Journal of The Electrochemical

Society

You may also like

- Structural, hydrogen bonding and in situ
Computational Fluid Dynamics Modeling of Proton studies of the effect of hydrogen dilution
on the passivation by amorphous silicon of
Exchange Membrane Fuel Cells n-type crystalline (1 0 0) silicon surfaces
H Meddeb, T Bearda, Y Abdelraheem et
al.
To cite this article: Sukkee Um et al 2000 J. Electrochem. Soc. 147 4485 - Amorphous silicon oxide passivation films
for silicon heterojunction solar cells studied
by hydrogen evolution
Kazuyoshi Nakada, Shinsuke Miyajima
and Makoto Konagai
View the article online for updates and enhancements. - Effect of the Structural Change of
Hydrogenated Microcrystalline Silicon Thin
Films Prepared by Hot-Wire Chemical
Vapor Deposition
Shuichi Hiza, Wataru Matsuda, Akira
Yamada et al.

This content was downloaded from IP address 103.158.42.8 on 03/12/2023 at 09:29

 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000) 4485
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

Computational Fluid Dynamics Modeling of Proton Exchange

Membrane Fuel Cells
Sukkee Um,a C.-Y. Wang,a,*,z and K. S. Chenb,*
aGATE Center for Advanced Energy Storage, Department of Mechanical and Nuclear Engineering, The Pennsylvania State
University, University Park, Pennsylvania 16802, USA
bEngineering Sciences Center, Sandia National Laboratories, Albuquerque, New Mexico 87185-0834, USA

A transient, multidimensional model has been developed to simulate proton exchange membrane fuel cells. The model accounts
simultaneously for electrochemical kinetics, current distribution, hydrodynamics, and multicomponent transport. A single set of
conservation equations valid for flow channels, gas-diffusion electrodes, catalyst layers, and the membrane region are developed
and numerically solved using a finite-volume-based computational fluid dynamics technique. The numerical model is validated
against published experimental data with good agreement. Subsequently, the model is applied to explore hydrogen dilution effects
in the anode feed. The predicted polarization curves under hydrogen dilution conditions are in qualitative agreement with recent
experiments reported in the literature. The detailed two-dimensional electrochemical and flow/transport simulations further reveal
that in the presence of hydrogen dilution in the fuel stream, hydrogen is depleted at the reaction surface, resulting in substantial
anode mass transport polarization and hence a lower current density that is limited by hydrogen transport from the fuel stream to
the reaction site. Finally, a transient simulation of the cell current density response to a step change in cell voltage is reported.
© 2000 The Electrochemical Society. S0013-4651(00)08-104-0. All rights reserved.

Manuscript submitted August 22, 2000; revised manuscript received August 27, 2000.

Proton exchange membrane fuel cell (PEMFC) engines can poten- en12 further demonstrated the important roles played by water and
tially replace the internal combustion engine for transportation be- heat management in maintaining high performance of PEMFCs.
cause they are clean, quiet, energy efficient, modular, and capable of Most recently, Gurau et al.11 presented a two-dimensional model
quick start-up. Since a PEMFC simultaneously involves electrochem- of transport phenomena in PEM fuel cells. This work illustrated the
ical reactions, current distribution, hydrodynamics, multicomponent utility of a multidimensional model in the understanding of the inter-
transport, and heat transfer, a comprehensive mathematical model is nal conditions of a fuel cell such as the oxygen and water distribu-
needed to gain a fundamental understanding of the interacting elec- tions. In a separate development, Yi and Nguyen13 formulated a two-
trochemical and transport phenomena and to provide a computer- dimensional model to explore hydrodynamics and multicomponent
aided tool for design and optimization of future fuel cell engines. transport in the air cathode of PEMFCs with an interdigitated flow
Performance of a fuel cell is measured by its current-voltage rela- field. Recent efforts have also been made to model two-phase flow
tion (i.e., the polarization curve). At a particular current, the voltage and transport in the air cathode, a critical issue that has repeatedly
drop is mainly caused by (i) overpotentials of electrochemical reac- been emphasized in the literature for fuel cells operated under high
tions (mainly on the cathode), (ii) the ohmic drop across the current densities.14,15
ionomeric membrane, and (iii) the mass transport limitations of reac- The objective of this study is two-fold. One goal is to develop a
tants and products. At high current densities of special interest to transient, multidimensional model for electrochemical kinetics, cur-
vehicular applications, excessive water is produced within the air rent distribution, fuel and oxidant flow, and multicomponent trans-
cathode in the form of liquid, thus leading to a gas-liquid two-phase port in a realistic fuel cell based on a finite-volume-based computa-
flow in the porous electrode. The ensuing two-phase transport of tional fluid dynamics (CFD) approach. The CFD method was first
gaseous reactants to the reaction surface, i.e., the cathode/membrane adapted to electrochemical systems by Gu et al.16 and has since been
interface, becomes a limiting mechanism for cell performance, par- applied successfully to a variety of battery systems including lead-
ticularly at high current densities (e.g., >1 A/cm2). On the anode acid, nickel-cadmium, nickel-metal hydride, lithium-ion, and the
side, when reformate is used for the feed gas, the incoming hydro- primary Li/SOCl2 cell.16-22 The work here is intended to extend the
gen stream is diluted with nitrogen and carbon dioxide. The effects efficient single-domain CFD formulation previously developed for
of hydrogen dilution on anode performance, particularly under high batteries to PEMFCs. The second goal, and one of practical impor-
fuel utilization conditions, are significant.1,2 tance, is to explore hydrogen dilution effects on PEMFCs running on
Excellent reviews of hydrogen PEMFC research up to the mid- reformate gas. Such effects were most recently investigated experi-
1990s were presented by Prater3 and Gottesfeld.4 Modeling and mentally but have not been modeled.
computer simulation of hydrogen fuel cells have been attempted by The following section describes a transient, multidimensional
a number of groups with the common goal of better understanding mathematical model for electrochemical and transport processes
and hence optimizing fuel cell systems. Notable work includes that occurring inside a PEMFC. Model validation against the experimen-
of Bernardi and Verbrugge5,6 and Springer et al.7,8 whose models are tal data of Ticianelli et al.23 is presented in the Results and Discus-
essentially one dimensional. Fuller and Newman,9 Nguyen and sion section along with a detailed, two-dimensional study of hydro-
White,10 Garau et al.,11 and Yi and Nguyen12,13 developed pseudo gen dilution effects. The last section summarizes the major conclu-
two-dimensional models accounting for composition changes along sions from this study and identifies further research based on the pre-
the flow path. While such models are useful for small single cells, sent CFD modeling framework.
their applicability to large-scale fuel cells, particularly under high
Numerical Model
fuel utilization conditions, is limited. Nevertheless, the one-dimen-
sional models of Bernardi and Verbrugge5,6 and Springer et al.7,8 Figure 1 schematically shows a PEMFC fuel cell divided into
provided a fundamental framework to build multidimensional mod- seven subregions: the anode gas channel, gas-diffusion anode, anode
els that followed. The pseudo two-dimensional models developed by catalyst layer, ionomeric membrane, cathode catalyst layer, gas-dif-
Fuller and Newman,9 Nguyen and White,10 and later Yi and Nguy- fusion cathode, and cathode flow channel. Distinct from previous
models, the present model considers the anode feed consisting of
* Electrochemical Society Active Member. hydrogen, water vapor, and nitrogen to simulate reformate gas,
z E-mail: [email protected]
whereas humidified air is fed into the cathode channel. Hydrogen
4486 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000)
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

∂( eX k )
1 = ?( euX k ) 5 =?( Dkeff = X k ) 1 Sk [5]
∂t
=?(seeff=Fe) 1 SF 5 0 [6]
Here, u, p, Xk, and Fe denote the intrinsic fluid velocity vector,
pressure, mole fraction of chemical species k, and the phase poten-
tial of the electrolyte membrane, respectively. The diffusion coeffi-
cient of species k and ionic conductivity of the membrane phase in
Eq. 5 and 6 are effective values modified via Bruggman correlation
to account for the effects of porosity and tortuosity in porous elec-
trodes, catalyst layers, and the membrane. That is
Dkeff 5 e1.5
m Dk [7]
s keff 5 e1.5
m sk [8]
where em is the volume fraction of the membrane phase. Other sym-
bols in Eq. 3 through 8 can be found in the List of Symbols.
It is worth further explaining the mole fraction of oxygen ap-
pearing in Eq. 5 because oxygen is a gaseous species in the cathode
Figure 1. Schematic diagram of a proton exchange membrane fuel cell flow channel and gas-diffusion electrode but becomes a species dis-
(PEMFC). solved in the electrolyte in the catalyst layer and membrane regions.
Our definition is given by
oxidation and oxygen reduction reactions are considered to occur c g / ctot
k in the gas phase
only within the active catalyst layers where Pt/C catalysts are inter- Xk 5  [9]
ck / ctot
mixed uniformly with recast ionomer. Other aspects of hydrogen fuel e
in the electrolyte
cell modeling can be found in the works of Bernardi and Verbrugge6
and Springer et al.8 where ck is the molar concentration of species k and superscripts g
The fuel and oxidant flow rates can be described by a stoichio- and e denote the gas and the electrolyte phases, respectively. Thus,
metric flow ratio, z, defined as the amount of reactant in the cham- Xk is a true mole fraction in the gas phase but is a pseudo mole frac-
ber gas feed divided by the amount required by the electrochemical tion when species k is in the dissolved form. In addition, there is a
reaction. That is discontinuity in the value of Xk at the interface between the gas-dif-
fusion electrode and the catalyst layer due to the following thermo-
p1 4 F
z1 5 XO0 2 q1
0 [1] dynamic relation
RT IA
RT g
0 p2 2 F cke,sat 5 c 5 H ′ckg [10]
z2 5 XH0 2 q2 [2] H k
RT IA
where q0 is the inlet volumetric flow rate to a gas channel, p and T where H is the Henry’s law constant equal to 2 3 105 atm cm3/mol
the pressure and temperature, R and F the universal gas constant and for oxygen in the membrane.5 The dimensionless Henry’s constant,
Faraday’s constant, I the current density, and A the electrode surface H9, is thus approximately equal to 0.15 at 808C.
area. The subscripts (1) and (2) denote the cathode and anode In spite of the discontinuity in Xk across the interface between the
sides, respectively. For convenience, the stoichiometric flow ratios catalyst layer and gas-diffusion electrode, the single-domain formu-
defined in Eq. 1 and 2 are based on the reference current density of lation, Eq. 5, intrinsically maintains the balance of species fluxes;
1 A/cm2 here so that the ratios can also be considered as dimension- namely
less flow rates of the fuel and oxidant. ∂X k ∂X k
2( Dkeff )2 5 2( Dkeff )1 [11]
Model assumptions.—This model assumes (i) ideal gas mixtures; ∂n 2 ∂n 1
(ii) incompressible and laminar flow due to small pressure gradients
and flow velocities; (iii) isotropic and homogeneous electrodes, cat- where the symbols (2) and (1) stand for the left and right sides of
alyst layers, and membrane; (iv) constant cell temperature; and (v) the interface. Substituting the definition of Xk given in Eq. 9 into
negligible ohmic potential drop in the electronically conductive solid Eq. 11 yields the following interfacial balance of species fluxes
matrix of porous electrodes and catalyst layers as well as the current
collectors. ∂cke ∂ckg
2 Deeff 5 2 Dgeff [12]
Governing equations.—In contrast to the approach of Gurau ∂n ∂n
2 1
et al.,11 which employs separate differential equations for different
subregions, we take the single-domain approach used in our previ- Notice also that three source terms, Su, Sk, and SF, appear in
ous battery models in which a single set of governing equations valid momentum, species, and charge conservation equations to represent
for all subregions is used. As a result, no interfacial conditions are various volumetric sources or sinks arising from each subregion of a
required to be specified at internal boundaries between various re- fuel cell. Detailed expressions of these source terms are given in
gions. Generally, fuel cell operation under isothermal conditions is Table I. Specifically, the momentum source term is used to describe
described by mass, momentum, species, and charge conservation Darcy’s drag for flow through porous electrodes, active catalyst lay-
principles. Thus, under the above-mentioned assumptions, the model ers, and the membrane.6,11,16 If there is very small hydraulic perme-
equations can be written, in vector form, as19 ability (e.g., 10214 cm2) in the membrane region, the pore velocity
becomes so small that the inertia and viscous terms in Eq. 4 drop off,
∂( re) and the momentum equation reduces to the well-known Darcy’s law
1 = ?( eru) 5 0 [3]
∂t as used by Bernardi and Verbrugge.6 In addition, electro-osmotic drag
arising from the catalyst layers and the membrane is also included.
∂( reu) Either generation or consumption of chemical species k and the
1 = ?( eruu) 5 2e = p 1 = ?( em eff =u) 1 Su [4]
∂t creation of electric current (see Table I) occurs only in the active cat-
 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000) 4487
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

Table I. Source terms for momentum, species, and charge conservation equations in various regions.

Su Sk SF

Gas channels 0 0 N/A

m 2
Backing layers 2 e u 0 0
K
ja
2 for H 2
2 Fc tota

m k jc
Catalyst Layers 2 e m e mc u 1 F z f cf F = F e for O 2 j
kp kp 4 Fc totc

jc
2 for H 2 O
2 Fc totc

m k
Membrane 2 e m u 1 F z f cf F = F e 0 0
kp kp

alyst layers where electrochemical reactions take place. The Sk and side at the cathodic current collector. Thus, the surface overpotential
SF terms are therefore related to the transfer current between the given by Eq. 15 is only dependent on the membrane phase potential,
solid matrix and the membrane phase inside each of the catalyst lay- which is to be solved from Eq. 6.
ers. These transfer currents at anode and cathode can be expressed as The species diffusivity, Dk, varies in different subregions of the
follows4 PEMFC depending on the specific physical phase of component k.
1/ 2 In flow channels and porous electrodes, species k exists in the
ref
 XH   aa 1 ac  gaseous phase, and thus the diffusion coefficient takes the value in
ja 5 aj0,a 
2
  ?F?h [13]
 X H 2 , ref   RT  gas, whereas species k is dissolved in the membrane phase within the
catalyst layers and the membrane, and thus takes the value corre-
sponding to dissolved species, which is usually a few orders of mag-
ref
 XO   acF  nitude lower than that in gas (see Table II). In addition, the diffusion
jc 5 2aj0,c 
2
 exp 2 ?h [14]
 O 2 , ref 
X RT  coefficient is a function of temperature and pressure,25 i.e.
3/ 2
The above kinetics expressions are derived from the general Butler- T  po 
Volmer equation based on the facts that the anode exhibits fast elec- D(T ) 5 D0     [18]
 To   p
trokinetics and hence a low surface overpotential to justify a linear
kinetic rate equation, and that the cathode has relatively slow kinet- The proton conductivity in the membrane phase has been corre-
ics to be adequately described by the Tafel equation. In Eq. 13 and lated by Springer et al.7 as
14, the surface overpotential, h(x, y), is defined as
 1 
s e (T ) 5 100 exp 1268 2  (0.005139 l 2 0.00326)
h(x, y) 5 Fs 2 Fe 2 Voc [15] 1
  303 T 
where Fs and Fe stand for the potentials of the electronically con-
ductive solid matrix and electrolyte, respectively, at the elec- in S/cm [19]
trode/electrolyte interface. Voc is the reference open-circuit potential
of an electrode. It is equal to zero on the anode but is a function of where the water content in the membrane, l, depends on the water
temperature on the cathode,24 namely activity, a, according to the following fit of the experimental data

Voc,1 5 0.0025T 1 0.2329 [16]

0.043 1 17.18a 2 39.85a 2 1 36.0 a 3 for 0 < a # 1
l5 [20]
where T is in kelvin and Voc is in volts. Notice that Voc is not the true 14 1 1.4( a 2 1) for 1 # a # 3
open-circuit potential of an electrode, which would then depend
upon reactant concentrations according to the Nernst equation. In- The water activity is in turn calculated by
stead, Voc is only the constant part of the open-circuit potential. Its
concentration-dependent part in logarithmic form can then be moved XH 2 O p
a5 [21]
out of the exponent in the Butler-Volmer equation and now becomes p sat
the concentration terms in front of the exponent in Eq. 13 and 14, re-
spectively. Equation 13, which is a rewritten form of the Nernst where the saturation pressure of water vapor can be computed from
equation, precisely describes the effect of decreasing transfer current Springer et al.7
under hydrogen dilution. Based on the experimental data of Partha-
sarathy et al.,24 the dependence of the cathodic exchange current log10 psat 5 22.1794 1 0.02953 (T 2 273.15) 2 9.1837
density on temperature can be fitted as 3 1025 (T 2 273.15)2 1 1.4454 3 1027 (T 2 273.15)3 [22]
i0 (T ) The above calculated saturation pressure is in bars. Apparently, the
5 exp[0.014189 (T 2 353)] [17]
i0 (353 K ) water mole fraction XH2O, water activity, and water content all vary
spatially in the membrane layer, and thus the membrane conductivi-
Under the assumption of a perfectly conductive solid matrix for ty se is a variable.
electrodes and catalyst layers, Fs is equal to zero on the anode side For this multicomponent system, the general species transport
at the anodic current collector and to the cell voltage on the cathode equation given in Eq. 5 is applied to solve for mole fractions of
4488 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000)
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

Table II. Physical parameters and properties at 353 K.

Quantity Value

Gas channel length, L 7.112 cm

Gas channel width 0.0762 cm
Backing layer width 0.0254 cm
Catalyst layer width 0.00287 cm
Membrane width 0.023 cm
Fixed charge concentration, cf 1.2 3 1023 mol/cm3
Oxygen diffusivity in gas 5.2197 3 1022 cm2/s
Hydrogen diffusivity in gas 2.63 3 1022 cm2/s
Dissolved oxygen diffusivity in active layer and membrane 2.0 3 1024 cm2/s
Dissolved hydrogen diffusivity in active layer and membrane 2.59 3 1026 cm2/s
Faraday constant, F 96487 C/mol
Hydraulic permeability of membrane, kp 1.8 3 10214 cm2
Permeability of backing layer, K 1.76 3 1027 cm2
Electrokinetic permeability, kF 7.18 3 10216 cm2
Universal gas constant, R 8.314 J/mol K
Fixed site charge, zf 21
Cathodic transfer coefficient ac 2
Anodic transfer coefficient, aa 2
Backing layer porosity, e 0.4
Membrane water porosity, em 0.28
Volume fraction membrane in catalyst layer, emc 0.4
Inlet nitrogen-oxygen mole ration, XN2/XO2 0.79/0.21
Air-side inlet pressure/fuel-side inlet pressure 5/3 atm
O2 stoichiometric flow ratio, z1 3.0
H2 stoichiometric flow ratio, z2 2.8
Relative humidity of inlet air/fuel (anode/cathode) 100/0%
Reference exchange current density 3 area of anode, ajref
0,a 5.0 3 102 A/cm3
Reference exchange current density 3 area of cathode, aj0,c
ref 1.0 3 1024 A/cm3
Total mole concentration at the anode side, ctota 66.817 3 1026 mol/cm3
Total mole concentration at the cathode side, ctotc 17.808 3 1026 mol/cm3

hydrogen, oxygen, and water vapor. The mole fraction of nitrogen is the outlets, both channels are assumed sufficiently long that veloci-
then obtained by the following constraint ty and species concentration fields are fully developed. The bound-
ary condition for the electrolyte phase potential is no-flux every-
XN2 5 1 2 XH2 2 XH2O on the anode side [23a] where along the boundaries of the computational domain.
XN2 5 1 2 XO2 – XH2O on the cathode side [23b] Numerical procedures.—The conservation equations (Eq. 3-6)
Once the electrolyte phase potential is determined in the mem- were discretized using a finite-volume method26 and solved using a
brane, the local current density along the axial direction can be cal- general-purpose CFD code. Details of the numerical solution proce-
culated as follows dure and the code have been given in previous work.16,18 Worthy of
mention here is that although some species are practically nonexist-
∂F e ing in certain regions of a fuel cell, the species transport equation can
I ( y) 5 2 s eff
e [24]
∂x x 5I.F. still be applied throughout the entire computational domain by using
the large source term technique originally proposed by Voller.27 For
where I.F. means the interface between the membrane and cathode example, there is virtually no hydrogen in the cathode catalyst layer,
catalyst layer. The average current density is then determined by gas-diffusion cathode, and cathode gas channel. Therefore, in these
L subregions, a sufficiently large source term is assigned to the hydro-

∫ I( y) dy
1 gen transport equation, which effectively freezes the hydrogen mole
Iavg 5 [25]
L fraction at zero.
0 The jump condition in the mole fraction across the catalyst/back-
where L is the cell length. ing layer interface as expressed by Eq. 9 can be numerically treated
by a standard interfacial resistance approach which is detailed in the
Boundary conditions.—Equations 3-6 form a complete set of Appendix.
governing equations for (m 1 5) unknowns, where m is the physical Stringent numerical tests were performed to ensure that the solu-
dimension of the problem: u, p, XH2, XO2, XH2O, and Fe. Their tions were independent of the grid size. A 35 3 90 mesh provides
boundary conditions are required only at the external surfaces of the sufficient spatial resolution. The coupled set of equations was solved
computational domain due to the single-domain formulation used. simultaneously, and the solution was considered to be convergent
These are no-flux conditions everywhere except for the inlets and when the relative error in each field between two consecutive itera-
outlets of the flow channels. At the fuel and oxidant inlets, the fol- tions was less than 1025. A typical simulation involving approxi-
lowing conditions are prescribed mately 22,000 unknowns required about 10 min of central process-
uin,anode 5 U0, uin,anode 5 U10 [26] ing unit time on a 600 MHz PC.

XH2,anode 5 XH2,2, XO2,cathode 5 XO2,1, XH2O,anode Results and Discussion

5 XH2O,2, XH2O,cathode 5 XH2,1 [27] Experimental validation.—To validate the numerical model pre-
sented in the preceding section, comparisons were made to the
The inlet velocities of fuel and oxidant can also be expressed by their experimental data of Ticianelli et al.23 for a single cell operated at
respective stoichiometric flow ratios, i.e., z2 and z1 at 1 A/cm2. At two different temperatures. The parameters used in the following
 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000) 4489
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

Figure 4. Local current density distributions in the axial distance (y direc-

tion) for X 0H2O,1 5 0 and T 5 353 K.
Figure 2. Comparison of predicted and measured cell polarization curves at
two operating temperatures. The experimental data were adapted from
Ticianelli et al.23 fields within the porous cathode and its adjacent flow channel for the
experimental fuel cell at 353 K. The velocity exhibits parabolic pro-
files in both anode and cathode gas channels and reduces essentially
simulations were taken from Bernardi and Verbrugge6 and are sum- to zero in the regions of porous structure. The water vapor mole frac-
marized in Table II. tion increases along the cathode gas channel while the mole fraction
Figure 2 compares the computed polarization curves with the of oxygen decreases, due to the electrochemical reaction within the
measured ones. The calculated curves show good agreement with the cathode catalyst layer, where oxygen is reduced by protons migrat-
experimental data for both temperatures. The product of the specific ing through the membrane from the anode side to produce water. A
interfacial area and the exchange current density for the anodic reac- detailed oxygen mole fraction profile along the x direction in the
tion used in these simulations is so large that the predicted surface midlength plane (i.e., y 5 L/2) of the cathode is displayed in Fig. A-2
overpotential prevailing in the anode catalyst layer is practically neg- of the Appendix, demonstrating that the jump condition in oxygen
ligible, which is in accordance with the experimental measurement. concentration at the interface between the catalyst and backing lay-
While no data of velocity and concentration fields were provided ers is successfully captured by the present computer model.
in the experiments of Ticianelli et al.,23 these detailed results can Figure 4 displays the local current density distributions at various
also be generated from the present model to shed light on the inter- average current densities along the axial direction of fuel cells. The
nal operation of the fuel cell. Figure 3 shows the predicted flow field current density distributions are uniform at all cell average current
inside the entire cell and the water vapor and oxygen mole fraction densities, probably due to the small cell and high stoichiometric flow
ratios employed in these experiments.
In the present model, it is assumed that water only exists in the
vapor state. However, if the electrochemical reaction rate is suffi-
ciently high, the amount of water produced is condensed into the liq-
uid phase. In this situation, two-phase flow and transport may have to
be considered. Figure 5 displays water vapor mole fraction profiles
along the interface between the catalyst layer and gas diffuser on the
cathode side where there is the highest content of water vapor. When
the cell current density is higher than about 0.6 A/cm2, the water
vapor mole fraction along the catalyst layer/gas diffuser interface

Figure 3. (a) Computed velocity profiles in anode and cathode flow chan-
nels, (b) water vapor mole fraction, and (c) oxygen mole fraction in the cath-
ode gas diffuser and flow channel for Vcell 5 0.6 V, X 0H2O,1 5 0, and T 5
353 K. The left boundary in (b) and (c) is the interface between catalyst layer Figure 5. Water vapor mole fraction profiles along the interface between the
and gas-diffusion cathode. cathode catalyst layer and gas diffuser for X 0H2O,1 5 0 and T 5 353 K.
4490 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000)
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

Figure 6. Hydrogen dilution effect on the cell average current density for
X 0H2O,1 5 0, Vcell 5 0.6 V, and T 5 353 K. Figure 7. Effect of the inlet hydrogen mole fraction on cell polarization
curves for X 0H2O,1 5 0.

already exceeds the saturated level, i.e., X Hsat2O 5 0.093541 at a pres-

sure of 5 atm and a cell temperature of 353 K. Hence, it suggests that ered, particularly in the range below 50%. The cell polarization
the two-phase flow regime already starts at intermediate current den- curves under pure and hydrogen dilution conditions are displayed in
sities. Nonetheless, a single-phase analysis such as the present model Fig. 7, showing qualitatively good agreement with the most recent
and all existing models6-11 still provides a good first approximation experiments of Rockward et al.1 A quantitative comparison is not
of cell performance for current densities above 0.6 A/cm2. This is true possible due to a lack of specific information about the experiments
because the water distribution affects the electrochemical process and that were only reported in a meeting abstract.1
oxygen transport in the air cathode primarily in two ways. Figure 8 shows predicted profiles of the membrane phase poten-
First, the presence of liquid water affects the water content in the tial across the anode catalyst layer, the membrane, and the cathode
membrane and thus slightly alters its ionic conductivity. More im- catalyst layer at various axial locations for Vcell 5 0.85 and 0.6 V.
portant, liquid water present in the gas diffusion cathode hampers The membrane phase potential is in reference to the anode solid
oxygen transport to the catalyst layer. However, recent two-phase potential of zero. For pure hydrogen, nearly zero overpotential is
calculations15,28 indicated that there is only 5% of the liquid water predicted within the anode catalyst layer, indicating that the hydro-
saturation at current densities as high as 1.5 A/cm2, making the gen oxidation kinetics on the anode is much faster than the cathode
effect of liquid water existence on the oxygen transport likely mini- reaction. Furthermore, the membrane phase potential profile remains
mal. For these reasons, the present single-phase analysis is still ap- similar at different axial locations. However, in the presence of large
plied to predict cell performance for current densities greater than hydrogen dilution, the overpotential for the hydrogen oxidation
0.6 A/cm2 in the following without consideration given to the water within the anode catalyst layer significantly increases as the hydro-
distribution. Rigorous two-phase analysis, however, is the subject of gen mole fraction at the reaction surface is drastically reduced. This
future work. is easily explained by Eq. 9 where the transfer current is proportion-
al to the square root of the hydrogen mole fraction. At low operating
Hydrogen dilution effects.—When reformate gas is used as the current densities (e.g., < 0.2 A/cm2), hydrogen is not being depleted
anode feed, the hydrogen mole fraction at the anode inlet is signifi- in the catalyst layer or along the flow channel (since H2 stoichiome-
cantly lower than that in the pure hydrogen condition (i.e., hydrogen try is 2.4 at 1 A/cm2). Hence, the overpotential changes negligibly
plus water vapor only). Consequently, hydrogen dilution has a strong for hydrogen bulk concentrations ranging from one hundred to ten
effect on cell performance. A series of simulations for different percents. This can be seen clearly from Fig. 8 left side. However, at
hydrogen inlet fractions at the anode were carried out to illustrate large current densities, the anode gas is progressively depleted of
this effect. In all these simulations, the stoichiometric flow ratios of hydrogen downstream along the gas channel as it is consumed by the
fuel and oxidant are fixed at 2.8 and 3.0 at 1 A/cm2, respectively. electrochemical reaction, and thus the overpotential required to sus-
Figure 6 and 7 show that the cell current density at a cell voltage of tain a particular current density becomes substantially larger. There
0.6 V decreases significantly as the inlet hydrogen content is low- are also appreciable differences in the membrane phase potential

Figure 8. Phase potential distributions in

the transverse direction (x direction) for
X 0H2O,1 5 0, T 5 353 K, and z2 5 2.8 at
1 A/cm2 for pure H2.
 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000) 4491
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

profile at various locations along the channel, as seen from Fig. 8

right. A limiting current density may occur when hydrogen is com-
pletely depleted at the reaction surface, thus leading to an infinitely
large anode polarization loss. The prediction that there is no perfor-
mance change under hydrogen dilution for low current densities dis-
agrees with the experimental observations of Rockward et al.,1 indi-
cating possible contributors other than the mass-transfer limitation
modeled here.
Figure 9 shows the two-dimensional contours of hydrogen mole
fraction in the anode gas channel and porous electrode under Vcell 5
0.6 V for 10% hydrogen dilution and for no dilution. Under the hy-
drogen dilution condition, the hydrogen mole fraction decreases
along the flow channel as well as across the porous anode. These
trends are more apparent from Fig. 10, which displays the hydrogen
mole fraction profiles at several important boundaries. The down-
the-channel effect becomes particularly significant in the high fuel
utilization conditions that are needed for the highest possible fuel
efficiency. However, this effect can be alleviated by using large fuel
flow rates, i.e., at the expense of fuel utilization. For example, the Figure 10. Hydrogen mole fraction profiles at (a) the flow channel wall, (b)
down-the-channel effect is not significant in the case displayed in the interface between the anode flow channel and gas-diffusion electrode,
Fig. 10. The substantial drop of H2 mole fraction across the anode is and (c) the interface between the gas-diffusion anode and catalyst layer for
caused by diffusion of hydrogen from the anode stream to the reac- X 0H2O,1 5 0, Vcell 5 0.6 V, and T 5 353 K.
tion surface, a limiting step for the cell current density in this case.
On the contrary, the hydrogen supply from the fuel stream to the step change from 0.6 to 0.55 V. The figure shows that the cell cur-
reaction surface appears adequate in the case of no dilution, as seen rent density abruptly adjusts to a new value because the charge-
in Fig. 9b. transfer reaction and membrane potential equation respond instanta-
The above simulations provide a good illustration of the capabili- neously. Subsequently, the concentration fields begin to evolve,
ties of the present model. Although the results for hydrogen dilution gradually reducing the cell current density to the steady-state value
effects are realistic, more detailed analyses and parametric studies are corresponding to Vcell 5 0.55 V. This transient simulation demon-
desirable to fully investigate hydrogen dilution and high fuel utiliza- strates the interesting phenomenon of current overshoot as well as
tion effects. These studies are the subject of work in the near future. the fuel cell transient response time of the order of a couple of sec-
The present CFD model can also simulate the transient response onds. Detailed transient results are reported in a separate publication
of a fuel cell to dynamically varying operating conditions. Figure 11 due to the space limitation of this paper.
shows a preliminary example in which the cell voltage undergoes a
Conclusion
A single-domain formulation was developed to comprehensively
describe electrochemical kinetics, current distribution, hydrodynam-
ics, and multicomponent transport in hydrogen PEMFCs. A finite-
volume-based CFD technique was successfully adapted to simulate
multidimensional behaviors of the fuel cell. The CFD model was
able to predict not only the experimental polarization curves of
Ticianelli et al.23 but also the detailed reactant and product distribu-
tions inside the cell. In addition, the CFD model was used to under-
stand the hydrogen dilution effect when the reformate gas is used as
the anode feed. Hydrogen dilution leads to a much lower cell current
density that is limited by the diffusive transport of hydrogen to the
reaction site. Efforts are presently underway to undertake three-di-
mensional and two-phase analyses of PEMFCs.

Figure 9. Hydrogen mole fraction distributions in the anode flow channel

and gas diffusion electrode for X 0H2O,1 5 0, Vcell 5 0.6 V, T 5 353 K, and
z2 5 2.8 at 1 A/cm2. (a) 10% hydrogen X 0H2,2 5 0.08441) and (b) pure Figure 11. Transient response of current density from Vcell 5 0.6 to 0.55 V
hydrogen X 0H2,2 5 0.8441). for X 0H2,1 5 0, T 5 353 K, and z2 5 2.8 at 1 A/cm2 for pure H2.
4492 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000)
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

Acknowledgments J
hint 5 [A-4]
This work was partially supported by Sandia National Laborato-  ( dx )cw2 J 
ries under contract no. BF-6597. Sandia is a multiprogram laborato- (1 2 H ′ )c P 2 
ry operated by Sandia Corporation, a Lockheed Martin Company,  Dg 
for the United States Department of Energy under contract DE- The overall conductivity at the interface for use in a CFD model is tradition-
AC04-94AL85000. The work is also supported by NSF under grant ally defined as
no. DUE-9979579.
k int
J 5 (c P 2 c W ) [A-5]
The Pennsylvania State University assisted in meeting the publication ( dx ) W
costs of this article.
Therefore, one obtains the following from Eq. A-1
Appendix
( dx ) W ( dx ) w1 ( dx ) w2 1
Numerical Treatment of Concentration Discontinuity at Catalyst 5 1 1 [A-6]
k int De Dg hint
Layer/Backing Layer Interface
Consider the concentration profile in the two control volumes surround- Substituting Eq. A-4 into A-6 yields
ing the interface between the catalyst and backing layers, as shown in
Fig. A-1. Following Henry’s law, the concentration undergoes a discontinu- 21
f (1 2 fw )  cP (1 2 fw ) 
ity across the interface because it represents the value in the membrane phase kint 5  w 1 1 (1 2 H ′ )? 2  [A-7]
within the catalyst layer but stands for the value in the gas phase within the  De

Dg  J (δx ) W Dg 

backing layer. In Fig. A-1, subscripts in capital letters denote central nodal
points of control volumes, whereas the corresponding subscript in lowercase where
represents the face of the control volume.
The total diffusion flux across the interface can be written as follows ( dx ) w1
fw 5 [A-8]
( dx ) W
(c 2 cw2 ) (c 2 cW )
J 5 Dg P 5 De w1 5 hint (cw2 2 cw1 ) [A-1]
( dx ) w2 ( dx ) w1 Equation A-7 can be simply implemented in a CFD code to simulate the
jump condition at the catalyst/backing interface. Note that if H9 51 (i.e., in
where hint is an effective interfacial mass-transfer coefficient introduced to the absence of a jump condition), the last term in Eq. A-7 vanishes and Eq. A-
numerically model the concentration discontinuity. Its value can thus be de- 7 identically reduces to the conventional geometrical mean formula. Howev-
termined by er, if H9 < 1, an additional resistance arises at the interface, effectively simu-
lating the jump condition in the concentration. Note that Eq. A-7 must be iter-
J J J ated to update the flux, J; however, calculations indicated that this iteration
hint 5 5 5 [A-2]
cw2 2 cw1 cw2 2 H ′cw2 (1 2 H ′)cw2 is quickly convergent.
A representative oxygen mole fraction profile in the midlength plane (i.e.,
Since y 5 L /2) across the entire air cathode, including the membrane, catalyst
( dx ) w2 J layer, and backing layer, is shown in Fig. A-2. The cell operating conditions
cw2 5 c P 2 [A-3] are identical to those corresponding to Fig. 3. The oxygen mole fraction starts
Dg from about 0.17 at the outer face of the backing layer due to gradual deple-
tion along the flow channel, further reduces to approximately 0.14 in the gas
Equation A-2 can be rewritten as phase at the catalyst/backing interface due to mass-transport resistance with-
in the backing layer, and undergoes a discontinuous drop at the interface to
about 0.02, a value for the membrane phase in accordance with Henry’s law.
Subsequently, all oxygen in the membrane phase is consumed by the oxygen
reduction reaction, which occurs within a short distance from the cata-
lyst/backing interface. The general trends are consistent with those predicted
by the one-dimensional model of Bernardi and Verbrugge.5

List of Symbols
a effective catalyst area per unit volume, cm2/cm3
A superficial electrode area, cm2

Figure A-1. Schematic of concentration profiles in the two control volumes Figure A-2. Oxygen mole fraction profile across the membrane, catalyst
in the vicinity of the interface between the catalyst layer and the backing layer, and backing layer on the cathode side at y 5 L/2. Cell conditions are
layer in the cathode. identical to those in Fig. 3.
 Journal of The Electrochemical Society, 147 (12) 4485-4493 (2000) 4493
S0013-4651(00)08-104-0 CCC: $7.00 © The Electrochemical Society, Inc.

c molar concentration, mol/cm3 References

D mass diffusivity of species, cm2/s 1. T. Rockward, T. Zawodzinski, J. Bauman, F. Uribe, J. Valerio, T. Springer, and
I current density, A/cm2 S. Gottesfeld, Abstract 566, The Electrochemical Society Meeting Abstracts,
j transfer current, A/cm3 Vol. 99-1, Seattle, WA, May 2-6, 1999.
p pressure, Pa 2. J. K. Hong, L. A. Zook, M. Inbody, J. Tafoya, and N. E. Vanderborgh, Abstract 570,
The Electrochemical Society Meeting Abstracts, Vol. 99-1, Seattle, WA, May 2-6,
R gas constant, 8.314 J/mol K
1999.
S source term in transport equations 3. K. B. Prater, J. Power Sources, 51, 129 (1994).
t time, s 4. S. Gottesfeld, in Advances in Electrochemical Science and Engineering, C. Tobias,
T temperature, K Editor, Vol. 5, p. 195, John Wiley & Sons, New York (1997).
u velocity vector, cm/s 5. D. M. Bernardi and M. W. Verbrugge, AIChE J., 37, 1151 (1991).
U inlet velocity, cm/s 6. D. M. Bernardi, and M. W. Verbrugge, J. Electrochem. Soc., 139, 2477 (1992).
V cell potential, V 7. T. E. Springer, T. A. Zawodinski, and S. Gottesfeld, J. Electrochem. Soc., 136, 2334
X mole fraction of species (1991).
8. T. E. Springer, M. S. Wilson, and S. Gottesfeld, J. Electrochem. Soc., 140, 3513
Greek (1993).
l membrane water content, H2O/mol SO2
3
9. T. F. Fuller and J. Newman, J. Electrochem. Soc, 140, 1218 (1993).
10. T. V. Nguyen and R. E. White, J. Electrochem. Soc., 140, 2178 (1993).
F phase potential, V 11. V. Gurau, H. Liu, and S. Kakac, AIChE J., 44, 2410 (1998).
h overpotential, V 12. J. S. Yi and T. V. Nguyen, J. Electrochem. Soc., 145, 1149 (1998).
m viscosity, kg/m s 13. J. S. Yi and T. V. Nguyen, J. Electrochem. Soc., 146, 38 (1999).
n kinematic viscosity, cm2/s 14. T. V. Nguyen, Paper 880 presented at The Electrochemical Society Meeting, Seat-
q volumetric flow rate, cm3/s tle, WA, May 2-6, 1999.
r density, kg/cm3 15. C. Y. Wang, Z. H. Wang, and Y. Pan, in Proceedings of ASME Heat Transfer Divi-
s ionic conductivity, S/cm sion-1999, HTD-Vol, 364-1, ASME, New York, p. 351 (1999).
16. W. B. Gu, C. Y. Wang, and B. Y. Liaw, J. Electrochem Soc., 144, 2053 (1997).
z stoichiometric flow ratio, Eq. 1
17. W. B. Gu, C. Y. Wang, and B.Y. Liaw, J. Power Sources, 75/1, 154 (1998).
Superscripts 18. W. B. Gu, C. Y. Wang, and B. Y. Liaw, J. Electrochem Soc., 145, 3418 (1998).
19. C. Y. Wang, W. B. Gu, and B.Y. Liaw, J. Electrochem. Soc., 145, 3407 (1998).
o gas channel inlet value
20. C. Y. Wang, W. B. Gu, R. Cullion, and B. Thomas, in Proceedings of IMECE99,
eff effective value ASME, New York (1999).
sat saturation value 21. W. B. Gu, C. Y. Wang, S. Li, M. M. Geng and B. Y. Liaw, Electrochim. Acta, 44,
Subscripts 4525 (1999).
22. W. B. Gu, C. Y. Wang, J. Weidner, R. Jungst, and G. Nagasubramanian, J. Elec-
a anode trochem. Soc., 147, 427 (2000).
c cathode 23. E. A. Ticianelli, C. R. Derouin, and S. Srinivasan, J. Electroanal. Chem., 251, 275
e electrolyte (1988).
k species 24. A. Parthasarathy, S. Srinivasan, and A. J. Appleby, J. Electrochem Soc., 139, 2530
m membrane (1992).
oc open circuit 25. R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena, John Wiley
& Sons, New York (1960).
ref reference value
26. S. V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere, New York
s solid phase of electrode (1980).
u momentum equation 27. V. R. Voller, Numer. Heat Transfer, Part B, 17, 155 (1990).
F potential equation 28. T. V. Nguyen, in Tutorials in Electrochemical-Engineering Mathematical Modeling,
1 cathode R. F. Savinell, J. M. Fenton, A. West, S. L. Scanlon, J. Weidner, Editors, PV 99-14,
2 anode p. 222, The Electrochemical Society Proceedings Series, Pennington, NJ (199).