See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223269919

Mathematical modelling of ambient air-breathing fuel cells for portable

devices

Article  in  Electrochimica Acta · March 2007

DOI: 10.1016/j.electacta.2006.11.002

CITATIONS READS

59 303

2 authors:

Shawn Litster Ned Djilali

Carnegie Mellon University University of Victoria
211 PUBLICATIONS   4,728 CITATIONS    312 PUBLICATIONS   12,437 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Wall Flows friction turbulence and roughness effects View project

Hydrogen peroxide detection in PEM Fuel cells View project

All content following this page was uploaded by Ned Djilali on 29 September 2017.

The user has requested enhancement of the downloaded file.

 This article was originally published in a journal published by
Elsevier, and the attached copy is provided by Elsevier for the
author’s benefit and for the benefit of the author’s institution, for
non-commercial research and educational use including without
limitation use in instruction at your institution, sending it to specific
colleagues that you know, and providing a copy to your institution’s
administrator.
All other uses, reproduction and distribution, including without
limitation commercial reprints, selling or licensing copies or access,
or posting on open internet sites, your personal or institution’s
website or repository, are prohibited. For exceptions, permission
may be sought for such use through Elsevier’s permissions site at:

http://www.elsevier.com/locate/permissionusematerial
 Electrochimica Acta 52 (2007) 3849–3862

Mathematical modelling of ambient air-breathing

py
fuel cells for portable devices
S. Litster 1 , N. Djilali ∗

co
Institute for Integrated Energy Systems, and Department of Mechanical Engineering, University of Victoria, Victoria, BC V8W 3P6, Canada
Received 27 November 2005; received in revised form 2 November 2006; accepted 5 November 2006
Available online 30 November 2006

Abstract
Competitive costs, instant recharge, and high energy density make fuel cells ideal for supplanting batteries in portable electronic devices. In this

al
study, we derive a semi-analytical model to elucidate the transport of ions, heat and mass in air-breathing fuel cells. The model includes an empirical
correlation for membrane conductivity that improves accuracy when modelling membrane dry-out. A detailed comparison with experimental
data demonstrates that the model accurately predicts fuel cell performance through detailed accounting of catalyst layer specifications, including
on
variable width, and electrochemical parameters. A comprehensive parametric study resolves the trends associated with a variety of design
specifications and operating conditions. Membrane dry-out is identified as the primary limitation on current density, and is shown to be strongly
dependent on heat transfer. The study also identifies some unique effects of coupling between ambient air temperature and humidity on the
performance of air-breathing fuel cells.
rs

© 2006 Elsevier Ltd. All rights reserved.

Keywords: Fuel cells; Analytical models; Portable devices; Convective heat transfer; Membrane dry-out
pe

1. Introduction is that even if the capacity of Li-ion batteries grows at 10%

Techno-economic analyses have shown that portable
consumer electronics, currently powered by rechargeable
batteries, present a more accessible market for proton exchange
r's
per year, they will still not be capable of meeting the power
requirements of future devices.
The envisaged implementation of PEM fuel cells for portable
devices is in a device integrated form. This entails placing the

membrane (PEM) fuel cells than transportation applications in
the immediate future, because a higher cost per unit of energy is
acceptable at these smaller scales [1,2]. For example, Dyer [1]
stated that the cost tolerance for fuel cells in portable equipment
o
fuel cell either on, or within portable electronics. For example,
fuel cells can be mounted on the exterior surfaces of mobile
phones, as illustrated in Fig. 1, and in notebook PCs. Based
on the predicted planar area available on a mobile phone, the
power density requirement of the fuel cell is 60 mW cm−2 [2].

is two orders of magnitude greater than that for automotive

applications. Another key advantage of fuel cells for portable The cathode gas diffusion layer is fed ambient air, rather than
th

applications over batteries is longer continuous operation with
almost instantaneous refueling.
An increasingly more prevalent driver of fuel cell develop-
ment for portable devices is the ever-increasing energy density
Au
conditioned air through flow channels, and the fuel is stored
and distributed below the fuel cell. One advantage of an air-
breathing cathode is that it eliminates the need for manifolding.
In addition, the open surface facilitates heat transfer from the

requirements of new consumer electronics, such as mobile fuel cell.

phones with digital broadcast reception [2]. A key argument for
the integration of fuel cells into next-generation mobile phones
∗ Corresponding author. Tel.: +1 250 721 6034; fax: +1 250 721 6323.
E-mail address: ndjilali@uvic.ca (N. Djilali).
1 Present address: Department of Mechanical Engineering, Stanford Univer-
sity, Stanford, CA 94305, USA.
A schematic of such an air-breathing fuel cell is shown in
Fig. 2. This arrangement incorporates multiple methodologies
for reducing the size and number of ancillary components:
1. To alleviate the need for air distribution channels, along with
the necessary pumps and fans, the cathode gas diffusion layer
is in direct contact with the ambient air.

0013-4686/$ – see front matter © 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electacta.2006.11.002
3850 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862
Fig. 1. Illustration of an ambient air-breathing fuel cell as a power source for a
mobile phone with video playback.
2. External humidification systems are eliminated and the fuel
cell relies on the ambient relative humidity and water produc-
tion in the cathode for the humidification of the membrane.
3. The circulating ambient air facilitates the cooling of the fuel
cell in lieu of a dedicated heat management system.
4. The anode is fed dry hydrogen at roughly ambient tem-
perature and pressure, eliminating hydrogen conditioning
rs
components.
5. Electrical current is distributed across the cell in an edge-
collection arrangement.
pe
Four illustrative examples of planar fuel cell designs that
passively breathe ambient air have been presented in the open
literature by Wainright et al. [3], Schmitz et al. [4], Hahn et al.
[5], and most recently Fabian et al [6]. The Wainright et al. [3]
fuel cells were assembled on ceramic substrates with thick-film
fabrication techniques. The design featured hydrogen storage
r's
in a reservoir wafer located directly below the anode; this is
similar to the arrangement shown in Fig. 2. Current was dis-
tributed in porous metal gas diffusion layers in which the current
o
th
Au
Fig. 2. Schematic of the planar air-breathing fuel cell.
flowed through the gas diffusion layer parallel to the mem-
brane. Schmitz et al.’s [4] planar micro-fuel cell was produced
using printed circuit board (PCB) technology. In this design, the
surface of the cathode gas diffusion layer was largely open to
ambient air. A plate with rectangular openings placed over the
gas diffusion layer provided improved current conduction and
allowed the passage of air to the cathode. The active area of this
cell was 10 cm2 , producing a power density of 100 mW cm−2 at
a cell voltage of 0.5 V. The second Fraunhofer planar micro-fuel
py
cell, discussed by Hahn et al. [5], was fabricated with wafer-
level patterning technologies and had a planar area of 1 cm2 . The
cell was able to operate in a stable manner at a power density of
80 mW cm−2 using dry hydrogen as the fuel. The Fabian et al. [6]
co
cell had an active area of 9 cm2 in a similar arrangement to that
of Schmitz et al. This fuel cell was fabricated using printed cir-
cuit board technology and was able to maintain a current density
of 400 mA cm−2 under most environmental conditions.
Researchers and designers are increasingly turning to fuel cell
models to improve fundamental understanding of the transport
al
phenomena present in PEM fuel cells and to use these models
for design optimization. The usefulness of such models when
considering device integrated fuel cells requires an approach and
on
model parameters that are not specific to overly narrow operating
conditions or design parameters.
A number of previous PEM fuel cell models have employed
analytic or semi-analytic solution techniques [7–16]. These
models contrast those requiring discretization in conjunction
with advanced numerical methods such as computational fluid
dynamics [17–23]. Although analytical models typically require
fairly extensive simplifying assumptions, the closed form math-
ematical solution they yield can often provide significant insight
into the expected physical behaviour of the system, without
requiring the time consuming and computationally intensive
parametric studies that must be conducted with numerical
models. This can be particularly advantageous in a design envi-
ronment requiring rapid analysis and assessment cycles.
In the present work we attempt, through modelling, to
understand the coupling between key physical processes in air-
breathing fuel cells, and to refine design objectives and identify
constraints for such fuel cells. In the next section, we present
the derivation and validation of a semi-analytical MEA model.
We then combine this model with heat and mass transfer models
for air-breathing conditions, and with a membrane conductivity
correlation that is suitable for low humidity operation. The pre-
dictive capabilities of the model are validated against relevant
experimental data. The model is finally used to analyze the per-
formance of air-breathing fuel cells for a range of heat transfer
rates, ambient conditions, and catalyst layer specifications.
2. Analytical MEA model
2.1. Introduction
In this section, we present a one-dimensional model of the
membrane electrode assembly in a PEM fuel cell suitable for
analysis and optimization of micro-fuel cells that passively
breathe ambient air. Passive air-breathing operation frequently
 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862 3851

following expression for the potential summation through the

MEA:
WMem
(EOC − Ecell ) − ηC − ηA − iC (ηC ) = 0 (1)
σMem
which is subject to the constraint of current conservation:
iC (ηC ) − iA (ηA ) = 0 (2)

py
The reversible potential (Er ) results from the Gibbs energy
of formation (G):
G
Er = − (3)
nF

co
where n is the moles of product per electron transferred (n = 2)
and F is Faraday’s constant. In this section, the correlation from
Parthasarathy et al. [28] is used to obtain G as function of
temperature for the formation of liquid water. In Section 3,
Fig. 3. Schematic of the one-dimensional analytical MEA model.
where an air-breathing fuel cell is considered, standard ther-
modynamic tables determine G for the formation of water
entails low-humidity conditions and lower current densities due

al
vapour. The Nernst relation accounts for the activities of oxy-
to increased electrolyte resistance [6,24,25]. Thus, we expect gen and hydrogen (aO2 and aH2 ) and provides the open-circuit
Ohmic losses in the electrolyte to have a greater relative impor- voltage:
on
tance than the mass transfer limitations. Consequently, the model  
derivation focuses on the Ohmic losses in the electrolyte rather RT 1
EOC = Er + ln(aH2 ) + ln(aO2 ) (4)
than mass transport in the cathode. nF 2

2.2.2. Catalyst layer model

rs
2.2. Model
The slow oxygen reduction reaction (ORR) distinguishes the
The model domain, shown in Fig. 3, includes both cath- cathode catalyst layer from the anode. When oxygen transport to
ode and anode catalyst layers and the membrane. As shown in the catalyst layer is insufficient, the ORR occurs predominantly
pe

the Figure, the MEA model consists of five main components.
The potential summation algorithm calculates the distribu-
tion of losses through the MEA. Individual potential losses
in the cathode, anode, and membrane are calculated using a
finite-thickness catalyst layer model, an interface catalyst layer
model, and a linear membrane model with uniform conductiv-
ity, respectively. The finite-thickness model is implemented with
r's
near the GDL interface, whereas when the oxygen supply is
abundant and the electrolyte conductivity is low, the ORR con-
centrates by the membrane interface because of Ohmic losses in
the catalyst layer. Due to a much higher electrical conductivity,
the electric potential is approximately uniform in comparison
[29].
In the finite-thickness catalyst layer model derived here, we

parameters evaluated from a macrohomogeneous catalyst layer
model that correlates the required volume average quantities
with the microstructure of the catalyst layer. A novel feature
of the present macrohomogeneous model is the variable cata-
o
neglect the gradient in oxygen concentration that is due to dif-
fusion. This is acceptable when considering low current density
operation typical of air-breathing fuel cells and in cases where
Ohmic losses are dominant. Eikerling and Kornyshev [30] pre-

lyst layer thickness, which is determined from the catalyst layer sented a parameter that determines when these conditions prevail
specifications. based on the Tafel slope (RT/αF), ionic conductivity (σ), and a
th

diffusion parameter (Θ):

2.2.1. Potential summation algorithm αF Θ
The potential summation algorithm is similar to the voltage- g= (5)
RT σ
Au

to-current methods employed by Nguyen et al. [19] and Sivertsen

and Djilali [21]. The algorithm allocates the difference between where
open-circuit voltage and cell potential (EOC − Ecell ). This poten- 4FPair DO
CL
tial difference is the sum of the activation overpotentials in the Θ= 2
(6)
RTWCL
cathode and anode (ηC , ηA ) and the potential drop through the
membrane. We neglect the potential drop through the electrical in which DO CL is the oxygen diffusivity in the catalyst layer and
2
pathways due to its minimal contribution to the overall potential WCL is the thickness of the catalyst layer. For the parameter
drop. The potential drop through the membrane is the product values considered in this analysis g  1, and oxygen transport
of the area-specific membrane resistance (WMem /σ Mem ) and the is rapid in comparison to proton conduction. Our catalyst layer
current density (i). Taking current to be a result of the over- model is similar to that of Eikerling and Kornyshev [30] for the
potential in the cathode catalyst layer (iC (ηC )), we obtain the limit of rapid oxygen transport. It is also similar to the model
3852 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862

derived by Gurau et al. [13] for their analytical half-cell model [13,30,33,34]. We propose a new implementation that accounts
of a PEM fuel cell, but with the important distinction that Gurau for variable catalyst layer thickness.
et al. assumed a uniform overpotential within the catalyst layer
to solve oxygen distribution. 2.2.3.1. Catalyst layer composition. The volume of the catalyst
Our derivation begins with the governing equation for proton layer is composed of four components: platinum, Nafion, car-
conduction, Ohm’s law: bon, and void space. By employing the standard catalyst layer
dφ specifications and a known catalyst layer thickness, each vol-
i = −σ (7) ume fraction can be determined [34]. The typical specifications,
dx
ranges, and units are [35]:

where φ is the electrolyte potential. Cast in a conservative form
with a source term for the current generation (jT ) the equation
becomes:
py
1. Platinum loading, mPt (0.05 − 5 mg Pt cm−2 ).
2. Platinum/carbon ratio, yPt (20 − 40% Pt/C).
3. Nafion content, yN (20 − 60 wt.%).

co
σ ∇ 2 φ = jT (8)
The rate of the ORR is locally dependent on the overpotential The volume fraction of platinum is simply the loading divided
(the difference between the potential of electrolyte and electric by the density of platinum (ρPt ) and the catalyst layer thickness
potential in the platinum η = φ − φs ). Following Perry et al. [31] (WCL ):
and many other studies, we determine the reaction rate, assuming
1
high overpotentials (η  RT/αF), with the Tafel equation: εPt = mPt (12)
 γ ρPt WCL

al
 
c̄O2 αF
jT = j0 ref exp η (9) The volume fraction occupied by carbon is a function of the
cO RT carbon loading (mC ) that results from the specified platinum
2
on
loading and the platinum/carbon ratio:
In addition, the reaction rate depends on the average oxygen
concentration in the layer (c̄O2 ). The concentration dependence 1 mPt (1 − yPt ) 1
εC = mC = (13)
exponent (γ) specifies the sensitivity of the reaction to oxygen ρC WCL yPt ρC WCL
concentration. We solve Eqs. (7)–(9) implicitly to obtain:
rs

  As shown in Eq. (14), we evaluate the volume fraction of

1 bA exp(bη0 ) Nafion from the area loading of the Nafion (mN ) divided by
η(x) = ln tan2 (x − WCL ) + . . . the Nafion density (ρN ) and the thickness of the catalyst layer.
b 2
pe

Nafion loading is a function of three commonly documented

√ 

exp(bη1 ) − exp(bη0 ) catalyst layer properties: Nafion content (weight percentage of

arctan √ + 1 + η0 the catalyst layer that is Nafion), platinum loading, and plat-
exp(bη0 )
(10) inum/carbon ratio:
1 yN mPt 1
where εN = mN = (14)
 γ ρN WCL yPt (1 − yN ) ρN WCL
j0 c̄O2 αF
r's

A= and b= , and since the volume fractions must add to unity, the void frac-
σ cOref RT
2 tion (εV ) is:
where x is the location in the layer, WCL the thickness of the layer,
εV = 1 − (εPt + εC + εN ) (15)
and η0 and η1 are the overpotentials at the GDL and membrane
o

interfaces, respectively. η1 corresponds to ηC in Eq. (1). The

derivation of Eq. (10) is presented in ref. [32]. Eq. (10) is similar, 2.2.3.2. Variable-thickness implementation. Previous studies
of catalyst layer composition optimization based on similar rela-
th

but not identical, to Eikerling and Kornyshev’s [30] expression.

The derivative of overpotential multiplied by the conductivity tions [34,36,37] have typically assumed the void fraction to vary
provides an expression for the protonic current into the cathode: with catalyst layer specifications and the catalyst layer thick-
 ness to remain constant regardless of specifications. However,
Au

2A  the experimental results of Inoue et al. [38] demonstrate that the

i=σ exp(bη1 ) − exp(bη0 )
b
2.2.3. Catalyst composition and variable thickness
We utilize a macrohomogeneous catalyst layer model to
investigate the effect of catalyst layer composition. This model
considers the catalyst layer microstructure as a homoge-
neous medium with properties that reflect the heterogeneous
microstructure. As noted by Eikerling and Kornyshev [30], the
macrohomogeneous model has been around for decades and
variations in derivation and implementation continue to emerge
(11) Nafion content and platinum loading have a significant effect on
catalyst layer thickness. Their results show that the catalyst layer
thickness is proportional to platinum loading and that increasing
Nafion content from 10 to 50 wt.% results in a 100% increase in
thickness.
In contrast with previous work [34,36,37], in our analysis
the thickness of the catalyst layer is calculated from the catalyst
layer specifications and we hold the void fraction constant in
accordance with the experimental results of Inoue et al. [38].
We conjecture that the volume fraction of solvent in the catalyst
 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862
Fig. 4. Schematic of the influence of changing Nafion content or platinum loading has on the microstructure and thickness of the catalyst layer.
layer prior to the evaporation step in the MEA fabrication is con-
sistent over a range of catalyst layer specifications and produces
a constant void fraction.
Fig. 4 presents a schematic of the effect of changing two
different catalyst layer specifications (Nafion content and plat-
inum loading), and illustrates how increasing the Nafion content
increases the thickness of the catalyst. In addition, for constant
platinum loading, an increase in Nafion content increases the
mean distance between catalyst particles. Increased Nafion con-
rs
tent, resulting in a higher Nafion volume fraction, improves the
ability of ions to reach the catalyst sites. However, the increased
thickness also increases the length scales for oxygen diffu-
pe
sion and ionic conduction, and, correspondingly, reduces their
transport rates. Our hypothesis is supported by experiments doc-
umenting greater mass transport limitations with higher Nafion
contents [35]. In contrast, when the platinum loading is increased
there is no change in volume fractions. The only change is
the thickness of the layer. Thus, the benefit of increasing the
platinum content, namely the increased total electrochemically
r's
active area (surface area of Pt × catalyst utilization), is countered
by increased diffusion and conduction length scales.
The thickness of the catalyst layer is determined by solving
the previous set of volume fraction equations (Eqs. (12)–(15))
o
for a constant void fraction:
 
1 mPt (1 − yPt ) 1 yN mPt
th
1 1
WCL = mPt + +
1 − εV ρPt yPt ρC yPt (1 − yN ) ρN
Au
We use the volume fraction of Nafion in the catalyst layer to
calculate the layer’s effective ionic conductivity. The bulk ionic
conductivity is determined from the humidity and temperature
in the layer. The Bruggemann correction, which accounts for
porosity and tortuosity, determines the effective conductivity:
σ= ε1.5
N σBulk
The Bruggemann exponent of 1.5 was obtained from a study
on the electrical conductivity of dispersions [39]. The Brugge-
mann correction has been employed in this form by many
3853
py
co
researchers [13,36,37,40], and is suitable here because of the
disperse catalyst agglomerates.
al
2.2.3.3. Volumetric exchange current density. When solving
on the Tafel equation in the finite-thickness model, the main electro-
chemical parameter is the volumetric exchange current density
(j0 ). As shown in Eq. (18), the volumetric exchange current
density is the product of three factors. The dominant one is
the exchange current density of the platinum/Nafion interface
Pt/N
(i0 ), the second is the area of platinum/Nafion interface per
unit volume (APt ), and the third is the platinum utilization (υPt ),
which is the percentage of the platinum area that is electrochem-
ically active.
Pt/N
j0 = APt υPt i0 (18)
The platinum surface area per unit volume (APt ) is a function
of the platinum loading (mPt ), platinum surface area per unit
mass (sPt ), and the thickness of the catalyst layer (WCL ):
mPt sPt
APt = (19)
WCL
To account for the utilization of the total platinum surface
area, we employ an empirical relationship between the Nafion
content in the catalyst layer and utilization. The utilization data
of Sasikumar et al. [41] was obtained using the ratio of elec-
trochemical surface area, measured with cyclic voltammetry, to
the theoretical value for a platinum/carbon ratio of 20% Pt/C.
(16)
A third-order polynomial fit to their platinum utilization (υPt )
data is presented in Eq. (21). The optimum Nafion content (yN )
from a utilization standpoint is approximately 35–40 wt.%; this
is similar to other results [35]:
υPt = 1.833 × 10−5 yN
3
− 2.807 × 10−3 yN
2
+ 1.332 × 10−1 yN − 1.476 (20)
(17)
An initial increase in utilization with higher Nafion content
[41] could be a result of increased connectivity of the electrolyte.
We hypothesize that the subsequent decline in utilization with
3854 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862

Table 1 catalyst layer as an interface using the Butler–Volmer equation:

Catalyst layer properties and parameters     
αa F A αc F A
Property Value i = iA
0 exp η − exp − η (22)
RT RT
Air pressure, Pair 1 atm
Fuel pressure, Pf 1 atm
Fuel cell temperature, T 353 K (80 ◦ C)
2.2.5. Solution approach
Relative humidity 100% We solve the equations presented in the previous sec-
Membrane thickness (Nafion 115), WMem 125 ␮m tion without resorting to spatial numerical discretization. We
Catalyst layer void fraction, εV 55% [46] solve the one-dimensional MEA model with the bisection

py
Membrane conductivity, σ Mem, Bulk 6.94 S m−1 [46] algorithm. The function we solve is the difference between
Platinum/carbon ratio, yPt 20% Pt/C
Platinum surface area, sPt 1120 cm2 mg−1
the current calculated in the cathode and the current cal-
Platinum density, ρPt 21500 kg m−3 culated in the anode (f = iC (ηC ) − iA (ηA )). The independent
Nafion density, ρN 1900 kg m−3 variable is the cathode overpotential at the membrane inter-

co
Carbon density, ρC 2267 kg m−3 face (ηC ). The overpotential in the anode is determined from
Cathode transfer coefficient, α −1
1 [42,43] the expression ηA = (EOC − Ecell ) − ηC − WMem σMem iC (ηC ).
Anode transfer coefficient, αa and αc 1 [40]
O2 concentration, CO2 5 mol m−3
The sub-routine for solving the implicit finite-thickness cata-
H2 concentration, CH2 34.3 mol m−3 lyst layer equation employs the Newton–Raphson method. In
O2 concentration exponent, γO2 1.03 [42] this sub-routine the Newton–Rhapson function is the difference
H2 concentration exponent, γH2 0.5 [40] between the previous solution of the overpotential at the catalyst
O2 reference concentration, cO ref
113 mol m−3 [42] layer/GDL interface (x = WCL ), and that determined by Eq. (10);

al
2
H2 reference concentration, cH ref
40.88 mol m−3 [40]
2 f = η0 − η(WCL ).
Anode exchange current density, iA 0 6 × 103 A m−2 [26]
Gibbs energy of formation G◦ = −295800 − 33.5T ln T
+ 388.7T [28]
on
2.3. MEA model evaluation

To evaluate the present model, we use the experimental results

further increase in Nafion content results from reduced electrical of Sasikumar et al. [41] because of the special attention they
connectivity between the catalyst particles. paid to the catalyst layer specifications and the wide scope of
rs

We employ the exchange current density of the plat- their parametric study. In addition, Sasikumar et al. used oxygen
inum/Nafion interface and its temperature dependence as instead of air, which is better for validating the MEA model since
determined in the experimental work of Beattie et al. [42]. We this eliminates the ambiguities associated with mass transport
pe

take the interfacial exchange current density to vary exponen- limitations.

tially with temperature: Table 1 lists the operating conditions and fuel cell parameters.
The model operating conditions correspond to the experimen-
= 6.379 × 10−14 exp(6.782 × 10−2 T )
Pt/N
i0 (21)
tal ones [41]. A key parameter is the void fraction, which is
This temperature dependence, which is a consequence of the specified as 55% (based on the results of Navessin [46]). The
overall electron transfer process being an activated process, can molar density of oxygen at a pressure of 3 atm (113 mol m−3 )
be obtained in rather general form from transition state theory is used to correlate the exchange current density of Beattie et
r's

(e.g. Bockris et al. [44]). The parameters in this equation (pref-
actor and activation energy) are treated as empirical parameters.
2.2.4. Interface model of the anode catalyst layer
o
al. [42] to other gas pressures and concentration polarization. In
lieu of a comprehensive mass transfer model, we set the oxy-
gen concentration in the catalyst layer to a reduced value of
5 mol m−3 .

The faster kinetics of anode reaction in PEM fuel cells in con- In the following, the platinum loading varies between 0.10
junction with the low conductivity of the electrolyte drives the and 0.25 mg Pt cm−2 and the Nafion content is specified in the
th

anode reaction into a thin strip next to the membrane. Kornyshev range of 20–60 wt.%. Table 2 presents the resulting catalyst layer
and Kulikovsky’s [45] expression for the thickness of the reac- parameters for four points in this spectrum of specifications.
tive strip in a catalyst layer supports this assumption. Therefore, As discussed in the development of the model, a change in
Au

in contrast with the cathode, it is acceptable to model the anode platinum loading impacts only the catalyst layer thickness. In

Table 2
Catalyst layer properties resulting from the macrohomogeneous model
Configuration (Pt Volumetric exchange Catalyst layer Nafion volume fraction (%)
loading|Nafion content) current density (A m−3 ) thickness (␮m) (effective conductivity (S m−1 ))a

0.10 mg Pt cm−2 |40 wt.% 13355 6.9 25 (0.88)

0.25 mg Pt cm−2 |40 wt.% 13355 17.4 25 (0.88)
0.25 mg Pt cm−2 |20 wt.% 6825 13.7 12 (0.29)
0.25 mg Pt cm−2 |60 wt.% 6088 27.1 36 (1.57)
a Bulk conductivity = 6.94 S m−1 .
 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862
Fig. 5. Comparison between polarization curves obtained experimentally by
Sasikumar et al. [41] and those calculated by the present model for platinum
loadings of 0.10 and 0.25 mg Pt cm−2 .
contrast, Table 2 shows that changing the Nafion content has a
on
significant impact on all three parameters, and there is a maxi-
mum exchange current density at approximately 40 wt.%. The
decrease in exchange current density at higher Nafion contents
is due to the diminishing platinum concentration as the layer
rs
expands.
Fig. 5 presents the polarization curves of two validation
cases with platinum loadings of 0.10 and 0.25 mg Pt cm−2 ,
with a 40 wt.% Nafion content in both cases. There is excel-
pe
lent agreement with the experimental curves, especially for
0.10 mg Pt cm−2 , and the variable-thickness macrohomoge-
neous model satisfactorily reproduces the influence of platinum
loading.
Fig. 6 depicts the effect Nafion content on the polarization
curve. Again, the model is in good agreement with experimen-
r's
o
th
Au
Fig. 6. Comparison between polarization curves obtained experimentally by
Sasikumar et al. [41] and those calculated by the present model for Nafion
contents of 20, 40 and 60 wt.%
3855
tal observations. In the experimental data, increasing the Nafion
content from 20 to 40 wt.% significantly improves performance,
and a further increase from 40 to 60 wt.% marginally reduces
performance. The model captures this effect. Although the cat-
alyst layer conductivity is almost doubled at a Nafion content
of 60 wt.%, the fuel cell voltage drops due to reduced utiliza-
tion and a longer conduction distance. These results demonstrate
that the macrohomogeneous model with the variable-thickness
formulation can correlate the catalyst layer specifications to the
py
layer’s microstructure, and allows subsequent translation of the
microstructure to volume-average parameters for modelling.
co
3. Planar air-breathing fuel cells
3.1. Model
Having established the validity of the MEA model, we
now expand this model to include additional phenomena for
investigating air-breathing fuel cells. The additional phenomena
al
include:
1. The transport of oxygen and water vapour between the sur-
face of the GDL and the ambient air.
2. The diffusion of gases through the GDL.
3. The diffusion of oxygen into the catalyst layer.
4. The influence of temperature and humidity on the conduc-
tivity of the membrane.
5. The heat transfer from the fuel cell to the ambient air.
3.1.1. Assumptions
In addition to the assumptions used in the derivation of the
analytical MEA model, the following lists the major assumptions
we use in developing the model.
1.Transport in the fuel cell is considered one-dimensional.
2.The fuel cell operates in steady state conditions.
3.Water vapour is produced in the cathode.
4.The water exists only as vapour.
5.There is zero net transport of water through the membrane.
6.The hydrogen supply is “dead-ended” and is not recycled.
7.The hydrogen supply is dry (zero humidity).
8.Water does not accumulate in the hydrogen storage system.
9.The water activity is uniform across the electrolyte in the
membrane and catalyst layers and is equal to the water activ-
ity at the cathode interface of the gas diffusion layer and the
catalyst layer.
10. Potential losses due to electron conduction are neglected.
11. The anode side of the fuel cell is perfectly insulated against
heat fluxes and heat only exits the fuel cell through the
cathode surface.
3.1.2. Gas diffusion
The flux of gases through the gas diffusion layer results from
the consumption of reactants and the supply of products in the
catalyst layer. Single-phase conditions are assumed to prevail.
In the cathode, the consumption of oxygen is a function of the
3856 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862

current density: In addition, we approximate the oxygen mass fraction in the

catalyst layer to capture the onset of reactant depletion, using
M O2 i
S O2 = − (23) an approach similar to Berg et al. [15]. For the purpose of
4F approximating this oxygen mass fraction, we assume a homo-
and the local source of water due to the reaction is: geneously distributed reaction rate through the catalyst layer,
MH2 O i which is reasonable in the intermediate regime when the cata-
S H2 O = (24) lyst layer thickness is close to optimal [30]. The resulting oxygen
2F
concentration is used to correct the reaction rate in the Tafel
The governing equation for gas transport in the GDL, assum- equation.

py
ing negligible advection, is Fick’s law. When expressed in terms
of a mass flux ṅA , the differential form of Fick’s law is expressed GDL/CL MO2 iw2
yO
CL
= yO2 − (32)
as: 2
32FρDCL

co
ṅA = −ρDA
eff
∇yA (25) 3.1.3. Water transport through the membrane
where DA
eff is the effective molecular diffusivity of species A in Since we consider steady state operation and a dead-ended
the porous media and yA is the mass fraction of species A. We supply of dry hydrogen, we assume zero net water transport
correct the density of air (ρ) for temperature with the ideal gas across the membrane. Kulikovsky [16] made a similar assump-
law. We also correct the diffusivity with the porosity ε and the tion. Moreover, this assumption is supported by the experimental
tortuosity factor τ, which accounts for the longer diffusion path results of Janssen and Overvelde [49] who report a very low
effective drag coefficient across the membrane (−0.01) at a

al
in the porous gas diffusion media (see [48]):
low current density (400 mA cm−2 ) with dry air and hydrogen
ε at low stoichiometries. We also assume that the water activity
DA
eff
= DA (26)
τ
on
is uniform across the membrane, which is justifiable for thin
The form of Fick’s law employed herein is typically used for membranes [32].
binary mixtures, and is justifiable given the low water concentra-
tions, and the similar molecular weights of oxygen and nitrogen. 3.1.4. Membrane conductivity
rs

We calculate the binary diffusivity as a function of temperature The potential drop through the electrolyte contributes signifi-
using the empirical method presented by Cussler [48]: cantly to the polarization of the fuel cell. We model the influence
√ of humidity on ionic conductivity with an empirical fit to the data
T 1.75 (1/MA ) + (1/MB ) presented by Sone et al. [47] for Nafion 117 without heat treat-
DAB =
pe

(27)
P 1/3
φA + φB
1/3
ment. The polynomial for ionic conductivity as a function of
activity at 303 K is:
where φA and φB represent molecular volumes [48]. The con-
vective mass transfer rates at the interface of the GDL with the σ303 K = 3.46a3 + 0.0161a2 + 1.45a − 0.175 (33)
ambient air is quantified in terms of mass transfer coefficients
(hO2 , hH2 O ), which are functions of the mass transfer Nusselt We model the influence of temperature with the expression
number (hA = NuDA /l). We calculate the interface mass fractions of Springer et al. [7]:
r's

by balancing the species production/consumption MA i/nA F with   

1 1
the convective mass transfer ρhA (yA 0 − yAmb ):
A σ = exp 1268 − σ303 K (34)
303 T
1 M O2 i
yO
0
= yO
Amb
− (28) However, the influence of temperature on ionic conductivity
o

2 2
ρhO2 4F is secondary in comparison to humidification. We also note that
1 M H2 O i almost all conductivity is lost below a relative humidity of 20%
th

yH
0
2O
= yH
Amb
2O
+ (29) when using the correlation in Eq. (33). We chose this correlation
ρhH2 O 2F
because the widely used model of Springer et al. [7] has poor
where i is the current density of the fuel cell and yA
Amb is the accuracy in low humidity conditions [50,51], which is a critical
Au

mass fraction in the ambient air. regime for air-breathing fuel cells.
Considering the GDL as a homogeneous and isotropic porous
medium, we derive the oxygen and water mass fractions at the 3.1.5. Heat transfer
GDL/CL
interface between the GDL and the catalyst layer (yA ) from Temperature assumes a critical role in the performance of
Eq. (26): PEM fuel cells because of its impact on relative humidity through
the saturation pressure and reaction kinetics. Temperature is also
GDL/CL WGDL MO2 i
yO2 = yO
0
− (30) significant for air-breathing fuel cells because they passively
2
ρDO
GDL 4F
2 dissipate the heat produced by the fuel cell without relying on
WGDL MH2 O i the cooling systems of larger PEMFC stacks. For a body with a
GDL/CL
yH2 O = yH
0
2O
+ (31) sufficiently uniform temperature distribution, a lumped system
ρDH
GDL 2F
2O analysis that considers variation of temperature with time only
 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862 3857

is adequate, and the characteristic response time of temperature Table 3

to changes in heat production is given by: Operating conditions for the planar fuel cell
Property Value
V ρc
τ= (35)
A h Heat/mass transfer length scale, l 1 cm
Air pressure, Pair 1 atm
where V and A are the volume and surface area of the body, ρ the Fuel pressure, Pf 1 atm
density of the body, c the specific heat, and h is the heat trans- Ambient air temperature, T∞ 293 K (20 ◦ C)
Ambient relative humidity 60%
fer coefficient. This shows that the heat transfer time constant
Nusselt number, Nu 10
varies proportionally with the characteristic length scale given

py
GDL thickness, WGDL 250 ␮m
by the volume to surface area ratio (V/A). Following standard GDL porosity, εGDL 0.45
heat transfer practice, a lumped system analysis is appropriate GDL tortuosity, τ GDL 2.5
when the Biot number (ratio of convective to conductive heat Platinum loading, mPt 0.25 mg cm−2
Nafion content, yN 45 wt.%
transfer) is small, i.e.

co
Catalyst layer void fraction, εV 55% [46]
hδ Platinum/carbon ratio, yPt 20% Pt/C
Bi = < 0.1 (36)
ksb
where δ is the conduction length scale and ksb is the solid body 3.1.6. Solution procedure
conductivity. Using conservative parameters [32], we estimate We solve for the current density at a given cell potential using
a Biot number of 0.03 for a 1 cm2 fuel cell, which meets the the classical bisection method. We prefer this robust method over

criterion for lumped system analysis. This assumption is further
supported by CFD modelling, which predicts minimal temper-
ature variation within such fuel cells [27].
on
The heat generated in the fuel cell is a combination of the
heat of reaction and that due to the various irreversibilities. The
heat of reaction is the product of the temperature and the change
in entropy (s). The sum of the irreversibilities of the fuel cell
rs
al
more rapidly converging approaches, such as Newton–Raphson,
because the system of equations is highly sensitive to tempera-
ture, causing other methods to fail for some cases. The function
we solve is the difference between the bisection midpoint and the
current density calculated for the temperature and concentrations
associated with the bisection midpoint.

is equal to the difference between the reversible open-circuit

3.2. Results and discussion
potential and the cell potential (EOC − Ecell ). Thus, we compute
the overall heat production (qp ) from:
3.2.1. Baseline properties and parameters
pe

Table 3 lists the baseline parameters. The dimensions of
T (−s)
qp = + (EOC − Ecell ) i (37) the fuel cell are similar to that developed by Hahn et al. [5]
2F
(1 cm × 1 cm). We select an ambient air temperature of 293 K
We assume that all the heat generated by the fuel cell is (20 ◦ C) and an ambient relative humidity of 60% because of
rejected by either forced or free convection from the exterior their broad applicability. The prescribed Nusselt number of 10
surfaces of the fuel cell. The convective heat transfer to the is moderately higher than that feasible by natural convection
environment (qt ) is: alone to represent the effect of ambient air currents.
r's

AA
qt = h(T − T∞ ) (38) 3.2.2. Comparison with experiment
AS
We use the recent experimental results of Fabian et al. [6]
where AA is the electrochemically active planar area, AS the sur- to validate the model. The experimental MEA consists of a
o

face area, and T∞ is the temperature of the ambient environment. carbon cloth GDL, a catalyst layer with 1 mg Pt cm−2 , and a
For planar fuel cells with heat transfer from the cathode surface Nafion 112 membrane. We assume a typical value of 33% for
th

only, the ratio AA /AS is equal to one. Again following standard the Nafion content of the cathode catalyst layer [35]. All other
heat transfer practice, the convective heat transfer coefficient h parameters were given by Fabian et al. [6], with the exception of
is determined from an appropriate value of the Nusselt number the total heat transfer coefficient hAS , which we estimate to be
Au

(Nu): 0.0993 W K−1 from the experimental GDL temperature data. By

Nu k
h= (39)
l transfer coefficients for the GDL/air interface.
where l is the characteristic length of the system and k is the air’s
thermal conductivity. Herein, the characteristic length is that of
the shortest side of the fuel cell. We also employ this Nusselt
number to calculate the mass transfer coefficients at the cathode
surface in Eqs. (28) and (29). In this case, the same length scale
is used, but the thermal conductivity is replaced by the respective
diffusivities of the gases.
approximating the Nusselt number from the heat transfer coef-
ficient and estimated surface area, we can calculate the mass
Fig. 7 compares the theoretical and experimental polariza-
tion curves. A limiting current density is evident in both set of
curves, and although this presents like a reactant mass trans-
fer limitation, it is in fact the result of membrane dry-out.
This is demonstrated by the membrane conductivity data pre-
sented in Fig. 8. The Fabian et al. membrane conductivity data
was obtained with AC impedance spectroscopy. We estimate
3858 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862
Fig. 7. Polarization curves obtained experimentally by Fabian et al. [6] and those
predicted by the model for ambient temperatures of 20 and 30 ◦ C.
the membrane conductivity by subtracting the 11 m electrical
resistance [6] and accounting for the membrane thickness and
the fuel cell area. The membrane dehydration phenomenon is
directly related to the rise in temperature associated with greater
current density, as shown in Fig. 9. The rate of temperature
increase outweighs water production, resulting in a decrease
rs
of the internal relative humidity and, consequently, a drop in
membrane conductivity.
The above comparisons demonstrate the predictive capabili-
ties of the model without any a priori performance information,
pe
fitting of electrochemical parameters or “tuning” of other model
parameters. The model predicts quite satisfactorily the cur-
rent density that marks the transition to membrane dry-out
and the accompanying rapid rise in fuel cell temperature. The
small deviations in predicting the exact current limitation lev-
els is not surprising given the lack of precise data for some of
r's
o
th
Au
Fig. 8. Membrane conductivity vs. current density obtained experimentally by
Fabian et al. [6] and those predicted by the model for ambient temperatures of
20 and 30 ◦ C.
py
co
Fig. 9. Fuel cell temperature vs. current density obtained experimentally by
Fabian et al. [6] and those predicted by the model for ambient temperatures of
20 and 30 ◦ C.
al
the transport parameters, and the simplifications made in the
on
model.
In noting the remarkable agreement between the experi-
mental data and the model, it should be emphasized that the
procedure for acquiring polarization data is critical. Fabian et
al. obtained each data point by allowing the membrane to equi-
librate over a 2-h period. Long equilibration times are required
to accurately depict membrane dehydration because the time
scales can be much larger than other relevant time scales in the
fuel cell. For example, Moxely et al. [52] show that the equili-
bration of membrane water content can take anywhere from a
few minutes to many hours.
3.2.3. Nusselt number
We investigate the influence of the air currents surrounding
the fuel cell by conducting a parametric study of the Nusselt
numbers. The variation of the Nusselt number with the magni-
tude of air current velocity can be determined from a number
of correlations for forced convection over various geometries
and flow directions [53]. We present results for a wide range of
Nusselt numbers to ensure broad applicability.
Fig. 10 presents the polarization curves for Nusselt numbers
between 10 and 40. It is evident that the rate at which heat
dissipates from the fuel cell can dramatically affect performance.
For a Nusselt number of 10 (approximating free convection) the
current density is considerably restricted. An increase in the
Nusselt number from 10 to 40 increases the maximum current
density by 350%.
Although the limiting current density shown in the polariza-
tion curves of Fig. 10 resemble a reactant mass transfer limitation
pattern, the underlying process is in fact different as the change
in the oxygen concentration from the ambient air to the catalyst
layer is insignificant at these low current densities. At the max-
imum current of 520 mA cm−2 the oxygen concentration at the
catalyst layer/GDL interface was still 82% that of the ambient
air. Fifty percent of the oxygen depletion occurs in the cathode
 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862
Fig. 10. Polarization curves for a range of Nusselt numbers.
electrode, the other 50% is depleted at the GDL/air interface.
Nevertheless, this reduction in concentration has a negligible
effect on fuel cell performance.
Fig. 11 depicts the influence of irreversibilities on heat pro-
on
duction. The increased Joule heating at the point where cell
voltage drops sharply due to membrane dry-out is evidenced by
the rapid rise in temperature near the end of each temperature
curve. Fig. 12 presents the competing effects of water produc-
rs
tion and heat generation on the water activity in the membrane.
At low currents, and therefore low fuel cell temperatures, the
activity increases with current density because of water produc-
pe
tion. However, the water activity decreases as the temperature
rises with increasing current density. Fig. 13 serves as guideline
for operating air-breathing fuel cells by demonstrating that for a
given heat transfer rate there is an optimal operating voltage for
which the membrane conductivity is maximized. The optimal
voltage based on membrane conductivity ranges between 0.7
and 0.85 V and decreases with greater heat transfer coefficients.
r's
As shown in Fig. 13, increasing the heat removal rate decreases
the voltage achievable without membrane dry-out.
o
th
Au
Fig. 11. Temperature vs. current density over a range of Nusselt numbers.
3859
py
co
Fig. 12. Water vapour activity in the membrane vs. current density for a range
of Nusselt numbers.
al
3.2.4. Ambient condition
To quantify the impact of ambient air conditions, we vary
the temperatures from 283 K (10 ◦ C) to 303 K (30 ◦ C) and
relative humidity from 30 to 90%. The resulting polarization
curves, in Fig. 14, show that the highest current densities occur
when the ambient air is highly humid or relatively cold. The
improved performance at high ambient relative humidity of 90%
increases performance directly through the membrane’s humid-
ification requirement as in more conventional fuel cells. The
more interesting and less intuitive result is the improved per-
formance at lower temperatures. At these lower temperatures,
self-humidification is enhanced because less water vapour is
required to saturate the air.
3.2.5. Platinum loading
To investigate the potential for minimizing Platinum loading,
we consider loadings of 0.05, 0.25, 0.5, and 1.00 mg Pt cm−2 .
Fig. 15 depicts the corresponding polarization curves. The effect
Fig. 13. Membrane conductivity vs. cell voltage over a range of fixed Nusselt
numbers.
3860 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862
Fig. 14. Polarization curves for five different ambient air conditions.
of platinum loading is most pronounced at low current densities
where the activation of the oxygen reduction reaction is domi-
nant. The limiting current does not change significantly because
on
it is a result of membrane dry-out. The optimum platinum load-
ing is approximately 0.25 mg Pt cm−2 (based on diminishing
returns beyond this value). This value is in good agreement with
that found by Qi and Kaufman [54].
rs
3.2.6. Nafion content
Since ionic resistance is a major component of polarization
pe
in ambient air-breathing fuel cells, we expect in general an opti-
mum Nafion content for the catalyst layer [35]. We analyzed the
effect of varying Nafion content from 20 to 60 wt.% as shown in
Fig. 16. The polarization curve for a Nafion content of 20 wt.%
indicates a significant drop in platinum utilization and ionic
conductivity in the cathode catalyst layer. The reduced ionic con-
ductivity in the catalyst layer is evidenced by the earlier onset
r's
of the membrane dry-out. For Nafion contents between 33 and
o
th
Au
Fig. 15. Polarization curves for a range of platinum loadings in the cathode
catalyst layer.
py
co
Fig. 16. Polarization curves for a range of Nafion contents in the cathode catalyst
layer.
al
60 wt.%, there is little difference in performance. The optimum
value is approximately 45 wt.%.
4. Conclusion
Fuel cells have the potential to supplant batteries in portable
electronic devices because of increased cost tolerances for
fuel cells at this scale, their rapid recharges, and their higher
energy densities. In this paper we presented a model of pla-
nar air-breathing PEM fuel cells for portable devices. The key
component of our model is a 1-D semi-analytical model of the
membrane electrode assembly (MEA). The model incorporates
an improved membrane model correlation that is more suitable
for representing membrane dry-out. We performed parametric
studies to identify the influence of key design specifications and
operating conditions on the performance of air-breathing fuel
cells. The insights we obtain from the analysis can supplement
and guide the design of air-breathing fuel cells. The analysis
revealed some trends that are quite different from those observed
in fuel cells with conditioned reactant streams. Of particular
note is the elucidation of the significant effect of heat trans-
fer on membrane dry-out in ambient air-breathing fuel cells.
The study suggests that improved performance of air-breathing
fuel cells can be achieved by increasing the heat removal
rate and thus promoting higher relative humidity levels in
the GDL.
Acknowledgements
The authors gratefully acknowledge Tibor Fabian of Stanford
University for providing experimental data for the validation
of the model. Funding for this research was provided by the
Natural Sciences and Engineering Research Council of Canada
(NSERC) in the form of a Canada Graduate Scholarship and
a supplement from the Canadian Space Agency to SL. This
research was also funded in part by an NSERC Discovery Grant
to ND and the Canada Research Chairs Program.
 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862 3861

Appendix A. Nomenclature CL catalyst layer

eff effective value
GDL/CL GDL and catalyst layer interface
a activity H2 hydrogen
AA active area (m2 ) H2 O water
AS planar surface area (m2 ) Mem membrane
Bi Biot number N Nafion
c specific heat of solid; also concentration (J kg−1 K−1 ; O2 oxygen
mol m−3 ) OC open-circuit

py
Cp specific heat of gas (J kg−1 K−1 ) Pt platinum
D diffusion coefficient (m2 s−1 ) ref reference value
E potential
F Faraday’s constant (96485 C mol−1 )

co
References
G Gibbs energy of formation (J mol−1 )
h heat transfer coefficient (W m−2 K−1 ) [1] C.K. Dyer, J. Power Sources 106 (2002) 31.
hA mass transfer coefficient (kg m−2 s−1 ) [2] H. Kariatsumari, H. Yomogita, Nikkei Electronics Asia (February 2005).
i local current density (A m−2 ) [3] J.S. Wainright, R.F. Savinell, C.C. Liu, M. Litt, Electrochim. Acta 48 (2003)
j0 volumetric exchange current density (A m−3 ) 2869.
k thermal conductivity (W m−1 K−1 ) [4] A. Schmitz, S. Wagner, R. Hahn, H. Uzun, C. Hebling, J. Power Sources
127 (2004) 197.

al
l convection length scale (m) [5] R. Hahn, S. Wagner, A. Schmitz, H. Reichl, J. Power Sources 131 (2004)
m specific mass loading (kg m−2 ) 73.
M molar mass (kg mol−1 ) [6] T. Fabian, J.D. Posner, R. O’Hayre, S.-W. Cha, J.K. Eaton, F.B. Prinz, J.G.
specific mass flux (kg m−2 s−1 )
on
ṅ Santiago, J. Power Sources 161 (2006) 168.
[7] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, J. Electrochem. Soc. 138
Nu Nusselt number
(1991) 2334.
P pressure (Pa) [8] D.M. Bernardi, M.W. Verbrugge, J. Electrochem. Soc. 139 (1992) 2477.
qp flux of heat produced (W m−2 K−1 ) [9] T.F. Fuller, J. Newman, J. Electrochem. Soc. 140 (1993) 1218.
flux of heat transferred (W m−2 K− )1
rs

qt [10] T.V. Nguyen, R.E. White, J. Electrochem. Soc. 140 (1993) 2178.
R universal gas constant (8.3145 J mol−1 K−1 ) [11] J.C. Amphlett, R.M. Baumert, B.A. Peppley, P.R. Roberge, J. Electrochem.
sPt specific platinum surface area (m2 kg−1 ) Soc. 142 (1995) 1.
[12] J.S. Yi, T.V. Nguyen, J. Electrochem. Soc. 145 (1998) 1149.
T temperature (K)
pe

[13] V. Gurau, F. Barbir, H. Liu, J. Electrochem. Soc. 147 (2000) 2468.

V volume (m3 ) [14] R. Bradean, K. Promislow, B. Wetton, Num. Heat Transfer, A 42 (2002)
WCL catalyst layer thickness (m) 121.
y species mass fraction [15] P. Berg, K. Promislow, J. Pierre St., J. Stumper, J. Electrochem. Soc. 151
(2004) A341.
yN Nafion content
[16] A.A. Kulikovsky, Electrochem. Comm. 6 (2004) 969.
yPt platinum/carbon ratio [17] T. Berning, N. Djilali, J. Power Sources 124 (2003) 440.
[18] D. Natarajan, T.V. Nguyen, J. Power Sources 115 (2003) 66.
r's

Greek letters [19] P.T. Nguyen, T. Berning, N. Djilali, J. Power Sources 130 (2004) 149.
α transfer coefficient [20] S. Shimpalee, S. Greenway, D. Spruckler, J.W. Van Zee, J. Power Sources
135 (2004) 79.
δ heat transfer length scale (m)
[21] B. Sivertsen, N. Djilali, J. Power Sources 141 (2005) 65.
Δs entropy change (J mol−1 K−1 ) [22] S. Um, C.Y. Wang, J. Power Sources 125 (2004) 40.
ε porosity
o

[23] L. Wang, H. Liu, J. Power Sources 134 (2004) 185.

φ electrolyte phase potential (V) [24] T. Hottinen, M. Mikkola, P. Lund, J. Power Sources 129 (2004) 68.
γ concentration exponent [25] C. Ziegler, A. Schmitz, M. Tranitz, E. Fontes, J.O. Schumacher, J. Elec-
th

trochem. Soc. 151 (2004) A2028.

η overpotential (V)
[26] T. Berning, D.M. Lu, N. Djilali, J. Power Sources 106 (2002) 284.
ρ density (kg m−3 ) [27] S. Litster, J.G. Pharoah, G. McLean, N. Djilali, J. Power Sources 156 (2006)
σ conductivity (S m−1 ) 334.
Au

τ tortuosity factor [28] A. Parthasarathy, S. Srinivasan, A.J. Appleby, C.R. Martin, J. Electrochem.
υPt platinum utilization Soc. 139 (1992) 2856.
[29] C.Y. Du, P.F. Shi, X.Q. Cheng, G.P. Yin, Electrochem. Comm. 6 (2004)
435.
Subscripts/superscripts [30] M. Eikerling, A.A. Kornyshev, J. Electroanal. Chem. 453 (1998) 89.
a anodic [31] M.L. Perry, J. Newman, E.J. Cairns, J. Electrochem. Soc. 145 (1998) 5.
A anode [32] S. Litster, Mathematical modelling of fuel cells for portable devices,
air air M.A.Sc. Thesis, University of Victoria, 2005.
[33] L. Pisani, M. Valentini, G. Murgia, J. Electrochem. Soc. 150 (2003) A1549.
Amb ambient
[34] Q. Wang, M. Eikerling, D. Song, Z. Liu, T. Navessin, Z. Xie, S. Holdcroft,
c cathodic J. Electrochem. Soc. 151 (2004) A950.
C cathode [35] S. Litster, G. McLean, J. Power Sources 130 (2003) 61.
cell fuel cell [36] C. Marr, X. Li, J. Power Sources 77 (1999) 17.
3862 S. Litster, N. Djilali / Electrochimica Acta 52 (2007) 3849–3862

[37] W. Sun, B.A. Peppley, K. Karan, Electrochim. Acta 50 (2005) 3359. [47] Y. Sone, P. Ekdunge, D. Simonsson, J. Electrochem. Soc. 143 (1996)
[38] H. Inoue, H. Daiguji, E. Hihara, JSME Int. J. Ser. B 47 (2004) 228. 1254.
[39] R.E. De La Rue, C.W. Tobias, J. Electrochem. Soc. 106 (1959) 827. [48] E.L. Cussler, Diffusion–Mass Transfer in Fluid Systems, Cambridge Uni-
[40] H. Ju, C.Y. Wang, J. Electrochem. Soc. 151 (2004) A1954. versity Press, New York, NY, 1997.
[41] G. Sasikumar, J.W. Ihm, H. Ryu, Electrochim. Acta 50 (2004) 598. [49] G.J.M. Janssen, M.L.J. Overvelde, J. Power Sources 101 (2001) 117.
[42] P.D. Beattie, V.I. Basura, S. Holdcroft, J. Electroanal. Chem. 468 (1999) [50] J. Fimrite, B. Carnes, H. Struchtrup, N. Djilali, J. Electrochem. Soc. 152
180. (2005) A1815.
[43] L. Zhang, C. Ma, S. Mukerjee, J. Electroanal. Chem. 568 (2004) 273. [51] T. Thampan, S. Malhotra, H. Tang, R. Datta, J. Electrochem. Soc. 147
[44] J.O.M. Bockris, A.K.N. Reddy, M. Gamboa-Aldeco, Modern Electrochem- (2000) 3242.
istry 2A: Fundamentals of Electronics, Kluwer/Plenum, New York, 1998. [52] J.F. Moxley, S. Tulyani, J.B. Benziger, Chem. Eng. Sci. 58 (2003) 4705.

py
[45] A.A. Kornyshev, A.A. Kulikovsky, Electrochim. Acta 46 (2001) 4389. [53] N.V. Suryanarayana, Engineering Heat Transfer, West Publishing Com-
[46] T. Navessin, S. Holdcroft, Q. Wang, D. Song, Z. Liu, M. Eikerling, J. pany, 1995.
Horsfall, K. Lovell, J. Electroanal. Chem. 567 (2004) 111. [54] Z.G. Qi, A. Kaufman, J. Power Sources 113 (2003) 37.

co
al
on
rs
pe
o r's
th
Au

View publication stats