three-dimensional numerical simu-
lation of a straight gas flow channel in a PEMFC was performed by
Dutta et al. [2] using a commercial CFD software FLUENT® . Hon-
∗ Corresponding author. Tel.: +1 519 253 3000x2630; fax: +1 519 973 7007. tanon et al. [3] also employed FLUENT® to implement their 3D,
E-mail address: [email protected] (B. Zhou). stationary gas flow model. A study exploring the steady-state gas

0378-7753/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jpowsour.2008.03.047
198 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Nomenclature
e electrochemical reaction
a water activity eff effective
As heat transfer surface area, m2 g gas phase
Asurf reactive surface area, m2 H2 hydrogen
cf concentration of sulfonic acid ions (HSO3 − ), H2 O water
kmol m−3 i species i
cp specific heat capacity, J kg−1 K−1 in inlet
Ci species concentration i, kmol m−3 l liquid phase
Di diffusion coefficient of species i in gas mixture, m membrane phase
m2 s−1 out outlet
F Faraday constant, 9.6487 × 107 C kmol−1 N2 nitrogen
h convective heat transfer, W m−2 K−1 O2 oxygen
hH2 O the enthalpy of formation of water vapor, N m kg−1 ref reference
I current density, A m−2 s solid phase
Iave average current density, A m−2 sat saturated
J mass flux, kg m−2 s−1 surf surface
keff effective thermal conductivity, W m−1 K−1 w water vapor
Mi molecular weight of species i in gas mixture,
kg kmol−1
nd electro-osmotic drag coefficient transport phenomena in micro-scale parallel flow channels was
nf charge number of the sulfonic acid ion conducted by Cha et al. [4] in which oxygen concentration along
P pressure, Pa a single gas flow channel and other flow patterns that may affect
Pi partial pressure of species i, Pa fuel cell performance were discussed. Similarly, gas concentration
R universal gas constant, 8314 J kmol−1 K−1 of a steady-state flow along fuel cell flow channels was obtained
Q̇ heat rate, W numerically by Kulikovsky [5]. However, in all the studies men-
Rcat , Ran volumetric current density, A m−3 tioned above, the effects of liquid water were neglected. Yi et al. [6]
S source term pointed out that water vapor condensation was inevitable on both
t time, s the anode and cathode sides of a PEMFC, and they discussed a liquid
T temperature, K water removal technique that used a water transport plate to lead
u, v, w velocities in X, Y, and Z directions, respectively, m s−1 excess liquid water to the coolant flow channels by a pressure dif-
Voc open-circuit potential, V ference. Wang et al. [7] conducted a two-phase model on PEMFC
Vcell cell potential, V cathode to address the liquid water saturation. You and Liu [8]
Vref reference potential, V also considered liquid water saturation in a straight channel on the
volume, m3 cathode side. Both the Refs. [7,8] showed the importance for consid-
Xi mole fraction of species i ering liquid water in numerical modeling of PEMFCs. A 3D model
Yi mass fraction of species i by Natarajan and Nguyen [9] introduced the effects of flow rate,
inlet humidity and temperature on the liquid water flooding in the
Greek symbols
PEMFC cathode. The numerical results of this model were validated
˛ transfer coefficient
with experimental data. Berning and Djilali [10] presented a multi-
ˇ the factor accounts for energy release
phase, multi-component 3D model with heat and mass transfer,
 concentration dependence
where liquid water transport in GDLs was numerically modeled by
 p,  T exponent factors
using viscous and capillary effects. This method was also used in
ε porosity
a 3D model by Mazumder and Cole [11]. Other approach to deal
w contact angle, ◦
with two-phase flow in GDLs based on thermodynamic equilib-
 overpotential, V
rium conditions was given by Vynnycky and Vynnycky [12]. In this
 surface curvature
approach, the location of interface between one phase and the other
p hydraulic permeability, m2
phase was pointed out from its numerical results. In recent years,
 electrokinetic permeability, m2
more two-phase models have been published [13–15], these sim-
 water content
ulations predicted water flooding inside PEMFCs, and the liquid
dynamic viscosity, kg m−2 s−1
water effects on PEMFC performances. Large-scale simulations for
density of gas mixture, kg m−3
complex flow field were also performed with experimental valida-
i density of species i, kg m−3
tions [15–18].
phase conductivity, −1 m−1
All the above research papers have significantly contributed to
gaseous permeability, m2
PEMFC research and development. However, by far, to the authors’
 reaction rate, kmol m2 s−1
knowledge, most of the two-phase numerical models have not con-
 phase potential, V
sidered the interface tracking between liquid water and gas. The
ϕ relative water content
detailed behaviors of liquid water transport inside the PEMFCs were
 surface tension coefficient, N m−1
rarely discussed except for the present authors’ previous study [19],
ω excess coefficient
which only dealt with part of serpentine channels—the single U-
shaped channel. Recently, Zhou et al. [20–22] also conducted two
Subscripts and superscripts
more studies that dealt with liquid water in serpentine and straight
an anode
parallel fuel cell stacks. The results showed that different designs of
cat cathode
gas diffusion layers (GDLs) and flow channels will affect the liquid
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 199

water flow patterns significantly, thus influencing the performance

of PEMFCs. Djilali and co-workers [23] presented a 2D, numerical
investigation of the dynamic behavior of liquid water entering a
PEMFC channel through a small hole that was assumed as a GDL
pore. In these studies, the water behaviors can give very useful
insights about water management only in the cathode channels
although the electrochemical reactions were not considered.
Numerical models that considered liquid water in the porous
mediums such as GDL were developed in several ways. Nam and
Kaviany [24] developed a two-dimensional, two-phase numeri-
cal model by considering random carbon fiber mats as the GDL.
Single- and two-layer diffusion mediums were both considered to
investigate the effective diffusivity and water saturation. Two main
factors influencing on two-phase transport in GDL of a PEMFC such
as permeability and capillary pressure were also determined via
a capillary network model developed by Djilali and co-workers
[25]. In addition, the GDL considered as a network of pore bod-
ies and pore throats used to address the liquid water behavior in
PEMFC were presented by Fowler and co-workers [26] or Sinha
and Wang [27]. A study on the interaction between the GDL and
the flow field was performed by Dohle et al. [28] numerically and
experimentally. Other models that considered the porous medi-
ums also mainly focused on the porosity of the carbon fiber paper
that could influence the performances of PEMFCs [29,30]. How-
ever, the detailed flow patterns that liquid water exhibits across
the porous medium of GDL was rarely discussed. Based on their
previous efforts on water behaviors inside cathode channels, the
present author’s group further investigated water behaviors inside
innovative GDLs [31,32].
Based on the above literature reviews, it is evident that there
is an urgent need to develop a general mathematical model Fig. 1. (a) Schematic diagram of PEMFC and (b) computation domain.
with all detailed physics included, e.g., multi-phase with VOF
interface-tracking between gas-phase and liquid–water-phase,
multi-components, heat and mass transfer, electrochemical reac- Table 1
tions, and water-phase-change effects. In this paper, a general Physical properties and parameters
model is presented with all parts of PEMFC including membrane, Physical properties or parameters Value
catalyst layers, GDLs, gas flow channels and current collectors.
Total channel length 0.115 m
Channel width 0.001 m
2. Mathematical model Channel height 0.001 m
Membrane thickness 50 × 10−6 m
GDL thickness 300 × 10−6 m
Fig. 1 schematically shows a disassembling diagram of a PEMFC Catalyst layer thickness 10 × 10−6 m
comprising of two distinct current collectors with flow channels Anode inlet pressure, Pan,in 2 atm
formed on them at both the cathode and anode; an MEA consists Cathode inlet pressure, Pcat,in 2 atm
of a proton exchange membrane sandwiched between two catalyst Anode and cathode inlet temperature 300 K
Anode inlet excess coefficient, ωa 3
layers. Between the current collectors and the MEA, there are two
Cathode inlet excess coefficient, ωc 3
GDLs on both sides. For the sake of simplicity and popularity, the Relative humidity of anode inlet 20%
flow channels are in serpentine form. In this paper, the mathemati- Relative humidity of cathode inlet 100%
cal model considers all the parts shown in Fig. 1a, the computation Open-circuit voltage, Voc 1.15 V
Faraday constant, F 96,487,000 C kmol−1
domain is shown in Fig. 1b, with geometrical parameters listed in
Gas constant, R 8314 J kmol−1 K−1
Table 1. The computation domain consists of current collectors, flow Anode volumetric reference exchange current 7 × 1010 A kmol−1
channels, GDLs, catalyst layers and membrane. The channels are density/reference hydrogen concentration,
ref /(C ref )an
Ran
serpentine, total path length for a channel of 115 mm with 1 mm H
2
(width) × 1 mm (height) in the cross-section flow area. The species Cathode volumetric reference exchange current 7 × 105 A kmol−1
considered are hydrogen, oxygen, water vapor and nitrogen. density/reference oxygen concentration,
ref /(C ref )cat
Rcat O
2
Anode transfer coefficient, ˛an 0.5
2.1. Model assumptions Cathode transfer coefficient, ˛cat 0.5
Anode concentration dependence,  an 0.5
Cathode concentration dependence,  cat 1.0
The assumptions used in developing the model are as follows:
Factor accounts for energy release, ˇ 0.5
Membrane porosity, εmem 0.5
1. Ideal gas law was employed for gaseous species. Diffusion layer porosity, εGDL 0.5
Catalyst layer porosity, εcatalyst 0.5
2. The fluid flow in the fuel cell was laminar due to the low flow
Permeability of porous media, 1.76 × 10−11 m2
velocities and the small size of gas flow channels. Contact angle,  w 90◦
3. The porous media including membrane, catalyst layers and GDLs Surface tension,  0.065 N m−1
were considered to be isotropic.
200 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

4. The fuel cell cooling was controlled by forced convection heat Table 2
The source terms of governing equations
transfer.
Governing equation Volumetric source terms and location of
2.2. Governing equations application

Conservation of mass For all parts: Sm = 0

2.2.1. Mass conservation equation
The continuity equation is expressed as follows: Volume fraction  Ss = rw , rw =
For all parts: 
cr max (1 − sl ) PwvRT
−Psat
MH2 O , −sl l
∂ε
+ ∇ · (ε v ) = Sm (1)
∂t Conservation of momentum For gas channels: Sv = 0
For multi-phase flow, the mixture density is defined as follows For GDLs and void of catalyst layers:
2 ∇s
[33,34]: Sv = g − ε2 v +  ( + l )
l g

For membrane:
= sl l + (1 − sl ) g (2) 2 ∇s 
Sv = g − ε2 v +  ( + l ) + p cf nf F∇m
l g
sl + sg = 1 (3)
I2
Conservation of energy For current collectors: ST = s

2.2.2. Momentum conservation equation For gas flow channel: ST = 0

The momentum equation is dependent on the volume fractions I2
For GDL: ST = eff
+ rw hL
of all phases: s
I2
For membrane: ST =
∂ eff
m
(ε v ) + ∇(ε v v ) = −ε∇p + ∇[ε ∇ v ] + Sv (4) For catalyst layer:
∂t 
1 1
ST = Ran,cat + I 2 + + rw hL
where = sl l + (1 − sl ) g (5) eff
s
eff
m

MH
For a mixture of liquid and gas phases, the local mass average Hydrogen transport For anode catalyst layer: SH2 = − 2F
2
Ran
velocity v is defined as [33,34]:
MO
sl For cathode catalyst layer: SO2 = − Rcat
l vl+ sg g vg
2
Oxygen transport 4F
v= (6)
sl l + sg g
Water vapor transport For anodecatalyst layer:

The source term for the momentum equation used in the model nd M H O
SH2 O = − F
2
Ran − rw
describes the flow of the fluid through a porous media by using
For cathode catalystlayer: 
Darcy’s drag force. The gravity and surface tension forces were also MH O n d MH O
considered in the momentum equation source term. The source SH2 O = 2
2F
Rcat + F
2
Rcat − rw

terms for different regions of the fuel cell are given in Table 2.
Conservation of Charge For anode catalyst layer:
Ses = −Ran ; Sem = Ran
2.2.3. Volume fraction equation
For cathode catalyst layer:
The tracking of the interface between the phases was accom-
Ses = Rcat ; Sem = −Rcat
plished by the solution of a continuity equation for the volume
fraction of one (or more) of the phases. The motion of the interface For other parts: Ses = 0; Sem = 0

between two immiscible liquids (namely, gas mixture and liquid

water) of different density and viscosity in the VOF method was
defined by volume fraction of liquid water sl and volume fraction of water, Ohmic heating caused by the Ohmic resistance of solid
of gaseous phase sg [33]. phases and the heat due to phase change. For different zones, the
For volume fraction of liquid water: source terms are expressed as in Table 2.

∂
(εsl l )
l ) + ∇(sl l v = Ss (7) 2.2.5. Species transport equations
∂t
The model predicts the local mass fraction of each species, Yi ,
2.2.4. Energy conservation equation through the solution of a convection–diffusion equation for the
In this model, the energy balance in terms of temperature ith species. The species transport equations are generally in the
change was also considered. In the multi-phase model, the energy following form:
equation is also shared among the phases [33,35]:
∂
(ε Yi ) + ∇ · (ε v Yi ) = Di,m ∇ 2 ( Yi ) + Si (12)
∂T ∂t
( cp )eff + ( cp )eff (v ∇T ) = ∇(keff ∇T ) + ST (8)
∂t
where Di,m is the diffusion coefficient for species i in the mixture
where T is the temperature (K) that is defined as [34].
sl l Tl + (1 − sl ) g Tg  P p  T T
T= (9) Di,m = ε1.5 (1 − sl )rs Di,m
0 0
(13)
sl l + (1 − sl ) g P T0
To specify these parameters in porous media, the effective prop- 0
where Di,m is the diffusion coefficient for species i in the mix-
erties were determined:
ture at reference temperature and pressure. The source term in the
( cp )eff = ε f cp,f + (1 − ε) s cp,s (10) transport equation is shown in Table 2.
During the operation, the H+ move from the anode to the cath-
keff = εks + (1 − ε)kf (11)
ode and also pull water molecules with them, this is known as
The source term of energy equation for two-phase model may the electro-osmotic drag effect. Physically, the water transport rate
consist of heat from electrochemical reactions, heat of formation through the membrane from anode to cathode by electro-osmotic
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 201

drag is computed as 2.3. Water transport and its effect on the properties of porous
media
nd MH2 O
ṁH2 O = Rcat (14)
F
Water is formed in the cathode catalyst layer by electrochem-
where nd is the drag coefficient and is proposed by Springer et al. ical reactions—an amount of oxygen is consumed and an amount
[36] for Nafion membrane: of water is produced. Due to the proton movement from the anode
2.5 to the cathode through the membrane, water molecules are pulled
nd =  (15) with the protons by a force called electro-osmotic drag [30]. Addi-
22
tionally, water may diffuse through the membrane due to the
where  is the water content inside the polymer membrane [36]. concentration differences.
In the present study, the membrane water diffusivity and mem-
2.2.6. Conservation of charge brane ionic conductivity were calculated by the expressions given
The current transport of electrons through the solid phase and by Springer et al. [36]. It is almost a standard model that has been
ions through the membrane phase was represented by the follow- well explained in the recent textbook [37].
ing equations [33,35]:

∇ ·( 2.4. Volume-of-fluid (VOF) model

s ∇s ) = Ses (16)

∇ ·( m ∇m ) = Sem (17) The VOF technique was implemented in the channels and porous
media (including GDLs and catalyst layers) [33]. Interface between
The volumetric source terms Ses and Sem are defined as volu-
gas and liquid (two-phase flow) is tracked by the volume fraction of
metric transfer currents. The source term in the equation of charge
liquid water in the computational cell volume. In the VOF approach,
is shown in Table 2.
the source term of continuity and momentum equations used in
The source terms representing transfer current were calculated
porous media include the effects of surface tension and wall adhe-
by using Butler–Volmer equation [33]:
sion and capillary water transport phenomenon, therefore, was also
 an
CH2
  ˛ F   ˛ F  considered.
ref an an cat cat
Ran = Ran exp − exp − (18)
ref
CH RT RT
2 2.4.1. Geometric reconstruction scheme
 cat The geometric reconstruction PLIC scheme (piecewise linear
CO2
  ˛ F   ˛ F 
ref
Rcat = Rcat exp −
cat cat
− exp
an an
(19) interface construction) was employed because of its accuracy and
COref RT RT applicability for general unstructured meshes, compared to other
2
methods such as the donor–acceptor, Euler explicit, and implicit
The relation between species concentration and species mass schemes. A VOF geometric reconstruction scheme is divided into
fraction is two parts: a reconstruction step and a propagation step. Details
Yi can be found in Refs. [33,38].
Ci = (20)
Mi
2.4.2. Implementation of surface tension
The volumetric transfer current R is driven by the activation
The addition of surface tension to the VOF method is modeled
overpotential , which is the potential difference between solid and
by a source term in the momentum equation. The pressure drop
membrane phases [33,35]:
across the surface depends upon the surface tension coefficient 
 = s − m − Vref (21) [33,38]:
1 1

The reference potential of the electrode Vref is 0 on the anode p =  + (26)
side and is equal to the open circuit voltage on the cathode side: R1 R2

Vref = 0, on anode side; Vref = Voc , on cathode side (22) where R1 and R2 are the two radii, in orthogonal directions, to
measure the surface curvature.
The cell potential is then the difference between the cathode The surface tension can be written in terms of the pressure jump
and anode solid phase at the terminal collectors (two ends of the across the interface, which is expressed as a volumetric force F
cell collectors which are connected to the external electric circuit added to the momentum equation:
from the both electrodes).
∇sl
F =  (27)
Vcell = s,cat − s,an (23) ( l + g )/2

In order to satisfy the conservation of charge, the total current where  denotes the surface tension coefficient and  is the surface
of either electrons or protons coming out from the anode catalyst curvature. The source terms for different regions of the fuel cell are
layer must be equal to the total current coming into the cathode given in Table 2.
catalyst layer and must be equal to the total current caused by the The surface curvature  in Eq. (27) can be defined in terms of the
proton movement through the membrane [35]: divergence of the normal unit vector of the interface n̂:

 = ∇ · n̂ = ∇ · (n̂w cos w + t̂w sin w ) (28)

(24)
where n̂ is the unit vector normal to the interface between two
The average current density is the total current generated in the phases near the walls, n̂w is the unit vector normal to the walls, t̂w
fuel cell divided by the geometric area: is the unit vector tangential to the walls, and w is the static contact
angle at the walls shown in Fig. 2. For the electrode surfaces with
different wettabilities, different static contact angles were assigned,
(25) and different contact angles could result in different surface ten-
202 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Iave Asurf
UO2 ,in = ωO2 MO2 (32)
4F O2 Ain

2.6.2. Outlet of flow channels

The fully developed condition was assumed to be applied for
velocity field and species concentrations.

Fig. 2. Static contact angle at the walls.

2.6.3. The terminal collectors
There were two terminal collectors in which the anode and cath-
sions (F), thus influencing on water transport. The values of surface ode collectors were connected to the external electric circuit. It was
tension and contact angle used in this model are listed in Table 1. necessary to determine the boundary conditions for phase poten-
tials on these interfaces.
2.5. Mass transfer

2.6.4. External boundaries

Liquid water is considered to appear in the fuel cell when the
External boundaries are defined as all outside surfaces of PEMFC
water vapor pressure reaches its saturated value at the cell oper-
except the terminals (the outside surfaces of current collectors
ating temperature. Contrarily, liquid water is evaporated when the
where the current entry to or departure from). Zero-current-flux
water vapor pressure is smaller than its saturated value. Water con-
condition was applied for the external boundaries due to no cur-
densation and evaporation phase change is also an important factor
rents coming or leaving.
to determine the presence of liquid water in multi-phase model. In
For the sake of simplicity, the cooling channel was not included
the present study, the variation of water mass flow rate was due to
in the present model. However, it is necessary to control the cell
condensation (or evaporation) as described here [33]:
temperature. Consequently, the heat transfer was assumed to take
MH2 O (Pw − Psat ) place between the all outside surfaces and the ambient environ-
ṁ = rw (29)
RT ment by forced convection as shown:

2.6. Boundary conditions Q̇convection = hAs (Ts − T∞ ) (W) (33)

To close the equation systems including conservation equa- where h is the convection heat transfer coefficient (W m−2 K−1 ), As
tions of continuity, momentum, energy, species, and charge with is heat transfer surface area (m2 ), Ts and T∞ are the temperatures
unknowns: u, v, w, p, T, Yi and , the boundary conditions are of the surfaces and the free stream, respectively.
required. The summary of boundary conditions is listed in Table 3. Noted, heat is produced when the fuel cell operates. For the case
Below a brief description is provided. in which water finally ends in vapor form, this heat generation rate
is defined as [39]:
2.6.1. Inlet of flow channels Q̇generated = I(1.25 − Vcell ) (W) (34)
Inlet velocities, fuel and oxidant temperatures and mass concen-
tration of species were set as given parameters listed in Table 3. Inlet In order to ensure that the forced convection dissipates the heat
velocities can be preliminarily calculated by the inlet flow rates converted from electricity, the following condition must be satis-
based on the average current density (Iavg ) and excess coefficient ω fied:
that is defined as
mH2 ,supply mO2 ,supply Q̇convection = Q̇generated (35)
ωH2 = ; ωO2 = (30)
mH2 ,consumption mO2 ,consumption then the convective coefficient h was chosen as
Then, inlet velocity expression is given by the following equations: I(1.25 − Vcell )
h= (36)
Iave Asurf As (Ts − T∞ )
UH2 ,in = ωH2 MH2 (31)
2F H2 Ain
2.7. Solution procedure
Table 3
Boundary conditions
The above-coupled set of governing equations and relative equa-
tions were implemented into FLUENT® 6.2 by developing our own
Locations of application Boundary conditions user-defined-functions (UDFs) based on the general FLUENT® pack-
Inlet of the anode flow channel u = Uan,in ; YH2 = YH2 ,in ; YH2 O age that does not include the PEMFC module developed by Fluent
= YH2 O,an,in ; Tan,in = TH2 ,in Inc. The developed own UDFs are written in C language with
Inlet of the cathode flow channel u = Ucat,in ; YO2 = YO2 ,in ; YH2 O about 2000 statements. There were totally 279,500 grid cells in
= YH2 O,cat,in ; Tcat,in = Tair/O2 ,in the computation domain used to simulate the physical and elec-
∂uout,an ∂YH ∂YH O,an trochemical phenomena in the fuel cell. The solution procedure for
Outlet of the anode flow channel ∂x
= 0; ∂x2 = 0; 2
∂x
∂Tan,out
pressure–velocity coupling was based on PISO algorithm [40].
= 0; =0
∂x In order to save the calculation time, an unsteady single-phase
∂uout,cat ∂YO ∂YH O,cat
Outlet of the cathode flow channel = 0; ∂x2 = 0; 2 model was simulated from the time t = 0 s to 0.5 s. At the time
∂x ∂x
= 0;
∂Tcat,out
=0 t = 0.5 s, an initialization of a series of liquid water droplets was
∂x
initially suspended on the cathode channel at the time t = 0.50 s
∂m
The anode terminal s = 0; ∂y
=0 and a general model of PEMFC was called to further investigate
∂m two-phase flow behavior, especially liquid water behavior across
The cathode terminal s = Vcell ; ∂y
=0
∂s ∂s ∂m ∂m
porous media, together with electrochemical reaction, heat and
External boundaries = 0; = 0; = 0; =0
∂x ∂z ∂x ∂z mass transfer.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 203

in situ the production and distribution of water in a PEMFC. Fein-

del et al. [44] developed a nuclear magnetic resonance (NMR)
microscopy system to investigate the distribution of water through-
out an operating H2 /O2 PEMFC. The areas investigated include
the flow fields, current collectors, membrane electrode assem-
bly (MEA), and the membrane surrounding the catalysts. Another
method is neutron radiography and tomography that has been suc-
cessfully applied for investigation of liquid water evolution and
transport in low temperature PEMFCs by Manke et al. [45] and
Davey et al. [46].
Now let us make the comparisons. The water distribution in
the channel and porous media of the fuel cell could be obtained
from the numerical model. To validate the numerical results pre-
sented in this study, qualitative comparisons were made by using
some available experimental visualization data. Fig. 4a presents the
computed water distribution in a Y–Z cross-sectional plane of this
fuel cell model. Fig. 4b depicts the water distribution captured by
using Neutron Radiography method [46]. Note that in Fig. 4b, there
Fig. 3. A series of water droplets was initially suspended into the cathode channel. was more water (black = no water; blue/white = more water) accu-
mulated near the outlets than near the inlets. The water near the
3. Results and discussion

Fig. 3 shows the arrangement of the initial liquid water distri-

bution. There were 15 equidistant spherical droplets (r = 0.4 mm)
freely suspended along the cathode channel. The cell voltage was
constantly set to be 0.5 V. The operating condition and fuel cell
parameters are listed in Table 1.

3.1. Water distribution inside the fuel cell—comparison of

numerical simulation and experimental visualization

Before we have comparisons, let us have a quick look at the

experimental investigations in the available literature. Experimen-
tal studies to probe detailed liquid water transport from the GDL
into the gas flow channels have been performed by Yang et al. [41]
and Zhang et al. [42]. Three transparent PEMFCs with different flow
fields, including parallel, interdigitated and cascade, were intro-
duced by Liu et al. [43]. As presented in Liu’s paper, the effects
of flow field, cell temperature, cathode gas flow rate and opera-
tion time on water build-up and cell performance were presented,
respectively. However, only the water flooding and two-phase flow
in the transparent channel is visualized. Recently, there are some
innovative visualization methods that were applied to investigate

Fig. 5. (a) Volume fraction distribution of liquid water in the middle-plane of

the cathode channel at t = 0.502 s; (b) neutron tomogram showing the three-
Fig. 4. (a) Water concentration distributions in the porous media on the cross- dimensional water distribution in the cathode channel of a serpentine PEMFC
sectional planes (Y–Z plane) at X = 0.01 m; (b) water profile in two 2.25 cm2 active (Manke et al. [45]): (b–a) right before shut up, (b–b) several hours after a tomography
area cells with 50% RH, 1.1 stoich and 172 kPa at the anode and 2.2 stoich and 172 kPa was performed, (b–c) quotient of image (b–a) and (b–b), (b–d) neutron tomogram
at the cathode at 1.4 A cm−2 and 100% RH at the cathode inlet (Davey et al. [46]). showing the three-dimensional water distribution.
204 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Fig. 6. The motion and deformation of droplets vs. time on the cathode channel: (a) t = 0.500 s; (b) t = 0.50025 s; (c) t = 0.5005 s; (d) t = 0.50075 s; (e) the velocity distribution
and the volume-of-fluid of liquid water in the X–Z plane at Y = 0.002 m at t = 0.50075 s; (f) the pressure distribution and the volume-of-fluid of liquid water in the X–Z plane at
Y = 0.002 m at t = 0.50075 s; (g) the water and velocity distribution in the different cross-sections along the main flow direction at t = 0.50075 s, corresponding to (a) X = 0.001 m,
(b) X = 0.006 m, (c) X = 0.014 m and (d) X = 0.019 m, respectively.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 205

Fig. 6. (Continued ).

outlet was more concentrated in the cathode GDL, especially the droplets that enter the turning area from the straight channels hit
GDL above the land area. This is consistent with the generation of the turning-wall surface. As a result, the deformed droplets were
water at the cathode and its removal by the gas flow in the channels. further elongated and the strong “turning” gas flow breaks these
The water generated above the land area needs to diffuse out later- droplets into even smaller ones sticking on the turning-wall sur-
ally to the channel before it can be removed through the flow fields. face. Due to the effect of wall adhesion and surface tension, the
The numerical result in Fig. 4a qualitatively shows good agreement smaller water droplets sticking on the turning-wall would slowly
with the experimental data in Fig. 4b made by Davey et al. [46]. In move forward while the other water droplets from the channel is
addition, more liquid water concentrated in the turning area of ser- continuously coming to the turning, and further broken into the
pentine channel fuel cell is also presented by the computed model other smaller ones. Such these water droplets would coalesce, then
in this study shown in Fig. 5a and observed in Fig. 5b that was cap- expanding and deforming into “water bands” due to the shear stress
tured by using neutron tomography method in the experiments of from “turning” gas flow and pressure drop, as shown in Fig. 6c and
Manke et al. [45]. d. It is noted that since the coalescence of water droplets is formed
In summary, by comparing the numerical results with avail- in the turning area resulting in a high concentration of liquid water
able experimental results, we can conclude that the present general volume, a substantial blockage in the channel would take place. This
model is qualitatively in good agreement with the available experi- blockage induces a decrease of the gap between the water droplets
mental results. For this research, qualitative comparisons could not and the channel wall. In addition, the water droplets hitting the
be conducted at the present time due to the sizes of the fuel cell in turning wall would stick onto it due to the wall adhesion. Fig. 6e
the cited experimental research are not known. The present authors shows the velocity distribution and the volume fraction of liquid
are currently developing a bench experimental set-up to further water in the X–Z plane at Y = 0.002 m at t = 0.50075 s. The velocity
validate the general model presented here and detailed results will field of airflow is accelerated through the gaps that were formed by
be published in future. the blockage of liquid water in the turning area, and then is decel-
erated after leaving the turning area in which the blockage occurs.
3.2. The motion and deformation of liquid water inside the In the other views shown in Fig. 6g, one can see that the water
cathode channel droplets hit the turning wall and expanded to the surroundings of
the landing points. The droplets could induce high-pressure zones
Fig. 6 shows the motion and deformation of droplets versus time in the channel, as shown in Fig. 6f, especially in the turning area.
on the cathode channel of the PEMFC. At t = 0.5 s, the droplet was As the time progressed to t = 0.501 s, as shown in Fig. 7a, the
placed at the channel inlet and was in its initial spherical shape water bands in the turning walls keep moving forwards along
(Fig. 6a). As time progressed, droplet deformation (elongation) in the main flow direction. One can see in Fig. 7a and b that, the
the flow direction, which is due to a combination of droplet sur- water bands have a tendency of elongating without breaking away
face tension and shear stress from surrounding airflow, could be under the impact of shear force from the gas flow. However, the
observed. In other words, the enlargement of droplet surface area water bands were broken again into smaller droplets distributed
increased the surface tension such that the force balance on the over the straight channel wall since these bands left the turn-
droplet could be achieved. Fig. 6b indicates that, after hitting the ing area. Subsequently, the formed small water drops are easily
turning surface, these deformed droplets had the tendency of frag- removed and distributed everywhere on the straight channel walls,
menting and then entering the central airflow due to the dragging as it can be seen in Fig. 7c and d at t = 0.504 s and 0.506 s. Fig. 7e
effect from the turning flow. Obviously, in the turning area, the shows that the airflow is rapidly accelerated at the turning area
shear force from the airflow is much stronger than the droplet due to the water blockage as earlier mentioned. Hence, the shear
surface tension, thus together with the inertia force, resulting in stress from the airflow in the straight channel is significant to
significant droplet deformation and complex liquid water distri- break the water bands into small, discrete liquid droplets rather
bution. As the time progressed after milliseconds, the deformed than deforming or elongating such water bands. This also results
206 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Fig. 7. The motion and deformation of droplets vs. time on the cathode channel: (a) t = 0.501 s; (b) t = 0.502 s; (c) t = 0.504 s; (d) t = 0.506 s; (e) the velocity distribution and
the volume-of-fluid of liquid water in the X–Z plane at Y = 0.002 m at t = 0.506 s; (f) the pressure distribution and the volume-of-fluid of liquid water in the X–Z plane at
Y = 0.002 m at t = 0.506 s.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 207

Fig. 8. The motion and deformation of droplets vs. time on the cathode channel: (a) t = 0.520 s; (b) t = 0.522 s; (c) t = 0.524 s; (d) t = 0.526 s; (e) the velocity distribution and
the volume-of-fluid of liquid water in the X–Z plane at Y = 0.002 m at t = 0.526 s; (f) the pressure distribution and the volume-of-fluid of liquid water in the X–Z plane at
Y = 0.002 m at t = 0.526 s.
208 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Fig. 9. The motion and deformation of droplets vs. time on the cathode channel: (a) t = 0.560 s; (b) t = 0.562 s; (c) t = 0.564 s; (d) t = 0.566 s; (e) the velocity distribution and
the volume-of-fluid of liquid water in the X–Z plane at Y = 0.002 m at t = 0.566 s; (f) the pressure distribution and the volume-of-fluid of liquid water in the X–Z plane at
Y = 0.002 m at t = 0.566 s; (g) the liquid water in the Y–Z plane at X = 0.001 m, 0.002 m, 0.018 m and 0.019 m.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 209

Fig. 9. (Continued ).

in a decrease of pressure drop in comparison with the value results in an evident decrease in pressure drop distribution which
at t = 0.50075 s. is presented in Fig. 8f.
At t = 0.52 s as shown in Fig. 8a, a portion of water droplets was Most of the water droplets were moved out of the channel by the
removed from the cathode channel outlet by the gas flow. The small gas flow at t = 0.56 s. As the time proceeded to t = 0.566 s, it could
droplets tend to scatter everywhere in the channel rather than coa- be observed that there were only few small droplets remaining
lescing or forming water bands. Even in the turning area, it can in the exit channel. However, an amount of liquid water pene-
be seen that the water bands either are tiny or are not formed at trated through the gas diffusion layer seemed to be retained at
all. The wall adhesion of the small droplets previously coalesced the same position in the layer regardless the droplets in the chan-
at the turning areas is insufficient to resist the shear stress from nel continued to be removed outwards. It can be obviously seen
the gas flow. In addition, one can see that there are some droplets that the water amount kept in the diffusion layer is marked by cir-
penetrating to the gas diffusion layer. It can be realized that this cles as shown in Fig. 9a–d. Noticeably, the marked circles in Fig. 9
phenomenon takes place mostly in the turning areas where the evoke that the penetrated water usually occur in the turning area
water is not easy to be removed rather than in the straight chan- in which the liquid water is more difficult to be removed and have
nel as previously discussed. As shown in Fig. 8b–d that as the time a tendency of passing through the gas diffusion layer instead of
proceeds to t = 0.526 s, the water continues to be taken away from moving along the channel to the outlet. The liquid water in the
the channel outlet except the water amount remaining in the gas gas diffusion layer is also clearly illustrated in Fig. 9g since the
diffusion layer. At t = 0.526 s, as shown in Fig. 8e, the velocity field cross-sectional planes along X-direction at X = 0.001 m, 0.002 m,
is observed to be “more stable” since a large amount of liquid water 0.018 m and 0.019 m are extracted. As time progresses, the more
was taken away and then the blockage caused by the small droplets liquid water gradually move out of the channel, the less the liquid
less influences on the gas flow movement. Consequently, this also water blocks the flow field in the channel. Therefore, the block-
210 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Fig. 10. (a) The distribution of volume fraction of liquid water in the cross-sectional planes (Y–Z plane) at X = 0.0015 m, 0.010 m and 0.0185 m (Case 2); (b) distribution of
volume fraction of liquid water in the middle-plane of the cathode flow channel (Y = 0.0015 m) and catalyst layer (Y = 0.002305 m) (Case 2).
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 211

Fig. 11. Velocity vectors in the channel: on the middle-plane of cathode channel (Y = 0.0015 m) and the middle-plane of anode channel (Y = 0.00317 m): (a) Case 1 and (b)
Case 2; on the cross-sectional planes (Y–Z plane) at X = 0.0015 m, 0.010 m and 0.0185 m: (c) Case 1 and (d) Case 2.
212 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Fig. 11. (Continued ).

Fig. 12. Velocity vectors in the porous media on the cross-sectional planes (Y–Z plane) at X = 0.0015 m, 0.010 m and 0.0185 m: (a) Case 1 and (b) Case 2.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 213

age insufficiently influences on the velocity and pressure fields in 3.3. Effects of liquid water on the fuel cell characteristics
the fuel cell channel. It can be seen in Fig. 9e and f, the velocity
and pressure fields at t = 0.566 s are more stable, have no sudden In order to investigate the influences of liquid water on the cell
changes and have a smooth distribution along the channel, similar characteristics and operating condition, two cases (Cases 1 and 2)
to those when compared with the case without liquid water in the are compared in this study. The first case would be considered
channel. at the time instant t = 0.53 s when the liquid droplets were not

Fig. 13. Velocity vectors and pressure distributions on the middle-plane of catalyst layers (0.002305 m): (a) Case 1 and (b) Case 2.
214 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

introduced in the cathode channel and therefore the fuel cell oper- effects. In Case 2, fuel cell characteristics such as velocity, pressure,
ates in the condition without liquid water. The second case would temperature, species concentrations and the fuel cell performance,
be presented at the time instant t = 0.53 s since a series of water etc. are considered as taking the appearance of water droplets into
droplets had been introduced at t = 0.5 s into the cathode channel account. Hence, the numerical model at the first case was consid-
and evolved for 0.03 s that can be used to show the liquid water ered as single-phase model and the second case was considered

Fig. 14. Velocity vectors and pressure distributions on the middle-plane of anode channel (Y = 0.00317 m) and catalyst layers (0.002365 m): (a) Case 1 and (b) Case 2.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 215

as the general model that includes two-phase interface tracking X = 0.0015 m, 0.010 m and 0.0185 m and Fig. 10b depicts in the
together with all other detailed sub-models. middle-plane of cathode flow channel and catalyst layer, respec-
tively. Note that the liquid water was not taken into account in
3.3.1. Volume fraction of liquid water single-phase flow that was applied for the fuel cell prior to t = 0.50 s.
In Case 2, Fig. 10a shows the distribution of volume frac- Therefore, the presence of volume fraction of liquid (liquid water
tion of liquid water in the cross-sectional planes (Y–Z plane) at distribution) is not applicable to Case 1. It could be observed that

Fig. 15. Mass fraction distributions of oxygen on the middle-plane of cathode channel (Y = 0.0015 m) and catalyst layers (0.002305 m): (a) Case 1 and (b) Case 2.
216 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

the liquid droplets distributed in the fuel cell at a specified time gressed after 0.03 s at t = 0.53 s. The change of the water occupation
were presented via volume fraction of water. To investigate the fraction in the channel and liquid water transport from the cath-
motion and interaction of liquid droplets inside the fuel cell and ode channel to the porous media were studied. By tracking the
the influence of liquid phase on the fuel cell operating condition, presence of volume fraction of water, one can determine where
the results of this simulation would be analyzed as the time pro- the liquid water occurs and the effects of presence of liquid water

Fig. 16. Mass fraction distributions of hydrogen in the middle-planes of anode channel (Y = 0.00317 m) and catalyst layers (0.002365 m): (a) Case 1 and (b) Case 2.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 217

on the physical parameters and characteristics of the fuel cell. 3.3.2. Liquid water effects on velocity field and pressure
Fig. 10a and b shows that most of the liquid water concentrated distribution
in the channel and moved through porous media is in the turning The velocities at the cathode and anode inlets are 7.0 m s−1 and
area of the serpentine-shaped flow channel. As a result, this would 4.5 m s−1 , respectively. Fig. 11a and b shows that the gases (air
especially influence the other flow parameters of the turning area and water vapor at cathode; and hydrogen and water vapor at
also. anode) are supplied from the inlet, moving along each branch of the

Fig. 17. Mass fraction distributions of water vapor on the middle-plane of cathode channel (Y = 0.0015 m) and catalyst layers (0.002305 m): (a) Case 1 and (b) Case 2.
218 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

serpentine-shaped channel for Case 1 and Case 2, respectively, dra- media had no change on the same cross-section for every channel
matically varied their direction when approaching the turn, then branch.
go to the next branches. In Case 1 (Fig. 11c), the velocity vectors on Fig. 13a and b shows velocity vectors and pressure distribu-
Y–Z planes on cross-sectional planes of channel branches shown in tions on the middle-plane of catalyst layers (0.002305 m) for Cases
Fig. 11c seem to have similar variation to each other. In other words, 1 and 2, respectively. Now, the flow field on X–Z planes could be
these velocity distributions on the same cross-section through all observed. In both cases, there remained alternate velocity distribu-
channel branches are the similar. This is because that liquid water tions having “triangle-shaped” on the porous media in which the
was not considered in Case 1 and therefore there are no liquid magnitude of velocity vectors increased or decreased along the X-
droplets effects on the flow field. In Case 2 (Fig. 11d), the velocity direction. This is the nature of fluid flow. It could be noticed that
and pressure distributions are quite different. Due to the presence the high velocity vectors were located in the region in the land
and blockage of water droplets over the channel as previously men- area between the two corners of two sequent branches, also having
tioned, the flow field was not similar to each other depending on the largest pressure difference. The flow in the corner of a branch,
where the droplets located. It is noticeable that the pressure drop therefore, tends to move towards the corner of the next branch at
in Case 2 was observed to be higher than that in Case 1, especially the same side through the porous media instead of moving along
around the location of the droplets, the pressure drop reaches high- the channel. This is the shortest way and easiest way as well. The
est values due to the momentum resistance caused by liquid water magnitude of such velocity vectors decreases along the channel due
in airflow. to the decrease of pressure difference and it is lowest at the corner
The velocity in the gas channels on both sides shown in Fig. 10a at opposite side where the flow freely moves in the channel turn.
and b were very high (the order of magnitude is 1) by comparing Moreover, the chaos of velocity distribution caused by the effect of
to the low velocity in the porous media shown in Fig. 12a and b liquid water corresponding to the location of droplets in the porous
(the order of magnitude is 10−2 ) due to the flow resistance in the media is more realized as shown in Fig. 13b.
porous media (including GDLs and catalyst layers). Although the Fig. 14a and b shows velocity vectors and pressure distributions
porous media is solid, there still exists the flow field due to pres- on the middle-plane of anode channel (Y = 0.00317 m) and catalyst
sure gradient distribution in the void of such regions where reactant layers (0.002365 m) for Cases 1 and 2. Note that the flow field in
gases and water can move through. It can be seen in Fig. 12a and b the fuel cell may vary at the different times. Because the flow con-
for both cases that the flow in Y–Z planes had a tendency of moving ditions at the anode side reached the steady state at Case 1 and the
through the porous media from one channel branch to the next, time period between Case 1 (at t = 0.50 s) and Case 2 (at t = 0.53 s)
especially in the land area (the porous zones underneath the col- is almost negligible, then the velocity and pressure distributions
lector ribs). The difference of pressure on the same cross-section were almost the same for the both cases due to the absence of liquid
in all branches is the driving force to such flow motion. Similarly, water at the anode side.
it can be clearly seen in Fig. 12b how the blockage of liquid water
in the porous media influenced the velocity fields. On the contrary, 3.3.3. Liquid water effects on concentrations
for the case where liquid water was absent in the fuel cell as shown In the both cases, it could be observed that the oxygen fraction
in Fig. 12a, it is most likely that the flow in Y–Z planes in the porous decreased along the cathode channel from the inlet to the outlet

Fig. 18. The distribution of local current density at the catalyst/GDL interfaces at the cathode and anode: (a) Case 1 and (b) Case 2.
 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 219

and from the channel to the catalyst layer due to the oxygen con- oxygen mass fraction distribution in the porous media underneath
sumption on the catalyst layer. Mass transport of oxygen from the the turning area could be seen to have a uniform, smooth U-shaped
channel to the GDL was mainly driven by its concentration gradi- as shown in Fig. 15a. Differently in Case 2, the shape was various,
ent (diffusion) and this effect was also partially influenced by the unsmoothed and arbitrary. This is because the liquid water occu-
velocity component (convection) inside the porous media. The flow pied the void in which oxygen was supposed to occupy. The higher
in cathode channel enhanced the gas transport in the region under volume fraction of liquid water occupies, the lower the oxygen mass
the channel, rather than in the region under the rib (the land area). fraction could be observed.
As shown in Fig. 15a and b, high oxygen region under the cath- Similar to the oxygen mass fraction distribution in the cathode,
ode channel was observed. On the other hand, the land area had a the hydrogen gradually decreased from the inlet to the outlet in
low amount of oxygen. There are two reasons accounting for this; both the anode flow channel and porous media as shown in Fig. 16a
the first is that mass transport of oxygen in such area was low due and b. However, the decrease of hydrogen mass fraction along the
to weak influence of convection and diffusion. The second is that anode channel was insignificant due to absence of liquid water.
a large amount of oxygen was consumed due to a high reaction For the both cases, the water vapor mass fraction increased along
rate—the chemical reaction had strongly taken place in the land the cathode channel by two sources: by-product water was gen-
area which was directly connected to the collector ribs that the erated from the electrochemical reaction at the cathode and the
current density vectors mainly went through. This phenomenon migration of water from the anode to the cathode due to the electro-
will be discussed in later section. The presence of liquid water sig- osmotic drag. As shown in Fig. 17 that there was a high water vapor
nificantly influences the oxygen concentration in the channel and content remaining in the land area at the cathode as a result of
porous media as well. This is very important in water management water build-up due to the higher reaction rate. Furthermore, due to
of a fuel cell. As mentioned in many studies, liquid water occurring the small velocity field in the land area nearby the edge (shown in
in the fuel cell causes the flooding phenomenon and further blocks Figs. 11 and 12), the accumulated water in such area was not easy to
the oxygen transport through the GDL and catalyst layer, then sig- be removed. Obviously, water concentration is highest in the edge
nificantly influences the fuel cell performance. In Case 2, there was adjacent to the outlet channel branch due to water build-up and
a noticeable blockage of liquid water in oxygen concentration, espe- low convection effect. In Case 2, the presence of liquid droplets also
cially in the land region underneath the turning area. In Case 1, the influenced the distribution of water vapor mass fraction, similar to

Fig. 19. (a) Temperature distribution in the cross-sectional planes (Y–Z plane) at X = 0.0015 m, 0.010 m and 0.0185 m (Case 2); temperature distributions on the middle-plane
of cathode channel (Y = 0.0015 m) and catalyst layers (0.002305 m): (b) Case 1 and (c) Case 2.
220 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

Fig. 19. (Continued ).

 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222 221

that on oxygen transport. As can be seen in Fig. 17b, the higher the
volume fraction of liquid water occupied, the larger the water vapor
mass fraction observed.

3.3.4. Liquid water effects on current density distributions

Fig. 18a and b shows the distribution of local current density at
the catalyst layer/GDL interfaces on the cathode and anode sides
for the both cases, respectively. Note that the local current density
at the interfaces was generated by the electron movement while
the local current density at the mid-plane of the membrane by
the proton movement through the membrane. The current density
peaked at the land area in the porous media because the current
flow tended to pass through the land area of GDLs/catalyst layers
which were in direct contact with the collector ribs. Hence, the
current flow could move through a circuit from the anode to the
cathode in the shortest and fastest way. Note that the current den-
sity is proportional to the reaction rate. This also accounts for the
reason why reaction rate or reactant gases consumption was high-
est in the land area. For unsteady state, the average current density
may vary with respect to time. Namely, the average current den-
sity at t = 0.50 s (Case 1) was 0.624 A cm−2 and at t = 0.53 s (Case 2)
was 0.616 A cm−2 . Due to a small change of average current den-
sity, it was not easier to observe the differences on current density
distributions for both cases.

3.3.5. Liquid water effects on temperature distributions

The lowest temperature could be observed at the flow inlet
where the flows with ambient temperature come in, and the
temperature increases along the positive flow direction. The heat
generation was due to the electrochemical reaction and Ohmic
heating. Therefore, the highest temperature concentrates in the
catalyst layers in which the reactions take place and the current
is generated. In addition, the temperature distribution through the
solid materials (including flow plates/collectors and solid phase
of porous media) is also affected by the heat conduction process. Fig. 20. The average current density of the cell vs. time (Case 2): (a) from t = 0 s to
Depending on different solid materials, the thermal conductivity 0.50 s and (b) from t = 0.50 s to 0.58 s.

may change, resulting in different temperature distributions in dif-

ferent solid layers as shown in Fig. 19a. Note that the temperature shown in Fig. 20a. As the time progressed, the average current den-
distribution in the collectors seemed to be uniform due to a high sity slightly increased after the fuel cell operation was started. In
thermal conductivity of the collector material. By controlling the this stage, liquid water was not taken into account and hence it did
convective coefficient h (Eq. (36)) to cool the cell, the average tem- not influence the increase of average current density. Since a stable
perature was in the range of 334–340 K. flow and species parameters was obtained, it could be observed that
Fig. 19b and c shows temperature distributions on the middle- the average current density reached a constant value at 6246 A m−2
plane of the cathode channel (Y = 0.0015 m) and catalyst layers at t = 0.4 s. Fig. 20b shows that when the water droplets were added
(0.002305 m) for Cases 1 and 2, respectively. It was observed that into the channel at t = 0.5 s, the average current density of the cell
the temperature distribution in Case 1 was higher than that in Case was no longer constant at a value of 6246 A m−2 . In contrast, the
2. The presence of liquid water induced a decrease of temperature at average current density tended to decrease. It could be realized
the location occupied by the droplets. Initially, water droplets were that the presence of liquid water in the fuel cell is a main reason
suspended in the channel where the temperature was the lowest in to decrease the current density of the cell. As mentioned, water
the fuel cell due to the convection effect. The temperature of initial in the form of liquid may hinder the gas transport, resulting in a
droplets was set to be equal to the value of ambient temperature concentration loss and therefore cause a reduction of the current
(300 K). As the time progressed, these droplets moved forward as density. However, this decrease was only maintained until the time
the airflow direction in the channel and a portion of liquid water t = 0.508 s at which the liquid water was mostly moved out of the
was transported through the porous media. It is noticeable that channel except for some kept in the porous media. The absence of
the temperature in the porous media was highest due to chemical liquid water led to a gradual increase in the current density as the
reactions taking place on the catalyst layer. The low temperature of time progressed, as shown in Fig. 20b.
liquid water transferred to and the heat generated from the cath-
ode catalyst layer resulted in an uneven temperature distribution 4. Conclusion
on the porous media.
In this study, a three-dimensional, unsteady, multi-phase, multi-
3.4. Average current density with respect to time component PEMFC model with VOF interface tracking technique
was solved using FLUENT® by developing own UDFs to customize
The average current density of the cell is presented with respect the governing and relative equations. Numerical simulations were
to time as shown in Fig. 20a and b. For the time period from 0 s to conducted for a single PEFMC with serpentine flow channels. This
0.5 s, the fuel cell operated without an addition of liquid water as model considered all the necessary components consisting of a unit
222 A.D. Le, B. Zhou / Journal of Power Sources 182 (2008) 197–222

PEMFC and coupled the fluid flow (momentum transport), species perature of the channel and porous media with the presence of
transport, energy transport, electron and proton transport, with liquid water was lower than that without liquid water.
volumetric electrochemical reactions. The membrane, gas diffusion
layer and catalyst layers were treated as volumetric domain with Acknowledgments
grid points, and the electrochemical reactions in the catalyst layers
as volumetric reactions. Especially, by using VOF interface tracking The authors are grateful for the support of this work by the
algorithm, the flow behaviors of liquid water and temperature field Auto21TM Networks of Centers of Excellence (Grant D07-DFC),
were also investigated. The main interpretations and conclusion of the Natural Sciences and Engineering Research Council of Canada
this study are summarized as follows: (NSERC), the Canada Foundation for Innovation (CFI), the Ontario
Innovation Trust (OIT), and the University of Windsor.
1. The motion and behaviors of liquid water in a full 3D, unsteady
PEMFC was completely described by using VOF interface tracking
References
algorithm combined with solving the fluid flow, species trans-
port, energy transport and electrochemical governing equations. [1] S. Um, C.Y. Wang, K.S. Chen, J. Electrochem. Soc. 147 (2000) 4485.
2. The motion of droplets in the PEMFC model was illustrated at [2] S. Dutta, S. Shimpalee, J.W. Van Zee, J. Appl. Electrochem. 30 (2000) 135.
different time steps in numerical simulation. Firstly, the droplets [3] E. Hontanon, M.J. Escudero, C. Bautista, P.L. Garcia-Ybarra, L. Daza, J. Power
Sources 86 (2000) 363.
freely moved in the channel then deformed (elongation) in the [4] S.W. Cha, R. O’Hayre, Y. Saito, F.B. Prinz, J. Power Sources 134 (2004) 57.
flow direction, which was due to a combination of droplet surface [5] A.A. Kulikovsky, Electrochem. Commun. 3 (2001) 460.
tension and shear stress from surrounding airflow. After hitting [6] J.S. Yi, J.D. Yang, C. King, AIChE J. 50 (2004) 2594.
[7] H. Ju, G. Luo, C.Y. Wang, J. Electrochem. Soc. 154 (2007) B218.
the turning surface, these deformed droplets had the tendency [8] L. You, H. Liu, Int. J. Heat Mass Transfer 45 (2002) 2277.
of fragmenting and then entering the central airflow due to the [9] D. Natarajan, T.V. Nguyen, J. Power Sources 115 (2003) 66.
dragging effect from the turning flow. Since the coalescence of [10] T. Berning, N. Djilali, J. Electrochem. Soc. 150 (2003) A1589.
[11] S. Mazumder, J.V. Cole, J. Electrochem. Soc. 150 (2003) A1510.
water droplets was formed in the turning area resulting in a high [12] M. Vynnycky, M. Vynnycky, Appl. Math. Comput. 189 (2007) 1560.
concentration of liquid water volume, a substantial blockage in [13] U. Pasaogullari, C.Y. Wang, J. Electrochem. Soc. 152 (2005) A380.
the channel occurred. After leaving the turning area, a significant [14] H. Meng, C.Y. Wang, J. Electrochem. Soc. 152 (2005) A1733.
[15] Y. Wang, C.Y. Wang, J. Electrochem. Soc. 153 (2006) A1193.
shear stress from the airflow in the straight channel broke the
[16] H. Meng, C.Y. Wang, Chem. Eng. Sci. 59 (2004) 3331.
coalescent droplets into smaller, discrete liquid droplets rather [17] H. Ju, C.Y. Wang, J. Electrochem. Soc. 151 (2004) A1954.
than deforming or elongating such water bands. Finally, liquid [18] H. Ju, C.Y. Wang, S. Cleghorn, U. Beuscher, J. Electrochem. Soc. 153 (2006) A249.
droplets were gradually removed out of the channel under a [19] Y. Wang, C.Y. Wang, J. Power Sources 153 (2006) 130.
[20] P. Quan, B. Zhou, A. Sobiesiak, Z. Liu, J. Power Sources 152 (2006) 131.
strong flow field. [21] K. Jiao, B. Zhou, P. Quan, J. Power Sources 154 (2006) 124.
3. The presence of liquid water in the channel significantly influ- [22] K. Jiao, B. Zhou, P. Quan, J. Power Sources 157 (2006) 226.
ences the flow field parameter. Namely, due to the blockage of [23] X. Zhu, P.C. Sui, N. Djilali, J. Power Sources 172 (2007) 287.
[24] J.H. Nam, M. Kaviany, Int. J. Heat Mass Transfer 46 (2003) 4595.
water droplets, the gas flow would become unevenly distributed, [25] B. Markicevic, A. Bazylak, N. Djilali, J. Power Sources 171 (2) (2007) 706.
the high pressure-drop regions would occur around the location [26] J.I. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, J. Power Sources 173 (1)
of liquid droplets. (2007) 277.
[27] P.K. Sinha, C.Y. Wang, J. Electrochem. Acta 52 (28) (2007) 7936.
4. The structure of the channel will significantly influence the [28] H. Dohle, R. Jung, N. Kimiaie, J. Mergel, M. Muller, J. Power Sources 124 (2003)
distribution of liquid water inside the cell. As simulated in a 371.
serpentine channel, the liquid water is difficult to be removed [29] U. Pasaogullari, C. Wang, Electrochem. Acta 49 (2004) 4359.
[30] S. Freni, G. Maggio, E. Passalacqua, Mater. Chem. Phys. 48 (1997) 199.
in the turning area. Therefore, a portion of liquid water tends [31] K. Jiao, B. Zhou, J. Power Sources 169 (2007) 296.
to move through the gas diffusion layer in the porous media. [32] K. Jiao, B. Zhou, J. Power Sources 175 (2008) 106.
By simulating different types of the channels, the character- [33] Fluent® 6.2 Documentation, Fluent Inc., 2005.
[34] R.B. Bird, W.E. Steward, E.N. Lightfoot, Transport Phenomena, John Wiley & Sons,
istics of liquid water removal for different designs could be
New York, 1960.
investigated. [35] F. Barbir, PEM Fuel Cells, Elsevier Academic Press, 2005.
5. By quantitatively and qualitatively analyzing the behaviors of liq- [36] T. Springer, T. Zawodzinski, S. Gosttesfeld, J. Electrochem. Soc. 138 (1991) 2334.
uid water and the average current densities at different periods [37] R. O’Hayre, S.-W. Cha, W. Colella, F.B. Prinz, Fuel Cell Fundamentals, John Wiley
& Sons, New York, 2006.
of time, it could be observed that the liquid water would hinder [38] C.S. Peskin, J. Comput. Phys. 25 (1997) 220.
the gas transport in the fuel cell, resulting in a high concentration [39] J. Larminie, A. Dicks, Fuel Cell Systems Explained, 2nd ed., John Wiley & Sons,
loss and directly decreasing the current density—in other words, England, 2003.
[40] J.H. Ferziger, M. Peric, Computational Methods for Fluid Dynamics, 3rd ed.,
severely affect the fuel cell performance. Springer, 2002.
6. The local current density contours are also presented, show- [41] X.G. Yang, F.Y. Zhang, A. Lubawy, C.Y. Wang, Electrochem. Solid-State Lett. 7
ing that the high magnitude of the vectors mostly distribute in (2004) A408.
[42] F.Y. Zhang, X.G. Yang, C.Y. Wang, J. Electrochem. Soc. 153 (2006) A225.
the region underneath the channel where high reaction rates [43] X. Liu, H. Guo, C. Ma, J. Power Sources 156 (2006) 267.
exist. [44] K.W. Feindel, L.P.-A. LaRocque, D. Starke, S.H. Bergens, R.E. Wasylishen, J. Am.
7. By solving the energy equation, the temperature distribution Chem. Soc., Commun. 126 (37) (2004) 11436.
[45] I. Manke, Ch. Hartnig, M. Grünerbel, J. Kaczerowski, W. Lehnert, N. Kardjilov, A.
over the fuel cell was described. The temperature was controlled Hilger, W. Treimer, M. Strobl, J. Banhart, Appl. Phys. Lett. 90 (2007) 184101.
at around 333 K by using a suitable forced convective coefficient [46] J.R. Davey, R. Mukundan, T. Rockward, J.S. Spendelow, D.S. Hussey, D.L. Jacobson,
applied on the fuel cell. The model also showed that the tem- M. Arif, R.L. Borup, 212th ECS Meeting, Washington, October 2007.