A study of direct-conversion SOFC with n-butane at higher fuel utilization

Article abstract of DOI:10.1149/1.1574807

The performance of solid-oxide fuel cells (SOFCs) with Cu-ceria yttria-stabilized zirconia anodes in n-butane at 973 K has been studied as a function of fuel conversion. In order to simulate the local anode environment at high fuel utilization, n-butane was oxidized with O₂ in a separate reactor before sending the fuel to a small 0.45 cm² cell. When n-butane oxidation was carried out over a ceria catalyst at 973 K, the main oxidation products were CO₂ and H₂O. By comparing the performance of the cell in the partially oxidized fuel to the performance of the cell in n-butane diluted in He, it was demonstrated that CO₂ and H₂O have only minimal effect on cell performance other than to dilute the fuel. This dilution of the fuel results in a significant decrease in performance at higher fuel conversions. When n-butane oxidation was carried out over a Pd-ceria catalyst, significant amounts of H₂ were generated by steam reforming of hydrocarbons with the steam generated by hydrocarbon oxidation. As a result, the maximum power density generated by the cell increased with fuel conversions up to 30% and remained very high at fuel conversions up to at least 70%. Based on these results, it is suggested that the inclusion of a steam-reforming catalyst within the anode compartment of direct-conversion SOFC should improve their performance at high fuel utilization.

Full text of DOI:10.1149/1.1574807

A858
Journal of The Electrochemical Society, 150 ͑7͒ A858-A863 ͑2003͒
0013-4651/2003/150͑7͒/A858/6/$7.00 © The Electrochemical Society, Inc.
A Study of Direct-Conversion SOFC with n-Butane at Higher
Fuel Utilization
,z
*
*
O. Costa-Nunes, J. M. Vohs, and R. J. Gorte
Department of Chemical and Biomolecular Engineering, University of Pennsylvania, Philadelphia,
Pennsylvania 19104, USA
The performance of solid-oxide fuel cells ͑SOFCs͒ with Cu-ceria yttria-stabilized zirconia anodes in n-butane at 973 K has been
studied as a function of fuel conversion. In order to simulate the local anode environment at high fuel utilization, n-butane was
oxidized with O2 in a separate reactor before sending the fuel to a small 0.45 cm2 cell. When n-butane oxidation was carried out
over a ceria catalyst at 973 K, the main oxidation products were CO2 and H2O. By comparing the performance of the cell in the
partially oxidized fuel to the performance of the cell in n-butane diluted in He, it was demonstrated that CO2 and H2O have only
minimal effect on cell performance other than to dilute the fuel. This dilution of the fuel results in a significant decrease in
performance at higher fuel conversions. When n-butane oxidation was carried out over a Pd-ceria catalyst, significant amounts of
H2 were generated by steam reforming of hydrocarbons with the steam generated by hydrocarbon oxidation. As a result, the
maximum power density generated by the cell increased with fuel conversions up to 30% and remained very high at fuel
conversions up to at least 70%. Based on these results, it is suggested that the inclusion of a steam-reforming catalyst within the
anode compartment of direct-conversion SOFC should improve their performance at high fuel utilization.
© 2003 The Electrochemical Society. ͓DOI: 10.1149/1.1574807͔ All rights reserved.
Manuscript submitted November 4, 2002; revised manuscript received January 7, 2003. Available electronically May 12, 2003.
Work in our laboratory has recently demonstrated the direct elec-
carbons in our application; however, it has been demonstrated that
ceria-supported precious metals, especially Pd/ceria and Pt/ceria
catalysts, are very active for steam reforming of higher hydrocar-
bons, including aromatics, and are much more stable toward coking
than conventional Ni catalysts.11,12 Furthermore, even when
precious-metal catalysts deactivate at low H2O:C ratios, the deacti-
vation is not catastrophic as in the case of Ni catalysts. One does not
observe large carbon deposits on the precious-metal catalysts and
they are easily regenerated in the presence of steam.
In the work described in this paper, we examined the likely effect
on cell performance that higher fuel utilizations would have by feed-
ing a partially oxidized fuel to a model SOFC. To produce the prod-
ucts that would be obtained at higher conversions in direct conver-
sion of hydrocarbons on an anode composed of Cu, ceria, and yttria-
stabilized zirconia ͑YSZ͒, we preoxidized n-butane with varying
amounts of O2 on a ceria catalyst in a tubular reactor at 973 K
before feeding the products to the SOFC. To determine the effect of
having a reforming catalyst in the anode, we preoxidized the fuel
over a Pd/ceria catalyst. The results demonstrate that the addition of
a reforming catalyst to the anode compartment could significantly
change the composition of the fuel and should be effective for im-
proving performance at high fuel utilizations.
trochemical conversion of a variety of hydrocarbon fuels in solid-
oxide fuel cells ͑SOFCs͒ using Cu-based anodes.1-4 Direct conver-
sion, without the addition of steam or air to internally reform the
fuel, is made possible by the fact that Cu, unlike Ni, does not cata-
lyze the formation of carbon in the presence of dry hydrocarbons at
973 K.5,6 Fuel cells that can produce electricity by direct electro-
chemical conversion of hydrocarbons could have significant effi-
ciency advantages over conventional systems that require the fuel to
first be reformed, especially when using hydrocarbons larger than
methane. Because steam reforming of higher hydrocarbons requires
the use of high H2O:C ratios to avoid coke formation,7-9 it is usually
necessary to do autothermal or oxy-reforming of these fuels.8,9 In
addition to the loss of fuel efficiency that occurs with oxidation, the
addition of air to the fuel results in dilution of the fuel by N2 . For
example, even ideal oxy-reforming of C4H10 to CO and H2 with air
results in a fuel that has a mole fraction of N2 that is 47%. Since the
ultimate fuel utilization that can be achieved is limited by the con-
centration of fuel remaining at the stack exit, this N2 dilution lowers
overall cell efficiency.
However, fuel dilution can also be a problem in direct-
conversion fuel cells because each fuel molecule produces a larger
number of product molecules. For example, in the case of butane, 9
mol of CO2 or H2O are formed for each mole of butane that reacts.
These product molecules dilute the fuel so that the concentration of
butane is only 10% at 50% fuel conversion and only 3.6% at 75%
fuel conversion. At low fuel concentrations, gas-phase diffusion will
likely limit cell performance, even though each fuel molecule pro-
duces 26 electrons. One solution to the dilution problem is to use the
steam produced by oxidation of the hydrocarbon at the electrolyte
interface to reform the remaining hydrocarbon molecules. The re-
forming process increases the fuel concentration by consuming H2O
and producing H2 . Unfortunately, the same catalytic properties that
make Cu inert to carbon formation in dry hydrocarbons make it a
very poor steam-reforming catalyst, so one should not expect high
reforming activity from the anode itself.
Experimental
The detailed procedures used to prepare the fuel cells in this
study, as well as a detailed description of the structure of the cells
used in these studies, have been reported elsewhere.5 Briefly, a
yttria-stabilized zirconia ͑YSZ͒ wafer, with one side porous and the
other dense, was prepared by tape-casting methods, with graphite
and poly-methyl methacrylate ͑PMMA͒ pore formers used to intro-
duce porosity. The thickness of the dense electrolyte layer was 60
␮m and this was supported by the 600 ␮m porous layer. The cathode
material, a 50 wt % YSZ and LSM (La0.8Sr0.2MnO3 , Praxair Sur-
face Technologies͒ mixture, was applied as a paste onto the dense
side of the wafer, then calcined to 1400 K. To prepare the anode, the
porous YSZ layer was impregnated with aqueous solutions of
Ce͑NO3)3•6H2O and Cu͑NO3)2•3H2O and calcined to 723 K to
form the oxides. The final composition of the anodes was 20 wt %
Cu and 10 wt % ceria. Electrical contacts to the electrodes were
formed using Pt mesh and Pt paste at the cathode and a Au mesh and
Au paste at the anode. Each cell, having a cathode area of 0.45 cm2,
was sealed onto 1.0 cm alumina tubes using Au paste and a zirconia-
based adhesive ͑Aremco, Ultra-Temp 516͒. The entire cell was then
placed inside a furnace and heated to 973 K at 2 K/min in flowing
H2 . Before taking any electrochemical data, the cell was condi-
In fuel-cell stacks that require internal reforming of methane, it is
common practice to include a reforming catalyst within the stack,
separate from the anode material itself.10 Because large H2O:C ra-
tios are required to maintain stability of the catalyst to coking, it is
not practical to use conventional Ni-based catalysts for larger hydro-
* Electrochemical Society Active Member.
z E-mail: gorte@seas.upenn.edu

Journal of The Electrochemical Society, 150 ͑7͒ A858-A863 ͑2003͒
A859
tioned by exposing the anode to flowing n-butane for 30 min at 973
K.13 No evidence for decreased performance or deactivation was
observed under any conditions used in the measurements performed
in this study.
Even at the maximum current in our SOFC, the conversion of
n-butane was less than 1% for a flow rate of 20 mL/min, making it
impossible for us to study the effect of fuel utilization by changing
the flow rate of the fuel. Therefore, the pure n-butane fuel was
partially burned before sending it into the fuel cell. Preoxidation was
carried out at 973 K in a quartz tubular reactor, 0.6 mm diam. The
partial pressures of the reactants were established by varying the O2
flow rate, and maintaining a fixed flow rate for n-butane of 20 mL/
min ͑at room temperature͒ and a total pressure of 1 atm. In this
paper, the rate of O2 fed to the reactor is reported as the fraction
needed for complete combustion of the n-butane according to the
reaction C4H10 ϩ 6.5O2 → 4CO2 ϩ 5H2O. To avoid explosions,
the O2 was fed to the reactor at three different points along the
catalyst bed, with 100 mg of catalyst placed after each entry point.
This amount of catalyst was sufficient to completely consume O2 in
each stage. Finally, all tubing between the oxidation reactor and the
fuel cell were maintained at approximately 400 K in order to avoid
condensation of products.
Figure 1. Cell potential and current densities for a cell operating at 973 K in
͑᭿,ᮀ͒ H2 and ͑᭹,᭺͒ n-butane.
The outlet of the oxidation reactor could be analyzed using an
on-line gas chromatograph equipped with a thermal-conductivity de-
tector and Hayesep Q packed column. To avoid problems with water
in the GC, the combustion products were passed through a trap held
at 273 K before sending them into the GC, so that the composition
was obtained in a dry basis.
since a number of hydrocarbon products were detected in the efflu-
ent, primarily from C-2 to C-7 hydrocarbons, although most of the
n-butane did remain intact.
To simulate normal conversion of n-butane in the SOFC, the
catalyst used in the oxidation reactor was a ceria powder prepared
by decomposition of Ce͑NO3)3•6H2O in air at 873 K so that it
would be similar to the ceria in the SOFC anode. ͑A few experi-
ments were also performed using 10 wt % Cu on ceria to simulate
the composition of the SOFC anode, but the results were similar to
those obtained for pure ceria, demonstrating again that Cu is not an
effective reforming catalyst.͒ To simulate the performance that
would be obtained when a reforming catalyst is present in the anode
compartment, the ceria catalyst was replaced with a 1 wt % Pd-ceria
catalyst in which the Pd was added to the ceria by wet impregnation
using an aqueous solution of Pd͑NH3)4(NO3)2 ͑Aldrich, 99.99%͒,
followed by drying at 383 K and calcination at 873 K.11,12
In order to study the effect of fuel conversion on performance, as
shown in Fig. 2, it was necessary to partially oxidize the fuel before
sending it to the anode. In the first set of experiments, the catalyst in
the oxidation reactor was ceria, chosen because ceria is also added
to the anode where it appears to act in part as a catalyst.13,14 The fuel
conversions were calculated as the amount of O2 added to the oxi-
dation reactor as a fraction of the amount of O2 that would be
required for complete combustion of the fuel. Figure 2 shows the
cell potential and power density as a function of the current density
for different fuel conversions, and Table I shows the mole fractions
that would be expected in the fuel at each conversion if combustion
were the only reaction to take place.
The data indicate that there is a slight drop in the OCV, from 1.04
to 0.9 V, as the fuel conversion goes from 0 to 70%. This cannot be
explained on the basis of a change in the pO , because the pO
Results
2
2
The performance characteristics of one of the cells used in this
study are shown in Fig. 1 when either pure H2 or pure n-butane are
fed to the anode at 973 K. With H2 as fuel, the open-circuit voltage
͑OCV͒ was 1.12 V and the maximum power density was 0.22
W/cm2. The OCV is in reasonable agreement the expected Nernst
potential, demonstrating that there are no significant leaks in the
cell. With n-butane as fuel, the OCV was 1.04 V and the maximum
power density was 0.13 W/cm2. These values are similar to what has
been reported previously for hydrocarbons in similar cells.13,14 The
fact that the OCV is less than the Nernst standard potential of 1.12 V
has been explained as being due to equilibrium with partially oxi-
dized species on the anode surface.13,14 Indeed, an earlier report
from our laboratory has indicated that a small quantity of carbon-
aceous material is formed in the anode at steady state during expo-
sure to pure n-butane and that the OCV in n-butane changes with
cell pretreatment, decreasing in value when the cell is held at OCV
for prolonged periods and increasing when the cell is shorted.13 The
decreased power density with n-butane compared to H2 must be due
to reaction limitations with n-butane.
values only change from 2.7 ϫ 10Ϫ25 atm ͑at 10% conversion͒ to
9.51 ϫ 10Ϫ25 atm ͑at 70% conversion͒. The lower OCV observed at
higher fuel conversion is probably due to a change in the nature of
the carbonaceous species present at the anode surface.13 Figure 2
also indicates that the maximum power density decreases continu-
ously with fuel conversion, going from 0.13 W/cm2 at 0% conver-
sion to 0.08 W/cm2 at 70% conversion. We demonstrate that the
decrease in performance is primarily due to dilution of the fuel in
H2O and CO2 .
One way to avoid the effect of fuel dilution is to reform the
remaining hydrocarbon using the steam generated in the cell. To
simulate the effect of adding a reforming catalyst to the stack, we
replaced the ceria catalyst with a Pd-ceria catalyst ͑1 wt % Pd͒ in the
oxidation reactor. Pd-ceria is equally effective as ceria in oxidizing
the n-butane; however, Pd-ceria is also a very active steam-
reforming catalyst.11,12 Figure 3 shows the performance curves as a
function of fuel conversion with the Pd-ceria oxidation catalyst, re-
sults analogous to Fig. 2. The major difference in the results when
using the Pd-ceria oxidation catalyst is that the power densities for
10% and 30% fuel conversions are actually higher than we observe
for pure n-butane. It is also noteworthy that the Pd-ceria catalyst was
entirely stable during these experiments. Even after passing pure
n-butane over the Pd-ceria catalyst for more than 1 h, the catalyst
exhibited high activity as soon as even relatively small amounts of
O2 were added to the fuel.
As discussed in the Experimental section, the calculated conver-
sion of n-butane at the maximum current density of 550 mA/cm2
was less than 1%. Indeed, GC analysis of products leaving the anode
compartment was unable to detect oxidation products in significant
amounts at the flow rates we used in this study. GC analysis did
demonstrate the importance of gas-phase, free-radical chemistry,

A860
Journal of The Electrochemical Society, 150 ͑7͒ A858-A863 ͑2003͒
Figure 2. Performance curves for a cell operating at 973 K in partially
oxidized n-butane at the following calculated conversions: ͑᭿͒ 0, ͑᭡͒ 10,
͑᭹͒ 30, ͑ᮀ͒ 50, and ͑᭝͒ 70%. The oxidation reactor operated with a ceria
catalyst at 973 K.
Figure 3. Performance curves for a cell operating at 973 K in partially
oxidized n-butane at the following calculated conversions: ͑᭿͒ 0, ͑᭡͒ 10,
͑᭹͒ 30, ͑ᮀ͒ 50, and ͑᭝͒ 70%. The oxidation reactor operated with 1%
Pd-ceria catalyst at 973 K.
The reason for the improved performance of the cell when feed-
ing the fuel oxidized by the Pd-ceria catalyst is easily understood
from the composition of the fuel leaving the oxidation reactor,
shown in Fig. 4. The compositions are given on a water-free basis,
with the results from the ceria catalyst shown in Fig. 4a and that for
the Pd-ceria catalyst in Fig. 4b. Some thermal cracking is observed
in the reactor in the absence of added O2 . The unspecified products
are primarily cracking products for n-butane ͑methane, ethane, eth-
ylene, propane, and propylene͒, along with products of higher mo-
lecular weight that must be formed by oligomerization. Considering
the experiments with the ceria catalyst ͑Fig. 4a͒, the main product,
other than free-radical cracking products, is CO2 on a water-free
basis, and the amount of CO2 increases linearly with the amount of
added O2 . While there is an increase in the quantity of other hydro-
carbon products at low feed rates of O2 , this probably has a minimal
effect on cell performance. It is important to notice that only small
amounts of H2 were formed at all fuel conversions.
Table I. Calculated compositions leaving the oxidation reactor
assuming only total combustion.
Molar Fraction of Gases
n-C4H10
By contrast, when the Pd-ceria catalyst was used to oxidize the
fuel ͑Fig. 4b͒, significant amounts of H2 , more than 20% at some
conversions, were formed in the presence of O2 , and the amount of
CO2 formed at low fuel conversions was noticeably higher. These
two observations are explained by the fact that H2O generated by
oxidation of the initial n-butane must be reforming some of the
Utilization ͑%͒
n-C4H10
CO2
H2O
0
10
30
50
70
1
0.5
0.21
0.10
0.05
0
0
0.22
0.35
0.40
0.42
0.28
0.44
0.50
0.53

Journal of The Electrochemical Society, 150 ͑7͒ A858-A863 ͑2003͒
A861
Figure 5. A plot of the maximum power density achieved in the cell at 973
K as a function of the calculated conversion of the fuel entering the anode
compartment. The data are given for oxidation of n-butane by ͑᭿͒ ceria and
͑᭡͒ by 1 wt % Pd-ceria. The maximum power densities obtained when
n-butane was diluted by He to a concentration equivalent to that which
would be obtained by simple oxidation, as shown in Table I, are shown by
͑᭺͒.
Pd-ceria for steam reforming under severe conditions, even with
hydrocarbons that are liquids at room temperature, has been dis-
cussed elsewhere.11,12
In the absence of reforming, it is important to determine whether
CO2 and H2O simply dilute the fuel concentration or are chemically
involved in the oxidation process at the three-phase boundary of the
anode. To determine the effect of fuel dilution in the absence of
chemistry, we tested a cell in n-butane diluted with He. The trends
we observed with fuel dilution were similar to the data in Fig. 2.
While the OCV did not change as much with dilution of the
n-butane with He, the maximum power densities decreased signifi-
cantly. Figure 5 plots the maximum power density of a cell as a
function of fuel conversion using a fuel that has been partially oxi-
dized either over a ceria catalyst or a Pd-ceria catalyst. Also plotted
in Fig. 5 is the maximum power density of a cell with n-butane
diluted in He to give the concentrations in Table I. The maximum
power densities achieved following He dilution are almost identical
to what we observe following oxidation of the fuel over the ceria
catalyst. Therefore, the data in Fig. 5 indicate that the oxidation
products, CO2 and H2O, affect cell performance primarily by dilut-
ing the fuel.
Figure 4. The molar composition of the fuel leaving the oxidation reactor as
a function of the conversion calculated from the O2 :C4H10 ratio fed to the
reactor. Results are shown for ͑A͒ when the catalyst was pure ceria and ͑B͒
when the catalyst was 1 wt % Pd-ceria. The compositions are given on a dry
basis for: ͑᭿͒ n-butane, ͑᭝͒ CO2 , ͑ᮀ͒ H2 , and ͑᭺͒ other hydrocarbon
products.
remaining n-butane, forming additional CO2 along with the H2 ac-
cording to the reaction
Discussion
While traditional SOFCs exhibit a number of favorable charac-
teristics, one unfavorable property is the fact that oxidation products
form at the anode and remain in the gas phase on the anode side of
the membrane, unavoidably leading to dilution of the fuel at high
fuel utilization. When the fuel is syngas produced by steam reform-
ing of methane or auto-thermal reforming of higher hydrocarbons,
there is an additional dilution due to the presence of steam or N2 that
is carried along with the fuel. A more complete discussion of con-
centration polarization at the anode has been given by others.15-17
SOFCs that convert hydrocarbons directly without requiring ad-
ditional steam or air to be added to the fuel could prevent some of
the difficulty associated with dilution. However, when direct conver-
sion is applied to larger hydrocarbons, fuel dilution by the oxidation
products does limit the performance, as the work in this paper has
demonstrated. To understand the role of fuel dilution, it is instructive
to consider the diffusive flux of n-butane within the anode. Assum-
C4H10 ϩ 8H2O ϭ 4CO2 ϩ 13H2
As shown in Fig. 1, cell performance is better in H2 than it is in
n-butane; therefore, the maximum power density increases with fuel
conversion when using the Pd-ceria catalyst for fuel oxidation, due
to the presence of H2 produced by reforming.
Previous work on steam-reforming of n-butane over Pd/ceria has
shown that the catalyst deactivates when the H2O:C ratio is less than
one.11,12 It is therefore somewhat surprising that exposure of the
catalyst to pure n-butane at 973 K did not irreversibly poison it.
Apparently, minimal amounts of carbon form on the catalyst during
deactivation in the presence of hydrocarbons, and this carbon is
quickly removed when the catalyst is exposed to steam or O2 . It is
also noteworthy that the product ratios shown in Fig. 4b were found
to be stable with time for at least several hours. The high stability of

A862
Journal of The Electrochemical Society, 150 ͑7͒ A858-A863 ͑2003͒
ing n-butane is the only fuel and that no reforming occurs, the cur-
rent density, i, can be equated to the difference in concentrations of
n-butane at the anode surface, Csurface , and at the electrolyte-anode
interface, CTPB , by the following equation
would reduce the flexibility one would have in choosing the catalyst
formulation and placement within the anode compartment.
In previous applications where a steam-reforming catalyst has
been added to the anode compartment, steam is added with the fuel.
Since the addition of steam is not desirable, the reforming catalyst in
direct-conversion SOFCs would preferably use steam generated by
the electrochemical reaction. Therefore, catalyst should be placed at
positions within the stack where significant amounts of steam have
been generated, not near the entrance to the stack where fuel con-
versions are low and cell performance is reasonable.
i ϭ
D
C
surface-CTPB͒ / ␦•26•F͒
͓1͔
͓
͑
͔ ͑
C H
10
4
The most crucial property of the reforming catalyst for this ap-
plication would be its stability at high H2O:C ratios. Standard, Ni-
based reforming catalysts are inappropriate because of their ten-
dency to catalyze the production of carbon fibers.19-23 For example,
conventional, Ni-based catalysts tend to form carbon fibers through
a mechanism that involves carbon formation on the Ni surface, car-
bon dissolution into the bulk Ni, and precipitation of graphitic car-
bon from some part of the Ni after the Ni becomes supersaturated in
carbon.20,23 Unless sufficient amounts of steam or oxygen are
present along with the hydrocarbon in order to remove carbon from
the Ni surface at a rate faster than that of carbon dissolution and
precipitation, the catalyst will be destroyed. While thermodynamic
calculations are often used to predict the conditions under which
carbon formation is favored on Ni,24 carbon fibers form for kinetic
reasons, even when carbon is not thermodynamically predicted. This
is particularly true with hydrocarbons larger than methane, because
these molecules are more reactive and therefore decompose rapidly
to form carbon on the Ni surface. For example, in the case of steam-
reforming of butane, a Ni catalyst was found to coke instantly for a
H2O:C ratio of 2,11 conditions which should be thermodynamically
stable. Catalysts based on Co, Fe, and Ru would likely suffer similar
limitations, given their high activities for C-C bond formation, as
demonstrated by their use as Fischer-Tropsch catalysts. Catalysts
based on Pt, Rh, and Pd tend to be much more tolerant toward
carbon formation at the low H2O:C ratios to which the catalyst will
be exposed and show less tendency to deactivate in the same cata-
strophic manner. This has been discussed in some detail in other
publications.11,12
where DC H is the effective diffusivity for n-butane in the porous
4
10
anode, ␦ is the thickness of the anode ͑600 ␮m͒, F is the Faraday
constant, and 26 is the number of electrons produced by oxidation of
each n-butane molecule.
The estimation of DC H requires some assumptions but can
4
10
probably be determined within a factor of approximately five. First,
one can neglect Knudsen diffusion because the characteristic pore
size in the anode, 2 ␮m, is significantly larger than the molecular
mean-free path at 1 atm. Second, based on the Chapman-Enskog
equation,18 the binary diffusion coefficients at 973
K
for
), and
2
n-butane in n-butane (Dn-C H
), CO2(Dn-C H
-CO
4 10
-n-C H
4
10
4
10
H2O͑Dn-C H -H O) vary between 0.5 and 0.9 cm2/s, so that it is
4
10
2
reasonable to assume an average value of 0.7 cm2/s for the diffusiv-
ity. Finally, equating the effective, fractional area for diffusion to the
porosity of the anode, 0.53, and using the suggested value of three
for the tortuosity in porous media,18 DC H must be approximately
4
10
0.1 cm2/s.
Using this value and Eq. 1, (Csurface-CTPB) is approximately 1
ϫ 10Ϫ7 mol/cm3 at a current density of 0.5 A/cm2, implying that
diffusional limitations should become apparent when Csurface ap-
proaches 1 ϫ 10Ϫ7 mol/cm3, because CTPB begins to deviate sig-
nificantly from the composition of the vapors above the anode at this
point. Because the molar density of an ideal gas at 973 K is 1.2
ϫ 10Ϫ5, diffusional limitations are severe when the mole fraction
of fuel is a few percent.
This result provides insight into two additional issues. First, as
shown by the experiments with He dilution of n-butane, the cell
performance has decreased significantly at fuel concentrations well
above a few percent. While it is possible that our calculation has
overestimated the diffusivity of n-butane, it seems more likely that
diffusion alone is not limiting performance at intermediate fuel uti-
lization. Since the electrochemical reaction in the three-phase
boundary ͑TPB͒ of the anode likely depends on the partial pressure
of the fuel at the TPB, the effect of fuel dilution is more complex
than simply to cause concentration gradients.
Second, assuming the diffusivity of H2O and CO2 are at least as
high as that determined for n-butane, the above-described calcula-
tion provides confidence that the simulated studies in which the fuel
is oxidized before passing it to the anode should indeed provide
similar results to that which would be obtained when the fuel is
oxidized on the anode, by the oxygen ions passing through the elec-
trolyte. For a diffusion coefficient equal to 0.1 cm2/s, concentration
gradients for CO2 and H2O are not important, except at fuel conver-
sions near 0% or 100% and for fuel cells that exhibit significantly
higher performance than that which we obtained.
The addition of a steam-reforming catalyst like Pd/ceria to the
anode could significantly improve the performance of direct-
conversion SOFCs at higher fuel utilization. The inclusion of a re-
forming catalyst within the anode compartment is standard practice
for molten-carbonate fuel cells.10 A similar addition of a reforming
catalyst within an SOFC could utilize the steam generated by oxi-
dation of the hydrocarbon at the electrolyte interface to produce a
gas mixture that allows better performance by the SOFC. While one
could add the catalyst directly to the anode, this would probably be
undesirable, because there would be danger of forming an alloy of
the expensive precious metal with the Cu used for currently collect-
ing, and because having the catalyst be an integral part of the anode
Clearly, the operating conditions for direct-conversion SOFCs
that can generate electrical power from larger hydrocarbons without
either external or internal reforming will need to be very different
from the operation of traditional SOFCs. Everything from the en-
ergy balances to the way in which fuels are introduced to the anodes
needs to be rethought. What we have attempted to demonstrate in
this paper is that fuel-utilization issues can probably be solved in a
straightforward manner. We expect that similar developments can
and will be made in other aspects of operating direct-conversion
SOFCs.
Conclusions
Model studies of the effect of higher fuel conversions on the
performance of direct-conversion SOFCs operating on n-butane in-
dicate that performance decreases primarily due to dilution of the
fuel in CO2 and H2O. The addition of a steam-reforming catalyst
that is stable at low H2O:C ratios, like Pd/ceria, can improve cell
performance by increasing the concentration of fuel and generating
H2 .a
Acknowledgments
This work was supported by the Office of Naval Research.
The University of Pennsylvania assisted in meeting the publication costs
of this article.
a In a paper that appeared after submission of this paper, Hibino and co-workers
have also reported that the addition of a reforming catalyst could enhance performance
of SOFCs that directly utilize hydrocarbon fuels ͑Ref. 25͒.

Journal of The Electrochemical Society, 150 ͑7͒ A858-A863 ͑2003͒
A863
15. H. Yakabe, M. Hishinuma, M. Uratani, Y. Matsuzaki, and I. Yasuda, J. Power
Sources, 86, 423 ͑2000͒.
16. J. Kim, A. V. Virkar, K. Fung, K. Metha, and S. C. Singhal, J. Electrochem. Soc.,
146, 69 ͑1999͒.
17. W. Lehnert, J. Meusinger, and F. Thom, J. Power Sources, 87, 57 ͑2000͒.
18. E. L. Cussler, Diffusion Mass Transfer in Fluid Systems, 2nd ed., Chap. 5 and 6,
Cambridge, MA ͑1997͒.
19. R. Aiello, J. E. Fiscus, H. C. zur Loye, and M. D. Amiridis, Appl. Catal., A, 192,
227 ͑2000͒.
References
1. S. Park, R. Cracium, J. M. Vohs, and R. J. Gorte, J. Electrochem. Soc., 146, 3603
͑1999͒.
2. S. Park, J. M. Vohs, and R. J. Gorte, Nature (London), 404, 265 ͑2000͒.
3. H. Kim, S. Park, J. M. Vohs, and R. J. Gorte, J. Electrochem. Soc., 148, A693
͑2001͒.
4. S. Park, R. J. Gorte, and J. M. Vohs, Appl. Catal., A, 200, 55 ͑2000͒.
5. R. J. Gorte, S. Park, J. M. Vohs, and C. Wang, Adv. Mater. (Weinheim, Ger.), 12,
1465 ͑2000͒.
6. S. Park, R. J. Gorte, and J. M. Vohs, J. Electrochem. Soc., 148, A443 ͑2001͒.
7. R. J. Farrauto and C. H. Batholomew, Fundamentals of Industrial Catalytic Pro-
cesses, pp. 341-357, 1st ed., Blackie, London ͑1997͒.
20. R. T. K. Baker, P. S. Harris, J. Henderson, and R. B. Thomas, Carbon, 13, 17
͑1975͒.
21. R. T. K. Baker, G. R. Gadsby, and S. Terry, Carbon, 13, 245 ͑1975͒.
22. C. W. Keep, R. T. K. Baker, and J. A. France, J. Catal., 47, 232 ͑1977͒.
23. M. L. Toebes, J. H. Bitter, A. J. van Dillen, and K. P. de Jong, Catal. Today, 76, 33
͑2002͒.
24. Fuel Cell Handbook, 5th ed., pp. 9-23 and 9-24, U.S. Department of Energy
͑www.netl.doe.gov͒, Morgantown, WV ͑2000͒.
25. T. Hibino, A. Hashimoto, K. Asano, M. Yano, M. Suzuki, and M. Sano, Electro-
chem. Solid-State Lett., 5, A242 ͑2002͒.
8. N. Armor, Appl. Catal., A, 176, 159 ͑1999͒.
9. T. Maillet, J. Barbier, and D. Duprez, Appl. Catal., B, 9, 251 ͑1996͒.
10. Fuel Cell Handbook, 5th ed., pp. 6-29, U.S. Department of Energy
͑www.netl.doe.gov͒, Morgantown, WV ͑2000͒.
11. X. Wang and R. J. Gorte, Catal. Lett., 73, 15 ͑2001͒.
12. X. Wang and R. J. Gorte, Appl. Catal., A, 224, 209 ͑2002͒.
13. S. McIntosh, J. M. Vohs, and R. J. Gorte, J. Electrochem. Soc., 150, A470 ͑2003͒.
14. S. McIntosh, J. M. Vohs, and R. J. Gorte, Electrochim. Acta, 47, 3815 ͑2002͒.