A three-dimensional numerical model of a single-chamber

solid oxide fuel cell

Naveed Akhtar
a,b,
*, Stephen P. Decent
a
, Daniel Loghin
a
, Kevin Kendall
b
a
School of Applied Mathematics, University of Birmingham, Birmingham B15 2TT, UK
b
Department of Chemical Engineering, University of Birmingham, Birmingham B15 2TT, UK
a r t i c l e i n f o
Article history:
Received 21 June 2009
Received in revised form
28 July 2009
Accepted 31 July 2009
Available online 25 August 2009
Keywords:
Single-chamber
Solid oxide fuel cell
Numerical model
Fueloxidant mixture
COMSOL Multiphysics
a b s t r a c t
The aim of this work is to analyze the hydrodynamic/electrochemical performance of
a solid oxide fuel cell operating on nitrogen diluted hydrogen/oxygen mixture. In this
respect, a three-dimensional numerical model of a single-chamber solid oxide fuel cell
(SC-SOFC) is developed. The model incorporates the coupled effects of uid ow in
a rectangular duct with mass transport in porous electrodes, selective electrochemical
reactions (i.e. hydrogen oxidation on anode and oxygen reduction on cathode) on indi-
vidual electrodes while operating on nitrogen diluted hydrogenoxygen mixture. Results
show the effect of depletion of gaseous mixture due to hydrogen and oxygen consumption
along the ow direction. The model can predict hydrodynamic/electrochemical effects by
varying the porosity of the gas diffusion electrodes/catalyst layers. The model is formu-
lated in COMSOL Multiphysics 3.4, a commercial Finite Element Method (FEM) based
software package.
2009 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved.
1. Introduction
A conventional solid oxide fuel cell consists of two electrodes
(i.e. an anode and a cathode, both of which are porous), an
electrolyte(whichis non-porous, andsandwichedbetweenthe
two electrodes), an interconnect (for current collection and
stacking purpose) and channels (also called bipolar plates) for
fuel and air supply. Conventional solid oxide fuel cells (SOFCs),
also known as dual-chamber SOFCs, due to their separate
compartments for fuel and air are operated inthe temperature
range of 7001000

C [1]. This high temperature requirement
limits the applicability of SOFCs as power sources for electric
vehiclesandportabledevices, andposes seriousproblemswith
their structural design. Due to the operation at this elevated
temperature, high thermal stresses become more likely,
leading to thermal strains which can destroy the cell. In short,
conventional SOFCs face a number of technical issues,
including:
Bulky geometrical design due to the large weight of the
bipolar plates.
Slow start-up time.
Large thermo-mechanical stresses.
Requirement of gas tight sealing to prevent fuel and air
leakage.
Onestrategyfor addressingtheaboveissues istousesingle-
chamber, solidoxidefuel cells (SC-SOFCs) inwhichthefuel and
air are premixed; anode and cathode reactions can take place
within the same physical volume. Due to elimination of sealing
between the anode and cathode, the SC-SOFC greatly simplies
the system design. An added advantage of using SC-SOFCs is,
* Corresponding author. Department of Chemical Engineering, University of Birmingham, Birmingham B15 2TT, UK.
Tel.: 44 7726 126748; fax: 44 121 4143389.
E-mail address: nxa675@bham.ac.uk (N. Akhtar).
Avai l abl e at www. sci encedi r ect . com
j our nal homepage: www. el sevi er . com/ l ocat e/ he
0360-3199/$ see front matter 2009 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2009.07.113
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3
they have to operate at relatively lowtemperatures (400600

C)
because selectivity of the electrodes decreases with increase in
temperature. This low temperature operation enhances the
thermal and mechanical shock resistance (much safer from
a structural point of view), thereby enabling rapid start-up and
shut-down. Not only this, the SC-SOFCs can be made much
lighter because the heavy bipolar plates (separate owchannels
for fuel and air) are not required [2].
The research work on SC-SOFCs is mainly experimentally
driven with a very few reports on modelling of SC-SOFCs [2].
These experiments give information about the overall cell
performance, a deeper understanding of the hydrodynamic
and electrochemical interaction is lacking. Modelling can be
a best tool to study all insights but unfortunately there is
a very little work in this direction. The models developed so
far are mainly limited to two-dimensional analysis, showing
that the theoretical studies in this area are at an early stage
yet [37]. The only developed three-dimensional model by
Chung et al. [8] does not provide any information on ow/
species distribution inside the gas-chamber. This model is
also lacking temperature distribution due to isothermal
assumption. The only two active groups (Hao et al., USA and
Chung et al., Korea) have reported initial numerical studies
mainly focused on understanding the causes of low perfor-
mance in SC-SOFCs. Very recently, Hao et al. [6] studied the
effect of various parameters such as ow rate, mixing ratio,
ow geometry, stack and balance gas on the performance of
conventionalplanar type SC-SOFC. Their main emphasis
was to study the causes of low fuel utilization (<1%) often
reported in experimental literature [2]. The main conclusion
of their study was that the low performance of SC-SOFCs is
because of short convection time scales and the use of
single-cells. They suggested that the electrode microstruc-
ture (activity and selectivity) and design of stack will be
advantageous in improving the cell performance and ef-
ciency [6].
The group of Chung et al. [5] numerically studied the co-
planar SC-SOFC design. They reported that the performance
of a co-planar type SC-SOFC system is found to be largely
dependent on the structural parameters, i.e. the distance
between the electrodes, the thickness of the electrolyte and
the inow direction of the gas mixture. While they suggested
possible improvements in the power density by reducing the
inter-distance between the electrodes and increasing the
electrolyte thickness, an optimumfor electrode inter-distance
was not discussed.
The numerical results of both of the above groups were
pointing on poor uid ow and electrochemical manage-
ment which is the main cause in lowering the cell perfor-
mance of an SC-SOFC. We tried to focus on these issues
and our modelling objective is to look more closely into the
hydrodynamic issues (related to the uid ow) and elec-
trochemical issues (related to the cell performance) of
a novel-planar SC-SOFC design. We extended our earlier
modelling studies on SC-SOFCs [9,10] to three-dimensional
analysis in order to capture the insights. The nal aim was
to suggest a suitable geometry that can minimize the
adverse effects of both of the above issues, directing
towards the realization of a better experimental setup, if
possible.
2. The mathematical model
As stated earlier, the SC-SOFCs operate ona mixture of fuel and
oxidant supply. Though, hydrocarbons (like methane, propane
and butane) are the most practical fuels due to their relatively
narrow explosive limit compared to hydrogen, their anode
chemistry is much complicated and not very well understood
yet [7]. Ontheother hand, useof hydrogenasafuel simpliesthe
overall anode chemistry and does not lead to coking problems
which are normally encountered in case of hydrocarbons. For
the sake of simplicity, our initial efforts will focus on hydrogen/
oxygen mixtures diluted with nitrogen for safety reasons.
The complete cell system is subdivided into individual
layers, andmodelling of eachlayer is presented below. The cell
consideredinthis studyis composedof nickel supportedanode
gas diffusion electrode, nickelyttria-stabilized zirconia (Ni
YSZ) anode catalyst layer, yttria-stabilized zirconia (YSZ)
electrolyte, yttria-stabilized zirconialanthanum strontium
manganite (YSZLSM) cathode catalyst layer and lanthanum
strontium manganite (LSM) as cathode electrode. In a rst
attempt tosimplifythemodel, it hasbeenassumedthat thecell
is fed with nitrogen diluted hydrogen/oxidant mixture instead
of hydrocarbon/oxidant mixtures. The addition of 96%
nitrogenisnecessarytoensurethat themixtureissafe[11]. The
simulation results shown in this study are on the basis of the
remainder 4%. Furthermore, a separate section (Section 5) is
devoted for safety discussion on hydrogen/oxygen mixtures.
The following reactions are assumed to occur in the elec-
trodes (full direct combustion is avoided due to the dilution
effect):
1) Reaction in the anode:
H
2
O
2
/H
2
O2e

(1)
2) Reaction in the cathode:
1
2
O
2
2e

/O
2
(2)
2.1. Model description
The fuel cell analyzed in the present study is of planar type
congurationwithaholeinthemiddleof thecell toallowradial
ow into and through the porous electrodes. It is composed of
seven sub-domains, namely the gas-chamber, two porous
electrodes (anode and cathode), two porous catalyst layers
(anode and cathode catalyst layer), the electrolyte and the hole
inthe cell. The geometrical dimensions of the gas-chamber are
listedinTable 1. The diffusionparameters are giveninTables 2
and 3. All physical properties such as density, heat capacity,
viscosity, thermal conductivity and diffusivity are considered
as temperature dependent (see Table 4). These properties do
not dependoncompositionproportionality becauseof the very
different behavior of the species considered in the mixture,
therefore these individual properties are further corrected for
gaseous mixture. Detailedformulationof mixture properties is
presented by Todd et al. [12].
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8646
2.2. Model assumptions
The ow is steady.
The electrodes are assumed to be ideally selective for the
respective electrochemical reactions, therefore, electro-
chemical oxidation of fuel and reduction of oxygen are
taking place in the anode and cathode catalyst layers,
respectively [8].
Ohmic heating (in the porous electrodes and catalyst layers)
due to electrical current transport is neglected because of
high electrical conductivity as compared to the ionic
conductivity [13].
The electrolyte is a non-porous (dense, solid) material.
The effect of radiative heat exchange between the cell and
furnace walls can be neglected because of negligible
temperature gradients. This assumption is made on the
basis that the real catalytic chemistry is not known due to
experimental limitations and the electrodes are assumed to
be ideally selective with reversible/irreversible loss as the
only source of heat.
The conduction heat transfer is dominant in the cell as
compared to the heat transport by convection because of
the low gas speed in the porous electrodes.
The outer surfaces of the anode and cathode diffusion
electrodes are used as current collectors, therefore the
effects of interconnects are neglected.
In following sections, detailed modelling strategy is pre-
sented for each sub-domain.
2.3. Computational domain
2.3.1. Gas-chamber
The gas-chamber consists of a rectangular duct with
a membrane electrode assembly also called positive electrode,
electrolyte and negative electrode (PEN) located in the center
of the duct [Fig. 1]. The PEN element has a hole in the center to
allow gaseous mixture to pass through the axial direction
thereby relieving the pressure that could be built-up in front of
the cell. The applicable equations are:
Continuity equation:
V$ru 0 (3)
Momentum equation:
ru$Vu Vp mV
2
u
1
3
mVV$u (4)
Species conservation equation:
V$ruY
i
V$j
i
(5)
Energy conservation equation:
V$
_
rC
p
uT
_
V$kVT (6)
The density of the mixture is calculated using [14]:
r
1

N
i1
Y
i
=r
i
(7)
The density of each species, r
i
is obtained from the perfect
gas law relation [14]:
r
i

pM
i
RT
(8)
Concentration of each species is calculated by:
c
i

pX
i
RT
(9)
where X
i
is the mole fraction of the ith species which is related
to the mass fraction Y
i
by the following relation:
X
i
Y
i
_
M
M
i
_
(10)
and
M

N
i1
X
i
M
i
(11)
In equation (5) the multicomponent diffusive mass ux
vector (j
i
) is described by the generalized Ficks law [15,16]:
j
i

N1
j1
rD
ij
VY
j
(12)
Table 1 Geometry dimensions.
Dimensions Values (mm)
Chamber length (x-axis) 150
Chamber height ( y-axis) 25
Chamber width (z-axis) 25
PEN width 20
PEN height 20
Anode thickness 70 10
3
Cathode thickness 50 10
3
Electrolyte thickness 20 10
3
Anode catalyst
layer thickness
5 10
3
Cathode catalyst
layer thickness
5 10
3
Hole diameter 2.6
Table 2 Diffusion volumes in FuellerSchettlerGiddings
correlation parameters.
Molecule Diffusion volume (cm
3
/mole)
H
2
7.07
O
2
16.6
H
2
O 12.7
N
2
17.9
Table 3 MaxwellStefan diffusion coefcients calculated
using values given in Table 2.
Molecular pair D
ij
(m
2
/s)
H
2
N
2
4.0176 10
4
H
2
O
2
4.1178 10
4
N
2
O
2
1.0982 10
4
H
2
H
2
O 4.6541 10
4
N
2
H
2
O 1.3972 10
4
O
2
H
2
O 1.3992 10
4
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8647
Table 4 Input parameters used in the model.
Property Symbol Value Units References
Working electrical
potential at anode
4
s,a
0 V [28]
Working electrical
potential at cathode
4
s,c
0.7 V [28]
Anodic pre-exponential coefcient g
a
1.6e
9
[13]
Cathodic pre-exponential coefcient g
c
3.9e
9
[13]
Anodic activation energy E
act,a
120 J mole
1
[13]
Cathodic activation energy E
act,c
120 J mole
1
[13]
Effective anode
ionic conductivity
s
eff
ea
0.29 S m
1
[27]
Effective cathode
ionic conductivity
s
eff
ec
0.24 S m
1
[27]
Effective anode
electronic conductivity
s
eff
sa
4800 S m
1
[27]
Effective cathode
electronic conductivity
s
eff
sc
1600 S m
1
[27]
Anode electronic conductivity s
sa
2.0 10
6
S m
1
[32]
Electrolyte ionic conductivity s
e
3.34 10
4
exp (10,350/T ) S m
1
[32]
Cathode electronic conductivity s
sc
(42 10
6
/T ) exp (1150/T ) S m
1
[32]
Inlet temperature T
0
773 K []
Anodic anodic
charge transfer coefcient
a
a
a
2 [27]
Anodic cathodic
charge transfer coefcient
a
a
c
1 [27]
Cathodic anodic
charge transfer coefcient
a
c
a
1.5 [27]
Cathodic cathodic
charge transfer coefcient
a
c
c
0.5 [27]
Faradays constant F 96487 C mole
1
[24]
Universal gas constant R 8.314 J mole
1
K
1
[24]
Porosity e 0.3 [13]
Tortuosity s 3.80 [24]
Electrochemically active
surface area
A
s
102,500 m
1
[33]
Permeability k 1.0 10
13
m
2
[28]
Hydrogen inlet
mass fraction
(of 4%)
Y
H2in
0.5 [5]
Oxygen inlet
mass fraction
(of 4%)
Y
O
2in
0.105 [5]
Nitrogen inlet
mass fraction
(of 4%)
Y
N
2in
0.38 [5]
Water inlet
mass fraction
(of 4%)
Y
H2O
in
0.015 [5]
Operating pressure p
o
1.013 10
5
Nm
2
[23]
Inlet velocity u
in
0.1 ms
1
[]
Average pore diameter d
p
1.0 mm [23]
Hydrogen viscosity m
H2
6.162e
6
1.145e
8
T Pa s [13]
Oxygen viscosity m
O2
1.668e
5
3.108e
8
T Pa s [13]
Water viscosity m
H2O
4.567e
6
2.209e
8
T Pa s [13]
Nitrogen viscosity m
N2
1.435e
5
2.642e
8
T Pa s [13]
Hydrogen specic heat C
p,H2
13960 0.950 T J kg
1
K
1
[13]
Oxygen specic heat C
p,O2
876.80 0.217 T J kg
1
K
1
[13]
Water specic heat C
p,H2O
1639.2 0.641 T J kg
1
K
1
[13]
Nitrogen specic heat C
p,N2
935.6 0.232 T J kg
1
K
1
[13]
Hydrogen thermal conductivity k
H2
0.08525 2.964e
4
T Wm
1
K
1
[13]
Oxygen thermal conductivity k
O2
0.01569 5.690e
5
T Wm
1
K
1
[13]
Water thermal conductivity k
H2O
0.01430 9.782e
5
T Wm
1
K
1
[13]
Nitrogen thermal conductivity k
N2
0.01258 5.444e
5
T Wm
1
K
1
[1]
Anode thermal conductivity k
a
3 Wm
1
K
1
[13]
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8648
where, Ntotal number of species in the mixture. D
ij
in
equation (12) is the multicomponent diffusion coefcient
which in general is not symmetric (D
ij
sD
ji
). Also, the multi-
component D
ij
does not have the physical signicance of the
binary Fick diffusivity in that the D
ij
do not reect the ij
interactions [17]. Multicomponent diffusioncoefcients D
ij
are
interrelated with MaxwellStefan diffusion coefcients D
ij
through matrix B, such that [17]:
D B
1
G (13)
For ideal gases the thermodynamic matrix G reduces to the
identity matrix and equation (13) becomes:
D B
1
(14)
where D is the multicomponent diffusion coefcient matrix
and B is a square matrix of order N1 with elements given by:
B
ii

Y
i
M
M
i
D
iN

N
k1;isk
Y
k
M
M
k
D
ik
(15)
B
ij

Y
i
M
M
i
_
1
D
ij

1
D
iN
_
; isj (16)
D in equations (15) and (16) is the MaxwellStefan diffusion
coefcient for binary pairs and is dependent on both
temperature and pressure [18]. For gas pressures up to about
10 atm at moderate to high temperatures, the diffusion coef-
cient for a binary mixture of gases i and j may be estimated
from the Fueller, Schettler and Giddings relation [19]:
D
ij

T
1:75
_
1=M
i
1=M
j
_
1=2
p
_
V
1=3
i
V
1=3
j
_
2
10
7
(17)
where D
ij
is the MaxwellStefan diffusion coefcient in m
2
s
1
,
T is the temperature in kelvin (K), p is the pressure in atmo-
spheres (atm), M
i
is the molecular weight of molecules in
g mol
1
, andV
i
is the molecular diffusionvolume incm
3
mol
1
.
The values of V
i
for different molecules are tabulated in [20].
Typical values of D
ij
m
2
s
1
for molecules commonly used in
fuel cells are calculated using equation (17) and are given in
Table 3 at a pressure of 1 atmand at anaverage temperature of
773 K, which is assumed to be the operating condition for the
single-chamber solid oxide fuel cell in this study.
Note that equation (17) is dimensionally inhomogeneous
(i.e. it is only valid for the stated unit system, and not for any
other unit system, unlike all other equations) and is a result of
regression analysis of 340 experimentally determined binary
Fig. 1 Gas-chamber with PEN element inside (not to scale).
Table 4 (continued)
Property Symbol Value Units References
Cathode thermal conductivity k
c
3 Wm
1
K
1
[13]
Electrolyte thermal conductivity k
e
2 Wm
1
K
1
[13]
Anode specic heat C
a
595 J kg
1
K
1
[1]
Cathode specic heat C
c
573 J kg
1
K
1
[1]
Electrolyte specic heat C
e
606 J kg
1
K
1
[1]
Anode density r
a
6870 kg m
3
[1]
Cathode density r
c
6570 kg m
3
[1]
Electrolyte density r
e
5900 kg m
3
[1]
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8649
diffusion coefcients [20]. Also, this equation can be easily
applied for pressure and temperature ranges used in a typical
fuel cell operation.
2.3.2. Gas diffusion electrodes
Gas diffusion electrodes consist of an anode and a cathode
which are porous media. The following equations model gas
diffusion electrodes:
Continuity equation:
V$ru S
m
(18)
S
m
is the source/sink term for the production/consumption
of gas molecules and is zero for gas diffusion electrodes
because the electrochemical reaction is assumed to take
place only in the catalyst layers. The reason for introducing
the source/sink term (given in Ref. [21]), is that sometimes
the reactive layers can extend into the electrodes.
Momentum equation:
In porous media ow where viscous forces dominate
convective ones, the momentum equation in the porous
media may be modied from the NavierStokes equation to
the Brinkman equation. In order to do so, the convective
termhas beenneglected and anadditional termfor pressure
drop in porous media, given by Darcys law, has been added.
This describes ow in porous media with a pressure
gradient as the only driving force [22].
Vp
m
k
u (19)
By inserting this term in the NavierStokes equation, we
have the Brinkman equation as:
m
k
u Vp mV
2
u
1
3
mVV$u (20)
Species conservation equation:
V$ruY
i
V$j
i
S
i
(21)
where, S
i
is the source/sink term for the production/
consumption of species, set to zero for gas diffusion elec-
trodes because of the above mentioned reason.
j
i
is the multicomponent diffusive mass ux vector in
porous media, given by:
j
i

N1
j1
rD
eff
DG
VY
j
(22)
where D
eff
DG
is the effective dusty gas diffusivity.
Diffusion in porous media is usually described by
a molecular (particle-particle collision) and/or a Knudsen
(particlewall collision) diffusion mechanism [23]. In order
to account for a detailed diffusion mechanism, both modes
have been considered by implementing the Dusty Gas
Model (DGM). The DGM is derived by considering the solid
matrix as large stationary spheres suspended in the gas
mixture as one of the species present. The DGMdiffusivity is
then given by [23]:
D
DG
i;j

D
ij
D
k;i
D
ij
D
k;i
(23)
where
D
k;i

1
3
d
p

8RT
pM
i

(24)
The values of D
ij
are calculated using equation (17) and the
DGM diffusitivities are further corrected using the following
expression [23]:
D
eff
DG
i;j

_
e
s
_
D
DG
i;j
(25)
Finally, the value of D
eff
DG
i;j
is used to calculate the matrix
D
eff
DG
i;j
, having elements:
B
eff
DG
ii

Y
i
M
M
i
D
eff
DG
i;N

N
k1;isk
Y
k
M
M
k
D
eff
DG
i;k
(26)
B
eff
DG
ij

Y
i
M
M
i
_
1
D
eff
DG
i;j

1
D
eff
DG
i;N
_
; isj (27)
Charge conservation equation:
Electrical current transport is described by a governing
equation for conservation of charge [13,23]:
V$s
s
Vf
s
A
s
S
f
s
(28)
S
f
s
, electrical current source term which is kept zero in the
anode and cathode gas diffusion electrodes.
Energy conservation equation:
Energy transport in gas diffusion electrodes is modelled
by considering porous nature of the electrodes. Instead of
using the thermal conductivity (k
s
), density (r
s
) and specic
heat (C
p,s
) of the solid matrix, effective thermal conductivity
(k
eff
), effective density (r
eff
) and effective specic heat (C
p,eff
)
have been used in the model [24].
V$
_
_
rC
p
_
eff
uT
_
V$
_
k
eff
VT
_
S
e
(29)
Effective properties of the porous media are given by
[24]:
_
rC
p
_
eff
1 er
s
C
p;s
erC
p
(30)
k
eff
2k
s

_
e
2k
s
k

1 e
3k
s
_
1
(31)
In equation (29), S
e
is the heat source term which is zero.
2.3.3. Catalyst layers
An additional ionic charge conservation equation in the
catalyst layer will represent the ionic current, and is given by
[13,23]:
V$
_
s
eff
e
Vf
e
_
A
s
S
f
e
(32)
S
f
e
, ionic current source term dened below.
Other transport equations in the catalyst layers are the
same as in the electrodes, except the source/sink terms are
activated in the conservation equations and the electrical
conductivity term (s
s
) in equation (28) is changed to effective
electrical conductivity (s
eff
s
) in order to account for mixed
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8650
ionicelectronic conductivity in the catalyst layers. The
source/sink terms in mass and species conservation equation
are given below [25,26].
At the anode side, hydrogen is consumed and water is
produced, so that:
S
m
S
i
S
H
2

M
H
2
2F
i
a
(33)
S
m
S
i
S
H
2
O

M
H
2
O
2F
i
a
(34)
At the cathode side, oxygen is consumed:
S
m
S
i
S
O
2

M
O
2
4F
i
c
(35)
Source/sink terms in the electrical and ionic charge
conservation equation are given below [27].
In the anode catalyst layer: electrical current density is
a sink term,
S
f
s
i
a
(36)
In the anode catalyst layer: ionic current density is a source
term,
S
f
e
i
a
(37)
In the cathode catalyst layer: electrical current density is
a sink term,
S
f
s
i
c
(38)
In the cathode catalyst layer: ionic current density is
a source term,
S
f
e
i
c
(39)
where i
a
anodic current density given by the ButlerVolmer
equation [27]:
i
a
i
0;a
_
exp
_
a
a
a
Fh
acta
RT
_
exp
_

a
a
c
Fh
acta
RT
__
(40)
i
c
cathodic current density given by [27]:
i
c
i
0;c
_
exp
_
a
c
a
Fh
actc
RT
_
exp
_

a
c
c
Fh
actc
RT
__
(41)
The exchange current densities, i
0,a
and i
0,c
are expressed as
a function of local partial pressure of the species [13]:
i
0;a
g
a
_
p
H
2
p
ref
_ _
p
H
2
O
p
ref
_
0:5
exp
_
E
act;a
RT
_
(42)
i
0;c
g
c
_
p
O
2
p
ref
_
0:25
exp
_
E
act;c
RT
_
(43)
where p
ref
is the reference pressure in the gas-chamber, i.e.
the total pressure of 1 atm. g
a
and g
c
are the anodic and
cathodic pre-exponential coefcients, E
act,a
and E
act,c
are the
anodic and cathodic activation energies, respectively.
The anode and cathode side activation overpotentials are
calculated by [28]:
h
acta
4
s=a;cl
4
e=a;cl
anode side activation overpotential:
h
actc
4
s=c;cl
4
e=c;cl
V
oc
cathode side activation overpotential:
where 4
s/a,cl
and 4
e/a,cl
are respectively the solid phase (elec-
tronic) and electrolyte phase (ionic) potential in the anode
catalyst layer, 4
s/c,cl
and 4
e/c,cl
are the electronic and ionic
potential in the cathode catalyst layer, and V
oc
is the open
circuit (Nernst) voltage, as expressed by [29]:
V
oc
V
o

RT
2F
ln
_
p
H
2
p
H
2
O
_

RT
4F
ln
_
p
O
2
_
(44)
Since the model is solved using Finite Element Method
(FEM), all unknown variables are calculated at each node. For
example, the operating voltage is dened at the cathode outer
surface and a zero voltage is set at the anode outer surface. By
implementing the charge conservation equation (i.e. Ohms
law), the ohmic loss is embedded while calculating the current
at each node (since at each node, all equations are solved in
FEM). The pressure distribution is calculated via mass trans-
port model, and, at each node, current is calculated corre-
sponding to that concentration. Therefore, there is no need for
explicit expressions for ohmic and concentration over-
potentials in FEM, rather they are an integral part of the model
being solved numerically.
The energy equation in the catalyst layers remains the
same as was in gas diffusion electrodes (equation (29)), except
the heat source term, S
e
, is activated because of chemical/
electrochemical heat generation in the catalyst layers. For
anode this term is given by [13]:
S
e
q
reva
q
irra
q
ohmea
(45)
where reversible heat generation in the anode [13,25]:
q
reva
T dsi
a
=zF (46)
ds
z
23:328 0:0042T (47)
irreversible heat generation in the anode [13]:
q
irra
h
acta
i
a
(48)
ohmic heat due to ionic resistance [13]:
q
ohmea
s
eff
ea
Vf
e
$Vf
e
(49)
Heat source term for the cathode is given by [13]:
S
e
q
irrc
q
ohmec
(50)
irreversible heat generation in the cathode [13]:
q
irrc
h
actc
i
c
(51)
ohmic heat due to ionic resistance [13]:
q
ohmec
s
eff
ec
Vf
e
$Vf
e
(52)
2.3.4. Electrolyte
The electrolyte is impermeable to gases and allows only ionic
charge transfer, so that:
V$s
e
Vf
e
0 (53)
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8651
Since there is no generation of ionic or electrical current
inside the electrolyte, the right hand side of the above equa-
tion is zero. Furthermore, the electrolyte is a dense, non-
porous material and thus only conduction is possible. The
only heat source term in electrolyte is the ohmic resistance
due to ionic current transfer [13,28].
V$k
s
VT S
e
0 (54)
S
e
s
e
Vf
e
$f
e
(55)
2.4. Boundary conditions
The boundary conditions for each layer are given below:
2.4.1. Gas-chamber
2.4.1.1. Inlet. At the inlet, the velocity in the x-direction is
prescribed to be:
u u
in
(56)
At the inlet, the mass fraction is dened as:
Y
i
Y
i
in
(57)
At the inlet, operating temperature is dened [13]:
T T
0
(58)
2.4.1.2. Walls. At the walls no-slip boundary condition is
applied. A no-slip condition means that the uids velocity is
equal to the boundary velocity, which is zero in the case of
a xed wall.
u 0 (59)
A mass insulation boundary condition is applied at the
walls, meaning that no mass ux is allowed to cross these
boundaries.
n$
_
rD
ij
VY
i
ruY
i
_
0 (60)
All of the four side walls (i.e. top, bottom, side and left) are
assumed to be at furnace temperature which is the operating
temperature.
T T
0
(61)
2.4.1.3. Outlet. The outow boundary condition is prescribed
as:
p p
o
(62)
The convective ux boundary condition is applied at the
outlet, meaning that at the outlet boundary, diffusion term is
negligible.
n$
_
rD
ij
VY
i
_
0 (63)
At the outlet, the heat transport is convection dominated.
Convective heat ux boundary condition ensures that at the
outlet boundary the only heat transport is by convection and
thus conduction heat transfer is negligible at this boundary
[13].
n$ kVT 0 (64)
2.4.2. Gas diffusion electrodes
2.4.2.1. Anode electrode. At all exterior surfaces of the anode
and at an interface between the anode electrode and anode
catalyst layer, continuity of ow is applied. This puts no
constraints on the velocity and mass ux. The current on the
anode side is collected from the outer surface which faces
inlet ow direction. Therefore, a zero voltage (ground)
boundary condition is applied at the anode outer surface.
f
s
0 (65)
All other outer surfaces of the anode electrode are insulated
to the electrical current, hence it is assumed that there is no
current ow across these boundaries.
n$ s
s
Vf
s
0 (66)
At the interface between the anode electrode and anode
catalyst layer, continuity of electrical current is maintained.
2.4.2.2. Cathode electrode. The same boundary conditions
which are described above for the anode will also hold for the
cathode, except the electrical potential boundary conditionfor
the cathode electrode surface which faces the outlet ow
direction has been modied to:
f
s
V
c
(67)
This boundary condition denes the operating cell voltage
and closes the current circuit. At all exterior and interior
boundaries of the gas diffusion electrodes, continuity of heat
ux is maintained. This boundary condition species that the
normal heat ux inside of the boundary is equal to the normal
heat ux outside of the boundary [13].
n$k
s
VT
in
n$k
s
VT
out
(68)
2.4.3. Catalyst layers
2.4.3.1. Anode catalyst layer. At all exterior surfaces of the
anode catalyst layer continuity of ow is applied, meaning
putting no constraints on the velocity and mass ux.
However, at the interface between the anode catalyst layer
and the electrolyte, gases are not allowed to enter into the
electrolyte because of its non-porous nature. Hence, normal to
the electrolyte boundary, both the velocity and mass ux are
zero.
n$u 0 (69)
n$
_
rD
eff
DG
i;j
VY
i
ruY
i
_
0 (70)
At all exterior surfaces of the anode catalyst layer and at
interface between the anode catalyst layer and the electro-
lyte, electrical insulation boundary condition has been
applied.
n$
_
s
eff
s
Vf
s
_
0 (71)
Ionic current cannot ow out of the catalyst layers. All
exterior surfaces of the anode catalyst layer and at interface
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8652
between the anode electrode and the anode catalyst layer,
ionic current is insulated by applying ionic current insulation
boundary condition.
n$
_
s
eff
e
Vf
e
_
0 (72)
2.4.3.2. Cathode catalyst layer. The same boundary conditions
as described above for the anode catalyst layer will also hold
for the cathode catalyst layer. Furthermore, continuity of heat
ux is maintained on all surfaces of the catalyst layers.
n$k
s
VT
in
n$k
s
VT
out
(73)
2.4.4. Electrolyte
Since the electrolyte is impermeable to gases, both the mass
ux and velocity normal to all surfaces of the electrolyte are
zero.
n$u 0 (74)
n$
_
rD
eff
DG
i;j
VY
i
ruY
i
_
0 (75)
Continuity of ionic current is maintained at interfaces
between the anode/cathode catalyst layers and the electro-
lyte. All outer surfaces of the electrolyte are insulated to ionic
current by applying an ionic insulation boundary condition.
n$ s
e
Vf
e
0 (76)
Continuity of heat ux is maintained at all exterior surfaces
of the electrolyte, i.e.
n$k
s
VT
in
n$k
s
VT
out
(77)
The hole in the PEN allows free ow in a similar way as in
the gas-chamber, therefore continuity of owis maintained in
the hole. However, we consider the electrolyte surface
boundaries consisting of a hole not to allow the gases to
permeate into the electrolyte, therefore insulation boundary
conditions have been applied at these boundaries.
3. Numerical implementation
The model equations are solved using COMSOL Multi-
physics 3.4, a commercial Finite Element Method (FEM)
based software package. The computations were performed
on a 32-node Linux cluster; 32 dual 3 GHz Intel Xeon Sun
Fire V60 servers each with 4 GB memory. The mesh consists
of 9398 triangular elements of good quality and is shown in
Fig. 2. In order to ensure convergence and accuracy of the
results in three-dimensional domain, the following were
employed:
Use of lower element order to reduce the number of
unknowns.
The mesh was kept more rened in the cell element where
higher resolution was needed to capture large gradients.
A direct solver such as UMFPACK is very stable but not good
for large problems, since it requires too much memory.
Therefore, an iterative solver (GMRES) with a preconditioner
(Incomplete LU) and tolerance of 0.005 were used.
The systemof equations was solved iteratively, i.e. rst the
charge balance, then the NavierStokes equation, then the
mass/species balance, while at each stage the solution was
stored and used as an initial condition for the subsequent
stage. The total computing time for all stages (with 65, 192
degrees of freedom) was approximately 13.5 min for one value
of voltage scan. The grid independency test was performed
using different mesh sizes. Fig. 3 shows that the velocity eld
becomes grid independent as we move above 7154 elements.
In order to keep balance between accuracy and computational
time, a mesh with 9398 elements was opted.
4. Results and discussion
The values of the electrochemical/hydrodynamic transport
parameters for the operating conditions are listed in Table 4.
The governing equations are summarized in Table 5. In Figs.
413, the results shown are based on parameters listed in
Table 4, whereas for the results shown in Figs. 1417, the value
of porosity has been changed to 0.4 in order to see the effect of
electrochemistry and hydrodynamics, with all other param-
eters remain the same as in Table 4. Figs. 1820 show
temperature, IV and IP characteristics based on parameters
listed in Table 4, respectively.
In Fig. 4, the velocity eld in the gas-chamber is shown. It
can be seen that as soon as the ow reaches the anode elec-
trode surface, some of the ow is diverted through the edges
of the electrode surface. It shows that the cell is an obstacle to
the uid ow with uid traveling a tortuous path in the gas
diffusion electrode leading to a reduced velocity in the cell.
Due to the diversion near the cell edges, an increased velocity
of the ow is observed because of the free passage available
for the ow as compared to the limited passage available in
the cell. Just behind the cell, the velocity is reduced and the
ow regains its center velocity after a distance of approxi-
mately 0.03 m from the anode electrode surface. This distance
is an important parameter to be considered in a case where
one plans to investigate several cells downstream.
Fig. 5 shows pressure drop in the gas-chamber. It repre-
sents another parameter to observe the ow behavior. The
pressure drop along the ow direction is responsible for
accelerating the ow. Moreover, near the anode electrode
surface there is a large pressure drop (due to the presence of
cell as an obstacle) followed by a Darcy pressure drop in the
gas diffusion electrodes which is strongly dependent on the
permeability of porous electrodes, that in turn depends upon
porosity and tortuosity of the porous media. The pressure
drop behind the cell is responsible for further acceleration of
the uid downstream.
Fig. 6 shows the hydrogen concentration distribution in the
gas-chamber. As can be seen, the hydrogen concentration
remains constant until the gaseous mixture reaches the anode
electrode surface as there is no electrochemical reaction
upstream in front of the anode electrode. As soon as the ow
reaches the anode catalyst layer, it starts reacting electro-
chemically (according to equation (1)). Since concentration is
pressure dependent, one observes a slight increase near the
cell front edges due to the increase of pressure, as discussed
before. Just behind the cell, hydrogen concentration is low but
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8653
quickly mixes with diverted (non-reacted) hydrogen coming
through the cell edges via convection and diffusion. Just after
approximately, 0.02 m from the cathode last edge, hydrogen
concentration attains another lower constant value of
4.765 mole m
3
. Again, this parameter is of interest when one
plans to put several cells downstream. Also, it is important to
note that the hydrogen available for the downstream cell may
be depleted partly due to consumption at the anode electrode
of the upstream cell and partly due to mixing with the by-
product water vapor. Therefore, additional supply of hydrogen
is suggested for uniform cell performance in a stack of several
cells. Alternatively, one could use a distributed feedfor all cells
in a stack e.g. in a parallel cell arrangement rather than serial.
Fig. 7 shows the oxygen concentration in the gas-chamber
where it can be seen that the oxygen due to electrochemical
reaction (according to equation (2)) starts consuming in the
cathode catalyst layer. The large concentration gradient is
along the cell due to electrochemical reduction of oxygen at
the cathode catalyst layer. Just behind the cell a lower oxygen
concentration is observed and attains another lower constant
value (0.195 mole m
3
) at approximately 0.038 mfromcathode
last edge. This distance is approximately double what is
Fig. 2 The computational mesh (9398 elements).
Fig. 3 The grid independency test.
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8654
needed for hydrogen mixing and the reason could be that
hydrogen has a high diffusion coefcient as compared to
oxygen, therefore hydrogen diffuses quicker than oxygen.
Also, hydrogen is the lightest molecule and the same
momentum can give a high velocity to a light molecule as
compared to heavy molecule like oxygen. Therefore mixing by
convection/diffusion for hydrogen is quicker than oxygen.
Fig. 8 represents water vapor concentration due to the
electrochemical reaction (according to equation (1)), water is
produced at the anode catalyst layer and this effect is clearly
Fig. 4 Velocity eld in gas-chamber.
Table 5 Computational domain and governing equations.
Domain Equations solved
U
GC,h
V$(ru) 0
ru$Vu Vp mV
2
u
1
3
mVV$u
V$(ruY
i
) V$j
i
V$(rC
p
uT ) V$(kVT )
U
a
V$(ru) 0
m
k
u Vp mV
2
u
1
3
mVV$u
V$(ruY
i
) V$j
i
V$(s
s
V4
s
) 0
V$((rC
p
)
eff
uT ) V$(k
eff
VT )
U
c
V$(ru) 0
m
k
u Vp mV
2
u
1
3
mVV$u
V$(ruY
i
) V$j
i
V$(s
s
V4
s
) 0
V$((rC
p
)
eff
uT ) V$(k
eff
VT )
U
a,cl
V$(ru) S
m
, S
m
S
H2
S
H2O

MH
2
2F
i
a

MH
2
O
2F
i
a
m
k
u Vp mV
2
u
1
3
mVV$u
V$(ruY
i
) V$j
i
S
i
, S
i
S
H2

MH
2
2F
i
a
, S
H2O

MH
2
O
2F
i
a
V$s
eff
s
Vf
s
A
s
S
f
s
V$s
eff
e
Vf
e
A
s
S
f
e
V$((rC
p
)
eff
uT ) V$(k
eff
VT ) S
e
, S
e
q
reva
q
irra
q
ohmea
U
c,cl
V$(ru) S
m
, S
m
S
O2

MO
2
4F
i
c
m
k
u Vp mV
2
u
1
3
mVV$u
V$(ruY
i
) V$j
i
S
i
, S
i
S
O2

MO
2
4F
i
c
V$s
eff
s
Vf
s
A
s
S
f
s
V$s
eff
e
Vf
e
A
s
S
f
e
V$((rC
p
)
eff
uT ) V$(k
eff
VT ) S
e
, S
e
q
revc
q
irrc
q
ohmec
U
e
V$(s
e
V4
e
) 0
V$(k
s
VT ) S
e
0, S
e
s
e
V4
e
$V4
e
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8655
observed near the anode electrode surface. The by-product
water mixes with hydrogen and oxygen and thus dilutes the
gaseous mixture downstream.
In Fig. 9, the legend description corresponding to Figs. 10
17 is shown. As can be seen the variables plotted in Figs. 1017
are along the channel length with varying values of the y-
coordinate. The z-coordinate is kept xed in the center of the
PEN in all these gures.
Fig. 10 shows the velocity eldplot inthe gas-chamber. Blue
and black curves are almost on top of each other showing the
same trends in velocity prole, these lines are plotted at
a distance of 2 mm below and above the cell in y-direction.
Initially the velocity is constant until it reaches the cell. Then,
due to diversionof the ow(due to the presence of the cell) the
velocity starts increasing (up to approximately 0.24 ms
1
) and
then settles behind the cell at a value of approximately
0.0996 ms
1
. Lavender and light green curves show the
velocity plot inside the cell at a distance of 3 mm from top and
bottom edge of the cell. Due to be on same distances they are
also on top of each other and show that the velocity remains
constant until it reaches the cell. Inclose vicinityof the cell, the
ow feels the presence of the obstacle (the cell), therefore the
Fig. 5 Pressure drop in gas-chamber.
Fig. 6 Hydrogen concentration in gas-chamber.
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8656
velocity decreases and becomes almost zero in the PEN. This
suggests that theowinthePENis nomoreconvective but only
diffusive. Just behind the cell the ow again accelerates and
attains a velocity of approximately 0.18 ms
1
. The red curve
shows the center line velocity andas canbe seenfromthe plot,
the center line velocityhas the maximumvalue just before and
after the cell. The center line velocity also feels the effect of the
obstacle and therefore drops to a value of 0.025 ms
1
at the
anode electrodesurface andthendue tothe presenceof hole in
the center of the PEN, the ow sees a free ow (accelerates
again). Due to the radial convection/diffusion in the hole
domain the velocity uctuates for a while and then the ow
comes out of theholewithagreater (nozzlelike) push. Finallyit
regains its maximum center line velocity at a distance of
approximately 0.03 m from the cathode electrode.
Fig. 11 shows the hydrogen mass fraction plot and it can be
seen that the mass fraction remains constant until the reac-
tants reach the cell, where the electrochemical reaction is
taking place. A mass fraction drop (though very little due to
high degree of dilution in the mixture) is observed in the cell
(light green, lavender and red curves), precisely at the anode
catalyst layer due to the consumption of hydrogen. Also, the
Fig. 7 Oxygen concentration in gas-chamber.
Fig. 8 Water vapor concentration in gas-chamber.
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8657
hole is not a completely non-reactive domain and there is
some consumption along radial direction in the hole. Only the
electrolyte is impermeable and; although it is possibly hard to
see in the plot the light green and lavender curves show
a discontinuity of mass fraction in the electrolyte.
Fig. 12 shows the oxygen mass fraction in the gas-chamber
and as discussed before for hydrogen consumption, the same
trend is seen for oxygen consumption but this time in the
cathode catalyst layer. Also, a discontinuity of the curves
(light green and especially of lavender) is clearly seen at the
electrolyte due to its impermeable nature.
Fig. 13 shows the water vapor mass fraction and it can be
seen that due to the water generated at the anode catalyst
layer, a sharp increase in the mass fraction is observed.
Clearly, light green and lavender curves show a discontinuity
as discussed before.
Figs. 1417 are the repetition of simulations with an
increased value of porosity (e 0.4). In contrast to Fig. 10, there
is some convective ow(Fig. 14) in the cell element due to ease
of the Darcy ow through the porous media. Increase in
porosity reduces the resistance in the uid ow and thus both
convective/diffusive uxes exist in the porous electrodes.
Although, it looks like that the increase in porosity will have
a better effect on hydrodynamics of the ow but Figs. 15 and
16 show a less utilization of reactants (compared with Figs. 11
and 12). This effect suggests that increase in porosity will
reduce the available active area for the electrochemical reac-
tion, possibly due to more open pores rather than connected.
Fig. 17 is a plot of the water vapor distribution and shows
a reduced value of water produced as compared with Fig. 13.
Therefore increasing the porosity could have an effect on
electrochemical performance of the cell.
Fig. 18showstemperatureprole(for valueslistedinTable4).
As can be seen, the anode has a higher temperature than
cathode and the reason is reversible heat generation in the
anodecatalyst layer duetoelectrochemical productionof water.
Althoughtheincreaseintemperatureisnot marginal, thisisdue
to the fact that the only heat source is due to inherent reversible
and irreversible effects. Furthermore, the full combustion of
hydrogen is neglected by assuming ideally selective electrodes
Fig. 9 Description of the legend used in Figs. 1017.
Fig. 10 Velocity along the channel length at different y-
coordinates for e [0.3. For interpretation of the references
to colour in this gure legend, the reader is referred to the
web version of this article.
Fig. 11 Hydrogen mass fraction along the channel length
at different y-coordinates for e [0.3.
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8658
and diluted gas mixture. It should be noted that the real elec-
trodes are not perfectly selective, and, one could observe full
combustion of hydrogen near the cell inlet if pure hydrogen/
oxygenmixture is used. It is therefore stronglyrecommendedto
never use pure hydrogen/oxygenmixtures inanSC-SOFCdue to
the above mentioned reason.
Figs. 19 and 20 show the calculated IV and IP character-
istics curves at 500

C for hydrogen/oxygen/nitrogen mixture
(with input parameters given in Table 4), respectively. As can
be seen the maximum power density obtained is
218.5 Wm
2
(i.e. 21.85 mWcm
2
). The observed low current is
due to the low operating temperature and dilution effect.
Furthermore, the power density is quite low as compared to
the dual-chamber conguration, this is because of extremely
low fuel utilization due to the presence of large amount of
inert gas in the mixture.
5. Safety issues
The safe operation is the prime issue for any device, not only
limited to SC-SOFCs. However, in case of SC-SOFCs the level of
safety is of much more concern because of the explosive
nature of fuel/oxidant mixtures. At this stage, it would be
appropriate to consider most commonly used fuels such as
hydrogen or hydrocarbons because of their well-known
application history in SOFCs. For instance, hydrogen/oxygen
mixtures could be the right candidates in terms of being less
problematic for commonly used nickel anodes (no coking
Fig. 12 Oxygen mass fraction along the channel length at
different y-coordinates for e [0.3.
Fig. 13 Water mass fraction along the channel length at
different y-coordinates for e [0.3.
Fig. 14 Velocity along the channel length at different y-
coordinates for e [0.4.
Fig. 15 Hydrogen mass fraction along the channel length
at different y-coordinates for e [0.4.
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8659
problem), however their safety concerns are exceptionally
very high. Hydrogen is considered as explosive in the am-
mability range of 475% by volume in air at standard condi-
tions (ambient pressure and temperature) and this limit
widens as the operating temperature increases. Moreover, the
auto-ignition temperature of a combustible hydrogen/oxygen
mixture is 585

C, far too low than a typical operating condi-
tion for an SOFC. For hydrogen/oxygen mixtures to be prac-
tical for SC-SOFCs, their safety window of operation is quite
narrow, for example either below 2% or above 93% hydrogen
(by volume in air) may be considered as safe at temperatures
up to 200300

C. Unfortunately, there is no practical
study available for hydrogen/oxygen mixture at temperatures
necessary for SC-SOFC operation. Even though, if we assume
that the mixture is safe at higher temperatures, it is still
impractical due to: 1) the use of rich fuels (>93% hydrogen)
may result in extremely low fuel utilization, because all the
fuel passed through the cathode will be nearly unutilized; 2)
the use of lean fuels (<2% hydrogen) will result in fuel de-
ciency and re-oxidation at the anode. Another problem in
working with rich fuels could be the fuel concentration may
fall into the explosive range because of utilization. Open tube
exhaust ow could lead into back diffusion problems of
ambient air, thus disturbing the operating conditions. Another
commonly used term quenching distance, dened by the
diameter of the gas-chamber tube, should be kept as low as
Fig. 16 Oxygen mass fraction along the channel length at
different y-coordinates for e [0.4.
Fig. 17 Water mass fraction along the channel length at
different y-coordinates for e [0.4.
Fig. 18 Temperature distribution in the gas-chamber.
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8660
possible in order to extinguish the ame in case of re.
Flashback arrestors and back ow preventing valves will
increase the systems safety level.
We must point out here that to our knowledge, only Isaiah
et al. [30] have reported diluted mixed gas (argon/hydrogen/
oxygen mixture) behavior for dual-chamber SOFC, thus simu-
lating the single-chamber conditions on individual electrodes.
The heat release (by direct combustion) in a concentrated fuel/
oxidant gas mixture will be much more higher than a diluted
fuel/oxidant mixture. Therefore, one should never use
concentrated hydrogen/oxygen mixture to be employed in an
SC-SOFC, as the presence of nickel on the anode side could
promote full combustion directly at the anode inlet causing
severe local overheating. This effect is observed by many
researchers when using hydrocarbon/air mixtures. Our
experimental studies [2,31] on methane/air mixture operated
SC-SOFC show a temperature rise of as high as 93

C, which
scales-up with ow rate and temperature. If hydrogen would
be used instead of methane, then for the same power output,
atleast 34 times higher ow rate of hydrogen will be required
(this caneasily be calculated onthe basis onelectronproduced
per electrochemical reaction in each case). Now by knowing
the fact that the hydrogenhas approx. 2.3 times higher specic
energy (kJ g
1
) than methane, the local heat produced by full
combustion of hydrogen could be much higher than the
methane/air mixtures. Therefore, it is not recommendedtouse
concentrated hydrogen/oxygen mixtures in an SC-SOFC for
safety reasons, also, neither they are practical in terms of
power production capacity, nor offering an acceptable elec-
trical efciency. Furthermore, 96% inert gas diluted hydrogen/
oxygenmixturecouldbe safebut still theyare not practical due
to very low OCV and power production capacity mainly
because of strong dilution effect.
6. Conclusions
A three-dimensional numerical model of a single-chamber
solid oxide fuel cell was developed. The model accounts for all
transport phenomena and cell potential. Results show that
varying one parameter (associated with the hydrodynamics)
can affect another parameter (associated with the electro-
chemical performance). Increase of porosity in the catalyst
layers can reduce the available active area for electrochemical
reaction, conversely, ne pores can cause a large pressure
drop and subsequently lead to a hydrodynamic problem. The
best design could be to stay with ne porosity of the electrodes
and optimize the hydrodynamic problem by reducing the
tortuous path of the ow inside the porous electrodes. This
could be done by varying the permeability of the porous
electrodes with a xed ne porosity. Instead of using dense
solid (non-porous) electrolyte, use of a fully porous cell
(including the porosity of the electrolyte) can give ease in ow
with reduced manufacturing cost. Another idea could be
instead of having holes inside the PEN element, the holes can
be provided at the circumference/edges of the interconnect
(or separator layer). This would not reduce the active area but
still provides better hydrodynamics. Distributed feed to
a number of cells placed in a stack can give a uniformreactant
utilization in each cell, therefore parallel feed with branches
at each cell could give a better performance both hydrody-
namically and electrochemically.
Although, experimental workonSC-SOFCs usinghydrogen/
oxygen mixtures (at temperatures necessary for SOFC opera-
tion) is never reported, this study would help the reader to
understand what are the major restrictions in doing so.
Acknowledgments
The authors would like to thank E.ON-UK for funding Mr.
Naveed Akhtar through Dorothy Hodgkin Postgraduate Award
(DHPA) scheme.
Fig. 19 IV characteristic curve for SC-SOFC. Fig. 20 IP characteristic curve for SC-SOFC.
Nomenclature
n normal vector
u velocity vector
p pressure
Y
i
mass fraction of the ith species
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8661
r e f e r e n c e s
[1] Singhal SC, Kendall K. High temperature solid oxide fuel
cells: fundamentals, design and applications. Kidlington
Oxford: Elsevier; 2003. pp. 224.
[2] Akhtar N, Decent SP, Loghin D, Kendall K. Mixed-reactant,
micro-tubular solid oxide fuel cells: an experimental study.
J Power Sources 2009;193:3948.
[3] Hao Y, Goodwin DG. Numerical modeling of single-chamber
SOFCs with hydrocarbon fuels. J Electrochem Soc 2007;154:
B20717.
[4] Hao Y, Shao Z, Mederos J, Lai W, Goodwin DG, Haile SM.
Recent advances in single-chamber fuel-cells: experiment
and modeling. Solid State Ionics 2006;177:201321.
[5] Chung CY, Chung YC. Performance characteristics of micro
single-chamber solid oxide fuel cell: computational analysis.
J Power Sources 2006;154:3541.
[6] Hao Y, Goodwin DG. Efciency and fuel utilization of
methane-powered single-chamber solid oxide fuel cells.
J Power Sources 2008;183:15763.
X
i
mole fraction of the ith species
N total number of species in the mixture
D
ij
binary diffusion coefcient for pair ij
j
i
mass diffusion ux of the ith species
x, y, z Cartesian coordinates in x, y and z direction,
respectively
M average (mixture) molecular weight
M
i
molecular weight of the ith species
R universal gas constant
T temperature
c
i
concentration of the ith species
B
1
matrix function of inverted binary diffusion
coefcients
B
ii
diagonal elements of inverted binary diffusion
coefcient matrix
B
ij
non-diagonal elements of inverted binary diffusion
coefcient matrix
D matrix of Fick diffusion coefcients
D
ij
MaxwellStefan diffusivity for pair ij
V
i
molecular diffusion volume
H
2
hydrogen
O
2
oxygen
H
2
O water
N
2
nitrogen
e
1
electron
O
2
oxygen ion
S
m
mass source term
S
i
species source term
D
eff
DG
Fick effective dusty gas diffusivity matrix
D
eff
DG
i;j
Fick effective dusty gas diffusivity for pair ij
D
DG
i;j
MaxwellStefan dusty gas diffusivity for pair ij
D
eff
DG
i;j
MaxwellStefan effective dusty gas diffusivity for
pair ij
D
k;i
Knudsen diffusivity
B
eff
DG
ii
diagonal elements of inverted binary effective dusty
diffusion coefcient matrix
B
eff
DG
ij
non-diagonal elements of inverted binary effective
dusty diffusion coefcient matrix
d
p
average pore diameter
A
s
electrochemically active surface area of the medium
per unit volume
S
4
current source term
i
a
anodic current density
i
c
cathodic current density
i
0,a
anodic exchange current density
i
0,c
cathodic exchange current density
V
c
cell voltage
V
oc
open circuit (Nernst) voltage
V
o
ideal (standard) voltage
z number of electrons participating per
electrochemical reaction
F Faradays constant
exp exponent
k
s
thermal conductivity of solid (i.e. cell components)
C
p,s
specic heat of solid (i.e. cell components)
k
eff
effective thermal conductivity of solid and gas phase
r
eff
effective density of solid and gas phase
C
p,eff
effective specic heat of solid and gas phase
k thermal conductivity of gas
C
p
specic heat of gas
q heat generation
ds change in entropy generation
E activation energy
Greek letters
V differential operator
V
2
Laplace operator
f
s
solid phase (electronic) potential
f
e
electrolyte phase (ionic) potential
m dynamic viscosity
r average (mixture) gas density
r
i
density of the ith species
r
s
density of solid (i.e. cell components)
G thermodynamics matrix
k permeability
e porosity
s tortuosity
p constant (3.14159)
s conductivity
a charge transfer coefcient
g pre-exponential coefcient
h overpotential
Subscripts
GC gas-chamber
h hole
a anode
e electrolyte
c cathode
cl catalyst layer
rev reversible
irr irreversible
ohm ohmic
act activation
eff effective
s solid phase (electronic)
ref reference
Superscripts
a anodic
c cathodic
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8662
[7] Hao Y, Goodwin DG. Numerical study of heterogeneous
reactions in an SOFC anode with oxygen addition.
J Electrochem Soc 2008;155:B66674.
[8] Chung CY, Chung YC, Kim J, Lee J, Lee HW. Numerical
modeling of micro single-chamber ceria-based SOFC.
J Electroceram 2006;17:95964.
[9] Akhtar N, Decent SP, Loghin D, Kendall K. Modelling of co-
planar type single-chamber solid oxide fuel cells (SC-SOFCs).
In: ASME 6th international fuel cell science, engineering and
technology conference, Denver, Colorado, U.S.A.; 1618 June
2008.
[10] Akhtar N, Decent SP, Loghin D, Kendall K. An isothermal
numerical model of single-chamber solid oxide fuel cell. In:
8th European SOFC forum, Lucerne, Switzerland; 30 June4
July 2008.
[11] Kumar RK. Flammability limits of hydrogenoxygen diluent
mixtures. J Fire Sciences 1985;3:24562.
[12] Todd B, Young JB. Thermodynamic and transport properties
of gases for use in solid oxide fuel cell modelling. J Power
Sources 2002;110:186200.
[13] Andreassi L, Rubeo G, Ubertini S, Lunghi P, Bove R.
Experimental and numerical analysis of a radial ow solid
oxide fuel cell. Int J Hydrogen Energy 2007;32:455974.
[14] Bessler WG, Gewies S. Gas concentration impedance of solid
oxide fuel cell anodes II. Channel geometry. J Electrochem
Soc 2007;154:B54859.
[15] Berning T, Lu DM, Djilali N. Three dimensional
computational analysis of transport phenomena in a PEM
fuel cell. J Power Sources 2002;106:28494.
[16] Sivertsen BR, Djilali N. CFD-based modelling of proton
exchange membrane fuel cells. J Power Sources 2005;141:
6578.
[17] Taylor R, Krishna R. Multi-component mass transfer. John
Wiley and Sons; 1993.
[18] Perry R, Green D. Perrys chemical engineering handbook. 6th
ed. McGraw-Hill; 1983.
[19] Sissom LE, Pitts DR. Elements of transport phenomena.
International student edition. McGraw-Hill; 1972.
[20] Coulson JM, Richardson JF. Fluid ow, heat transfer and mass
transfer. 5thed., vol. 1. Oxford: Butterworth-Heinemann; 1996.
[21] Morales JCR, Vazquez JC, Lopez DM, Irvine JTS, Nunez P.
Improvement of the electrochemical properties of novel
solid oxide fuel cell anodes LSCM-O and LSTMG-O, using
CuYSZ-based cermets. Electrochim Acta 2007;52:721725.
[22] Cheddie DF, Munroe NDH. Three dimensional modeling of
high temperature PEM fuel cells. J Power Sources 2006;160:
21523.
[23] Cui D, Chung YC, Liu L, Dong Y, Cheng M. Comparison of
different current collecting modes of anode supported
micro-tubular SOFC through mathematical modeling.
J Power Sources 2007;174:24654.
[24] Janardhanan VM, Deutschmann O. Numerical study of mass
and heat transport in solid-oxide fuel cells running on
humidied methane. Chem Eng Sci 2007;6:547386.
[25] Cheddie DF, Munroe NDH. A dynamic 1D model of a solid
oxide fuel cell for real time simulation. J Power Sources 2007;
171:63443.
[26] Serincan MF, Pasaogullari U, Sammes NM. Effects of
operating conditions on the performance of a micro-tubular
solid oxide fuel cell (SOFC). J Electrochem Soc 2008;155:
B111727.
[27] Gazzarri JI, Kesler O. Non-destructive delamination detection
in solid oxide fuel cells. J Power Sources 2007;167:43041.
[28] Guvelioglu GH, Stenger HG. Computational uid dynamics
modeling of polymer electrolyte membrane fuel cells.
J Power Sources 2005;147:95106.
[29] Liu S, Song C, Lin Z. The effects of the interconnect rib
contact resistance on the performance of planar solid oxide
fuel cell stack and the rib design optimization. J Power
Sources 2008;183:21425.
[30] Kellogg ID, Koylu UO, Petrovsky V, Dogan F. Effectiveness of
anode in a solid oxide fuel cell with hydrogen/oxygen mixed
gases. Int J Hydrogen Energy 2009;34:513843.
[31] Akhtar N, Decent SP, Kendall K. Cell temperature
measurements in micro-tubular, single-chamber, solid oxide
fuel cells (MT-SC-SOFCs). J Power Sources doi:10.1016/j.
jpowsour.2009.04.078.
[32] Shi Y, Cai N, Li C. Numerical modeling of an anode-
supported SOFC button cell considering anodic surface
diffusion. J Power Sources 2007;164:63948.
[33] Janardhanan VM, Deutschmann O. CFD analysis of a solid
oxide fuel cell with internal reforming: coupled interactions
of transport, heterogeneous catalysis and electrochemical
processes. J Power Sources 2006;162:1192202.
i nt e r na t i o na l j o ur na l o f hy d r og e n e ne r gy 3 4 ( 2 0 0 9 ) 8 6 4 5 8 6 6 3 8663