Oxidation of carbon monoxide over Au(1 1 0)-(1×2)

Article abstract of DOI:10.1016/j.susc.2004.06.067

We have studied the oxidation of carbon monoxide (CO) on an oxygen-precovered gold (1 1 0)-(1×2) surface. At low oxygen coverages, the reaction is first-order in the oxygen coverage and has an apparent activation energy of -1.8 kJ/mol. Assuming a Langmuir-Hinshelwood mechanism we evaluated a true activation energy of 57 kJ/mol. A complete reaction energy diagram of the CO oxidation on Au(1 1 0) is reported.

Full text of DOI:10.1016/j.susc.2004.06.067

Surface Science 566–568 (2004) 1112–1117
www.elsevier.com/locate/susc
Oxidation of carbon monoxide over Au(1 1 0)-(1 · 2)
*
J. Michael Gottfried , Klaus Christmann
Institut f €u r Chemie der Freien Universit €a t Berlin, Bereich Physikalische und Theoretische Chemie, Takustrasse 3,
Berlin D-14195, Germany
Available online 19 June 2004
Abstract
We have studied the oxidation of carbon monoxide (CO) on an oxygen-precovered gold (1 1 0)-(1 · 2) surface. At low
oxygen coverages, the reaction is first-order in the oxygen coverage and has an apparent activation energy of )1.8 kJ/
mol. Assuming a Langmuir–Hinshelwood mechanism we evaluated a true activation energy of 57 kJ/mol. A complete
reaction energy diagram of the CO oxidation on Au(1 1 0) is reported.
ꢀ
2004 Elsevier B.V. All rights reserved.
Keywords: Carbon monoxide; Carbon dioxide; Oxygen; Gold; Surface chemical reaction; Models of surface chemical reactions; Single
crystal surfaces
1
. Introduction
row. The chemical nature of the support (and tra-
ces of water) play a decisive role. It is assumed that
a positively polarized Au species on the junction of
A presently widely studied catalytic reaction is
the CO oxidation over oxide-supported gold par-
ticles [1–3]. Research is stimulated (a) by the sur-
prising fact that gold as the noblest metal shows
appreciable catalytic activity and (b) because of the
technically relevant low-temperature activity of
this type of catalysts. Despite the indefatigable re-
search efforts, the nature of the active species and
metal and oxide is essential for the O activation
2
([1] and references therein). On the other hand,
evidence has been obtained that negatively
ꢀ
dꢀ
charged, very small Au (or Au ) clusters catalyse
n
n
the CO oxidation even in the gas phase [4] or on
relatively inert substrates such as MgO or SiO
2
.
The activity depends strongly on the cluster size
and is highest for n ¼ 8. The support, however,
seems to play a minor role for this class of systems
[5,6]. The possible coexistence of several completely
different mechanisms for catalysis on (supported
and unsupported) gold particles makes the situa-
tion more complicated than originally thought, but
it may explain some contradictions in previous re-
sults (e.g. concerning the influence of reductive or
oxidative pretreatment on the activity).
the mechanism of the O activation are still con-
2
troversial. Apparently two major classes of systems
exist for which completely different reaction
mechanisms have been proposed. The one class of
systems includes Au particles of 2–5 nm diameter
supported on oxides of the first transition metal
*
Corresponding author. Tel.: +49-30-838-55377; fax: +49-
0-838-54792.
E-mail address: gottfr@zedat.fu-berlin.de (J. Michael Gott-
3
It has been pointed out that a further under-
standing of the nature of these gold-based catalysts
fried).
0
039-6028/$ - see front matter ꢀ 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2004.06.067

J. Michael Gottfried, K. Christmann / Surface Science 566–568 (2004) 1112–1117
1113
requires thorough quantitative kinetic measure-
ments [2]. While these are rare, there is also a lack
of UHV reference studies on gold single-crystal
surfaces. In one of the few previous investigations,
Outka and Madix [7] oxidized CO on an oxygen-
precovered Au(1 1 0) surface and evaluated an
apparent activation energy of (8 ± 4) kJ/mol for the
overall reaction. A true activation energy of the
surface reaction could not be determined because
the CO desorption energy remained unknown.
In one of our recent reports on oxygen
adsorption on Au(1 1 0)-(1 · 2) [8] we performed
reactive thermal desorption measurements of the
CO oxidation reaction and found a rate maximum
at 175 K. Above this temperature the rate has a
negative temperature coefficient. This result sug-
gests a negative apparent activation energy and is
thus at variance with the positive value reported by
Outka and Madix. To clarify this point we now
performed detailed kinetic measurements. The
data obtained in this and our recent publications
enable us to present a complete energy diagram of
the CO oxidation over Au(1 1 0).
Fig. 1. CO oxidation over O/Au(1 1 0)-(l · 2). CO
2
desorption
rate as a function of time for temperatures between 60 and 400
þ
K. Detected mass: m=z ¼ 44 (CO ). Inset: semi-logarithmic
2
plot.
2
. Experimental
all experiments the sample was heated to 450 K.
Thereafter, the sample temperature was set to a
constant value and the CO pressure was (almost)
A detailed description of our sample and its
ꢀ
6
preparation was given in Ref. [9], whereas the
experimental details of the kinetic measurements
can be found in Ref. [8].
instantaneously increased to 1 · 10 mbar. The
2
resulting CO desorption flux, which was mea-
sured with
a
mass spectrometer, decreases
The error of the pressure measurement amounts
to 20%. The different gauge sensitivities for the
particular gases were compensated by dividing the
approximately exponentially with time (Fig. 1).
First, we qualitatively consider the initial rates
(t ¼ 0): in the low-temperature range between 60
and 160 K, the initial rate increases with tempera-
ture due to the usual rise of the reaction rate con-
meter reading by the following factors: CO 1.1, O
0.8 (values recommended by the manufacturer of
the gauge).
2
stant k . Above 180 K, however, the rate decreases,
r
indicating a negative apparent activation energy.
The most likely reason is a decrease of the CO
equilibrium coverage with increasing temperature,
which is not compensated by the increase of kr.
Assuming a Langmuir–Hinshelwood mecha-
nism, which is strongly suggested by the behaviour
in the high-temperature range, we can derive a
mechanism that allows a quantitative analysis and
the determination of the true activation energy of
the surface reaction. The following mechanism is
proposed:
3
. Results and discussion
Fig. 1 displays the CO
2
desorption rate, RCO2 , as
a function of time for various fixed temperatures
between 60 and 400 K. In all measurements, the
following procedure was applied: chemisorbed
atomic oxygen (0.45 ML) was prepared by electron
bombardment of physisorbed O as described in
2
Refs. [9,10]. To ensure a uniform oxygen phase in

1114
J. Michael Gottfried, K. Christmann / Surface Science 566–568 (2004) 1112–1117
ꢀ
6
ka;O2
e.g., ꢂ 0.05 ML at 180 K and pCO ¼ 10 mbar
O
2
2
ðgÞ( +O
2
ðadÞ
ð1Þ
(
see Ref. [13]). Thus, we can neglect HCO in the
kd;O
2
denominator on the right hand side of Eq. (9).
Combining Eqs. (8) and (9) (simplified) under
elimination of HCO yields:
kdiss;O2
ðadÞ( +2OðadÞ
O
ð2Þ
ð3Þ
krec;O
dH
ka;COkr
kd;CO
O
R
CO2 ¼ ꢀ dt
¼
p
CO
H
O
ꢁ kexp
H
O
ð10Þ
ka;CO
COðgÞ( +COðadÞ
kd;CO
with the empirical rate constant kexp. Thus, the
mechanism in Eqs. (3)–(5) in combination with our
additional assumptions predicts a pseudo-first-
order reaction. Applying the Arrhenius equation
to the individual rate constants in Eq. (10) leads to
kr
COðadÞ þ OðadÞ ꢀ! CO
2
ðadÞ
ð4Þ
ð5Þ
kd;CO2
CO ðadÞ ꢀ! CO ðgÞ
2
2
the following expression for kexp
:
Eqs. (1) and (2) can be excluded from our con-
sideration, since O(ad) is not formed by sponta-
neous adsorption.
m
r
ꢀðErꢀEd;COÞ=RT
kexp ¼ ka;COpCO
e
ð11Þ
m
d;CO
with the CO desorption energy E , the activation
According to Eq. (5), the CO desorption rate is
2
d
energy of the oxidation step (Eq. (4)) E
respective frequency factors md;CO and m
shows that the apparent activation energy of the
overall reaction, E
r
, and the
. Eq. (11)
given by
r
dNCO2
RCO2
ꢁ
¼ kd;CO2 HCO2
ð6Þ
Adt
r
ꢀ Ed;CO, can be obtained from
exp
Above the CO desorption temperature on oxy-
2
the slope of a plot of ln k vs. 1=T.
gen-precovered gold, which is found to be ꢂ 120 K
In order to evaluate kexpðTÞ from the curves in
[
11,12], the CO
certainly very low and obeys the steady-state
2
equilibrium coverage, HCO , is
Fig. 1, we integrate the rate equation Eq. (10) and,
with this result, eliminate H from Eq. (10):
2
O
condition:
dH
O
ꢀ
kexpt
dHCO2
R
CO2 ¼ ꢀ dt ꢁ kexp
H
O;t¼0
e
ð12Þ
¼
ꢀkd;CO2
H
CO2 þ k
r
H
CO
H
O
ffi 0
ð7Þ
ð8Þ
dt
with the initial oxygen coverage H
Thus, kexp can be evaluated from the curves in
Fig. 1 by the relation:
.
O;t¼0
Eliminating HCO2 from Eqs. (6) and (7) yields:
dH
O
R
CO2 ¼ k
r
H
CO
H
O
¼ ꢀ dt
d lnRCO2
With an excess of gas phase CO, we can now
k
exp ¼ ꢀ
:
ð13Þ
dt
also eliminate HCO if we assume that the CO
adsorption/desorption equilibrium is only mini-
mally perturbed by the oxidation reaction, an
assumption that is well-justified [11]. Thus, the CO
coverage is given by the Langmuir isotherm, at
least for a small coverage range in which the
isosteric heat remains practically constant:
Accordiꢀng to Eqsꢁ. (11) and (13), the slope of a
d lnRCO
2
plot of ln ꢀ
vs. 1=T is ꢀðE
r
ꢀ Ed;COÞ=R,
dt
i.e., it is equal to the apparent overall reaction
activation energy. Such a plot for temperatures
between 200 and 400 K is displayed in Fig. 2
and yields ðE ꢀ Ed;COÞ ¼ ðꢀ1:8 Æ 0:9Þ kJ/mol, in
r
agreement with the negative temperature coeffi-
cient of the rate above ꢂ160 K (see Fig. 1).
k
a;CO
H
CO
K ꢁ k
¼ p
COð1 ꢀ HCO
Þ
d;CO
Determination of the true activation energy, E ,
r
H
CO
ꢂ p
for
H
CO ꢃ 1
ð9Þ
requires knowledge of the CO desorption energy,
CO
E
d;CO. In a recent paper [13], we evaluated the
isosteric heat of CO adsorption at zero coverage,
st ¼ ð59 Æ 2Þ kJ/mol, which is, for a mobile
At the given CO pressure and temperatures
above 160 K, the equilibrium CO coverage is low,
Q

J. Michael Gottfried, K. Christmann / Surface Science 566–568 (2004) 1112–1117
1115
O
solution contains an additional term in which H is
proportional to time t. Assuming steady-state
conditions for CO(ad) we obtain:
dHCO
¼
k
a;CO
pCOð1 ꢀ HCOÞ ꢀ kd;CO
H
CO
dt
ꢀ
k
r
H
CO
H
O
¼ 0
ð14Þ
Herein, we still assume a constant kd;CO, which
requires a coverage-independent CO desorption
energy, and a constant ka;CO, which requires that
the CO sticking coefficient is independent of tem-
perature.
With Eq. (14), we eliminate HCO from Eq. (8)
and obtain instead of Eq. (10):
dH
O
RCO2 ¼ ꢀ dt
ka;COkrpCO
¼ k
H
O
a;COp
CO
þ kd;CO þ k H
r
O
Fig. 2. Plot of lnðꢀd lnRCO2 =dtÞ vs. 1=T.
ꢁ kexpHO
ð15Þ
Eq. (15) reveals that the experimental rate con-
stant, kexp, is actually not independent of the oxy-
adsorbate layer, equal to the CO binding energy
11]. Since CO adsorption is non-activated (as
gen coverage H . Integration yields:
[
O
shown in Ref. [13]) we can identify the binding
energy with the desorption energy and thus obtain
HO
kr
ln
þ k
ðH
d;CO
O
ꢀ HO;t¼0Þ
H
þ k
O;t¼0
a;CO
a;CO
a value of E ¼ ð57 Æ 3Þ kJ/mol for the true acti-
r
k
r
k
p
CO
vation energy. This value is higher than the acti-
vation energy for the gas phase reaction between
CO and oxygen atoms, which is only 15 kJ/mol
¼ ꢀ k
t
ð16Þ
a;CO
p
CO þ kd;CO
The second term on the left hand side of Eq.
(16) leads to a deviation from the first-order
in the
[
r
14]. The difference between E and the gas phase
value can be interpreted as the energy expense for
the rehybridization of the adsorbate-substrate
bonds of CO and oxygen in the activated complex.
A closer evaluation of the curves in Fig. 1 re-
veals that the plots of ln RCO2 vs. 1=T actually
deviate somewhat from linearity, especially in the
low-temperature range below 180 K (Fig. 1, inset).
In the following, we consider several possible
reasons for these deviations which partly reflect
the simplifications we applied for deriving Eqs.
kinetics. The occurrence of ka;CO þ k
d;CO
denominator indicates that these deviations in-
crease when the CO adsorption and desorption
rates are small, i.e., at low-temperatures and low
CO pressures (as can be seen in Fig. 1, inset).
(b) The CO sticking coefficient SCO and the
desorption rate constant kd;CO may depend on
oxygen coverage H , which decreases during the
reaction.
O
(c) The Langmuir isotherm, which was used in
Eqs. (9) and (14), is only valid for very CO cov-
erages (e.g. at elevated temperatures) because the
isosteric heat of CO is coverage-dependent (see
Ref. [13]) and, thus, kd;CO itself is a function of the
CO coverage. Furthermore, ka;CO contains the CO
sticking coefficient SCO. If SCO is a function of
temperature, the temperature dependence of kexp is
not restricted to the exponential term in Eq. (11),
(11) and (13), and critically discuss further implicit
assumptions of our model.
(
a) We assumed that HCO / p (Eq. (9)), which
CO
is only valid for small HCO. Furthermore, we as-
sumed that the CO adsorption-desorption equi-
librium is not perturbed by the reaction of CO with
chemisorbed oxygen. Now we give up these sim-
plifications and demonstrate that the ‘complete’

1116
J. Michael Gottfried, K. Christmann / Surface Science 566–568 (2004) 1112–1117
Fig. 3. Quantitative energy diagram of the CO oxidation over Au(1 1 0)-(l · 2). All values concerning dioxygen adsorption are mul-
tiplied with 1/2 to account for the stoichiometry of the reaction which requires only 1/2 O for the oxidation of one CO molecule. All
dissociation energy of 498 kJ/mol) were taken from
2
values (apart from the standard gas phase reaction energy of 285 kJ/mol and the O
this and our recent publications.
2
which was implicitly assumed when we derived the
apparent activation energy from the plot in Fig. 2.
4. Energetics and kinetics of the CO oxidation on
gold
(
d) The oxygen phase may be less homogeneous
than the oxygen thermal desorption data in Ref.
10] suggest, leading to a variety of oxygen
In our recent publications, we extensively
studied the energetics and kinetics of the adsorp-
[
adsorption states with different reactivity towards
CO. This idea is supported by the UPS measure-
ments described in Ref. [11] and by the apparent
contradiction between the relatively high activa-
2
tion of oxygen [9,10,15], CO [13], and CO [11,12]
on Au(1 1 0). The respective data, complemented
by the activation energies as evaluated in this
work, enable us to report a complete energy dia-
gram (Fig. 3) of the CO oxidation on Au(1 1 0). It
is clear from Fig. 3 that the high activation barrier
for the dissociative oxygen chemisorption prevents
a direct catalytic reaction on the single-crystal
surface. In the case of a real gold-based catalytic
system, influences of the particle size or the sup-
port drastically reduce this barrier.
tion energy of E ¼ ð57 Æ 3Þ kJ/mol and the fact
r
that the reaction occurs already below 100 K.
Outka and Madix [7] reported an apparent
activation energy of (8 ± 4) kJ/mol for the CO
oxidation on an oxidized Au(1 1 0)-(l · 2) surface, a
result which disagrees with our value of ()1.8 ± 0.9)
kJ/mol. A possible reason for this deviation may be
the different methods for producing chemisorbed
oxygen. Outka and Madix employed thermal dis-
References
sociation of O on a hot Pt wire in front of the
2
sample. It cannot be excluded that a different
oxygen species than with our method is generated
in this way. Furthermore, impurities, especially Pt
from the hot wire, may influence the reactivity of
the chemisorbed oxygen (although no Pt contami-
nation was found with AES).
[1] G.C. Bond, D.T. Thomson, Cat. Rev.––Sci. Eng. 41 (1999)
319.
[
[
2] G.C. Bond, D.T. Thomson, Gold Bull. 33 (2000) 41.
3] R. Grisel, K.J. Weststrate, A. Gluhoi, B.E. Nieuwenhuys,
Gold Bull. 35 (2002) 39.
[4] D.M. Cox, R. Brickmann, K. Creegan, A. Kaldor, Z. Phys.
D 19 (1991) 353.

J. Michael Gottfried, K. Christmann / Surface Science 566–568 (2004) 1112–1117
1117
[
5] A. Sanchez, S. Abbet, U. Heiz, W.-D. Schneider, H.
H €a kkinen, R.N. Barnett, U. Landmann, J. Phys. Chem. A
[10] J.M. Gottfried, K.J. Schmidt, S.L.M. Schroeder, K.
Christmann, Surf. Sci. 525 (2003) 184.
1
03 (1999) 9573.
[11] J.M. Gottfried, Thesis, Freie Universit €a t Berlin 2003,
http://www.diss.fu-berlin.de/2003/133 (in English).
[12] J.M. Gottfried, K. Christmann, Surf. Sci., in preparation.
[13] J.M. Gottfried, K.J. Schmidt, S.L.M. Schroeder, K.
Christmann, Surf. Sci. 536 (2003) 206.
[
6] J. Hagen, L.D. Socaciu, M. Elijazyfer, U. Heiz, T.M.
Bernhardt, L. W o€ ste, Phys. Chem. Chem. Phys. 4 (2002)
1
707.
[
[
7] D.A. Outka, R.J. Madix, Surf. Sci. 179 (1987) 351.
8] J.M. Gottfried, K.J. Schmidt, S.L.M. Schroeder, K.
Christmann, Surf. Sci. 525 (2003) 197.
[14] M.A.A. Clyne, B.A. Thrush, Proc. R. Soc. London, A 269
(1962) 404.
[
9] J.M. Gottfried, K.J. Schmidt, S.L.M. Schroeder, K.
Christmann, Surf. Sci. 511 (2002) 65.
[15] J.M. Gottfried, N. Elghobashi, S.L.M. Schroeder, K.
Christmann, Surf. Sci. 523 (2002) 89.