Importance of gas-phase kinetics within the anode channel of a solid-oxide fuel cell

Article abstract of DOI:10.1021/jp037839w

Experiments using n-butane as a representative hydrocarbon fuel were conducted under gas-phase conditions similar to those expected in the anode channel of a solid-oxide fuel cell (SOFC). Butane conversion and product formation were monitored in quartz reactor experiments at P ～ 0.8 atm, τ ～ 5 s, and T = 550-800 °C. Three different fuel mixtures were used: neat n-butane, 50% n-C₄H₁₀/50% H₂O, and 50% n-C₄H₁₀/50% N₂. These experiments demonstrate that substantial gas-phase chemistry does occur and that this must be accounted for when predicting fuel cell efficiency. These data were compared to predictions using a plug-flow model that incorporated the experimentally measured temperature profile along the reactor. The reaction mechanism used for these simulations consisted of ～300 species and 2500 elementary reactions and included both pyrolysis and oxidation reactions. Comparisons of the model predictions to the experimental data show that the model, without any modifications, captures the observed strong temperature dependence of n-butane conversion and is also able to capture the changes in product selectivity with temperature for the neat butane and the diluted butane mixtures. The model also properly predicts the observed onset of deposit formation near 700°C. Both conversion and selectivity are shown to be sensitive to only a very small subset of the reactions in the mechanism. Comparison of the rate coefficients of this subset to literature values, where available, are generally reasonable and suggest that this kinetic model is adequate for describing the gasphase reactions of small hydrocarbons in the anode channels of a SOFC. Additional efforts are required to account for catalytic reactions on the surface of the porous anode.

Full text of DOI:10.1021/jp037839w

3772
J. Phys. Chem. A 2004, 108, 3772-3783
Importance of Gas-Phase Kinetics within the Anode Channel of a Solid-Oxide Fuel Cell
†
Chad Y. Sheng and Anthony M. Dean*
Department of Chemical Engineering, Colorado School of Mines, Golden, Colorado 80401
ReceiVed: December 12, 2003; In Final Form: February 27, 2004
Experiments using n-butane as a representative hydrocarbon fuel were conducted under gas-phase conditions
similar to those expected in the anode channel of a solid-oxide fuel cell (SOFC). Butane conversion and
product formation were monitored in quartz reactor experiments at P ∼ 0.8 atm, τ ∼ 5 s, and T ) 550-800
°
4 2 4
C. Three different fuel mixtures were used: neat n-butane, 50% n-C H10/50% H O, and 50% n-C H10/50%
2
N . These experiments demonstrate that substantial gas-phase chemistry does occur and that this must be
accounted for when predicting fuel cell efficiency. These data were compared to predictions using a plug-
flow model that incorporated the experimentally measured temperature profile along the reactor. The reaction
mechanism used for these simulations consisted of ∼300 species and 2500 elementary reactions and included
both pyrolysis and oxidation reactions. Comparisons of the model predictions to the experimental data show
that the model, without any modifications, captures the observed strong temperature dependence of n-butane
conversion and is also able to capture the changes in product selectivity with temperature for the neat butane
and the diluted butane mixtures. The model also properly predicts the observed onset of deposit formation
near 700 °C. Both conversion and selectivity are shown to be sensitive to only a very small subset of the
reactions in the mechanism. Comparison of the rate coefficients of this subset to literature values, where
available, are generally reasonable and suggest that this kinetic model is adequate for describing the gas-
phase reactions of small hydrocarbons in the anode channels of a SOFC. Additional efforts are required to
account for catalytic reactions on the surface of the porous anode.
Introduction
and allow the SOFC to be stackable and therefore scaleable.
Typical operating temperatures for SOFCs range from 700 to
Fuel cell technology presents the possibility of increased
energy efficiency with decreased detrimental environmental
impact over conventional combustion. Solid-oxide fuel cells
1
000 °C.
A complete SOFC model is very complex, requiring descrip-
tions of kinetics in multiple phases and coupling these kinetics
(SOFC), in particular, offers a very promising method for direct
2
to multiple transport processes. Chemical reactions within the
production of electrical energy from currently available fossil
fuels, thus postponing the need to build a hydrogen delivery
infrastructure. An added advantage of SOFC operation is the
ability to tolerate substantial quantities of CO in the feed stream,
unlike polymer electrolyte membrane (PEM) fuel cells. Hybrid
SOFCs, coupled with a heat recovery system such as a gas
turbine generator, have the potential for fuel-to-electricity
efficiencies approaching 75-80%.1
A typical SOFC design consists of two channels separated
by a “trilayer”. This trilayer consists of the anode, electrolyte
and a cathode. The anode and cathode are fabricated of
ceramic-metal (cermet) materials and the electrolyte consists
of a dense ceramic. The (hydrocarbon) fuel flows through the
anode channel, while the (air) oxidant flows through the cathode
channel. Oxygen anions are formed in the porous cathode and
diffuse through the (ion selective) electrolyte. Electrochemical
reactions occur in the three-phase region (electrolyte, anode,
fuel) to form CO2 and H2O, plus electrons. The electrons flow
through the anode and are collected by bipolar plates. The
bipolar plates provide an efficient method to conduct current,
separate the channels for the fuel and air, add physical strength
SOFC occur in three regions: in the anode channel (gas-phase
kinetics), on the surface of the porous anode (heterogeneous
catalysis), and at the three-phase boundary between the anode
and electrolyte (electrochemical catalysis). Thus, the homoge-
neous and heterogeneous kinetics have the potential to substan-
tially change the nature of the species that ultimately undergo
electrochemical oxidation. It is essential to account for such
kinetic modifications to properly characterize SOFC operation
and estimate the expected efficiency.
Another issue in SOFC operation is deposit formation. The
high operating temperatures of a SOFC can lead to fuel
degradation and formation of carbonaceous deposits within the
fuel channels and porous anode structures. SOFCs that run on
H2 as a fuel source often utilize nickel as the electrical conductor.
However, nickel is known to react with hydrocarbons to form
3
deposits. Several studies have reported on the use of different
4-10
materials to minimize deposit formation
as well as upstream
conversion of hydrocarbon fuels to syngas (CO and H2) as a
1
1-16
way to avoid deposit formation.
An important feature of SOFC operation is the electrochemi-
cal formation of H2O and CO2. These species form at the three-
phase boundary and will diffuse through the porous anode to
the anode channel, thus diluting the fuel stream. While still in
the anode, these species can participate in “internal reforming”
reactions, performing the same function of converting hydro-
carbon fuel to CO and H2 that might be accomplished by the
*
To whom correspondence should be addressed. E-mail: amdean@
mines.edu.
†
Current address: National Institute of Standards and Technology,
Gaithersburg, MD 20899.
1
0.1021/jp037839w CCC: $27.50 © 2004 American Chemical Society
Published on Web 04/03/2004

Importance of Gas-Phase Kinetics
J. Phys. Chem. A, Vol. 108, No. 17, 2004 3773
Figure 1. Experimental setup for study of n-butane pyrolysis under conditions similar to the anode channel of a SOFC.
upstream reforming discussed earlier. Under these conditions
it is likely that the water-gas shift reaction would also take place,
converting CO + H2O to H2 + CO2. These reactions can further
complicate the identity of the “fuel” that will participate in the
electrochemical reactions. In addition, steam gasification, to
remove any deposits already formed, might be catalyzed by the
metal in the anode cermet.
was a 24 cm length × 6 mm i.d. (12 mm o.d.) quartz tube,
which was housed within a 14.5 cm length single-zone clamshell
2.5 cm i.d. electric tube furnace equipped with a dedicated
Eurotherm model 2116 digital temperature controller. The
electric tube furnace had a heating zone of 11.5 cm. The product
stream flowed through a water trap, which contained Drierite
supplied by W. A. Hammond Drierite Co. Ltd. The dehumidified
product stream then flowed to a Hewlett-Packard 5890 Series
II+ gas chromatograph (GC) and then vented to a fume hood.
The system was kept near atmospheric pressure (∼0.8 atm).
(Refer to Figure 1 for experimental setup).
As a first step toward characterizing the complex kinetics,
we want to make sure that we properly understand the gas-
phase hydrocarbon chemistry. The effect of gas phase reactions
17
of methane has previously been reported. It was shown that
the electrochemical production of H2O and CO2 did indeed
significantly impact the predictions regarding deposit formation.
In this work, we extend this approach to butane, which is
significantly more reactive than methane and thus more
representative of a range of hydrocarbon fuels. Furthermore,
given the sparsity of available data on butane reactivity under
SOFC conditions, we thought it essential to generate the
requisite data to validate our kinetic model. Although numerous
studies on n-butane oxidation in combustion environments and
n-butane catalytic studies are available, our literature search
revealed very limited experimental studies on neat n-butane
pyrolysis. Hepp and Frey had performed experimental studies
Temperature profiles for the reactor were measured with an
OMEGA type K thermocouple probe that was moved axially
along the reactor length. Thermocouple errors caused by
radiation from the furnace wall were minimized by using a
grounded sheath and with a representative flow of inert gas.
The measured temperature profiles with a N2 flow through the
quartz reactor for each set point reactor temperature, i.e. 550,
6
00, 650, 700, and 800 °C, are shown in Figure 2a. The reactor
is kept near isothermal over the midrange of the heating zone.
The temperature gradients that occur during the 5 cm inlet and
the 5 cm outlet serve to preheat the reactants and cool the
products. Uncertainty in absolute temperature measurements is
estimated to be (1%, as specified by the thermocouple
manufacturer.
1
8
of n-butane at 550 °C and 160 atm in a flow reactor. Weisel
et al. performed experimental studies at low pressures, 10 Torr,
over the temperature range 600-1300 K, with approximately
Two other sets of temperature profiles were also measured.
Given that pyrolysis reactions are endothermic, additional
experiments at 700 °C were conducted to observe if there was
a decrease in temperature along the centerline of the reactor
resulting from reaction. The measured temperature profile along
the quartz reactor flowing 100% n-C4H10 is shown as the triangle
symbols in Figure 2b. The 100% n-butane system had a lower
temperature profile than the 100% N2 system. The largest
temperature difference between the 100% N2 and 100% n-C4H10
system is almost 30 °C near 13 cm. An expanded view of the
temperature profile is shown in Figure 2c. The measured
temperature profile for the 50% n-C4H10/50% N2 mixture at 700
2
1
ms residence time.19 Weisel et al. utilized a 10 cm length,
.5 mm i.d. alumina tubular reactor coupled to a molecular-
beam-sampling mass spectrometer. Modeling studies for neat
n-butane were also limited; the only study we found was by
2
0
Mallinson et al. Mallinson et al.’s modeling conditions were
over a temperature range of 200-600 °C with pressure ranging
from 1 to 1000 atm in a batch reactor with a residence time in
the hundreds of seconds.
Experiment Description
The experiments used n-butane as the fuel over a temperature
range of 550-800 °C, with a nominal 5 s residence time through
a tubular flow reactor operating at an ambient (high-altitude)
pressure of ∼0.8 atm. The reactor used in these experiments
°
C was, as expected, intermediate between those shown.
The n-butane gas is research grade (99.9%), supplied by Air
Liquide. Scott Specialty Gases were used as the calibration

3774 J. Phys. Chem. A, Vol. 108, No. 17, 2004
Sheng and Dean
in. stainless steel HayeSep DB column. The two different
columns were required to identify all the measured species.
Comparison of measurements made on both detectors indicated
an absolute uncertainty of ∼5%. Helium was used as the carrier
gas for both columns as well as the reference gas for both TCDs.
HP’s ChemStation software package was used to acquire and
quantitate the chromatograms from both detectors. Because of
the carrier gas type and reference gas in the TCD, hydrogen
could not be measured accurately, but the presence of hydrogen
was noted in some experiments.
Three different fuel mixtures were used: neat n-butane (100%
n-butane), a 50:50 n-butane:N2 mixture, and a 50:50 n-butane:
H2O(g) mixture. For brevity, the following terminologies will
be used for the remainder of this paper; the 50% n-butane +
5
0% N2 fuel mixture will be referred to as “nC4/N2” and the
5
0% n-butane + 50% H2O will be denoted as “nC4/H2O”. The
neat n-butane case is intended to simulate the situation near
the entrance of the anode channel where little CO2 and H2O
would be expected to be present. The nC4/H2O mixture
represents a condition further down the channel where the
products of the electrochemical reaction would be coming into
the fuel stream. The nC4/N2 mixture was used to account for
the effect of simple dilution with an inert in contrast to dilution
with steam, which could possibly be a reactant.
The mass flow controllers fixed the fuel mixture composition
for the nC4/N2 case. For the nC4/H2O case, a different config-
uration was required. The n-butane gas was bubbled through a
sparger in a heated stainless steel evaporator, acting as a
humidifier. The humidified effluent vapor composition from the
evaporator was determined with the assumption that the vapor-
liquid equilibrium (VLE) relationship between n-butane to water
is similar to the VLE of air to water. The temperature at which
the saturated equilibrium steam composition will be 50% at 0.8
atm is 77 °C. The flow rate of the n-butane through the sparger
-
1
was 7.3 mL min , with the displaced steam comprising the
-
1
other 7.3 mL min for a total volumetric flow rate of 14.6 mL
-1
min . A dedicated Eurotherm 2116e PID temperature controller
was used to regulate the vapor temperature within the humidifier
chamber. Transfer lines from the humidifier are heated above
1
00 °C to ensure the steam did not condense prior to entering
the reactor. A water trap at the exit of the reactor and prior to
entrance into the GC was used, since the GC temperature
program starts below 100 °C and removal of liquid water was
necessary to preserve the integrity of the packed columns.
Experimental runs were carried out for the three different
mixtures and the five different temperature profiles. For the
remainder of this paper, the temperature profiles will be referred
to by their “nominal” temperatures, e.g. 600 °C means that the
setting on the furnace controller was 600 °C.
Figure 2. (a) Experimentally obtained temperature profiles along the
quartz reactor with a representative N flow through the system. (b)
Comparison of temperature profiles at 700 °C with 100% N and 100%
n-C 10 flow through a quartz reactor. (c) Expanded view of part b to
illustrate the deviation between the temperature profiles.
2
2
4
H
gases. General Air supplied all the other gases used. All inert
gases used for the GC were 99.999% pure. MKS Instruments’
mass flow controllers were used to control each of the two gas
streams to the reactor. National Instrument’s LabView program
was used to develop a graphical user interface (GUI) and control
the mass flow controllers. The inlet volumetric flow rate into
Measurement of Conversion and Major Products
The experimental setup was designed to monitor the major
volatile products of n-butane pyrolysis as well as the loss of
butane. Helium was used as both the carrier gas and reference
gas for the TCD to optimize detector sensitivity. However, this
choice makes the detection of hydrogen in the product stream
very difficult, since the thermal conductivity of hydrogen and
helium are both large relative to the thermal conductivity of
-
1
the reactor for all experimental conditions was 14.6 mL min .
This flow rate provided a nominal 5 s residence time through
the reactor at 550 °C and ambient pressure, with the assumption
that the temperature profile is uniform and the inlet feed mixture
behaves as an ideal gas.
A Hewlett-Packard 5890 Series II+ GC with two thermal
conductivity detectors (TCD) was used to perform quantitative
analysis of the species emerging from the reactor. Two 10-port
sampling valves were used to introduce the samples into two
packed columns. Each of the sampling loop volumes was 1.0
21
other gases. Although the hydrogen sensitivity could have been
improved by using argon, this would have decreased sensitivity
for the other species, and we did not pursue this approach.
Comparison of the data for mixtures without steam to that
of the steam case requires accounting for the difference in the
fuel-steam experimental procedure. As pointed out earlier, this
1
mL. A 6 ft × /8 in. stainless steel column was packed with
1
HayeSep R, and the second packed column was a 30 ft × /8

Importance of Gas-Phase Kinetics
J. Phys. Chem. A, Vol. 108, No. 17, 2004 3775
product stream was passed through a water trap to remove the
water to protect the GC columns. Thus, it is misleading to
directly compare the measured GC mole fractions of the steam
mixture to those of the mixtures without steam. However, it is
straightforward to make the corresponding comparisons to model
calculations by first subtracting the predicted steam mole fraction
and then renormalizing the mole fractions of the remaining
species. These can then be compared directly to the GC mole
fractions measured for the steam case.
Before data were collected for a given condition, GC
measurements verified that the reactor had reached steady-state
operation. The effluent stream mole fractions were then
measured in several repeat runs. The typical relative error in
the replicate GC measurements was observed to be less that
any reaction that proceeds via formation of a new chemical
bond. In the context of free radical chemistry, this means that
one must account for the pressure dependence of radical addition
to unsaturated molecules (and the reverse reaction, â-scission
to form an unsaturated molecule and a radical) and insertion
reactions as well as radical recombinations. Another complica-
tion in analyzing reactions proceeding via energized adducts is
the possibility that additional dissociation channels to new
products will be available. Thus, radical addition, recombination,
and insertion reactions can manifest very complex temperature
and pressure behavior as the stabilization channel competes with
these multiple reaction channels. Substantial errors can result
when experimental measurements of these rate coefficients are
extrapolated to other regions of temperature and pressure without
accounting for the competition between unimolecular reactions
of the energized adducts and their bimolecular collisional
stabilization. For the mechanism used in this work, we estimated
these pressure dependencies using the methods described by
5%. Prior to the start of an experiment, the GC sensitivity was
checked by measurements on room-temperature butane; the
deviations from the mean were less than 5%.
The presence of unidentified peaks in the chromatograms was
observed under some high-temperature experimental conditions.
The unidentified peaks that were observed had retention times
that are indicative of either an unsaturated C3 or C4 species.
However, the peak areas of these were negligible compared to
the observed C3H6 and C4H10. Assuming that the response
factors for the unidentified peaks were comparable to those
measured for C3H6, C3H8, and C4H10, the unidentified peaks
accounted for less than 1% of the total mole fraction. Similarly,
there was no significant measurable mole fraction of any
molecular weight growth species. The sum of the mole fractions
for the major products observed, i.e. xCH + xC H + xC H +
25
Chang et al. The current mechanism consists of approximately
2
00 pressure-dependent reactions.
All reactions in the mechanism are treated as reversible, using
the computed equilibrium constant at the specified temperature.
The thermodynamic properties for the species in the mechanism
are obtained from the literature and/or estimation techniques,
2
6,27
such as group additivity
and hydrogen bond increments for
2
8
radicals. No adjustments were made to the rate constants to
improve the fits to the experimental data. The mechanism is in
a CHEMKIN format with the rate parameters A, n and Ea
expressed in the modified Arrhenius form, viz.
4
2
4
2 6
xC H , and that of the unreacted butane was generally measured
3
6
n
to be close to 100%, within experimental error. This suggests
that the unmeasured H2 mole fraction is less than 5%.
k ) AT exp(-E /RT)
a
The mechanism and the thermodynamic database are included
in the Supporting Information.
Reaction Mechanism and Thermodynamic Properties
The reaction mechanism that was utilized in this study
consists of 291 species and 2498 elementary reactions. It is an
Plug-Flow Computations
22
extension of a previous C4 mechanism. The extensions include
substantially more molecular weight growth chemistry, espe-
cially involving propargyl radical pathways, and explicit inclu-
sion of certain C6 hydrocarbon species, including n-hexane,
cyclohexane, and 2,3-dimethyl butane. The current mechanism
The calculations were performed using the gas-phase kinetic
model with the PLUG driver program in the CHEMKIN
29
Collection v. 3.6.2 suite programs. The flow conditions in our
experiments are expected to be similar to those in a SOFC anode
channel. The temperature profiles used for the modeling are
included in the Supporting Information. The other input
parameters that are required by the PLUG subroutine were
obtained directly from experimental conditions, i.e., molar
composition, initial velocity/volumetric flow rate, geometric
configuration of the reactor, etc.
1
7
is similar to the one recently used by Walters et al. in their
analysis of methane kinetics under SOFC conditions. The major
difference is that we now have updated the butane dissociation
reaction
C H S 2C H
5
4
10
2
Experimental Results
to reflect the recent measurement of the reverse reaction by
Shafir et al.23
The experimental results are summarized in Table 1. Under
all circumstances, there were only four major products ob-
served: methane, ethane, ethylene, and propylene. Note that,
within experimental error, the sum of the measured mole
fractions of these species, plus unreacted butane is generally
close to 100%, suggesting that the unmeasured hydrogen plus
any other hydrocarbons constitute less than 5% of the total
effluent. The higher sum observed at 700 °C suggests a small
systematic error in these measurements.
Abstraction reaction rate constants in the mechanism are
based on literature data when available. If not, estimates based
2
4
on the methodology described in detail by Dean and Bozzelli
are used. In this approach a reference reaction is selected as a
model for abstraction by a specific radical, and this is used as
the basis for assignment of the A factor (and preexponential
temperature dependence) for that radical. An Evans-Polanyi
analysis is used to estimate the energy of activation (Ea). The
current mechanism includes abstraction reactions by H, O, OH,
O2, and selected hydrocarbon radicals.
Kineticists have long known that it is necessary to consider
the pressure dependence of the rate coefficients of dissociation
and recombination reactions. Similar considerations apply to
5
50 °C. Very little conversion was observed at 550 °C, with
only 1-2% conversion for both neat butane and butane/steam
mixtures. In both mixtures we also see that the major products
are methane and propylene, while smaller amounts of ethane
and ethylene are formed, with no discernible differences between
the two mixtures.

3776 J. Phys. Chem. A, Vol. 108, No. 17, 2004
Sheng and Dean
TABLE 1: Predicted and Observed Species Mole Fractions (in %) at the Effluent of the 21 cm Length Quartz Reactor for the
Three Mixtures at the Nominal Temperaturesc
+
a
- b
inlet fuel
N
2
H
2
O
CH
4
C
2
H
4
C
2
H
6
C
H
3 6
n-C
4
H
10
H
2
C
5
C
4
Σx
i
5
50 °C
0.4
0.4 ( 0.1 0.3 ( 0.1
n-C
4
H
10
0.8
.6 ( 0.1
0.4
0.8
97.6
0.0 4.8E-05 1.7E-02
0.0 9.9E-06 5.9E-04
0
0.8 ( 0.1 98.8 ( 5.0
100.9 ( 5.0
100.2 ( 5.0
n-C
4
H10:H
2
O (0.5:0.5)
49.8
49.1
0.3
0.5)
0.1
(0.3)
0.1
(0.3)
0.3
(0.5)
49.4
(98.4)
(
0
0.3
.6 ( 0.1
0.4 ( 0.1 0.3 ( 0.1
0.7 ( 0.1 98.2 ( 5.0
0.3
n-C
n-C
n-C
4
4
4
H
H
H
10:N
2
(0.51:0.49) 49.8
0.1
0.1
49.4
0.0 9.7E-06 7.2E-03
0.1 8.1E-04 0.1
6
00 °C
10
3.1
1.7
1.6
3.1
90.3
2.4 ( 0.4
1.6 ( 0.2 1.1 ( 0.1
2.7 ( 0.3 93.6 ( 4.7
101.4 ( 4.7
10:H
2
O (0.5:0.5)
1.2
2.3)
0.7
(1.4)
0.6
(1.2)
1.2
(2.3)
47.2
(92.6)
0.1 2.0E-04 3.1E-02
(
1
.7 ( 0.3
1.2 ( 0.2 0.7 ( 0.1
0.7 0.6
1.0 ( 0.1 0.4 ( 0.1
50 °C
4.5
4.6 ( 0.8 2.2 ( 0.2
2.0 ( 0.2 93.7 ( 4.7
1.2 47.2
1.4 ( 0.2 48.2 ( 3.0
99.3 ( 4.7
99.9 ( 4.3
n-C
4
H10:N
2
(0.51:0.49) 49.1
1.2
1.8 ( 0.3
0.1 1.9E-04 3.7E-02
47.1 ( 3.0
6
n-C
H
4 10
8.3
7
5.0
8.2
73.3
0.5 8.9E-03 0.3
0.4 2.9E-03 0.1
.1 ( 1.0
6.5 ( 0.5 75.1 ( 4.0
95.5 ( 4.2
4 2
n-C H10:H O (0.5:0.5)
46.9
3.7
6.9)
2.4
1.8
(3.5)
3.6
(6.8)
41.1
(77.4)
(
(4.5)
5
.5 ( 0.8
4.1 ( 0.8 1.4 ( 0.1
2.3 1.9
3.8 ( 0.7 1.3 ( 0.1
00 °C
9.3
2.4 ( 1.5 15.7 ( 1.0 5.0 ( 0.5 16.4 ( 0.7 45.5 ( 3.0
5.5 ( 0.5 80.3 ( 4.0
3.6 41.1
4.5 ( 0.5 41.5 ( 3.0
96.8 ( 4.2
99.4 ( 4.4
n-C
4
H10:N
2
(0.51:0.49) 47.0
4.4 ( 3.0
3.7
3.9 ( 0.6
0.3 2.7E-03 0.1
4
7
n-C
H
4 10
16.9
2
11.5
16.1
43.8
1.6 8.2E-02 0.9
1.5 3.9E-02 0.5
105.0 ( 3.6
4 2
n-C H10:H O (0.5:0.5)
42.0
9.1
15.7)
6.9
4.3
(7.4)
8.5
(14.7)
27.2
(46.8)
(
(12.0)
2
5.8 ( 1.5 20.3 ( 1.5 5.0 ( 0.5 18.3 ( 0.7 35.0 ( 3.0
9.1 6.6 4.7 8.6 27.2
14.7 ( 1.0 10.2 ( 0.8 2.5 ( 0.3 9.7 ( 0.6 27.1 ( 3.0
50 °C
104.4 ( 3.8
105.3 ( 4.5
n-C
4
H10:N
2
(0.51:0.49) 42.1
1.1 ( 3.0
1.1 3.7E-02 0.5
4
7
n-C
4
H
10
24.9
19.7
12.1
20.7
16.2
4.2 0.5
4.0 0.3
1.9
1.2
n-C
4
H10:H
2
O (0.5:0.5)
35.5 15.8
14.2
6.0
12.4
10.7
(24.5)
(22.0)
(9.2)
(19.2)
(16.5)
n-C
n-C
n-C
4
4
4
H
H
H
10:N
2
(0.51:0.49) 35.8
15.8
13.4
6.8
12.7
10.7
3.3 0.3
7.3 1.5
6.6 1.2
1.2
2.6
1.9
8
00 °C
10
29.5
26.2
10.5
18.1
4.2
4.3 ( 0.4
36.3 ( 2.6 39.8 ( 2.5 5.6 ( 0.5 13.0 ( 0.7
99.0 ( 3.7
10:H
2
O (0.5:0.5)
31.4 20.3
29.7)
20.1
(29.3)
5.1
(7.4)
10.9
(15.9)
2.5
(3.7)
(
3
6.6 ( 2.6 42.3 ( 2.5 3.8 ( 0.4 11.2 ( 0.7
20.4 19.2 6.0 11.2
8.4 ( 0.6
4.1 ( 0.4
98.0 ( 3.7
n-C
4
H10:N
2
(0.51:0.49) 31.7
2.5
2.9 ( 0.3
6.0 1.2
1.9
31.3 ( 3.0
28.5 ( 2.0 28.8 ( 2.0 3.2 ( 0.3
103.1 ( 4.2
a
+
denotes the sum of species having five or more carbons. b
-
C
6
5
C
4
denotes the sum of species with four or less carbons that are not CH
4
, C
2
H ,
4
c
C
2
H
, C
H
3 6
, and n-C
4
H
10
.
Cf. text for model input parameters. Observed species mole fractions are shown in bold italics. Normalized results for
/(100 - xH2O) × 100%.
predicted mole fractions for fuel-steam mixtures are shown in parentheses. Normalization is accomplished via xi,normalize ) x
i
6
00 °C. Experiments for all three different fuel inlet condi-
although the ethylene yield is comparable for the diluted nitrogen
case. The ethylene yield is now substantially higher than that
of ethane.
tions were performed at 600 °C. The observed conversion is
noticeably higher, ranging from 4 to 6%. For all the mixtures,
the product selectivities are quite similar, with methane and
propylene as the dominant products. Here the ethylene yield is
noticeably higher than that for ethane.
700 °C. There is a substantial increase in conversion and a
noticeable tilt toward methane as the dominant product. The
trend toward increasing ethylene yield continues; in both diluted
cases it is now slightly larger than the propylene yield. The
relative amount of ethane formed continues to decrease.
800 °C. At this temperature, virtually all the butane has been
reacted. The trend in reaction yields continues, with ethylene
now the dominant product for all the mixtures. Methane is
the species with the next highest yield and is now much
higher than propylene. Ethane remains the species with the
lowest yield.
6
50 °C. At 650 °C, the decrease in n-butane concentration
begins to become more significant. (With a greater loss of
reactant, combined with the fact that the number of moles
increases as a result of reaction, the lower mole fraction observed
for butane results both from actual loss of butane as well as
simple dilution. A good illustration of this is the observed drop
in the mole fraction of N2 at higher temperatures.) For all
mixtures, methane and propylene are still the dominant products,

Importance of Gas-Phase Kinetics
J. Phys. Chem. A, Vol. 108, No. 17, 2004 3777
Figure 3. 6 mm i.d. quartz reactor after 4 h of operation at 700 °C
4 4 10
with a nominal 5 s residence time: (a) 100% n-C H10; (b) 50% n-C H :
5
0% N ; (c) 50% n-C 10:50% H O.
2
4
H
2
Deposit Formation. One of the major concerns with the use
Figure 4. Comparison of modeling results to experimental data: (a)
of hydrocarbon fuels in SOFCs is the propensity for deposit
formation. In our experiments, qualitative observations of
deposit formation were made by examining the quartz reactor
after running a specific mixture through the reactor for 4 h.
This approach indicated that there was no apparent deposit
formation within the reactor for any of the mixtures at 650 °C
and lower temperatures. As the reactor temperature was
increased to 700 °C, with all other experimental conditions the
same as at 650 °C, an obvious deposit on the inner walls of the
quartz reactor was observed with the neat butane mixture, cf.
Figure 3a. Less deposit was observed with the nC4/N2 mixture,
cf. Figure 3b. Interestingly, deposit formation was negligible
with the nC4/H2O mixture, cf. Figure 3c. The first inference
from these qualitative observations is that dilution does play a
role in minimizing deposit formation, as seen in comparing
Figure 3a and 3b. This is to be expected, since the lower
concentration of reactive species would be expected to decrease
the rate of molecular weight growth leading to deposits.
However, a qualitative comparison of the two diluted fuel cases,
i.e. the nC4/H2O and the nC4/N2, indicates a significant
difference in the amount of deposit formed. This suggests that
the steam is either actively participating in inhibition of
molecular weight growth or that it is participating in some type
of gasification of the deposit. All three mixtures show significant
deposit formation on the inner reactor walls at 800 °C.
neat n-butane and 50% n-C
Cf. text for normalization procedure used in part b.
4 2 4 2
H10/50% N ; (b) 50% n-C H10/50% H O.
excellent agreement with the observations for both the neat and
the nC4/N2 mixtures. The predictions clearly capture the
observed temperature dependence for both the neat and diluted
mixture cases. From both the model predictions and the observed
experimental data, it is also clear that the conversion for the
neat n-butane case is higher than the nC4/N2 case. This
difference is simply due to dilution of the butane for the nC4/
N2 mixture. This good agreement, achieved without any
modifications to the gas-phase kinetic mechanism, was very
encouraging and suggests that the model can account for
conversion of light hydrocarbons fuels under SOFC conditions.
The comparisons of the ratio of the final to initial n-butane
mole fractions for the butane-steam mixture are shown in Figure
4
b. As described earlier, the measured mole fractions exclude
the water that was condensed prior to introduction to the GC.
Thus, the modeling results were “normalized” by removing H2O
from the product profiles to allow for a direct comparison to
that measured. With the notable exception of the comparison
at 700 °C, the fit is very good; again capturing the strong
observed temperature dependence. The 700 °C case is quite
interesting; this was also the one case where no deposits were
observed. These observations might be connectedsperhaps both
are due to heterogeneous reactions in the presence of steam,
perhaps on the quartz surface or the presence of contaminants
(especially trace levels of metals) on the quartz surface. The
modeling results for the nC4/N2 mixture and the nC4/H2O
mixture results were identical. This indicates that steam is not
Comparison To Model Predictions
Butane Conversion. The ratio of the predicted conversion
of butane is compared to that observed for the neat and nC4/N2
mixture cases in Figure 4a. The modeling predictions are in

3778 J. Phys. Chem. A, Vol. 108, No. 17, 2004
Sheng and Dean
Figure 5. Comparisons between model predictions and experimental data for the selectivity for the major products from n-butane pyrolysis study.
Selectivity (%) ) x
/(xCH4 + xC2H4 + xC2H6 + xC3H6) × 100%.
i
participating to any significant extent in the gas-phase kinetics,
even though the gas-phase reactions such as
model. The propylene selectivity comparisons are shown in
Figure 5d. The model predictions are in generally good
agreement with the experimental data, with the exception that
only a portion of the sharp decline at 800 °C is predicted. Any
difference for the different mixtures is well within the experi-
mental error.
H O + n-C H S n-C H + OH
2
4
9
4
10
are included in the mechanism. (In the mechanism, this reactions
and others like it are written in the reverse direction).
Deposit Formation. Since the model contains a reasonably
complete description of gas-phase molecular weight growth, we
used these results to attempt to connect to the observed deposit
Product Selectivity. The predicted product distributions are
included in Table 1, and these are compared to the experimental
observations in Figure 5. For ease of comparison, the distribu-
tions are shown in terms of product selectivity. Figure 5a
compares the predicted and observed temperature dependence
of the selectivity to methane. Overall the agreement is encour-
aging, with the predictions capturing the slight increase with
temperature and showing little effect of mixture composition.
The ethylene selectivity comparisons are shown in Figure 5b.
Although the model predicts selectivities slightly lower than
observed, it clearly captures the sharp upswing between 700
and 800 °C. Both model and data show little difference in
selectivity among the three mixtures. The ethane selectivity,
shown in Figure 5c does not show a good match between the
model and experiment. The predicted ethane selectivity is too
high and, although it does predict the observed decline with
temperature, the predicted decline becomes large only above
formation. In particular, we considered all species above C4
+
(
which will be denoted as C5 ) as potential deposit precursors.
+
These results are included in Table 1. The predicted C5 values
for the steam and nitrogen dilution cases are very similar. This
is consistent with our observation that any gas-phase reactions
of steam are very slow under these conditions, and thus steam
is simply acting as a diluent, just like nitrogen. This is in contrast
to the experimental observations that steam diminished deposit
formation. Our modeling results suggest that this cannot be
explained by gas-phase reactions, and is probably due to
+
heterogeneous reactions. Note that the ratio of C5 predicted
for the neat butane case to that of the diluted mixtures decreases
from approximately five at 550 °C to only slightly greater than
unity at 800 °C. Thus, at the higher temperatures, the predictions
suggest that the inhibition of deposit formation by dilution is
much less effective than at lower temperatures.
700 °C, while the data demonstrate a more gradual decline
throughout the temperature range. It is interesting to note that
the small shift in selectivity with mixture composition, with the
highest selectivity for the neat butane case, is predicted by the
Figure 6 shows the predicted effect of temperature on
formation of C5 for the various mixtures. The dramatic increase
in deposit precursors near 700 °C is consistent with the observed
+

Importance of Gas-Phase Kinetics
J. Phys. Chem. A, Vol. 108, No. 17, 2004 3779
interest. Since CH4 is a product of R1, this is not surprising.
Another reason for the importance of this reaction is that the
other product, the secondary butyl radical, will undergo a rapid
â-scission to form CH3 + propylene, and this methyl radical
will form methane via hydrogen abstraction. The second most
positive coefficient is ethyl abstracting the secondary hydrogen
in butane (R3), which also forms a secondary butyl radical.
Lower positive sensitivities are calculated for the dissociation
reactions R5 and R6. In terms of inhibiting formation of
methane, note that all of these reactions (R7, R8, and R10) form
the allyl radical. Allyl is resonantly stabilized, and thus is quite
unreactive. Note that R7 inhibits methane formation even though
it forms a methyl radical. R8 and R10 have the effect of
replacing an active radical, either CH3 or C2H5, with allyl. Even
though the total number of radicals remains the same, the net
activity drops substantially.
Figure 6. Modeling results showing increase in mole fraction of C +
species with temperature.
5
The results for ethylene are shown in Figure 7b. These
illustrate the very important role that abstraction from butane
plays in determining the product distribution. The most sensitive
reactions for ethylene production are the abstractions from
butane by CH3 and C2H5. The two most positive contributions
to the formation of ethylene are R2 and R4, which produce
n-C4H9, while the most negative contributions are R1 and R3,
which produce s-C4H9. R2 and R4 favor ethylene production
since the n-C4H9 will undergo rapid â-scission to form C2H5 +
C2H4, a direct channel for ethylene formation. R1 and R3 inhibit
ethylene formation since the secondary butyl radical formed by
these abstractions is the primary pathway to the complementary
products methane and propylene. Thus, one might view methane
and propylene production as a marker for production of the
secondary butyl radical and ethane and ethylene production as
the corresponding marker for production of the n-butyl radical.
The sensitivities of the dissociation reactions R5 and R6 are
comparable to that seen for methane.
onset of deposits at this temperature, as shown in Figure 3. Also
consistent with the experimental observations, the predictions
indicate a lower temperature threshold for the neat butane case.
+
The major C5 species formed in all mixtures at 700 °C is
cyclopentene, with 1,3-cyclopentadiene having the next highest
mole fraction. Both are indeed likely deposit precursors, in that
they are stable unsaturated ring compounds, susceptible to
radical addition reactions leading to even more molecular weight
growth. Furthermore, 1,3-cyclopentadiene has an unusually weak
C-H bond, allowing relatively facile formation of the resonantly
stabilized cyclopentadienyl radical. There are several routes
leading from this species to higher molecular weight molecules.
Other Products. Table 1 also includes the predicted mole
-
fractions of H2 and the sum of minor C4 species, i.e., other
than those specifically listed in the table. Note that production
of H2 increases substantially with increasing temperature. At
The ethane sensitivity is shown in Figure 7c. The four most
important reactions are precisely those reported for ethylene,
confirming the importance of the rapid â-scission reactions for
the butyl radicals. The main difference here is that the ethyl
radical formed by â-scission of n-C4H9 must first abstract an H
atom from a stable hydrocarbon to form ethane. For this case,
the dissociation reaction producing ethyl radicals (R6) is more
important than R5, which does not produce C2 species. The new
reaction identified in the sensitivity analysis is the â-scission
of C2H5 to form ethylene (R9 running in reverse). This pathway
competes with the ethyl abstraction pathway, diminishing the
ethane yield and increasing that of ethylene. Increasing this rate
coefficient is one way to try to improve the overprediction of
ethane (and underprediction of ethylene) in the model.
8
00 °C, this mole fraction is ∼7%, suggesting that the sum of
mole fractions of the measured hydrocarbons (Table 1) might
be slightly too large at 800 °C. As discussed earlier, the
measured sum at 700 °C is likely slightly high. The computed
mole fractions of C4 species are generally consistent with the
experimental observations that such species are only observed
in trace amounts.
-
Sensitivity Analysis
As seen above, the predictions from the unadjusted kinetic
model are in reasonable agreement with the observed data.
However, this is obviously a very large mechanism and a
sensitivity analysis was performed to determine whether some
smaller subset of the reactions were responsible for much of
the observed kinetics. The sensitivity analyses were performed
using the SENKIN driver program from the CHEMKIN
Collection v. 3.6.2 suite programs using the experimental
The sensitivity analysis of propylene is shown in Figure 7d.
It is not surprising at this stage to see that the two reactions
with the highest positive sensitivity are R1 and R3, similar to
methane (cf. Figure 7a). The secondary butyl radical will
undergo â-scission to form propylene plus methyl. As seen
above, the next set of reactions are the two dissociation reactions
R5 and R6. We again see the inhibiting effect of formation of
allyl radicals (R7 and R8). Formation of the primary butyl
radical (R2) also inhibits propylene production, but interestingly
it becomes important at later times than allyl formation.
29
temperature profile at 700 °C.
The results from the sensitivity analysis are shown in Figure
, parts a-e, for the five species of interest, i.e. CH4, C2H4,
7
C2H6, C3H6, and n-C4H10. It is interesting to note that, although
the system consists of ∼2500 reactions, there are only 11
reactions whose rates significantly affect the formation or
consumption of the five species of interest. These reactions are
listed in Table 2.
Figure 7a shows the normalized sensitivity coefficients for
methane. The reaction with the largest positive coefficient is
R1. A positive sensitivity coefficient means that an increase in
the rate coefficient will lead to an increase in the species of
Figure 7e shows the sensitivity analysis for n-butane. Here
the sensitivity coefficients are smaller than for product forma-
tion. Since we are looking at consumption of butane, rather than
formation of products, a positive sensitivity coefficient for this
case means that an increase in the rate coefficient leads to a

3780 J. Phys. Chem. A, Vol. 108, No. 17, 2004
Sheng and Dean
TABLE 2: Rate Constant Parameters for the Most Sensitive Reactions for Formation and Consumption of Major Species in
n
3
n-Butane Pyrolysis, Where k ) AT exp(-E
a
/RT), with Units of mol cm s
reaction
A
n
E
a
(cal/mol)
k (1000 K)
•
R1
R2
R3
R4
R5
R6
R7
R8
R9
C
C
C
C
C
C
4
4
4
4
4
4
H
H
H
H
H
H
10 + CH
3
3
S CH
S CH
3
CH
CH
2
C HCH
3
+ CH
4
3.26E+06
4.89E+06
2.64E+04
3.96E+04
6.51E+20
1.04E+22
1.66E+80
7.89E+05
1.87
1.87
8908
10 630
9910
11 632
87 067
86 472
1.47E10
9.27E9
10 + CH
3
2
CH
C HCH
CH
+ CH
2
CH2• + CH
4
•
10 + C
10 + C
2
2
H
H
5
S CH
3
CH
CH
2
3
+ C
2
H
2
6
2.51
6.11E9
•
5
S CH
2
3
2
2
CH
2
+ C
H
6
2.51
3.85E9
•
10 S CH
3
CH
5
C H
2
3
-1.291
-1.737
-18.78
1.87
8.14E-3
8.05E-3
4.22E-1
5.64E9
10 S C
2
H
+ C
CH
S CH
+ H S C
+ C S CH
2
H
5
•
CH
2
dCH
2
CH
2
3
S CH
dCH
2
dCH
C H
2
C H
2
+ CH
4
3
110 973
8032
•
C
3
C
2
C
3
H
H
H
6
4
6
+ CH
3
2
2
2
+ CH
a
2
H
5
1.30E12
2.28E9
4.66E7
•
R10
R11
2
H
5
2
2
dCH
2
C H
2
+ C
H
2 6
6.34E+03
4.44E+11
2.51
-0.42
9034
12 438
•
CH
2
dCH
2
C H
2
+ C
H
4
S cyclopentene + H
a
Computed using Troe falloff parameters; value shown is for 0.8 atm.
Figure 7. Sensitivity Analysis for the five major species of interest in n-butane pyrolysis determined by SENKIN analysis at 700 °C: (a) CH
; (c) C ; (d) C ; (e) n-C
4
; (b)
C
2
H
4
H
2 6
H
3 6
4 10
H .
higher concentration of butane at a given time, i.e., the reaction
inhibits conversion. Thus, it is not surprising to see all of the
positive sensitivities involve reactions (R7, R8, and R10) that
form allyl radicals. Butane consumption is promoted by both
the abstraction reactions R1 and R2 as well as the dissociation
reactions R5 and R6.
The kinetic model predicted that cyclopentene is the major
deposit precursor. A rate analysis for this species revealed that

Importance of Gas-Phase Kinetics
one chemically activated reaction
J. Phys. Chem. A, Vol. 108, No. 17, 2004 3781
recent direct measurement of the reverse reaction by Shafir et
al., and that value is only slightly lower than that listed in
NIST for this reaction. As with R5, our rate coefficient is
0.7k∞, reflecting falloff effects. There is a substantial spread
in reported values for the unimolecular dissociation of 1-butene,
2
3
•
CH dCHC H + C H S cyclopentene + H (R11)
2
2
2
4
∼
is responsible for most of the cyclopentene formed. This reaction
represents a combination of events. Initially a linear unsaturated
radical is formed by addition. This energized radical has
sufficient energy to form the cyclopentyl radical by radical
addition to the double bond. This species (still energized) then
ejects a hydrogen atom via â-scission to form cyclopentene.
The sensitivity analysis showed that cyclopentene production
is most influenced by this rate coefficient. The other reactions
that influence formation of this deposit precursor are R2, R5,
and R6. R3 and R7 inhibit cyclopentene production. Thus, even
formation of the deposit precursors are dominated by the same
small subset of reactions. R2 produces n-C4H9, the primary
precursor of the reactant ethylene, while the butane dissociation
reactions R5 and R6 increase the radical pool. The fact that R7
is the main inhibiting reaction, even though it produces the
reactant allyl radical, indicates the importance of maintaining
an active radical pool for molecular weight growth; as discussed
above, formation of allyl, relative to more active radicals like
H, CH3 and C2H5 inhibits the overall reactivity of the system,
including molecular weight growth. R3 inhibits growth since
this reaction consumes C2H5, thus lowering the formation rate
of ethylene and H atoms via â-scission of ethyl.
3
0,32-34
R7.
The value that we used is close to the average of
those listed; at 1000 K, it is close to the high-pressure limit.
The much higher rate coefficient for R7, relative to R5 and R6,
is due to the substantially weaker C-C bond in 1-butene due
to formation of the resonantly stabilized allyl radical.
The â-scission of the ethyl radical is described in our
mechanism in terms of the reverse reaction R9. Our value (and
the pressure dependence) was taken directly from the GRI 2.11
3
5
mechanism. This reaction is well into the falloff region. The
GRI value of k∞ is ∼70% of the value suggested by Baulch et
3
6
al.
R8 and R10 are abstractions from propylene. Both of these
rate coefficients are in good agreement with the recommenda-
tions of Tsang that are included in the NIST database.3 There
are again no direct measurements of R11. This rate coefficient
estimate was based on a chemical activation analysis and is
approximately 2.5 times larger than that reported in NIST. The
NIST value is based on fitting to a complex mechanism.
In summary, the most sensitive reactions in our unadjusted
mechanism have rate coefficients that are reasonably consistent
with literature values. We are encouraged by this comparison,
but clearly recognize that the overall good agreement between
our predictions and the butane pyrolysis data cannot be taken
as evidence that each of our rate coefficient assignments is
correct. Such an assertion requires explicit measurements of
much less complex systems. The comparison to the available
literature data suggests that our abstraction reaction rate coef-
ficients are generally slightly higher. As discussed earlier, our
abstraction rate coefficients are based on an extensive analysis
of systems where reliable data are available. We have systemati-
cally extended these values to analogous systems, e.g., abstrac-
tion by ethyl, using standard thermochemical kinetics principles.
In particular, we have taken care to properly account for the
constraints imposed by microscopic reversibility to ensure that
forward and reverse estimates are thermodynamically consistent.
We have also explicitly accounted for the non-Arrhenius
behavior expected for abstraction reactions in the temperature
range of this study.
0,37
Rate Coefficient Comparison To Literature Values
Given that both butane conversion as well as formation of
the major and minor products are primarily governed by the
rates of only 11 reactions, the next step is to compare our values
for these rate coefficients with those available in the literature.
Earlier we made the point that it was possible to obtain a
reasonable agreement between our model and the data without
making any adjustments to the model. It is equally important
to demonstrate that the rate coefficients that we used for these
reactions are reasonable. In Table 2, we list the rate coefficients
in our kinetic model for the 11 reactions discussed above. We
will compare these expressions, evaluated at 1000 K, to those
that are available from the literature, e.g., to values that are
30
compiled in the NIST kinetics database.
For R1, there are no direct measurements listed at any
temperature in the NIST kinetics database. An earlier compila-
tion31 reports several values measured at lower temperatures.
Kinetics Overview
7
3
-1
-1
The average measured value at 500 K is 2.6 × 10 cm mol
-1
7
3
It appears that butane pyrolysis can be accurately described
in terms of a straightforward Rice-Herzfeld mechanism. After
initiation by breaking the C-C bonds in butane, the radicals
formed abstract from the parent to form n-C4H9 and s-C4H9
radicals. These radicals can then react as follows:
s . The coefficient used in this work is 4.6 × 10 cm mol
-1
s
at 500 K, in reasonable agreement with the measurements.
A similar situation exists for R2; there are no measurements
reported in the NIST database. The CRC tabulation reports 4.6
6
3
-1 -1
×
10 cm mol
s
at 500 K, while the value in our model is
7
3
-1 -1
1
.2 × 10 cm mol
R3 and R4 involve abstraction by ethyl radicals. NIST reports
a value for the total abstraction rate coefficient from butane in
s
at 500 K.
9
3
-1 -1
our temperature range, which is 5 × 10 cm mol
s
at 1000
K. Our combined value, including abstraction for both n-C4H9
1
0
3
-1 -1
and s-C4H9, is 1 × 10 cm mol s . We consider this to be
reasonable agreement.
Comparisons of the dissociation rate coefficients (R5, R6,
and R7) must take into account pressure-dependent falloff
effects. For R5, our value is just beginning to enter the falloff
regime, with k/k∞ ) 0.72 at 1000 K. The value we used for k∞
for R5 is approximately three times lower than the estimates in
the NIST database, but it is in good agreement with the recently
measured value of R(-6). The k∞ value for R6 is based on the
The primary radicals lead to the C2 products while the secondary
radicals lead to the C1 and C3 products. The increase in C2
products with increasing temperature, and the corresponding
decrease in production of methane and propylene, is due to the
higher activation energies of the abstraction reactions that form

3782 J. Phys. Chem. A, Vol. 108, No. 17, 2004
Sheng and Dean
the n-butyl radical. Thus, the selectivity to these two product
pairs can be traced to the selectivity of forming the two butyl
radicals. An interesting complication is that the propylene
product has an easily abstractable hydrogen, and the formation
of allyl acts to inhibit conversion by modifying the nature of
the free radical pool. This same allyl radical was also shown to
be responsible for the small amounts of molecular weight growth
observed since it can add to ethylene, forming cyclopentene
and H. This pathway is accelerated since it is chemically
activated, and the initially energized linear adduct can directly
form the cyclic species prior to collisional stabilization. An
essential component of this scheme is â-scission of the butyl
radicals, although these reactions do not exhibit high rate
coefficient sensitivities. The reason for this is that these
â-scission reactions are so fast at these temperatures that they
are not rate-limiting.
Acknowledgment. The authors would like to thank ITN
Energy Systems, Inc. for the use of their facilities, where the
experimental work was conducted under DARPA Contract
MDA972-01-C-0068. The authors appreciate Katie O’Gara’s
contributions to the experiment and data compilation. We also
would like to thank Hans-Heinrich Carstensen and Chitral Naik
for providing some of the elementary reaction rates used in the
current model. Chitral Naik also performed some of the
modeling calculations.
Supporting Information Available: The mechanism and
thermodynamic database, in CHEMKIN format, and tables of
the temperature profiles used for the modeling. This material
is available free of charge via the Internet at http://pubs.acs.org.
References and Notes
(
1) Brouwer, J. Solid-oxide Fuel Cell - Emergence of a Technology.
Implication of Results To SOFC Studies
Fuel Cell Catalyst 2000, 1, 1.
(
(
(
(
2) Zhu, H.; Kee, R. J. J. Power Sources 2003, 117, 61-74.
3) Wang, X.; Gorte, R. J. Catal. Lett. 2001, 73, 15-19.
4) Borowiecki, T. Appl. Catal. 1982, 4, 233.
These results show that it is important to take into account
the homogeneous reactions of hydrocarbon fuels in high
temperature fuel cells. In general, these kinetics can have major
impact into two distinct areas: (1) The species that will undergo
electrochemical reaction might be far removed from the parent
fuel delivered to the fuel cell. In practice, even bigger changes
in the nature of the species undergoing electrochemical reaction
will be expected since there will likely be catalytic reactions
with the metal in the porous anode as the species are being
transported to the three-phase boundary region. (2) These gas-
phase reactions have been shown to lead to molecular weight
growth, perhaps leading to deposit formation. An additional
experimental observation is that it appears that even a quartz/
steam system is enough to inhibit deposit formation. One might
expect this effect to be even more evident in a real fuel cell
with the anode metal acting as a catalyst for this reaction.
5) Borowiecki, T.; Golebiowski, A. Catal. Lett. 1994, 25, 309.
(6) Borowiecki, T.; Golebiowski, A.; Stasinska, B. Appl. Catal. A 1997,
53, 141.
7) Kepinski, L.; Stasinska, B.; Borowiecki, T. Carbon 2000, 38, 1845-
856.
1
(
1
(
(
8) Trimm, D. L. Catal. Today 1997, 37, 233.
9) Tracz, E.; Scholz, R.; Borowiecki, T. Appl. Catal. 1990, 66, 133.
(10) Trimm, D. L. Catal. Today 1999, 49, 3.
(11) Hickman, D. A.; Haupfear, E. A.; Schmidt, L. D. Catal. Lett. 1993,
7, 223.
12) Hickman, D. A.; Schmidt, L. D. Science 1993, 259, 343.
13) Hickman, D. A.; Schmidt, L. D. AIChE J. 1993, 39, 1164.
14) Otsua, K.; Wang, Y.; Sunada, E.; Yamanaka, I. J. Catal. 1998,
1
(
(
(
175, 152.
(
15) Tsang, S. C.; Claridge, J. B.; Green, M. L. H. Catal. Today 1995,
3, 3.
16) Krummenacher, J. J.; West, K. N.; Schmidt, L. D. J. Catal. 2003,
15, 332-343.
(17) Walters, K. M.; Dean, A. M.; Zhu, H.; Kee, R. J. J. Power Sources
003, 123, 182-189.
18) Hepp, H. J.; Frey, F. E. Ind. Eng. Chem 1953, 45, 410.
2
2
2
(
Summary
To explore this possibility of gas-phase chemistry in the anode
channel of a SOFC, experiments using n-butane as a representa-
tive hydrocarbon fuel were conducted under SOFC conditions.
Butane conversion and product formation were monitored in
quartz reactor experiments at P ∼ 0.8 atm, τ ∼ 5 s, and T )
(
(19) Weisel, M. D.; Chang, A. Y.; Zhong, X.; Dean, A. M. The Pyrolysis
and Oxidation of N-Butane & Isobutane: A Flow Tube Reactor Molecular
Beam Mass Spectrometry Study. Presented at the Fourth International
Conference on Chemical Kinetics, 1997, Gaithersburg, MD.
20) Mallinson, R. G.; Braun, R. L.; Westbrook, C. K.; Burnham, A.
K. Ind. Eng. Chem. Res. 1992, 31, 37-44.
(
5
50-800 °C. Three different fuel mixtures were used: neat
n-butane, 50% n-C4H10/50% H2O and 50% n-C4H10/50% N2.
These experiments demonstrate that substantial gas-phase
chemistry does occur, and that this must be accounted for when
predicting fuel cell efficiency. These data were compared to
predictions using a plug-flow model that incorporated the
experimentally measured temperature profile along the reactor.
Comparisons of the model predictions to the experimental data
show that the model, without any modifications, captures the
observed strong temperature dependence of n-butane conversion
and is also able to capture the changes in product selectivity
with temperature for the neat butane and the diluted butane
mixtures. The model also properly predicts the observed onset
of deposit formation near 700 °C. Both conversion and
selectivity are shown to be sensitive to only a very small subset
of the reactions in the mechanism. Comparison of the rate
coefficients of this subset to literature values, where available,
are generally reasonable and suggest that the kinetic model
employed is adequate for describing reactions of small hydro-
carbons in the anode channels of a SOFC. Thus, such models
should provide a useful tool to estimate the importance of gas-
phase chemistry for various proposed SOFC operating condi-
tions.
(21) CRC Handbook of Chemistry and Physics, 63rd. ed.; Weast, Robert,
C., Ed.; CRC Press: Boca Raton, FL, 1982.
(
22) Mims, C. A.; Mauti, R.; Dean, A. M.; Rose, K. D. J. Phys. Chem.
994, 98, 13357-13372.
23) Shafir, E. V.; Slagle, I. R.; Knyazev, V. D. J. Phys. Chem. A 2003,
107, 6804-6813.
24) Dean, A. M.; Bozzelli, J. W. In Gas-Phase Combustion Chemistry;
Gardiner, W. C., Ed.; Springer: Berlin, 2000; pp 125-341.
25) Chang, A. Y.; Bozzelli, J. W.; Dean, A. M. Z. Phys. Chem. 2000,
1
(
(
(
1
533-1568.
(26) Benson, S. W. Thermochemical Kinetics, 2nd ed.; Wiley-Inter-
science: New York, 1976.
(
(
27) Ritter, E.; Bozzelli, J. W. Int. J. Chem. Kinet. 1991, 23, 767-778.
28) Lay, T. H.; Bozzelli, J. W.; Dean, A. M.; Ritter, E. R. J. Phys.
Chem. 1995, 99, 14514-14527.
29) Kee, R. J.; Rupley, F. M.; Miller, J. A.; Coltrin, M. E.; Grcar, J.
F.; Meeks, E.; Moffat, H. K.; Lutz, A. E.; Dixon-Lewis, G.; Smooke, M.
D.; Warnatz, J.; Evans, G. H.; Larson, R. S.; Mitchell, R. E.; Petzold, L.
R.; Reynolds, W. C.; Caracotsios, M.; Stewart, W. E.; Glarborg, P.; Wang,
C.; Adigun, O. CHEMKIN Collection. [Release 3.6.2]. 2000. San Diego,
CA, Reaction Design.
(30) NIST Chemical Kinetics Database. Release 1.1, 2000. http://
kinetics.nist.gov/.
(
(
31) CRC Handbook of Bimolecular and Termolecular Gas Reactions;
Kerr, J. A., Moss, S. J. Eds.; CRC Press: Boca Raton, FL, 1981, Vol. 1
32) Dean, A. M. J. Phys. Chem. 1985, 89, 4600.
(33) Halstead, M. P.; Quinn, C. P. Trans. Faraday Soc. 1968, 64, 103.
(

Importance of Gas-Phase Kinetics
J. Phys. Chem. A, Vol. 108, No. 17, 2004 3783
(
34) Trenwith, A. B. Trans. Faraday Soc. 1970, 66, 2805-2811.
(36) Baulch, D. L.; Cobos, C. J.; Cox, R. A.; Frank, P.;.Hayman, G.;
Just, Th.; Kerr, J. A.; Murrells, T.; Pilling, M. J.; Troe, J.; Walker, R. W.;
Warnatz, J. J. Phys. Chem. Ref. Data 1994, 23, 847-1033.
(35) Bowman, C. T.; Hanson, R. K.; Davidson, D. F.; Gardiner, W. C.,
Jr.; Lissianski, V.; Smith, G. P.; Golden, D. M.; Frenklach, M.; Goldenber,
M. GRI-Mech 2.11. 4-16-1996. http://www.me.berkeley.edu/gri_mech/.
(37) Tsang, W. J. Phys. Chem. Ref. Data 1991, 20, 221-273.