Journal of Power Sources 101 (2001) 275±286
Modeling polymer electrolyte fuel cells: an innovative approach
G. Maggio*, V. Recupero, L. Pino
CNR-TAE Institute, Via Salita S., Lucia sopra Contesse 5, 98126 Santa Lucia, Messina, Italy
Received 18 December 2000; received in revised form 2 March 2001; accepted 5 March 2001
Abstract
In this paper, a mathematical simulation model is proposed to describe the water transport in proton conductive membranes, used in
polymer electrolyte fuel cells (PEFCs). The model, which includes the calculation of electrochemical parameters of a PEFC, represents a
quite innovative approach. In fact, it is based on the use of original mathematical relationships taking into account diffusional and ohmic
overpotentials for electrode ¯ooding and membrane dehydration problems.
The calculated performance of polymer fuel cells using a Na®on 117 membrane clearly demonstrates the model validation (3%
variation with respect to experimental data). Besides, analysis of model results allows a useful comparison of two different membranes
(Na®on 117, Dow) in order to de®ne the best membrane/electrode assembly. # 2001 Elsevier Science B.V. All rights reserved.
Keywords: Electrode ¯ooding; Membrane dehydration; Modeling; Polymer electrolyte fuel cells; Water transport
1. Introduction
The polymer electrolyte fuel cell (PEFC) operates at a low
temperature (70±908C), and promises to be one of the most
serious candidates for both stationary and automotive applications as a substitute of traditional systems (thermoelectric
power plants, internal combustion engine, etc.), due to wellknown advantages (high ef®ciency, no pollutant emission,
etc.).
The stringent requirements in terms of compactness, high
energy density, performance stability and low cost, move the
R&D in the direction of optimizing the different aspects of
the PEFC systems. One approach is the development of
theoretical analysis in order to identify and to solve problems related with the above-mentioned items.
In this respect, several mathematical models have been
proposed in literature for PEFCs [1±25]; all these models
aimed at determining the potential or cell voltage as a
function of the current density and, in general, of the
operating conditions.
One of the key objectives of the current research is an
adequate water content in the electrolyte to achieve a good
ionic conductivity. In fact, membrane dehydration causes
an increase in resistance to the proton ¯ow (the ionic
charges move through the sulfonic groups of the polymeric
*
Corresponding author. Tel.: 39-90-624227; fax: 39-90-624247.
E-mail address:
[email protected] (G. Maggio).
membrane surrounded by water molecules) and possible
deadhesion of the membrane from the electrode. On the
other hand, for excess water content, the cathode is ¯ooded
and the diffusion of the reactant gas through its pores is
hindered. Therefore, an accurate evaluation of the water
transport in PEFCs, often made by modeling approaches [1±
5], is desirable. Nevertheless, a formulation that directly
relates the performance of a polymer fuel cell with the
cathode ¯ooding and/or the dehydration problems at the
membrane/electrode interfaces has never been proposed.
As evidenced in literature [11], the cathode ¯ooding,
typical of high current density operation, involves a decrease
in the porosity available for gas diffusion in the electrode.
Likely, it can be inferred that the membrane dehydration
phenomenon, that especially occurs on the anode side at
high current densities, involves a decrease in ionic conductivity [7±9,11,26].
Starting from these considerations, a mathematical model
has been developed. It has been validated and applied
to polymer fuel cells based on Na®on 117 and Dow
membranes.
2. Model development
In this paragraph, a description of the mathematical model
developed for the solid polymer electrolyte fuel cell is
presented. The main assumptions of the model are: (i)
one-dimensional treatment; (ii) isothermal and steady state
0378-7753/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 7 7 5 3 ( 0 1 ) 0 0 7 5 8 - 3
276
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
Nomenclature
A
cH
d
DH
DO2 N2
EW
F
I
Ilim
Io
Io,Pt
kp
kF
ld
lm
M
M1,max
M2,max
OCV
p
psat
w
R
Rcell
Rmemb
SPt
T
UPt
v
Vcell
Vm
W
WPt
x
superficial electrode area (cm2)
fixed-charge concentration (mol cm 3)
membrane average pore diameter (cm)
protonic diffusion coefficient (cm2 s 1)
oxygen/nitrogen binary diffusion coefficient at
standard conditions (cm2 s 1)
membrane equivalent weight (g mol 1)
Faraday constant (96,487 C per eq.)
current density (A cm 2)
limiting current density (A cm 2)
exchange current density (A cm 2)
exchange current density per Pt surface
(A cm 2 Pt)
membrane hydraulic permeability (cm2)
membrane electrokinetic permeability (cm2)
cathode gas-diffusion layer thickness (cm)
wet membrane thickness (cm)
water amount to be added to the cell (mol s 1)
water amount corresponding to complete
flooding (mol s 1)
water amount corresponding to complete
dehydration (mol s 1)
open circuit voltage (V)
gas pressure (atm)
saturated vapor pressure (atm)
universal gas constant (8.3143 J mol 1 K 1)
cell resistance (O cm2)
membrane resistance (O cm2)
catalyst surface area (cm2 mg 1)
cell temperature (K)
catalyst utilization
water velocity in membrane pores (cm s 1)
cell voltage (V)
standard molar volume (22,414 cm3 mol 1)
water flow rate (mol cm 2 s 1)
catalyst loading (mg cm 2)
gas-phase mole fraction
Greek symbols
d
pore-water density (mol cm 3)
ddry
membrane dry density (g cm 3)
d
eg
gas porosity in gas-diffusion layer
em
membrane water porosity
w
Z
cell overpotential (V)
k
membrane ionic conductivity (O 1 cm 1)
m
pore-water viscosity (kg m 1 s 1)
membrane resistance/cell resistance ratio
tR
o
empirical constant for diffusional overpotential (O cm2 K 1)
x
membrane water-transport ratio
z
stoichiometric ratio
Subscripts and superscripts
act
activation
conv
dif
el
in
o
ohm
out
sat
tr
um
w
convective
diffusional
electrochemical reaction
cell inlet
initial or inlet to gas chamber
ohmic
cell outlet
saturated condition
water transport
humidification condition
water
cathode side
anode side
conditions; (iii) ideal and uniformly distributed gases; (iv)
electrode pores for gas ¯ow are separated from pores for
liquid water; (v) the inlet gas temperatures are equal to the
cell temperature.
A set of input data is used for model calculations that, for
convenience, have been divided into two distinct sections:
water balance calculations and electrochemical calculations.
It must be remarked that the gas porosity in the cathode
diffusional layer and the ionic conductivity of the membrane, which affect diffusional and ohmic overpotentials,
strictly depend on the water balance conditions calculated at
any current density value. The core of the calculations is
preceded by the determination of the membrane properties.
2.1. Membrane properties calculations
To calculate the membrane properties, the following
equations have been used:
cH
kF
ddry 1 em
w
;
EW
(1)
2
em
wd
;
80
(2)
m 1:002 10
1:3272 293 T 0:001053 293 T2 = T 168 3
;
(3)
d 3:014 10
2
4:179 10
DH DH
ko
22 C
4
2:216 10 T
10 3
7 2
5:842 10 T
T ;
(4)
T m 22 C
;
m 295
(5)
F2
D c ;
RT H H
(6)
where m(228C 0:9548 10 3 kg m 1 s 1 is the water
viscosity at 228C. In particular, the equations for fixedcharge concentration (Eq. (1)), electrokinetic permeability
(Eq. (2)), and initial ionic conductivity (Eq. (6)) have already
been reported in literature [2,27]. The equations for porewater viscosity (Eq. (3)) and density (Eq. (4)) have been
obtained by extrapolation of data reported in [28], in the
277
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
proper ranges of temperature (20±100 and 40±1008C,
respectively). Finally, the protonic diffusion coefficient
(Eq. (5)) has been derived on the basis of the relationship
Dm=T constant [2].
At the ®rst calculation loop, corresponding to OCV
conditions, it will be
k ko :
(7)
2.2. Water balance equations
The model makes a set of calculations in order to verify on
the basis of ®xed experimental conditions, the water transport characteristics of the membrane and to suggest some
solutions in order to achieve an optimal water transport in
the fuel cell.
As already mentioned, a proper water balance is critical
for good operation of a polymer fuel cell, also during
transient response. It is necessary to identify operative
conditions able to avoid electrode ¯ooding and membrane
dehydration problems which can drastically reduce the cell
performance.
In particular, at low current density, dehydration of the
cathode side of the membrane can occur: the water produced
within the cathode and that dragged through the membrane
from the anode to the cathode due to electro-osmosis (with
H protons) could be insuf®cient. On the contrary, at high
current density, the water production at the cathode can
cause a ¯ooding of the electrode. Moreover, the electroosmotic ¯ow of water increases with current, and the anode
side of the membrane is thus subjected to dehydration,
which is only partially counterbalanced by back diffusion
(from cathode to anode). Then, with the exception of the
intermediate current density region, at low and high current
densities, a proper water balance can be obtained only by
using appropriate expedients, e.g. by humidi®cation of the
gases entering the fuel cell or by supplying (removing)
liquid water.
Fig. 1 outlines the water ¯ows inside the cell. According
to [29], the amount of water produced by chemical reaction
due to cross-over of gases can be neglected.
The following equations have been used to determine the
water ¯ow rates related to the different phenomena:
Water transport :
Wtr
Electrochemical reaction :
Anode humidification :
I
x:
F
(8)
Wel
W um
I
:
2F
I
z
2F 1
(9)
xum
w
xum
w
:
(10)
Wum
I
z
4F 1
1
xoN2
xoO2
!
Anode outlet :
:
(11)
W out
I
z
2F
Cathode outlet :
"
!
xoN2
I
out
z 1 o
W
4F
xO 2
1
1
#
xsat
w
:
1 xsat
w
xsat
w
:
1 xsat
w
(12)
(13)
According to [1,3,4], the membrane water-transport ratio
can be calculated as
x em
wF
dv
;
I
(14)
where v is the water velocity in the membrane pores,
determined by the following equation:
1
F 2 c2H kF
kF cH FI kp dp
v
;
(15)
1
mk
mk
m dz
which establishes that the water transport in the membrane
depends on two contributions: the electro-osmotic transport
associated with the flow of the H protons, and the diffusive
transport due to a pressure gradient in the membrane (dp/dz).
This last term can be obtained by known relationships
[1,3,4].
The gas-phase mole fractions of water corresponding to
saturated condition have been calculated as
xsat
w
psat
w
;
p
(16)
xsat
w
psat
w
;
p
(17)
with
psat
exp
13:669
w
Cathode humidification :
xum
w
xum
w
Fig. 1. Schematic of water balance model Ð g.-d.l.: gas-diffusion layer;
c.l.: catalyst layer; W um , Wum : water entering the cell due to fuel and
oxidant humidification, respectively; W in , Win : water directly added (e.g.
by wicking) to the anode and cathode compartment, respectively; Wel:
water produced in the cell (on the cathode side) from the electrochemical
reaction; Wtr: water transported through the membrane from anode to
cathode due to drag (electro-osmosis) and diffusion (or back diffusion)
phenomena; W out , Wout : water outgoing from anode and cathode
compartment, respectively.
5096:23
:
T
(18)
Analogously, the mole fractions of water corresponding to
um
humidification condition (xum
w , xw ) can be determined by
278
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
considering the humidification temperatures, instead of the
cell temperature in Eq. (18).
The water balance is expressed by the following relationships:
At the anode :
W out W in W um
Wout
At the cathode :
Win
Wum
Wtr ;
(19)
Wel Wtr :
(20)
The amounts of water to be supplied at the anode and at
the cathode compartment to achieve the water balance
are
in
M W A;
(21)
M Win A:
(22)
To calculate the different overpotentials, the following
relationships have been considered:
RT
I
ln
:
(25)
Activation overpotential :
Zact
F
Io
ohmic overpotential :
Zohm IRcell :
(26)
Ilim
:
oTI ln
Ilim I
(27)
Diffusional overpotential :
Zdif
Convective overpotential :
Zconv FcH v
lm
:
k
(28)
As well known [30,31], under operative conditions, the
voltage of a fuel cell decreases, compared with the ideal or
reversible theoretical value, because of the irreversible
losses associated with activation, ohmic, and diffusional
polarizations. Once estimated accurately such effects, it is
possible to obtain a suf®ciently correct description of the cell
performance.
Cell voltage can be determined by subtracting the overpotentials related to the cell irreversibilities from the open
circuit voltage (OCV),
In Eq. (26), the cell resistance is considered to vary as a
function of the current density, depending on the water
balance conditions.
Eq. (27) for diffusional overpotential is different from that
commonly reported in most handbooks on fuel cells [30,31].
Here, the pre-logarithmic term is assumed to be dependent
on current density, according to [16,34], whose purpose was
to simulate the cell behavior in the entire current density
region through a rational description of mass-transport
problems. Moreover, unlike the classical approaches, the
limiting current density (Ilim) varies as a function of the cell
current density, and decreases in consequence of the ¯ooding of the cathode gas-diffusion layer (see Eq. (31)).
Eq. (28) provides the overpotential associated with proton
convection, which depends on the protonic concentration
and water velocity in the membrane pores [1±4].
Exchange current density per geometrical electrode surface, cell resistance, and limiting current density have been
determined by the following equations:
Vcell OCV
Io Io;Pt SPt WPt UPt ;
It is evident that when these last equations give negative
values, a removal, instead of addition, of the corresponding
water amounts is required to obtain the balance.
2.3. Electrochemical calculation equations
Zact
Zohm
Zdif
Zconv ;
(23)
where Zact is the activation overpotential, Zohm the ohmic
overpotential, and Zdif the diffusional overpotential; the last
term Zconv has been introduced on the basis of the following
considerations.
The ¯ux of a dissolved species in the membrane pores is
due to three effects: migration, diffusion, and convection
[2,32,33]. The ®rst one corresponds to the ohmic term
related to the membrane (already included in Zohm), the
second is associated with the so-called diffusion potential,
and the third to the effect of the potential gradient on the
velocity of the charged ¯uid. In the case of a PEFC, the only
mobile ions in the membrane pores are hydrogen ions,
whose concentration in the membrane is constant; thus,
there is no proton transport due to diffusion [2,33]. Therefore, only the proton transport for convection, to which the
overpotential Zconv is associated, has to be considered.
The OCV has been calculated by using the following
equation [1±4]:
OCV 1:23
0:9 10
3
T
298
RT
ln p2H2 pO2 :
4F
(24)
Rcell
Ilim
(29)
Rmemb
lm
;
tR
ktR
2FDO2 N2 edg 1:5 T=2730:823
ln 1
Vm ld
(30)
xO2 ;
(31)
where Io,Pt is the exchange current density referred to Pt
surface, SPt the catalyst surface area, WPt the catalyst loading
in the electrode, UPt the catalyst utilization coefficient,
tR 1 the membrane resistance/cell resistance ratio,
assumed to be constant (Eq. (31) from [11]).
Eq. (31) gives the dependency of the limiting current on
the gas porosity in the cathode diffusional layer, in agreement with the results obtained in [1±4] and with the considerations explained in [11]: gas transport limitations in the
cathode diffusional layer determine the limiting current of
the cell.
Finally, the gas porosity in the cathode diffusional layer
and the membrane ionic conductivity have been determined
from the following expressions:
jM j
d
d;o
eg eg 1
(32)
; if M < 0;
M1;max
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
8
>
>
>
ko 1
>
<
k
>
>
>
o
>
:k 1
M;
;
M2;max
M
1
M2;max
if M; > 0 and M
;
0;
M
; if M > 0 and M > 0;
M2;max
being egd;o the initial porosity in the gas-diffusion layer, ko the
initial membrane ionic conductivity calculated by Eq. (6),
M and M the amounts of water required for water balance
on the cathode and on the anode side, respectively, derived
from previous calculations (Eqs. (21) and (22)), M1,max the
water amount corresponding to a complete flooding, and
M2,max the water deficit corresponding to complete dehydration. As a first approximation, it has been assumed
M2;max M1;max ;
(34)
that is, the water deficit that results in complete dehydration
at the membrane/electrode interfaces is equal to the excess
water amount that induces the complete flooding of the
cathode gas-diffusion layer.
3. Innovative aspects of the proposed model
All the mathematical models for polymer fuel cells so far
proposed assume the gas porosity in the electrode diffusional
layer to be constant (for example, 40% in [1±4]). Nevertheless, Springer et al. [11] have previously evidenced that
this assumption does not explain the limiting current experimentally obtained and their dependence on gas composition
and pressure. As was observed, the average effective porosity of the gas-diffusion layer should be signi®cantly lower
(from 25 to 18.6% between 0 and 1.5 A cm 2 in [11]) than
the ``natural'' porosity (ca. 40%) of the carbon cloth electrode. The reason is a noticeable ``partial ¯ooding'' of the
cathode by the liquid water produced that slows down the
oxygen transport.
To express such an effect, we assumed that the gas
porosity in the diffusional layer decreases, in case of cathode
¯ooding (M < 0), according to Eq. (32). In other words, it
is assumed that the gas porosity in the diffusional layer
depends on current density; it decreases when the latter
increases, in agreement with [11].
In the mathematical model presented in [11], the effective
porosity (liquid water-free pores) could be ``adjusted''
and was considered either independent from the current
density or decreased in proportion to the rate of water
production at the cathode. To obtain reliable values of the
effective porosities corresponding to a given fuel cell, the
authors were forced to realize a simultaneous ®tting of
several polarization curves of the cell [11] recorded
under different operative conditions. Our approach, differently, allows direct calculation of the gas porosity through
Eq. (32).
279
(33)
As the cathode ¯ooding involves a decrease in the porosity available for the gas that diffuses through the electrode,
it can be argued that the membrane dehydration involves a
decrease in ionic conductivity. As a consequence, the overall
cell resistance, to which the membrane strongly contributes,
will correspondingly increase.
Even though such an effect is neglected in mathematical
models for fuel cells (except [9]), where a constant resistance is considered, it is evidenced in some experimental
works [7,8,11,26]: e.g. a resistance increase in the order of
10% going from 0 to 0.8 A cm 2 for a cell based on Na®on
117 is observed in [26].
Therefore, in analogy with the method used for the gas
porosity, we assumed that the membrane ionic conductivity
decreases, in case of cathode (M > 0) and/or anode dehydration (M > 0), according to Eq. (33). Eq. (33) is equivalent to the assumption that the ionic conductivity of the
membrane depends on current density, by decreasing when
the latter increases, in agreement with [26]. A decrease in the
membrane ionic conductivity and thus an increase in the cell
resistance at low current densities due to the incidental
dehydration at membrane/cathode interface (M > 0) is
also expected from Eq. (33).
The innovative character of the proposed model is evident. In particular, the more original aspect consists in the
correlation introduced between the performance of the
polymer fuel cell and the water balance conditions. In fact,
although it is evident that possible problems of electrode
¯ooding and/or membrane/electrodes dehydration cannot be
unin¯uential on the operation of this type of fuel cell, no
literature model explicitly expresses this. Such a result has
been obtained here through the use of two mathematical
relationships (Eqs. (32) and (33)) which state that the gas
porosity in the cathode diffusional layer and the membrane
conductivity vary with the current density, in agreement with
some literature considerations [11,26]. As a consequence,
the diffusional (Eq. (27)) and ohmic overpotentials (Eq. (26))
increase in case of ¯ooding and dehydration problems,
respectively.
This effect is shown, from a qualitative point of view, in
Figs. 2 and 3, where the overpotentials are reported in
arbitrary unities (A.U.). In particular, in Fig. 3 the membrane/cathode dehydration problems have been emphasized.
Appropriate humidi®cation conditions would let us exclude
the occurrence of this phenomenon. Nevertheless, as already
mentioned, if the humidi®cation conditions are not adequate, at low current densities, the water produced at the
cathode and that dragged from anode to cathode due to
280
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
Fig. 2. Qualitative effect of cathode flooding on diffusional overpotential, based on Eqs. (26), (31) and (32).
electro-osmosis could not be suf®cient to avoid dehydration
on the cathode side of the membrane.
Another innovative element of the proposed model is
represented by Eq. (27) used to calculate the diffusional
overpotential, which is partly based on some modeling
approaches of semi-empirical type [16,34]. In fact, the
relationship commonly reported in literature [30,31] in this
respect is very general, and does not seem to provide a good
description of the diffusional problems at high current
densities: the diffusional overpotential of the PEFC results
underestimated. Some models proposed for PEFCs are not
able to provide an adequate description of the cell behavior
in the whole range of current densities [12,13]. In fact, they
often result approximate in the high current density region,
just where the diffusional overpotentials (mass-transport
problems) prevail, especially when air is used as reactant
gas. On the contrary, the innovative elements introduced in
our model allowed to obtain an excellent agreement with the
experimental results of a polymer fuel cell.
4. Model input data
The proposed model has been validated through application to polymer fuel cells with a Na®on 117 membrane,
Table 1
Input data relative to the cell and the electrodes
Datum
Value
2
Current densities (A cm )
Cell temperature (8C)
Electrode surface (cm2)
Electrode gas-diffusion layer thickness (mm)
Electrode hydraulic permeability (cm2)
Cathode feed gas
Anode gas pressure (atm)
Cathode gas pressure (atm)
Anode inlet gas-flow rate (Nl min 1)
Cathode inlet gas-flow rate (Nl min 1)
Anode humidifying temperature (8C)
Cathode humidifying temperature (8C)
a
0.5 10 3 0.850a
70a (base-case)
50a
360a
4.73 10 15 [1]
Aira
2.5a (base-case)
3.0a (base-case)
0.53a
3.0a
85a (base-case)
75a (base-case)
Experimental data CNR-ITAE.
whose experimental results coming from our laboratory are
available. Afterwards, on the basis of literature data, the
model has also been applied to a fuel cell based on a Dow
membrane in order to make a comparison between the two
membranes.
The input data are summarized in Tables 1±3. As regards
the parameters required for electrochemical calculations
listed in Table 2, the initial porosity of the gas-diffusion
Fig. 3. Qualitative effect of membrane dehydration on ohmic overpotential, based on Eqs. (25), (30) and (33).
281
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
Table 2
Input data required for electrochemical calculations
Datum
Value
Initial gas porosity in gas-diffusion layer (%)
Catalyst surface area (cm2 mg 1)
Catalyst loading (mg cm 2)
Catalyst utilization
Exchange current density per Pt surface (A cm 2)
Membrane resistance/cell resistance ratio
Empirical constant for diffusional overpotential (O cm2 K 1)
Water amount for complete flooding (mol s 1)
30a
1100a
0.15a
0.5a
6.0 10
0.77d
3.4 10
3.0 10
9
(Nafion 117)b; 6.6 10
9
(Dow)c
4e
3f
a
Experimental data CNR-ITAE.
This value comes from elaboration of experimental cell potential by a program developed with the software package MathematicaTM [35].
c
From data reported in [36]: the ratio between the exchange current densities Dow/Nafion 117 is about 1.1 at 708C and cathode pressure 3 atm.
d
From experimental measurements carried out at CNR-ITAE, a membrane resistance (Nafion 117) of 0:16 0:02 O cm2 results, and the global cell
resistance 0:20 0:02 O cm2 in the intermediate current densities region.
e
From extrapolation of data reported in [34].
f
This value is fixed so as to have about a 10% increase in resistance from 0 to 0.8 A cm 2, in agreement with data reported in [26].
b
Table 3
Membrane properties required as input data
Property
Nafion 117, 7 mil
Dow, 7 mil
Wet thickness (mm)
Dry density (g cm 3)
Equivalent weight (g mol 1)
Hydraulic permeability (cm2)
Ê)
Average pore diameter (A
Water porosity (%)
H diffusion coefficient at 228C (cm2 s 1)
230 [1,37]
1.84 [38]
1100 [36,38]
1.8 10 14 [1]
55 [1,37]
28 [1,37]
1.4 10 5 [27]
230 [37]
2.143a
800 [36]
1.8 10
44 [1,37]
44 [1,37]
2.0 10
a
This value is derived from Eq. (1), based on a fixed-charge concentration of 1.5 10
reported in the table.
layer and the data relative to the catalyst refer to electrodes
manufactured in house; the other parameters (Io,Pt, tR, o,
and M1,max) have been determined according to the speci®cations accompanying the table. In general, when the
experimental data of cell potential are available, these model
parameters are ``®tting parameters'', which allow to adjust
the behavior of the theoretical curve. The exchange current
density per Pt surface (Io,Pt) in¯uences the low current
density region controlled by activation. The ratio between
membrane and cell resistance (tR) in¯uences the intermediate current density region, where the ohmic overpotentials
prevail. The empirical constant o and the water amount
corresponding to a complete ¯ooding (M1,max) in¯uence the
high current density region, dominated by diffusional problems; even if, on the basis of assumption (34), the parameter M1,max partly conditions the ohmic losses.
It must be observed that, apart from the value of exchange
current density per Pt surface, the values required for
electrochemical calculations (Table 2) have been assumed
to be equal for the two membranes, as a ®rst approximation.
Nevertheless, such a hypothesis should be better veri®ed in
further investigations.
For the membrane properties (Table 3), we made reference to literature data [1,27,36±38] relative to 7 mil
(175 mm) thick membranes in the dry, H state.
3
mol cm
3
14
5
[1]
[27]
[1], and on the porosity and equivalent weight values
5. Results and discussion
The properties calculated by the model for the two
membranes considered are reported in Table 4. The comparison shows that some properties (saturated vapor pressure, water viscosity and density), depending only on the
temperature of membrane operation (708C in this case), have
the same value for both membranes.
The electrokinetic permeabilities of the two membranes,
calculated by Eq. (2), are also equal; but in such a case, this
result comes from a balance between the volume of the
membrane pores and their size. In fact, the Dow membrane
Table 4
Membrane properties calculated at 708C
Property
Nafion
117, 7 mil
Dow,
7 mil
Saturated vapor pressure (atm)
Pore-water viscosity (kg m 1 s 1)
Pore-water density (mol cm 3)
H diffusion coefficient (cm2 s 1)
Initial ionic conductivity (O 1 cm 1)
Fixed-charge concentration (mol cm 3)
Electrokinetic permeability (cm2)
0.3046
4.04 10 4
0.0543
3.8 10 5
0.151
1.2 10 3
1.06 10 15
0.3046
4.04 10 4
0.0543
5.5 10 5
0.269
1.5 10 3
1.06 10 15
282
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
is characterized by a larger porosity (44 versus 28%), but it
Ê ). Other membrane
has a lower pore diameter (44 versus 55 A
characteristics calculated by the model show better values
for Dow, in terms of ®xed-charge concentration, protonic
diffusion, and ionic (initial) conductivity (78% higher than
Na®on 117), in agreement with literature [1].
Besides, calculations [39] based on a concentrated solution theory [40] showed that in spite of a higher frictional
coef®cient between water and polymer, the Dow membrane
takes advantage of a reduced friction of the hydrogen ions
inside the membrane. This result agrees with some studies
[37] that attribute to the Dow membrane, characterized by a
high protonic diffusion coef®cient (see Table 4), reduced
water ¯ow rates, and a better proton transport (compared
with Na®on 117).
A substantial difference between the two membranes is
represented by the water-transport number (Eq. (14)), whose
calculated values for Dow membrane are lower than Na®on
117 (the average difference is 0.31 H2O/H, and lies from a
minimum of 0.26 H2O/H at 0.250 A cm 2 to a maximum
of 8.04 H2O/H at OCV), in agreement with literature
[41,42]. This difference leads to different water requirements (M and M) to achieve water balance conditions. So
that, in any case, the ¯ooding and/or dehydration problems
will be more signi®cant for the fuel cell based on Na®on 117
membrane.
In Figs. 4±9, the model results corresponding to the
electrochemical calculations are presented. In particular,
Fig. 4 shows the behavior of diffusional and ohmic overpotential calculated for the fuel cell based on Na®on 117
Fig. 4. Influence of flooding and dehydration problems on diffusional and ohmic overpotentials calculated by the model in the case of a Nafion 117
membrane.
Fig. 5. Overpotential contributions calculated by the model in the case of a Nafion 117 membrane.
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
283
Fig. 6. Percentage of overpotential contributions calculated for Nafion 117 membrane at four current density values: (a) 0.2 A cm 2; (b) 0.4 A cm 2; (c)
0.6 A cm 2; (d) 0.8 A cm 2.
membrane (base-case), according to two different hypotheses: in the absence of ¯ooding (edg constant: 30%) and
dehydration phenomena (Rcell constant: 0.198 O cm2), and
in the presence of such problems (edg and Rcell vary with
current density). In compliance with the expectations anticipated in Figs. 2 and 3, it can be observed that the diffusional
overpotential is enhanced by the ¯ooding problems of the
cathode gas-diffusion layer, going from 0.094 V (in the
absence of ¯ooding) to 0.263 V at 0.850 A cm 2. Analogously, the ohmic overpotential suffers from the dehydration
of the anode side of the membrane; in fact, it increases
from 0.168 to 0.189 V at 0.850 A cm 2. Vice versa, there
is no evidence of any dehydration of the cathode side
of the membrane (possible, sometimes, at low current
densities); but this is justi®ed by the humidi®cation conditions chosen.
As already mentioned, the behaviors presented in Fig. 4
are due to the two negative effects: a decrease, caused by
¯ooding of the gas porosity in the diffusional layer; an
increase, caused by dehydration of the overall cell resistance. In fact, the gas porosity passes from the initial 30% (at
OCV) to 22.7% (24% less!) at 0.850 A cm 2; while, the
cell resistance goes from 0.198 O cm2 (at OCV) to
0.222 O cm2 at 0.850 A cm 2, with a rise of about 11%.
As Fig. 4 reveals, both these effects become evident (>5 mV)
only for current densities larger than about 0.5 A cm 2.
It must be observed that the value of the (initial) ionic
conductivity calculated by Eq. (6) allows to obtain (Eq. (30))
Fig. 7. Polarization curves: comparison between model and experimental results for the base-case of Nafion 117 membrane.
284
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
Fig. 8. Comparison between experimental (bullets) and theoretical (dashed lines) polarization curves under different operative conditions (Nafion 117
membrane).
an overall cell resistance (0.204 O cm2 at 0.4 A cm 2) in
excellent agreement with the value measured in the intermediate current density region, equal to 0:20 0:02 O cm2.
Besides, as it will be shown later, it has not been necessary
to artfully change the value of the membrane ionic conductivity to achieve a good ®tting of the experimental
results. Instead, a conductivity equal to 0.07 O 1 cm 1 at
808C has been considered in [2] to reach a satisfactory ®tting
of the polarization experimental curves, even though the
value calculated by Eq. (6) was signi®cantly higher (i.e.
0.17 O 1 cm 1).
From Fig. 5, it is possible to distinguish the different
overpotential contributions calculated by the model for the
fuel cell based on Na®on 117 membrane. At low current
densities (<0.1 A cm 2), the activation overpotential is
almost entirely responsible for the cell voltage losses. For
current densities > 0:3 A cm 2, the ohmic losses due to the
membrane and the electrodes become more important, and
the activation overpotential reaches a relatively constant
value. The results also show the effect due to the oxygen
mass-transport limitations, which involve an enhanced diffusional overpotential for current densities > 0:6 A cm 2.
The overpotential due to proton convection, which in Fig. 5
is coupled to the diffusional overpotential, is not noticeable:
it reaches a maximum of 11.2 mV at 0.850 A cm 2.
A fundamental difference, compared with analogous
results presented in literature [1,3,4], is an appreciable
limitation due to mass-transport phenomena. Actually, in
[1,3,4], the only parameter which in¯uences the high current
density region is the gas porosity in the diffusional layer edg,
but this is a datum of the model and does not vary with
current density. Two extreme cases were presented in
[1,3,4]: when the porosity edg 50%, no transport limitations were observed; for edg 11% instead, the diffusional
Fig. 9. Polarization and power density curves calculated by the model: comparison between results relative to Nafion 117 (base-case) and Dow membranes.
285
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
problems become evident, and the oxygen-limiting current
density reaches 0.750 A cm 2. However, the potential drop
related to the mass-transport problems appears to be too
sharp (the cell voltage is still very high, i.e. about 0.660 V at
0.650 A cm 2) compared with our predictions (Figs. 5±7).
Fig. 6 allows to discern the relative share of the calculated
overpotential contributions corresponding to four current
density values. The activation overpotential is 89.5% of the
total irreversibilities at 0.2 A cm 2, but decreases at 54.3%
for 0.8 A cm 2; the percentages of the other contributions,
instead, increases with current density. Thus, at 0.8 A cm 2,
ohmic and diffusional overpotential both represent more
than 21% of the total losses, while convective overpotential
does not exceed 1.3%. Is it signi®cant that for constant gas
porosity and membrane ionic conductivity, the percentages
at 0.8 A cm 2 become: 63.2% activation, 23.8% ohmic,
11.8% diffusional, and 1.2% convective overpotential.
Fig. 7 shows the comparison between the model and the
experimental results of the fuel cell based on a Na®on 117
membrane. The agreement obtained between the polarization curve predicted by the simulation model and the
experimental curve is very satisfactory. In fact, the mean
error is in the order of 5.7 mV, with two relative maxima of
26.6 mV at 0:5 10 3 A cm 2 (corresponding to 2.8%)
and of 15.6 mV at 0.7 A cm 2 (equal to 2.9%). Thus, the
maximum error for the entire curve is less than 3%.
Moreover, leaving out the ®rst point (0:5 10 3 A cm 2)
for which there are unquestionable measurement dif®culties
(recorded values for ®ve experimental curves range from
0.968 to 0.981 V!), the mean error between the two curves
reduces to 4.8 mV; i.e. in percentage terms, 0.7%. From
a more general point of view, apart from few points
(0:5 10 3 , 2:5 10 3 , 0.400, 0.800, and 0.850 A cm 2),
it can be noticed that the calculated cell voltage is always
slightly lower than the experimental one. However, peculiar
behaviors of the error corresponding to the three current
density regions controlled by activation, ohmic, and diffusional overpotentials are not observed. Then, as a matter of
fact, these three contributions seem to be described with a
good accuracy by the mathematical model here proposed,
whose validation has been consequently attained.
In fact, further simulations carried out at different operative conditions con®rmed the good agreement between the
model and the experimental results (Fig. 8).
Fig. 9 shows a comparison of the polarization and
power density curves calculated by the model for a fuel
cell having the same characteristics of the base-case, with
the two different types of membranes. The results, obtained
by considering the data reported in Tables 1±3, show a
superior performance for the Dow-based fuel cell. Such a
performance is justi®ed by several factors that can be easily
understood by comparing the overpotential values calculated at 0.850 A cm 2 for the two cases (Table 5).
In fact, the higher exchange current density associated with the Dow-based fuel cell, owing to its higher
protonic concentration, implies a slightly lower activation
Table 5
Comparison between calculated overpotentials at 0.850 A cm
2
Overpotential
Cell with Nafion
117 membrane
(mV)
Cell with
Dow membrane
(mV)
Difference
(mV)
Activation
ohmic
Diffusional
Convective
424.33
189.21
263.36
11.15
421.52
101.31
190.92
3.26
2.81
87.90
72.44
7.89
overpotential (3 mV less in all the current density range).
Moreover, the higher protonic concentration of the Dow
membrane results in a reduced overpotential for protonic
convection: about 8 mV less at 0.850 A cm 2. But, what
makes the big difference between the two fuel cells are the
ohmic and the diffusional overpotentials, respectively, 88
and 72 mV less at 0.850 A cm 2 for the Dow-based fuel cell.
Thus, altogether the cell voltage calculated at 0.850 A cm 2
results a good 0.171 V higher for the case of the fuel cell
with Dow membrane: 0.480 V against 0.309 V for the
Na®on-based cell.
This is clearly due to a lower cell (initial) resistance
(0.111 O cm2 against 0.198 O cm2 at OCV), i.e. a higher
ionic conductivity of the Dow membrane, but also, in
particular, to a less-pronounced effect of the dehydration
and ¯ooding problems. In fact, the resistance undergoes an
increase equal to 6.7%, and the gas porosity in the diffusional layer suffers a decrease equal to 19.8%, when we pass
from OCV to 0.850 A cm 2, versus 11.1 and 24.2%, respectively, for the Na®on-based fuel cell.
This is also related to the values of the water-transport
number (Eq. (14)), which in the case of the Dow membrane
are lower than those of the Na®on 117 membrane (e.g. 0.67
H2O/H versus 0.97 H2O/H at 0.850 A cm 2). Thus, the
water transport for electro-osmosis through the Dow membrane is slower, and it contributes to simultaneously reduce
the membrane dehydration (at the anode interface) and the
cathode ¯ooding.
Obviously, the power density of the fuel cell is also higher
for the cell using the Dow membrane, with a maximum of
about 0.438 W cm 2 at 0.750 A cm 2; whereas, for the
cell based on the Na®on 117 membrane, the maximum
power density is approximately equal to 0.370 W cm 2 at
0.700 A cm 2. The results obtained by the model are, from a
qualitative point of view, in agreement with literature
[1,3,4,14,21,26], where experimental and/or theoretical
polarization curves of PEFCs based on Na®on 117 and
Dow membranes are presented, with a clear superior
performance for the latter.
6. Conclusions
In this paper, a mathematical model for polymer fuel cells
is described; it represents an innovative approach compared
286
G. Maggio et al. / Journal of Power Sources 101 (2001) 275±286
with others reported in literature. In fact, it is based on the
introduction of some original relationships that envisage a
decrease in the cell resistance and in the porosity available
for the gas in the electrode, respectively, due to dehydration
on the anode side of the membrane and to cathode ¯ooding.
Once the operative conditions are ®xed, such relationships
allow to determine the ohmic and diffusional overpotentials
as a function of the water balance conditions of the fuel cell.
The cell voltage is accordingly calculated.
The model results here presented, for the case of a
polymer fuel cell using a Na®on 117 membrane, demonstrate an excellent agreement with the experimental voltage
values of a cell realized at the CNR-TAE Institute: the
maximum error is less than 3%. The simulations carried
out at different operative conditions con®rm this agreement.
Moreover, a superior performance (0.480 V against
0.309 V at 0.850 A cm 2) was predicted in the case of a
fuel cell, with the same characteristics (electrodes, operative, and humidi®cation conditions) based on a Dow membrane, whose relative ¯ooding and dehydration problems are
less critical.
In conclusion, the developed model represents a suitable
theoretical tool to provide a valid support for an advanced
design of the membrane/electrode assembly of PEFCs for
the realization of polymeric membranes at low cost and for
the optimization of cell-operating conditions.
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
References
[1] D.M. Bernardi, M.W. Verbrugge, General Motors Report GMR-7360,
15 May 1991.
[2] D.M. Bernardi, M.W. Verbrugge, AIChE J. 37 (1991) 1151±1163.
[3] D.M. Bernardi, M.W. Verbrugge, in: R.E. White, M.W. Verbrugge,
J.F. Stockel (Eds.), Proceedings of the Symposium on Modeling of
Batteries and Fuel Cells, Vol. 91-10, The Electrochemical Society,
Pennington, NJ, 1991, pp. 240±280.
[4] D.M. Bernardi, M.W. Verbrugge, J. Electrochem. Soc. 139 (1992)
2477±2491.
[5] D.M. Bernardi, J. Electrochem. Soc. 137 (1990) 3344±3350.
[6] T.E. Springer, S. Gottesfeld, in: R.E. White, M.W. Verbrugge, J.F.
Stockel (Eds.), Proceedings of the Symposium on Modeling of
Batteries and Fuel Cells, Vol. 91-10, The Electrochemical Society,
Pennington, NJ, 1991, pp. 197±208.
[7] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, in: R.E. White, M.W.
Verbrugge, J.F. Stockel (Eds.), Proceedings of the Symposium on
Modeling of Batteries and Fuel Cells, Vol. 91-10, The Electrochemical Society, Pennington, NJ, 1991, pp. 209±229.
[8] M.S. Wilson, T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, in:
Proceedings of the 26th Intersociety Energy Conversion Engineering
Conference, Vol. 3, Boston, 4±9 August 1991, pp. 636±641.
[9] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, J. Electrochem. Soc.
138 (1991) 2334±2342.
[10] T. Springer, S. Gottesfeld, M. Wilson, in: Proceedings of the Fall
Meeting of The Electrochemical Society, Toronto, 11±16 October
1992, Abstract 91.
[11] T.E. Springer, M.S. Wilson, S. Gottesfeld, J. Electrochem. Soc. 140
(1993) 3513±3525.
[12] E.A. Ticianelli, C.R. Derouin, A. Redondo, S. Srinivasan, in: S.
Srinivasan, S. Wagner, H. Wroblowa (Eds.), Proceedings of the
Second Symposium on Electrode Materials and Processes for Energy
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
Conversion and Storage, Vol. 87-12, The Electrochemical Society,
Pennington, NJ, 1987, pp. 166±181.
E.A. Ticianelli, C.R. Derouin, A. Redondo, S. Srinivasan, J.
Electrochem. Soc. 135 (1988) 2209±2214.
S. Srinivasan, D.J. Manko, H. Koch, M.A. Enayetullah, A.J. Appleby,
J. Power Sources 29 (1990) 367±387.
E.A. Ticianelli, J. Electroanal. Chem. 387 (1995) 1±10.
J. Kim, S.-M. Lee, S. Srinivasan, C.E. Chamberlin, J. Electrochem.
Soc. 142 (1995) 2670±2674.
J.C. Amphlett, M. Farahani, R.F. Mann, B.A. Peppley, P.R. Roberge,
in: Proceedings of the 26th Intersociety Energy Conversion Engineering Conference, Vol. 3, Boston, 4±9 August 1991, pp. 624±629.
J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley, P.R.
Roberge, T.J. Harris, J. Electrochem. Soc. 142 (1995) 1±8.
J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley, P.R.
Roberge, T.J. Harris, J. Electrochem. Soc. 142 (1995) 9±15.
S.J. Ridge, R.E. White, Y. Tsou, R.N. Beaver, G.A. Eisman, J.
Electrochem. Soc. 136 (1989) 1902±1909.
A. Cisar, in: Proceedings of the 26th Intersociety Energy Conversion
Engineering Conference, Vol. 3, Boston, 4±9 August 1991, pp.
611±618.
M.C. Kimble, N.E. Vanderborgh, in: Proceedings of the 27th
Intersociety Energy Conversion Engineering Conference, Vol. 3,
San Diego, 3±7 August 1992, pp. 413±417.
J.D. Canaday, T.A. Wheat, A.K. Kuriakose, A. Ahmad, Int. J.
Hydrogen Energy 12 (1987) 151±157.
J.-T. Wang, R.F. Savinell, Electrochim. Acta 37 (1992) 2737±2745.
X. Yu, R. Zhang, J.M. Fenton, in: Proceedings of the Spring Meeting
of The Electrochemical Society, Honolulu, 16±21 May 1993,
Abstract 1304.
M.S. Wilson, J.A. Valerio, S. Gottesfeld, Electrochim. Acta 40 (1995)
355±363.
A.G. Guzman-Garcia, P.N. Pintauro, M.W. Verbrugge, E.W.
Schneider, J. Appl. Electrochem. 22 (1992) 204±214.
R.C. Weast (Ed.), Handbook of Chemistry and Physics, 62nd Edition,
CRC Press, Boca Raton, 1981±82, Section F, pp. 11±42.
P. Staiti, Z. Poltarzewski, V. Alderucci, G. Maggio, N. Giordano, J.
Appl. Electrochem. 22 (1992) 663±667.
K. Kinoshita, F.R. McLarnon, E.J. Cairns, Fuel Cells Ð A Handbook, prepared by Lawrence Berkeley Laboratory for the US Department of Energy under contract no. DE-AC03-76SF00098, May 1988.
J.H. Hirschenhofer, D.B. Stauffer, R.R. Engleman, Fuel Cells Ð A
Handbook (Revision 3), prepared by Gilbert/Commonwealth Inc. for
the US Department of Energy under contract no. DE-AC0188FE61684, January 1994.
M.W. Verbrugge, R.F. Hill, J. Electrochem. Soc. 137 (1990) 886±893.
J. Leddy, N.E. Vanderborgh, Los Alamos National Laboratory Report
LA-UR 86-2072, 1986.
G. Squadrito, G. Maggio, E. Passalacqua, F. Lufrano, A. Patti, J.
Appl. Electrochem. 29 (1999) 1449±1455.
A.S. AricoÁ, A.K. Shukla, V. Antonucci, N. Giordano, J. Power
Sources 50 (1994) 177±186.
M. Wakizoe, O.A. Velev, S. Srinivasan, Electrochim. Acta 40 (1995)
335±344.
M.W. Verbrugge, R.F. Hill, J. Electrochem. Soc. 137 (1990)
3770±3777.
M.W. Verbrugge, R.F. Hill, J. Phys. Chem. 92 (1988) 6778±6783.
G. Maggio, V. Recupero, L. Pino, CNR-ITAE Report no. 14/98,
September 1998.
T.F. Fuller, J. Newman, in: R.E. White, A.J. Appleby (Eds.),
Proceedings of the Symposium on Fuel Cells, Vol. 89-14, The
Electrochemical Society, Pennington, NJ, 1989, pp. 25±38.
N.E. Vanderborgh, M.C. Kimble, J.R. Huff, J.C. Hedstrom, in:
Proceedings of the 27th Intersociety Energy Conversion Engineering
Conference, Vol. 3, San Diego, 3±7 August 1992, pp. 407±411.
T.A. Zawodzinski Jr., T.E. Springer, J. Davey, R. Jestel, C. Lopez, J.
Valerio, S. Gottesfeld, J. Electrochem. Soc. 140 (1993) 1981±1985.