In situ kinetics and mechanism of furan decomposition and desorption with CO formation on Pd(111)

Article abstract of DOI:10.1021/jp984783a

The work presented here uses kinetic studies under conditions of varying temperature, coverage, and isotopic substitution to elucidate the energetics and mechanism of furan decomposition and desorption on Pd(111). Laser-induced thermal desorption with Fourier transform mass spectrometry detection is used as a sensitive tool for measuring the in situ kinetics of furan decomposition simultaneously with the formation of CO. The kinetics of both furan loss and CO formation have been studied for 0.17 and 0.32 Langmuir exposures of furan on Pd(111) at 100 K. This corresponds to about 2 and 4% of a monolayer and approximately 15 and 30% of a saturation monolayer, respectively. Total observed furan loss follows two processes involving a competition between desorption and decomposition. The kinetic factors for each process are extracted yielding an activation barrier of approximately 60 kJ/mol and preexponential factor of 10⁷ s^-1 for decomposition, and an activation barrier of approximately 95 kJ/mol and preexponential factor of 10¹⁴ s^-1 for desorption. The parameters for decomposition show a substantial coverage dependence at significantly less than a saturation monolayer, whereas those for desorption are relatively independent of coverage. The kinetic parameters for CO formation match those for the decomposition of furan, indicating that the rate for both reactions is limited in the same processes. Results from coverage effects and isotopic labeling experiments suggest that the rate-determining step is not simple α-C-H bond cleavage.

Full text of DOI:10.1021/jp984783a

J. Phys. Chem. B 1999, 103, 7869-7875
7869
In Situ Kinetics and Mechanism of Furan Decomposition and Desorption with CO
Formation on Pd(111)
Tracy E. Caldwell† and Donald P. Land*
Department of Chemistry, UniVersity of California, DaVis, California 95616
ReceiVed: December 17, 1998; In Final Form: May 4, 1999
The work presented here uses kinetic studies under conditions of varying temperature, coverage, and isotopic
substitution to elucidate the energetics and mechanism of furan decomposition and desorption on Pd(111).
Laser-induced thermal desorption with Fourier transform mass spectrometry detection is used as a sensitive
tool for measuring the in situ kinetics of furan decomposition simultaneously with the formation of CO. The
kinetics of both furan loss and CO formation have been studied for 0.17 and 0.32 Langmuir exposures of
furan on Pd(111) at 100 K. This corresponds to about 2 and 4% of a monolayer and approximately 15 and
30% of a saturation monolayer, respectively. Total observed furan loss follows two processes involving a
competition between desorption and decomposition. The kinetic factors for each process are extracted yielding
an activation barrier of approximately 60 kJ/mol and preexponential factor of 107 s-1 for decomposition, and
an activation barrier of approximately 95 kJ/mol and preexponential factor of 1014 s-1 for desorption. The
parameters for decomposition show a substantial coverage dependence at significantly less than a saturation
monolayer, whereas those for desorption are relatively independent of coverage. The kinetic parameters for
CO formation match those for the decomposition of furan, indicating that the rate for both reactions is limited
in the same processes. Results from coverage effects and isotopic labeling experiments suggest that the rate-
determining step is not simple R-C-H bond cleavage.
Introduction
benzene proceeds via the coupling of two C3H3 fragments.6
From the standpoint of heteroatom removal, it is worth noting
Furan is often used in hydrodeoxygenation (HDO) studies
as a model for understanding the interactions between oxygen-
containing aromatic hydrocarbons and the catalytic surface of
transition metals. Heteroaromatic compounds, including the
sulfur- and nitrogen-containing analogues, are prevalent in crude
petroleum and in liquids derived from coal and biomass.1,2 Their
removal presents a challenge to the petrochemical industries
interested in converting alternative feedstocks into useful
chemicals.3,4 The harmful effects of these compounds (corrosion
of engine parts, degradation of catalysts, and contamination of
the environment) have made their removal an issue of impor-
tance. Late transition metals, such as Co or Ni, have been shown
to behave as promoters in the removal process, increasing
catalytic activity by about an order of magnitude.5 However,
little is known about the surface chemistry of heterocycles on
late transition metals, and even less is understood about the
effects of Pd, in particular.
that oxygen is efficiently removed from the Pd(111) surface in
the form of CO. In contrast to these Pd results, other late
transition metals, such as clean Cu(110)9 and Ag(110),10 have
been shown to adsorb furan reversibly with no other desorption
products observed during temperature-programmed reaction
(TPR). However, furan coadsorbed with oxygen on Ag(110)
has been shown to produce CO2 and H2O, both of which are
evolved starting at about 300 K during TPR.11 At higher
temperatures (520 K), very small amounts of partial oxidation
products, such as maleic acid, benzene, and bifuran, are
evolved.11 The Pd surface used in our study does not require
this highly oxidative environment to catalyze rapid deoxygen-
ation of furan at 300 K. Our results also show, however, that
significant amounts of carbonaceous material may be left behind
on the Pd surface. These species may act to poison further
catalytic reactivity unless removed, e.g., by hydrogen supplied
to the feed.
Previous studies by this group, using laser-induced thermal
desorption (LITD), revealed interesting and unique features in
the mechanism of furan decomposition on Pd(111), as illustrated
in Scheme 1.6,7 These results have been independently confirmed
in a subsequent article by Ormerod et al.8 The mechanism is
shown to proceed at 300 K via elimination of R-H and CO,
leaving a C3H3 species on the surface. Heating to 350 K causes
dimerization of some of the C3 species, forming benzene. This
implies that the C3H3 species are stable between 300 and 350K.
Deuterium labeling studies confirm that the formation of
For supported catalysts, Ni and Co are more effective
than Zn in the promotion of the most common Mo and W
HDO catalysts,12 particularly for hydrodesulfurization (HDS)
activity.13-16 Thus, it is reasonable to suggest that other group
VII and VIII metals, such as Pd, would also serve as effective
promoters for hydrotreating catalysts. Furthermore, Pd sulfides
have been shown to catalyze HDS more than an order of
magnitude more efficiently than sulfides of Co and Ni.17 It seems
reasonable to suggest then, on the basis of the comparison of
activities mentioned above, that Pd/W or Pd/Mo would have
improved activities as well, perhaps surpassing the quality of
the synergic contacts used most often today (i.e., Ni- and Co-
promoted sulfides of W and Mo). To the best of our knowledge,
* To whom correspondence should be addressed.
† NASA, Johnson Space Center, Houston, TX.
10.1021/jp984783a CCC: $18.00 © 1999 American Chemical Society
Published on Web 08/27/1999

7870 J. Phys. Chem. B, Vol. 103, No. 37, 1999
Caldwell and Land
SCHEME 1: Overview of Furan Decomposition Mechanism on Pd(111). Depiction of the Reaction Mechanism and Fate
of Products When Furan Is Heated on Pd(111). H2(g) and CO(g) Are Liberated during the Reaction, while Most of the
Carbon Remains on the Surface.
though, investigations of Pd promotion of W or Mo sulfides
have not yet been published.
LITD with Fourier transform mass spectrometry (FTMS) can
be used to monitor the surface composition prior to conventional
desorption, and is useful in determining the mechanistic and
kinetic details surrounding surface processes which are desorp-
tion-rate limited, i.e., where product evolution into the gas phase
would normally follow the kinetics of desorption. LITD uses a
focused, pulsed laser beam to generate a thermal spike (1000
K in 5 ns) over a 1 mm diameter spot on a surface. This
phenomenal heating rate (1011 K/s) can access high-energy
reaction pathways and typically results in desorption of species
adsorbed under the laser beam before they have a chance to
react.22-24 Many species which would react on the surface under
slow heating conditions can be desorbed intact under rapid laser
heating. FTMS is an ideal detection technique, because it is
inherently pulsed and results in a complete (typically 10-800
Dalton), high-resolution mass spectrum25-28 for each single laser
pulse.29-31
In this paper we have pursued a more thorough understanding
of the unique and unexpected reaction mechanism of furan
decomposition, namely in the kinetics of the initial stage leading
to CO formation. The results presented here not only help to
elucidate the mechanism, but also lend valuable insight into the
nature of the interactions of furan with the Pd surface. Moreover,
the conclusions drawn here provide the basis for an interesting
comparison between the reactivity of furan, thiophene (C4H4S),
and pyrrole (C4H4NH) on Pd(111).
Experimental Section
The experiments are performed in an ultrahigh-vacuum
(UHV) chamber (2 × 10-10 Torr) equipped with low-energy
electron diffraction, Auger electron spectroscopy, ion sputtering,
and adsorption/desorption measurements utilizing both conven-
tional and laser-induced heating methods. Details of the ap-
paratus were published elsewhere.18,19 The Pd(111) crystal
surface is cleaned by repeated cycles of Ar+ bombardment and
by heating in oxygen. Furan is purchased from Aldrich Chemical
Co. (99+%) and is further purified by several freeze-pump-
thaw cycles. The Pd sample is cooled to 100 K and is then
dosed with furan by backfilling the chamber. Gas-phase mass
spectra of furan during dosing shows that no benzene or other
contaminants are present. Exposures (Langmuir ) L ) 10-6
Torr‚s) have been corrected for ion gauge sensitivities. The
sensitivity factor for furan (3.29) is estimated using the value
listed for tetrahydrofuran, and the value for CO was assumed
to be unity.20
Kinetics experiments are performed as follows. The Pd
sample temperature is rapidly increased and then held constant
at a temperature at which the kinetics of the reaction can be
easily monitored. LITD mass spectra are obtained immediately
after the desired temperature has been reached. Ideally, the
sample temperature should be ramped at a sufficient rate so
that the extent of reaction is small for the first laser shots. The
laser is fired at a rate fast enough to accurately monitor the
rapid changes in surface composition caused by reaction. Over
time, the reaction slows and spectra can be taken less frequently.
Many more shots are needed in the beginning to fully model
the reaction rate laws and so that initial rate data can be used
to determine activation energies in the event that a reaction order
cannot be determined in the subsequent data analysis. The
reaction is allowed to proceed for up to a full hour, and is
monitored every 5-10 min to produce a clear indication that
the reaction has ceased. The maximum data acquisition rate is
one laser shot and mass spectrum per second. These time scales
define the range of rates accessible, and therefore the temper-
ature range.
For the kinetics experiments, corrected exposures of 0.17 and
0.32 L were used. These correspond to total fluxes of (MW )
68, T(gas) ) 300 K) 4.2 × 1013 and 7.9 × 1013 molecules/cm2.
Using a lattice parameter for Pd(111) of 2.276 Å,21 the surface
15
density is 2.23 × 10
atoms/cm2. Thus, these exposures,
assuming unit sticking probability, yield furan coverages of
about 2 and 4% of a monolayer, respectively. On Ag(110), the
saturation furan coverage corresponds to each molecule covering
an area of 26 Å2.10 This compares well to the area of the van
der Waals perimeter for furan, which is roughly a rectangle
measuring 6 by 5 Å,10 such that maximum packing would yield
3.3 × 1014 molecules/cm2 or 15% of a monolayer on Pd(111).
Thus, both of the coverages used here should be significantly
below a saturation monolayer. Note, however, that the reliability
of ionization gauges and the uncertainties involved with
pressure/flux gradients in the UHV chamber could affect the
accuracy of our exposure values by more than a factor of 2.
Further evidence supporting our claim that these are significantly
less than a saturation furan coverage comes from our thermal
desorption spectroscopy (TDS) experiments. Exposures as high
as 1.0 L produced no new peaks or any discernible evidence of
multilayer coverage.
Error bars in parameters calculated from the data are
determined by Kaleidagraph (Synergy Software, Reading, PA)
using a Marquardt algorithm.32 Global weighting factors are
applied to transformed data (such as wi ) yi2 for logarithmic
plots) and relative weighting factors (wi ) 1/σi2) are applied
for all data according to the uncertainties obtained in previous
calculations.
Results and Discussion
CO Formation Kinetics. The time-dependent signals for both
reactant (furan) and product (CO) are studied in this work. The
two processes, loss of reactant and formation of product, will
be addressed separately and then compared. Figure 1 shows the
mass intensities for both furan and CO plotted as a function of

Furan Decomposition on Pd(111)
J. Phys. Chem. B, Vol. 103, No. 37, 1999 7871
[Z]∞ is the concentration of product after the reaction has gone
to completion. The experimental data for the time evolution of
the furan signal is fit to eq 1, using both [X]0 and k as adjustable
parameters, and the data for the CO signal is fit to eq 2 using
both [Z]∞ and k as adjustable parameters. The curves in Figure
1 represent these first-order exponential fits. It is obvious from
the plots in this figure that the simple first-order assumption is
unsatisfactory. Figure 1(a) shows considerable deviation in the
furan signal at later time. Figure 1(b) is an expansion of the
region to the left of the arrow in Figure 1(a) with a fit to just
the initial data shown, and indicates that even the early stages
of the reaction do not follow the simple first-order curve. This
is not too surprising if one considers that, at these temperatures,
the desorption of furan is competing with the decomposition
(this topic will be addressed in the following section on Furan
Loss Kinetics). In addition, the buildup of products (CO, H,
and presumably C3H3) resulting from the decomposition could
significantly change the nature of the surface as a function of
time.
Although the kinetics are clearly not simple first-order, the
activation barrier for CO formation can be determined without
any knowledge of the details of the mechanism. Assuming an
arbitrary irreversible reaction,
xX + yY f Z
(3)
the initial rate of reaction (κi) equals the initial rate of formation
of Z and is simply the reaction rate constant (k) multiplied by
the initial concentrations of reactants, each raised to a power
appropriate for the reaction mechanism and kinetic dependence:
n
m
κi ) k[X]0 ‚[Y]0
(4)
Figure 1. Furan and CO concentrations vs time at 293 K. LITD/FTMS
signal magnitudes for furan (base peak is m/z 39) and CO (m/z 28) are
plotted as a function of time under isothermal conditions at 293 K for
a 0.17 L initial exposure of furan on Pd(111) at 100 K. This plot
illustrates the simultaneous loss of furan and formation of CO. The
curves are first-order exponentials, from which the data clearly deviate.
Plot (b) is an expansion of the region to the left of the arrow in (a) and
indicates that even the initial stages of the reaction do not follow simple
first order.
where [X]0 is the initial concentration of X, and the exponents
n and m may or may not be equal to the stoichiometric
coefficients x and y depending on the details of the reaction
mechanism. Substituting for
k ) Ae-E/RT
(5)
where A is the Arrhenius preexponential factor, E is the
activation barrier for the process, R is the gas constant, and T
is the absolute temperature, and taking the natural logarithm of
both sides gives:
time under isothermal conditions for a 0.17 L exposure of furan.
Mass 39 is the largest peak in the mass spectrum for furan and
is therefore followed to monitor the furan signal. However, all
masses are obtained simultaneously. All masses attributed to a
single component are checked to ensure that the rate of change
is similar. While CO is present in the background gases of the
UHV chamber, LITD with a pulsed electron beam timed to
coincide with the laser pulse can discriminate between back-
ground and laser-desorbed CO. Specifically, the electron beam
parameters are adjusted so that no CO is observed in the absence
of a desorption laser pulse. Figure 1(a) shows clearly the decay
in the furan signal simultaneously with the growth of CO. The
data were initially fit assuming each process followed simple
first-order kinetics. For a reaction converting species X into the
species Z, the loss of reactant follows
m
ln(κi) ) {ln(A) + ln([X]0n ) + ln([Y]0 )} - E/RT (6)
The three terms in brackets on the right in eq 6 should be
essentially independent of temperature for a small temperature
range. One must also assume that the reaction follows van’t
Hoff-Arrhenius behavior and that the exponents in the rate law
do not change significantly over the temperature range and
coverage range used. Thus, a plot of ln(κi) vs 1/T should yield
a straight line with a slope of -E/R. One can then calculate the
activation energy without having to make any assumptions about
the mechanism of the reaction. (However, the intercept can no
longer be used to obtain the preexponential factor unless the
details of the kinetic dependence and initial concentrations are
known.) Figure 2 is a plot of the first 25% of the data points
for a few of the isothermal kinetic experiments monitoring the
formation of CO from an initial exposure of 0.17 L of furan on
Pd(111). This plot illustrates that there is truly a temperature
dependence on the initial rate of reaction. An Arrhenius plot of
CO formation from 0.17 L exposure of C4H4O is shown in
Figure 3. From the slope of this weighted least-squares linear
regression, the activation energy was determined to be
[X]t ) [X]0 e-kt
(1)
and the evolution of product formation follows
[Z]t ) [Z]∞(1 - e-kt
)
(2)
where [X]t is the concentration of reactant at any time, t, [X]0
is the initial reactant concentration, k is the first-order rate
constant, [Z]t is the concentration of product at any time, t, and

7872 J. Phys. Chem. B, Vol. 103, No. 37, 1999
Caldwell and Land
Figure 4. Coverage dependence of CO formation on initial furan
exposure. The method of initial rates was used to obtain the rate data
in this Arrhenius plot for 0.32 L exposure of C4H4O on Pd(111) at 100
K. The data are fit using a weighted least-squares linear regression to
obtain Ea ) 69 ( 3 kJ/mol.
Figure 2. Initial rates of CO formation at several temperatures. Plot
of the first 25% of the data points for a few of the isothermal kinetic
experiments monitoring the formation of CO (m/z 28) from an initial
exposure of 0.17 L furan on Pd(111) at 100 K. The data are fit to
straight lines. This plot illustrates the temperature dependence of the
initial rate of reaction.
same kinetic experiments were run using d4-furan, also shown
in Figure 3, and an activation energy of 79 ( 3 kJ/mol was
obtained. A simple KIE could only account for a maximum of
about 5 kJ/mol difference between the two activation energies
(assuming that C-H(D) stretching is the reaction coordinate),
and the difference observed here is about a factor of 5 greater
than this. A KIE that is too large is somewhat problematic. In
H/D isotope effects, tunneling is often suspected when one
observes a rate for the proteo species that is significantly faster
than expected. Even, for example, for H + H2 vs D + D2, the
correction for tunneling at 300 K is a factor of 1.5 and at 200
K is only a factor of 1.7 in the rate,36 whereas we see a factor
in excess of 2.0 beyond the standard KIE calculated from zero
point energies alone at 290 K. In addition, if tunneling were
important, curvature should be observed in the Arrhenius plot.37
This is not the case. These results indicate that simple C-H
bond breaking is not the sole rate-determining step in the
decomposition of furan to yield CO. Because all four hydrogens
were replaced, a mechanism that somehow involves all or most
of them could perhaps be operative. More likely, however, is
that the reaction rate is not determined by a single slow step,
but instead is a composite of several slow steps, each contribut-
ing to the overall rate. Isotopic substitution might then result in
slowing of more than one of those steps, in addition to possibly
a change in the relative contributions of the different steps. This
will be discussed in more detail after the presentation of
additional results.
Figure 4 is an Arrhenius plot of CO formation from C4H4O
on Pd(111) at a higher initial exposure of furan. Initial rates
were again used for these Arrhenius-type plots. Compare the
Ea for the 0.17 L exposure of C4H4O (shown previously in
Figure 3) to that obtained for a 0.32 L exposure. The activation
energy for CO formation increases from 55 ( 3 to 69 ( 3 kJ/
mol with only a 2-fold increase in furan exposure at low
coverage. The increased barrier might arise from site blocking
or a crowding-induced tilt which inhibits CO formation at higher
exposures. To further elucidate this mechanism, more detailed
analyses of the kinetics of furan decomposition, in conjunction
with that of CO formation, are performed.
Figure 3. The Arrhenius plot gives the activation energy for CO
formation from the decomposition of C4H4O and C4D4O, each obtained
through separate experiments using a 0.17 L exposure of the respective
furan species on Pd(111) at 100 K. From the slope of this weighted-
least squares linear regression fit, an Ea of 55 ( 3 kJ/mol was obtained
for the per-proteo species. An Ea of 79 ( 3 kJ/mol was obtained for
the per-deutero furan.
55 ( 3 kJ/mol. Going back to the original question of how this
reaction proceeds, if R-C-H bond breaking is truly the rate-
determining step, then substituting each hydrogen for a deute-
rium should give rise to a measurable isotope effect in the
kinetics.
According to transition state theory, a kinetic isotope effect
(KIE) can arise from changes in vibrational zero point energies
in the reactants and transition state, which is manifest in Ea shifts
and differences in preexponential factors upon deuteration.33-35
As explained previously, no preexponential factor can be
extracted from the initial rate treatment, however, any changes
in Ea will give a good indication of whether a KIE exists. The
Furan Loss Kinetics. Recall from Figure 1 that the data for
furan exhibited a pronounced deviation from first-order kinetics.
It was discussed that the deviation might be due to a competition
between furan desorption and decomposition. To account for
this, a modified first-order fit was used which expresses the
rate of adsorbed furan loss (observed) as the sum of two

Furan Decomposition on Pd(111)
J. Phys. Chem. B, Vol. 103, No. 37, 1999 7873
Figure 7 shows Arrhenius plots for just furan decomposition
from both a 0.17 and 0.32 L exposure of C4H4O. As expected,
the activation energy increases significantly with only a 2-fold
increase in furan exposure and, in addition to having similar
magnitudes, follows the same trend as in the CO formation rates.
A slight increase in preexponential factor is also observed for
furan decomposition, which may be the result of a compensation
effect.38-40 Briefly, a decrease in activation energy can be
accompanied by energetic funneling to lower states. The loss
in entropy due to this funneling of energy can compensate for
the decrease in Ea for that state, hence, lowering the preexpo-
nential factor as well. The coverage dependence of both routes
to furan loss is summarized in Table 1.
If one compares the Ea and A for decomposition of C4H4O
and C4D4O (separate experiments) using a 0.17 L initial
exposure, again, one sees that the Ea for the proteo-species is
less than that of the deutero-species, though the magnitude for
both Ea and A are much higher for the deutero species than
expected (or even what is reasonable) based on the CO data.
The kinetics still do not exhibit a simple isotope effect and this
inconsistency in the KIE still indicates that some factor other
than simple C-H bond breaking is determining the rate of CO
formation.
Figure 5. Loss of furan at 293 K using first-order kinetics incorporating
two competing rate processes. The plot above contains the same furan
loss data as that found in Figure 1 for isothermal reaction at 293 K of
a 0.17 L initial exposure of furan on Pd(111) at 100 K. The data in
this case are fit using a sum of two first-order kinetics processes,
representing competing loss pathways, namely desorption and decom-
position.
Also note that the values for the activation barrier obtained
for the decomposition process in the loss of C4H4O agree well
with what is observed for CO formation under the same
conditions of initial coverage. This is summarized in Table 2.
(Recall that, because of the multiplex nature of LITD/FTMS,
the signal magnitudes for both species, CO and C4H4O, were
obtained simultaneously so reaction conditions are identical in
both sets of data.) Because the activation energies for both
C4H4O decomposition and CO formation are identical, it is
reasonable to hypothesize that both processes are moderated
by the same rate-determining step. No comparison between
preexponential factors can be made, however, because initial
rates were used to obtain Ea for CO and preexponential factors
are not obtainable.
Mechanistic Implications. Three intramolecular bonds must
break to form CO, H, and C3H3 from furan: C-C, C-O, and
C-H. It is not always the weakest bond that breaks first in a
catalyzed decomposition reaction; however, if one of these bonds
were significantly stronger than the other two, one would be
tempted to deflect attention to the other two first. The C-C
bond that must break is essentially a double bond in the free
molecule (bond length of 1.361 Å).41 As such, the dissociation
energy for this bond (∼580 kJ/mol) would be much higher than
that for the C-O or C-H bond (each around 400 kJ/mol).
However, rehybridization upon adsorption or during reaction
could make the barrier to C-C bond scission comparable to
either C-O or C-H bond scission. Thus, it is not possible to
categorically eliminate any of these as being the first step on
the basis of bond dissociation energies. However, the measured
activation energy of around 60 kJ/mol is consistent with
dehydrogenation barriers, whereas it is rather low for most
catalytic C-C bond breaking processes.42 Relatively little data
exist regarding the barrier for relevant catalytic C-O bond
scission, particularly because this C-O bond is part of an
aromatic ring. In general, however, the reactivity of the five-
membered heterocycles is dominated by reaction at the R-hy-
drogen, and dehydrogenation at this position is probably the
initial step in the reaction.43
Figure 6. Arrhenius plot of furan loss kinetics using first-order kinetics
with two competing rate processes. The Arrhenius plot was obtained
for an initial furan exposure of 0.17 L. The two rate constants represent
the desorption (circles) and decomposition (squares) pathways for furan
on Pd(111).
competing rate processes:
d
-
[furan]ads ) (kdes + kdecomp) [furan]ads ) kobs [furan]ads
dt
(7)
where kdes is the rate constant for desorption, kdecomp is the rate
constant for decomposition, and kobs is the observed rate constant
for loss of adsorbed furan. With a first-order/two-parameter
equation, improvement in the fit is clear, as shown in Figure 5.
Figure 6 is an Arrhenius plot of the two rate constants
obtained at each temperature from this two-parameter fit using
an initial furan exposure of 0.17 L on Pd(111). Notice that the
plots contain information regarding the preexponential factor,
A, in addition to Ea. The values calculated in this manner are
reasonable. Desorption processes are typically characterized by
preexponential factors in the range 1013-1018 s-1, whereas the
preexponential factor for surface reaction not involving desorp-
tion should typically have A < 1011.
In addition, adsorbate-Pd bonds may also need to be broken
to achieve the correct orientation. Angle-resolved ultraviolet
photoelectron spectroscopy (ARUPS) and high resolution

7874 J. Phys. Chem. B, Vol. 103, No. 37, 1999
Caldwell and Land
nearly flat adsorption for thiophene.45 This difference is evident
in the very different reactivity of the C4H4X moiety. For furan
and thiophene, ring opening and heteroatom elimination (via
Pd-CO or Pd-S, respectively) occurs readily at room temper-
ature. For pyrrole, however, these processes do not occur until
about 475 K. This is consistent with decomposition of the five-
membered ring requiring a nearly flat-lying geometry, a
geometry much harder to achieve for the covalently bonded
pyrrolyl than for furan or thiophene. The coverage-dependent
studies of furan described in this article are likely to be a more
subtle manifestation of a similar effectsinterference with the
ability of furan to lie flat.
Conclusions
The kinetics of both furan loss and CO formation have been
studied for 0.17 and 0.32 L exposures of furan on Pd(111) at
100 K. This corresponds to about 2 and 4% of a monolayer
and approximately 15 and 30% of a saturation monolayer,
respectively. Furan loss appears to follow first-order kinetics,
involving a competition between desorption and decomposition.
Desorption is observed to have an activation energy of about
90 kJ/mol and a preexponential factor of 1014 s-1, neither of
which are particularly sensitive to initial furan coverage. Furan
loss via decomposition and CO product formation both have
activation energies of about 55 kJ/mol at the lowest exposure
and about 70 kJ/mol at twice that exposure. The similarity in
activation barriers for the two processes implies that the same
slow step (or steps) dictate the rate for both processes. From an
initial rate treatment of isotope- and coverage-dependent data,
the following have been concluded:
(a) A larger-than-expected increase in Ea upon deuteration
implies that a primary KIE is not the sole factor affecting the
rate, implying that R-C-H bond breaking is not the sole rate-
determining step; and (b) an increase in Ea with higher exposure
suggests that increasing coverage hinders the formation of CO,
the barrier to which could be due to reorientation required for
furan.
Figure 7. Coverage dependence of furan decomposition. The Arrhenius
plot contains the Ea and Α for furan decomposition from two different
initial exposures of C4H4O on Pd(111) at 100 K: 0.17 L (circles) and
0.32 L (squares). The values for Ea are very close to those measured
for CO formation.
TABLE 1: Coverage Dependence of Furan Desorption and
Decomposition
initial furan
exposure
rate constant for furan
decomposition(s-1
rate constant for furan
desorption (s-1
)
)
0.17 L
0.32 L
107(0.7e-[(57(4)kJ/mol]/RT
109(1e-[(64(8)kJ/mol]/RT
1015(4e-[(99(25)kJ/mol]/RT
1013(2e-[(92(14)kJ/mol]/RT
TABLE 2: Coverage Dependence of CO Formation and
Furan Decomposition
initial furan
exposure
CO formation
Ea (kJ/mol)
furan decomposition
Ea (kJ/mol)
0.17 L
0.32 L
55 ( 3
69 ( 3
57 ( 4
64 ( 8
electron energy loss spectroscopy (HREELS) studies suggest
that the molecular plane of the furan molecule tilts to a
significant degree on the Pd(111) surface at relatively high
exposures (6-10 L).8 This reorientation may also contribute to
the kinetics of both furan decomposition and CO formation.
Consider if furan must lie flat on the surface to react. Increasing
the amount of furan could hinder this process if active sites are
occupied as a result (either through site blocking or lateral
repulsion/attraction). Because we have measured the desorption
barrier for furan to be about 90 kJ/mol, it is unlikely that the
diffusion barrier is significantly greater than about 30 kJ/mol,
thus it would seem that reorientation would play a minor role,
at best, in the overall determination of the rate of reaction.
Comparison with Thiophene and Pyrrole Decomposition
on Pd(111). The sulfur and nitrogen analogues of furan,
thiophene (C4H4S),7,44,45 and pyrrole (C4H4NH)7,46 have also
been studied on Pd(111), although no detailed kinetics have yet
been performed. Thiophene reacts somewhat similarly to furan,
decomposing at about room temperature in a ring-opening
reaction. However, because of somewhat different energetics,
formation of Pd-S is favored over CS, so that the products at
300 K are adsorbed S and C4H4.7,44 Pyrrole, however, has been
shown to lose the amino hydrogen at low temperatures (∼200
K).7,46 It is thought that the remaining pyrrolyl (C4H4N) adsorbs
with the molecular plane nearly perpendicular to the surface,
because the surface-adsorbate bond here is a covalent one
between N and Pd. This is in contrast to furan and thiophene,
for which valence photoelectron spectroscopy shows significant
bonding to the surface through the π system for both species
on Pd(111), hence leading to a tilted geometry for furan8 and
Acknowledgment. The authors thank the donors of the
Petroleum Research Fund, administered by the American
Chemical Society, and the National Science Foundation under
Grant CHE-9612732 for the partial support of this research.
T.E.C. also thanks the Committee on Research of UC Davis
for a Graduate Research Award, and the Patricia Roberts Harris
Foundation for a Graduate Fellowship Award. The authors also
acknowledge I.M. Abdelrehim for his help acquiring data.
References and Notes
(1) Furimsky, E. Catal. ReV. Sci. Eng. 1983, 25, 421.
(2) Cadey, W. E.; Seelig, H. S. Ind. Eng. Chem. 1952, 44, 2636.
(3) Speight, J. G. The Desulfurization of HeaVy Oils and Residua;
Marcel Dekker: New York, 1981; Vol. 4.
(4) Kane, S. M.; Rufael, T. S.; Gland, J. L.; Huntley, D. R.; Fischer,
D. A. J. Phys. Chem. B 1997, 101, 8486.
(5) Chary, K. V. R.; Rao, K. S. R.; Muralidhar, G.; Rao, P. K. Carbon
1991, 29, 478.
(6) Caldwell, T. E.; Abdelrehim, I. M.; Land, D. P. J. Am. Chem. Soc.
1996, 118, 907.
(7) Caldwell, T. E.; Land, D. P. Polyhedron 1997, 16, 3197.
(8) Ormerod, R. M.; Baddeley, C. J.; Hardacre, C.; Lambert, R. M.
Surf. Sci. 1996, 360, 1.
(9) Sexton, B. A. Surf. Sci. 1985, 163, 99.
(10) Solomon, J. L.; Madix, R. J.; Sto¨hr, J. J. Chem. Phys. 1990, 94,
4012.
(11) Crew, W. W.; Madix, R. J. J. Am. Chem. Soc. 1993, 115, 729.
(12) Patzer, J. F.; Montagna, A. A. Ind. Eng. Chem., Process Des. DeV.
1980, 19, 382.
(13) Furimsky, E. AIChEJ. 1979, 25, 306.

Furan Decomposition on Pd(111)
J. Phys. Chem. B, Vol. 103, No. 37, 1999 7875
(14) Furimsky, E. Catal. ReV.sSci. Eng. 1980, 22, 371.
(15) Bussell, M. E.; Somorjai, G. A. Catal. Lett. 1989, 3, 1.
(16) Ratnasany, P.; Sivasanker, S. Catal. ReV.sSci. Eng. 1980, 22, 410.
(17) Krastev, E. T.; Voice, L. D.; Tobin, R. G. J. Appl. Phys. 1996, 79,
6865.
(18) Land, D. P.; Abdelrehim, I. M.; Thornburg, N. A.; Sloan, J. T.
Anal. Chim. Acta 1995, 307, 321.
(31) Land, D. P.; Pettiette-Hall, C. L.; Sander, D.; McIver, R. T., Jr.;
Hemminger, J. C. ReV. Sci. Instr. 1990, 61, 1674.
(32) Marquardt, D. W. J. Soc. Ind. Appl. Math. 1963, II, 431.
(33) Wolfsberg, M. Annu. ReV. Phys. Chem. 1969, 20, 449.
(34) Mak, C. H.; Brand, J. L.; Koehler, B. G.; George, S. M. Surf. Sci.
1987, 188, 312.
(35) Madix, R. J.; Telford, S. G. Surf. Sci. 1992, 277, 246.
(36) Shavitt, I. J. Chem. Phys. 1959, 31, 1359.
(19) Abdelrehim, I. M.; Thornburg, N. A.; Land, D. P. ReV. Sci. Instr.
1997, 68, 4572.
(20) Bartmess, J. E.; Georgiadis, R. M. Vacuum 1983, 33, 149.
(21) Ohtani, H.; van Hove, M. A.; Somorjai, G. A. Surf. Sci. 1987, 187,
372.
(22) Cowin, J. P.; Auerbach, D. J.; Becker, C.; Wharton, L. Surf. Sci.
1978, 78, 545.
(23) George, S. M.; DeSantolo, A. M.; Hall, R. B. Surf. Sci. 1985, 159,
L425.
(37) Steinfeld, J. I.; Francisco, J. S.; Hase, W. L. In Chemical Kinetics
and Dynamics; Prentice Hall: Englewood Cliffs, NJ, 1989; p 319.
(38) Conner, W. C., Jr. J. Catal. 1982, 78, 238.
(39) Fichthorn, K. A.; Weinberg, W. H. Langmuir 1991, 7, 2539.
(40) Galwey, A. K. AdV. Catal. 1977, 26, 247.
(41) Bak, B.; Christensen, D.; Dixon, W. B.; Henson-Nygaard, L.;
(24) Hall, R. B.; Bares, S. J. In Chemistry and Structure at Interfaces:
New Optical Probes; Hall, R. B., Ellis, A. B., Eds.; VCH: Deerfield Beach,
FL, 1986; p 85.
(25) Gross, M. L.; Rempel, D. L. Science 1984, 226, 261.
(26) Wilkins, C. L.; Chowdhury, A. K.; Nuwaysir, L. M.; Coates, M.
L. Mass Spectrom. ReV. 1989, 8, 67.
(27) Russell, D. H. Mass Spectrom. ReV. 1986, 5, 167.
(28) Marshall, A. G. Acc. Chem. Res. 1985, 18, 316.
(29) Land, D. P.; Pettiette-Hall, C. L.; Hemminger, J. C.; McIver, R.
T., Jr. Acc. Chem. Res. 1991, 24, 42.
Andersen, J. R.; Schottla¨nder, M. J. Mol. Spectrosc. 1962, 9, 124.
(42) Somorjai, G. A. Introduction to Surface Chemistry and Catalysis;
Wiley: New York, 1994.
(43) Streitwieser, A., Jr.; Heathcock, C. H. In Introduction to Organic
Chemistry, 2nd ed.; Macmillan Publishing Company: New York, 1981; p
1069.
(44) Caldwell, T. E.; Abdelrehim, I. M.; Land, D. P. Surf. Sci. 1996,
367, L26.
(45) Netzer, F. P.; Bertel, E.; Goldmann, A. Surf. Sci. 1988, 199, 87.
(30) Land, D. P.; Pettiette-Hall, C. L.; McIver, R. T., Jr.; Hemminger,
J. C. J. Am. Chem. Soc. 1989, 111, 5970.
(46) Baddeley, C. J.; Hardacre, C.; Ormerod, R. M.; Lambert, R. M.
Surf. Sci. 1996, 369, 1.