Role of hydrocarbon deposits in the enhanced performance of direct-oxidation SOFCs

Article abstract of DOI:10.1149/1.1559064

We have examined the changes that occur in the performance of solid oxide fuel cells (SOFCs) with Cu-ceria-yttria-stabilized zirconia anodes at 973 K following exposure to various hydrocarbon fuels, including methane, propane, n-butane, n-decane, and tolu

Full text of DOI:10.1149/1.1559064

A470
Journal of The Electrochemical Society, 150 ͑4͒ A470-A476 ͑2003͒
0013-4651/2003/150͑4͒/A470/7/$7.00 © The Electrochemical Society, Inc.
Role of Hydrocarbon Deposits in the Enhanced Performance
of Direct-Oxidation SOFCs
,z
*
*
S. McIntosh, J. M. Vohs, and R. J. Gorte
Department of Chemical Engineering, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA
We have examined the changes that occur in the performance of solid oxide fuel cells ͑SOFCs͒ with Cu-ceria-yttria-stabilized
zirconia anodes at 973 K following exposure to various hydrocarbon fuels, including methane, propane, n-butane, n-decane, and
toluene. For cells with Cu contents of 20 wt % or less, large increases were observed in the power densities for operation in H2
after the anode had been exposed to any of the hydrocarbons except methane. The increased performance is completely reversible
upon oxidation of the anode and subsequent reduction in H2 . The enhancement decreases with increasing Cu content, implying
that the deposits improve the connectivity of the metallic phase in the anode. Impedance spectra taken on cells before and after
exposure to hydrocarbon fuels confirm that the conductivity of the anode improves after exposure. Temperature-programmed
oxidation and weight changes were used to show that the deposits that enhance performance correspond to ϳ1 wt % of the anode
and are probably not graphitic. Measurements of the open-circuit voltages in hydrocarbon fuels suggest that equilibrium is
established with partial oxidation products and that the chemical structure of the deposits change upon current flow. Finally, the
implications of these results for operation of SOFC on hydrocarbons without added steam and with low copper contents are
discussed.
© 2003 The Electrochemical Society. ͓DOI: 10.1149/1.1559064͔ All rights reserved.
Manuscript submitted June 14, 2002; revised manuscript received October 7, 2002. Available electronically February 28, 2003.
The materials used to make anodes in solid oxide fuel cells
pears that carbonaceous compounds are formed by larger hydrocar-
bons in the anodes and that these compounds help to provide con-
nectivity and increase performance. Even at open circuit, the
carbonaceous layer appears to reach a steady-state coverage and
does not lead to deactivation of the cell.
͑SOFCs͒ must meet several demanding requirements.1 First, the an-
ode must be electronically conductive at high temperatures, in re-
ducing environments, to remove electrons produced by electro-
chemical oxidation of the fuel. Second, the thermal coefficient of
expansion TCE for the anode must match that of the electrolyte so
as to prevent delamination upon heating and temperature cycling.
Third, the anode material must catalyze the electrochemical oxida-
tion of the fuel. The material most commonly used for SOFC anodes
is a composite of Ni and yttria-stabilized zirconia ͑YSZ͒. The YSZ
phase of this ceramic-metallic ͑cermet͒ composite provides a TCE
match to the YSZ electrolyte, supports the Ni particles, and inhibits
coarsening of the Ni at the operating temperature. The Ni phase of
the anode provides electronic conductivity and some hydrocarbon
catalytic-reforming activity. The Ni content is usually above 30 vol
% so as to maintain the required level of electronic conductivity.2
A major limitation to Ni cermet anodes is that Ni catalyzes
graphite formation in the presence of hydrocarbons unless there is
sufficient steam to simultaneously remove carbon as it forms.3,4
Therefore, it is not feasible to expose a Ni-based anode to dry hy-
drocarbons. Based on the catalytic properties of various electronic
conductors that could be used in the anode, we focused our attention
on developing Cu-based anodes for SOFCs.5-9 Compared to Ni, Cu
is not catalytically active for the formation of CuC bonds. Its melt-
ing temperature, 1083°C, is low compared to that of Ni, 1453°C;
however, for low temperature operation, Ͻ800°C, Cu is likely to be
sufficiently stable.
Experimental
The methods we have used for preparing and testing fuel cells
with Cu cermet anodes have been discussed in other papers.6,8 Be-
cause oxides of Cu melt at temperatures lower than that required for
sintering of the oxide components, the fabrication procedure in-
volved preparing a porous matrix of YSZ, impregnating this porous
matrix with Cu salt, and finally reducing the salt to metallic Cu.
In the first step, the dense electrolyte layer and the porous YSZ
matrix were prepared simultaneously by tape-casting methods. A
two-layer, green tape of YSZ, ͑Tosoh, 8 mol % Y2O3 , TZ-84͒ was
made by casting a tape with graphite and poly-methyl methacrylate
͑PMMA͒ pore formers over a green tape without pore formers. Fir-
ing the two-layer tape to 1800 K resulted in a YSZ wafer having a
dense side, 60 ␮m thick, supported by a porous layer, 600 ␮m thick.
The porosity of the porous layer was determined to be ϳ70% by
water uptake measurements.9 Next, a 50:50 mixture of YSZ and
La0.8Sr0.2MnO3 ͑LSM, Praxair Surface Technologies͒ powders was
applied as a paste onto the dense side of the wafer, then calcined to
1400 K to form the cathode. Third, the porous YSZ layer was im-
pregnated with an aqueous solution of Ce(NO3)3 • 6H2O and cal-
cined to 723 K to decompose the nitrate ions and form CeO2 . The
porous layer was then impregnated with an aqueous solution of
Cu(NO3)2 • 3H2O and again heated to 723 K in air to decompose
the nitrates. All of the cells used in this study were 10 wt % CeO2 ,
but the Cu content was varied between 5 and 30 wt %.
Electronic contacts were formed using Pt mesh and Pt paste at
the cathode and 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 . H2 , CH4 , propane, and
n-butane were fed to the cell undiluted, while toluene and decane
were fed as 75 mol % mixtures with N2 . All hydrocarbons, includ-
ing those that are liquids at room temperature, were fed directly to
the anode without reforming, as discussed elsewhere.10
Because Cu2O and CuO melt at 1235 and 1326°C, respectively,
temperatures below those that are necessary for densification of YSZ
electrolytes, it is not possible to prepare Cu-YSZ cermets by high
temperature calcination of mixed powders of CuO and YSZ, a
method analogous to that usually used as the first step to produce
Ni-YSZ cermets. Therefore, an alternative method for preparation of
Cu-YSZ cermets was developed in which a porous YSZ matrix was
prepared first, followed by addition of Cu and an oxidation catalyst
in subsequent processing steps.6,8 Because the Cu phase in the final
cermet must be highly connected, high metal loadings are necessary;
and, even then, connectivity between all Cu particles in the anode
structure is not assured.
In this paper, we will demonstrate that metallic connectivity can
limit the performance of the Cu cermet anodes. Fortunately, it ap-
The performance at 973 K for each cell was measured by its
voltage-current ͑V-I͒ curves with n-butane and H2 fuels, with im-
pedance spectra providing additional information on selected
* Electrochemical Society Active Member.
z E-mail: gorte@seas.upenn.edu
Downloaded on 2014-10-20 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 150 ͑4͒ A470-A476 ͑2003͒
A471
Figure 2. Performance curves in H2 for a cell with an anode consisting of
10% CeO2 and 15% Cu at 973 K. Results are shown ͑ࡗ͒ following the
initial heating in H2 ͑᭝͒ followed by 1 h in n-butane fuel, ͑छ͒ subsequent
exposure to 10% O2 in He, and ͑᭝͒ after exposure to n-butane.
Figure 1. Power density as a function of time at 973 K. The cell potential
was maintained at 0.5 V while the fuel was switched from H2 , to n-butane,
and back to H2 . The anode consisted of 10% CeO2 and 20% Cu.
the power density at 0.5 V is shown for a cell with 20 wt % Cu as a
function of time while changing the fuel from H2 , to n-butane, and
back to H2 . The anode had initially been exposed to H2 for a period
of several hours and the cell exhibited a power density of only
0.065 W/cm2. Upon changing the feed to pure n-butane, the power
density increased to a value of 0.135 W/cm2 following a brief tran-
sient period. After operating the cell in n-butane for 20 min, the feed
was switched to pure H2 and the power density increased to
0.21 W/cm2, a factor of 3.2 greater than the power density that had
been observed prior to exposing the anode to n-butane.
This enhancement of cell performance after exposure to n-butane
was found to be fully reversible upon reoxidation of the anode. In
Fig. 2 and 3, V-I curves and impedance spectra at 973 K are shown
after various pretreatments for a cell operating in pure H2 , with an
anode consisting of 10 wt % CeO2 and 15 wt % Cu. Data are shown
for the cell after the initial reduction of the anode in H2 , after
exposing the anode to pure n-butane for 60 min then after exposing
it to 15% O2 in He for 30 min and, finally, after a further 60 min
exposure to n-butane. ͑Following the oxidation cycle, the anode was
held in H2 for 30 min before recording the data.͒ Initially, Fig. 2
shows that the maximum power density in H2 was 0.045 W/cm2.
This increased to 0.16 W/cm2 after a 1 h exposure to n-butane.
samples. Since the cathodes and electrolytes were prepared in a
similar manner in all cases, changes in the fuel cell performance and
in the impedance spectra can be attributed to changes in the anode.
Since the fuel flow rates were always greater than 1 cm3/s at room
temperature, the conversion of the hydrocarbon fuels was always
less than 1%, so that water produced by the electrochemical oxida-
tion reactions was negligible. The impedance spectra were obtained
in galvanostatic mode at close to the open-circuit voltage ͑OCV͒,
using a Gamry Instruments, model EIS300.
The amount of carbon present in the SOFC anode after treatment
in n-butane was also measured. To accomplish this, anode cermet
samples were exposed to flowing n-butane in a quartz flow reactor at
973 K for various periods of time. We then measured either the
sample weight or the amount of CO and CO2 that formed upon
exposure to flowing O2 . In the weight measurements, the sample
temperature was ramped to 973 K in flowing He, exposed to flowing
n-butane for a limited period, and then cooled in flowing He. Fol-
lowing longer exposures, the samples were flushed in flowing He at
973 K for 24 h before cooling.
In the second method for measuring carbon contents in the an-
ode, samples were exposed to n-butane in the flow reactor at 973 K
and flushed with He. Then the sample was exposed to a flowing gas
consisting of a 15% O2-85% He mixture and the reactor effluent
was monitored with a mass spectrometer. The amount of carbon in
the sample was determined from the amounts of CO and CO2 leav-
ing the reactor. The type of carbon formed was also characterized by
TPO in a similar manner. In these measurements, a cermet sample
was exposed to flowing n-butane at 973 K for 30 min. The reactor
was cooled to 298 K in flowing He and again ramped to 973 K at a
rate of 10 K/min in a flowing gas mixture of 15% O2-85% He. In
principle, TPO experiments carried out with a mass spectrometer
would enable the calculation of carbon-to-hydrogen ratios because
the detector should be able to determine the amount of hydrogen in
the deposits; however, the background signal for water in our
vacuum system was too high to allow accurate measurement of this
quantity. A sample of 0.03 g of graphite powder ͑Alpha Aesar, con-
ducting grade 99.995%͒ was placed in an identical reactor and
heated in a 15% O2-85% He stream at 10 K/min for comparison.
Scanning electron microscope ͑SEM͒ measurements of the graphite
sample suggested that the particles were shaped as platelets, less
than 10 ␮m in thickness.
Figure 3. OCV impedance spectra in H2 from a cell with an anode consist-
ing of 10% CeO2 and 15% Cu at 973 K. Results are shown ͑ࡗ͒ following
the initial heating in H2 , ͑᭡͒ followed by 1 h in n-butane fuel, ͑छ͒ subse-
quent exposure to 10% O2 in He, and ͑᭝͒ after exposure to n-butane.
Results
The effect of treating the Cu cermet anodes in a hydrocarbon fuel
at 973 K is demonstrated by the results in Fig. 1. In this experiment,
Downloaded on 2014-10-20 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A472
Journal of The Electrochemical Society, 150 ͑4͒ A470-A476 ͑2003͒
Figure 4. V-I ͑closed symbols͒ and power/density ͑open symbols͒ curves for a cell with an anode consisting of 10% CeO2 with various Cu loadings. The
measurements used H2 as the fuel at 973 K. The data are shown ͑ࡗ͒ before and ͑᭡͒ after 30 min exposure to n-butane.
methane on Ni allowed conduction of electrons between metal par-
ticles. We suggest that similar arguments must be used to explain the
results in Fig. 1 to 3. Indeed, one must conclude that there is initially
poor connectivity between metal particles in the anode based on the
high initial ohmic resistance in Fig. 3. R⍀ should be less than
1 ⍀ cm2 for our cell based on literature values for the conductivity
of YSZ at 973 K and the thickness of our electrolyte. The fact that
R⍀ is initially much larger than this implies that part of the ohmic
resistance must be in the anode.
An obvious implication of the above conclusion is that increased
Cu contents should improve the initial performance and possibly
reduce the enhancement observed with treatment in hydrocarbon
fuels. This is indeed the case, as shown in Fig. 4 and 5. In Fig. 4a
through 4d, V-I curves, with pure H2 at 973 K, are reported for cells
containing 5, 10, 20, and 30% copper, before and after exposure to
n-butane for 30 min. The ceria content and YSZ structure were
identical in all of the cells. Fig. 4a and b indicate that the initial
performance for cells with a low Cu content is poor, but increases
dramatically upon exposure to n-butane. The maximum power den-
sity increased by a factor of 3.5 for these two cases. The data for the
cell with 20% copper, Fig. 4c, showed a more modest improvement,
with the maximum power density increasing by a factor of only 2.5
after treatment with n-butane. Finally, data for the cell with 30%
Following oxidation in 15% O2 and reduction in H2 , the perfor-
mance curve returned to its initial value. Finally, exposing the cell to
n-butane once again increased the performance curve to its higher
value.
Figure 3 shows the impedance curves, measured in H2 at OCV,
associated with the V-I curves in Fig. 2. The enhanced performance
upon exposure to n-butane and the reversibility upon re-oxidation
are evident from the total cell resistances, which are approximately
6 ⍀ cm2 before treatment in n-butane and 1.4 ⍀ cm2 after treatment
in n-butane. Of additional interest, the ohmic resistance of the cell,
R
⍀ , measured by the high frequency intercept with the real axis,
decreased from ϳ2.9 ⍀ cm2 to ϳ0.6 ⍀ cm2 after n-butane treat-
ment. Normally, R⍀ is associated with the conductivity of the elec-
trolyte. Migration of charged species in mixed-conducting anodes
and cathodes gives rise to an interfacial resistance, RI , taken to be
the difference between the high and low frequency intercepts with
the real axis. RI , too, decreases from more than 3 ⍀ cm2 to
ϳ1 ⍀ cm2 after treatment in n-butane.
Similar increases in performance and changes in impedance
spectra were reported for an SOFC made with Cu-Ni alloy cermets
after long exposures to methane.11 In that paper, it was suggested
that there was initially poor connectivity between the metal particles
in the cermet anode and that carbon fibers formed by reaction of
Downloaded on 2014-10-20 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 150 ͑4͒ A470-A476 ͑2003͒
A473
Figure 6. Temperature-programmed oxidation data for a sample consisting
of 10% CeO2 and 20% Cu in a YSZ matrix after exposure to n-butane for 30
min at 973 K. The dashed line is O2 (m/e ϭ 32) and the solid line is CO2
(m/e ϭ 44).
30 min, and 4.5% after 24 h. The carbon content based on the
production of CO and CO2 formed by reaction with the 15%
O2-85% He mixture was 2.1% after 10 min and 4.0% after 20 min,
but this number also includes any carbon formed on the reactor
walls. Since the performance increase following treatment in
n-butane occurs in much less than 10 min and is not lost upon
exposure to flowing H2 , the small carbon contents observed in these
measurements suggest that small amounts of hydrocarbon are
needed to increase the connectivity in the anode. This is particularly
interesting given that relatively large amounts of Cu need to be
added to achieve the same connectivity.
To determine how hydrocarbons other than n-butane would af-
fect the anode, we looked for enhanced performance of a cell made
with 20 wt % Cu and 10 wt % CeO2 in H2 at 973 K after exposing
it to methane, propane, n-decane, and toluene. Between measure-
ments, the cell was exposed to a 10% O2-90% N2 stream to reverse
any enhancements caused by the previous fuel. With n-decane and
toluene, enhanced performance was observed almost instantly after
exposing the fuel to the anode; and the performance enhancements
for n-butane, n-decane, and toluene were also indistinguishable. For
propane, we again observed a similar enhancement but the enhance-
ment occurred much more gradually. It was necessary to expose the
cell to propane for more than 10 min to achieve the maximum power
density. With methane, we observed no enhancement, even after
several hours. Because methane exhibits a much lower tendency to
undergo free radical reactions compared to the other hydrocarbons
examined, with propane the next least reactive, we interpret these
results as indicating that any fuel which causes hydrocarbons to
form in the anode will lead to similar performance enhancements.
We investigated the nature of the anode deposits using TPO car-
ried out in a He-O2 mixture. Figure 6 shows CO2 (m/e ϭ 44) and
O2 (m/e ϭ 32) signals from the TPO curves for a YSZ cermet
impregnated with 20% Cu and 10% CeO2 that was exposed to
n-butane for 30 min at 973 K before being cooled to room tempera-
ture in flowing He. The results show that CO2 is formed and O2
consumed in a narrow range of temperatures, between 623 and 723
K. An additional O2 consumption peak is observed at 773 K that we
assign as being due to reoxidation of bulk Cu, although some of the
O2 consumed in the lower peak must also correspond to Cu oxida-
tion. We were unable to observe water formation, but more O2 is
consumed than can be accounted for by CO2 production and Cu
oxidation. The additional O2 consumption is probably due to water
formation but is difficult to quantify. The likely formation of water,
Figure 5. OCV impedance spectra in H2 for cells having anodes with dif-
ferent copper loadings at 973 K: ͑᭿͒ 5% Cu, ͑᭡͒ 10% Cu, ͑ࡗ͒ 20% Cu, ͑᭹͒
30% Cu, ͑a͒ before and ͑b͒ after n-butane exposure.
copper showed only small changes in the performance curves after
exposure to n-butane.
Impedance spectra measured at OCV in H2 on the same cells are
shown in Fig. 5, with results before n-butane treatment given in Fig.
5a, and 5b shows the results after treatment. Prior to n-butane, there
is a steady decrease in both R⍀ and RI as the Cu content increased.
The changes in these values are particularly large in going from 10
wt % Cu to 20 wt % Cu. Even after treatment with n-butane, R⍀
decreases steadily, going from ϳ1.0 to ϳ0.5 ⍀ cm2. The changes in
R⍀ would suggest that connectivity of the electronic conductors in
the anode increase with both the addition of Cu and with n-butane
treatment, but that addition of Cu is more effective. However, it is
interesting to notice that RI in the 30 wt % Cu cell remains relatively
large after treatment in n-butane. Indeed, after treatment in n-butane,
the 30 wt % Cu cell had the largest RI of all the four cells investi-
gated.
Assuming that the enhanced anode conductivity is due to depo-
sition of hydrocarbons in the anode, we next examined the increase
in the mass of various samples after they had been heated in flowing
n-butane at 973 K in a tubular reactor. First, we observed no signifi-
cant differences in the mass changes for a pure YSZ matrix and for
a YSZ matrix having 20 wt % Cu and 10 wt % CeO2 added. For the
Cu cermet, the weight changes were 1.3% after 10 min, 2.1% after
Downloaded on 2014-10-20 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A474
Journal of The Electrochemical Society, 150 ͑4͒ A470-A476 ͑2003͒
Figure 9. A plot of the OCV as a function of time for a cell with an anode
consisting of 10% CeO2 and 20% Cu with n-butane. The cell was short-
circuited for 1 min at 4.45 h.
Figure 7. Temperature-programmed oxidation results for 0.03 g of graphite
powder. The dashed line is O2 (m/e ϭ 32) and the solid line is CO2 (m/e
ϭ 44).
0.85 V. The inset plot shows the transient performance in more
detail. Figure 10 shows the results for a similar experiment, except
that this time the cell potential was held at 0.5 V for 48 h before
measuring the OCV measurement. As shown in the inset, the OCV
started at greater than 1.0 V and fell to 0.85 V over a period of less
than 1 h. Similar behavior is evident in the transient behavior re-
ported in previous papers6 and was observed also for cells in which
ceria was not added.
Although we do not have a definitive explanation for the data in
Fig. 9 and 10, the data are consistent with the concept that there is a
hydrocarbon layer at the three-phase boundary ͑TPB͒ in the direct-
oxidation experiments. Since the OCV for these cells with H2 as the
fuel was 1.1 V, it seems unlikely that leaks can account for the low
OCV in n-butane at steady state. Also, the theoretical OCV for com-
plete combustion of n-butane to CO2 and H2O is 1.12 V at standard
conditions and 973 K. While the oxidation of carbon and most hy-
drocarbons should give an OCV of greater than 1 V, partial oxida-
tion reactions would result in lower standard potentials. For ex-
ample, the standard potential for oxidation of n-butane to n-butanal
is 0.87 V at 973 K. Other redox couples, such as oxidation of
Ce2O3 , cannot account for an OCV of 0.85 V. Therefore, the most
likely explanation for the OCV data shown here is that equilibrium
is established with partial oxidation reactions. The transients in the
OCV are probably due to slow changes in the chemical structure of
the hydrocarbon layer within the anode.
together with the fact that the deposits react at low temperatures,
strongly suggests that the deposits are not graphitic. In Fig. 7, a TPO
curve is shown for the graphite powder sample using the same ex-
perimental conditions. CO2 production does not occur until above
973 K, a value similar to that reported by others.12 Some of the
difference between the graphite and the anode deposits could be due
to surface area effects and the presence of ceria in the anode; how-
ever, neither the presence of a catalyst13 nor the increased surface
area would be expected to give a temperature increase of more than
300 degrees.
Finally, to determine whether the oxygen-ion flux through the
electrolyte might potentially ‘‘clean’’ the anode, we examined the
cell under OCV conditions at 973 K in the presence of 100% flow-
ing n-butane. In Fig. 8, the V-I curves are shown for a cell with 20
wt % Cu using n-butane as the fuel. The results are shown at the
beginning of the experiment and after the cell had been held at open
circuit for 24 h. There appears to be a slight decrease in the maxi-
mum power density after the 24 h exposure but the differences are
not significant.
During the course of this experiment, the OCV measurements
showed interesting trends, as shown in Fig. 9. Initially, the OCV in
n-butane was greater than 1.0 V but it fell quickly to a value of 0.85
V. After ϳ4 h, we briefly shorted the cell, then measured the OCV.
Again, the OCV started at more than 1.0 V and decreased rapidly to
Figure 8. V-I ͑closed symbols͒ and power density ͑open symbols͒ curves for
a cell with an anode consisting of 10% CeO2 and 20% Cu using n-butane as
fuel at 973 K. The data were taken on ͑ࡗ͒ a fresh sample and ͑᭡͒ after
holding that sample for 24 h in flowing n-butane at OCV.
Figure 10. Data showing the transient response for a cell consisting of 10%
CeO2 and 20% Cu after maintaining a cell potential of 0.5 V for 48 h.
Downloaded on 2014-10-20 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 150 ͑4͒ A470-A476 ͑2003͒
A475
schematic drawing of what we believe occurs in the region near the
TPB upon exposure of the Cu-based anodes to hydrocarbons. For
lower Cu contents, some of the Cu particles are initially not con-
nected to the outside circuit and are therefore unable to conduct
electrons away from the TPB. The addition of hydrocarbon ‘‘resi-
dues’’ likely fills the gaps between the metal particles and provides
sufficient conductivity to allow the flow of electrons.
What is surprising is that small amounts of hydrocarbon residue
are apparently sufficient to increase the conductivity substantially.
Although we do not know the chemical form of the residue, the
quantity necessary to significantly enhance performance appears to
correspond to no more than 2 wt %. If we assume a density for the
residue of 1 g/cm3, a value typical for hydrocarbons, the volume
fraction of this residue is less than 5%. If one assumes the density
for the residue is more similar to that of graphite, the volume occu-
pied by the residue would be lower. By comparison, the minimum
metal content for cermet anodes is reported to be 30 vol %.2 Obvi-
ously, the Cu contents in our samples were always much less than
this. Even the sample with 30 wt % Cu has a volume fraction of Cu
that is only 19%. The addition of an extra 5 vol % carbon would not
seem to be sufficient to increase the fraction of the electron-
conductive phase enough to make such a large difference in perfor-
mance. A partial explanation for the unexpected behavior may lie in
the structure of the sample anodes. Since Cu is added to the porous
YSZ matrix after the pore structure has been established, the anode
structure is likely to be much less random than cermets prepared by
more conventional methods. Therefore, the deposits may simply
coat the walls of the pores and enhance conductivity much more
effectively than would the random addition of an electron-
conductive phase.
We believe the anode deposits are tar-like, rather than graphitic.
Since we observed no noticeable difference in the amounts depos-
ited on pure YSZ and YSZ with Cu and ceria added, it would appear
that these deposits form through free radical decomposition, rather
than by any surface-catalyzed processes. Based on the TPO results,
the deposits are too reactive to be graphitic. Finally, because the
enhancements were similar for various hydrocarbon fuels, even
though the enhancements occurred rapidly, it would appear that the
detailed chemical structure of the deposits is not crucial. However,
since hydrocarbons are only electronic conductors when they con-
tain highly conjugated olefinic or aromatic groups, some dehydro-
genation of the tars must take place.
The low OCVs observed for n-butane at steady state appear to
provide indirect information on the mechanism for hydrocarbon oxi-
dation. Since the complete oxidation of n-butane to CO2 and H2O
involves 26 electrons, this reaction cannot occur in a single step. As
discussed earlier, the standard potential for the partial oxidation of
n-butane to n-butanal is 0.85 V, in good agreement with the ob-
served value at steady state, suggesting that the oxygen chemical
potential at the anode is established by partial oxidation reactions.
The fact that the OCV changes when current flows through the cell
would indicate that the species that are present on the surface near
the TPB can change depending on how the cell is treated.
Figure 11. Schematic to describe the changes in the three-phase boundary
͑a͒ before and ͑b͒ after operation with n-butane.
Discussion
Fuel cells that oxidize complex hydrocarbons directly, without
first reforming them to hydrogen and CO, are a relatively new dis-
covery. Although a number of groups have accomplished this with
methane,14-16 there is very little information in the literature on di-
rect oxidation of hydrocarbons with higher molecular weights. Be-
cause the oxidation of higher hydrocarbons must involve complex,
multistep reactions, it should not be surprising that complex inter-
mediates must exist on the anode surface. There are likely a wide
variety of intermediates formed at the anode/electrolyte interface;
hydrocarbon deposits are probable and should be expected to affect
performance. That these changes should be favorable was unex-
pected.
One important implication of the results in this paper is that it
may be possible to operate a direct-oxidation fuel cell with low
metal contents and still get reasonable performance. At low metal
contents, reoxidation of the Cu does not destroy the cell, as we have
demonstrated in this study. Also, it should be possible to counter the
effects of Cu sintering, which is likely to be a problem for operation
at higher temperatures due to the low melting temperature of Cu.
Therefore, the results of this study could prove to be useful, in
addition to scientifically interesting.
There could be significant advantages to fuel cells that can oxi-
dize hydrocarbons directly, without the need for reforming, both for
increasing the efficiencies and for simplifying the fuel cell system.
This area of investigation is still new, and there is a great deal that
needs to be done to prove the commercial viability of this concept.
The present study demonstrates that there is much we do not under-
stand about how direct oxidation works and what needs to be done
The performance enhancements observed here upon exposure of
the anodes to hydrocarbon fuels must be due to improved connec-
tivity in the electron-conducting phase based on the fact that the
addition of more Cu leads to similar enhancements. Figure 11 is a
Downloaded on 2014-10-20 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A476
Journal of The Electrochemical Society, 150 ͑4͒ A470-A476 ͑2003͒
to optimize the performance of SOFCs that oxidize hydrocarbons
directly. This area remains a fertile topic for investigation and we
expect to see significant advances in the years to come.
References
1. N. Q. Minh, J. Am. Ceram. Soc., 76, 563 ͑1993͒.
2. D. W. Dees, T. D. Claar, T. E. Easler, D. C. Fee, and F. C. Mrazek, J. Electrochem.
Soc., 134, 2141 ͑1987͒.
3. R. T. K. Baker, Carbon, 27, 315 ͑1989͒.
4. R. J. Farrauto and C. H. Batholomew, Fundamentals of Industrial Catalytic Pro-
cesses, 1st ed., p. 341, Blackie Academic & Professional, London ͑1997͒.
5. S. Park, J. M. Vohs, and R. J. Gorte, Nature (London), 404, 265 ͑2000͒.
6. R. J. Gorte, S. Park, J. M. Vohs, and C. Wang, Adv. Mater., 12, 1465 ͑2000͒.
7. S. Park, R. Craciun, J. M. Vohs, and R. J. Gorte, J. Electrochem. Soc., 146, 3603
͑1999͒.
8. S. Park, R. J. Gorte, and J. M. Vohs, J. Electrochem. Soc., 148, A443 ͑2001͒.
9. H. Kim, C. daRosa, M. Boaro, J. M. Vohs, and R. J. Gorte, J. Am. Ceram. Soc., 85,
1473 ͑2002͒.
10. H. Kim, S. D. Park, J. M. Vohs, and R. J. Gorte, J. Electrochem. Soc., 148, A693
͑2001͒.
11. H. Kim, C. Lu, W. L. Worrell, J. M. Vohs, and R. J. Gorte, J. Electrochem. Soc.,
149, A247 ͑2002͒.
Conclusions
Deposits formed in Cu-ceria-YSZ anodes of a SOFC operated
with various hydrocarbon fuels can lead to significant enhancements
in the performance of the fuel cells due to improved connectivity of
the electron-conductive phase. The deposits do not appear to be
graphitic and provide significant conductivity even when present at
a level of order 1 wt %. Open-circuit voltages suggest that the de-
posits likely contain partial oxidation products.
12. P. Wang, E. Tanabe, K. Ito, J. Jia, H. Morioka, T. Shishido, and K. Takehira, Appl.
Catal., A, 231, 35 ͑2002͒.
13. J. O.-Uchisawa, A. Obuchi, R. Enomoto, A. Ogata, and S. Kushiyama, J. Chem.
Soc. Chem. Commun., 1998, 2255.
14. B. C. H. Steele, I. Kelly, P. H. Middleton, and R. Rudkin, Solid State Ionics, 28,
1547 ͑1988͒.
Acknowledgments
This work was supported by the Office of Naval Research. We
are grateful for additional support from the Ford Motor Company.
The University of Pennsylvania assisted in meeting the publication costs
of this article.
15. O. A. Marina and M. Mogensen, Appl. Catal., A, 189, 117 ͑1999͒.
16. E. Perry Murray, T. Tsai, and S. A. Barnett, Nature (London), 400, 649 ͑1999͒.
Downloaded on 2014-10-20 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).