Oxidation of Pd(111)

Article abstract of DOI:10.1016/S0039-6028(00)00599-9

The oxidation of Pd(111) was characterized using scanning tunneling microscopy (STM), temperature-programmed desorption (TPD), and low energy electron diffraction (LEED). Initial exposure of Pd(111) to O₂ at temperatures between 300 and 575 K resulted in a (2×2) structure observable by both LEED and STM. The maximum coverage achieved by O₂ exposure at 300 K was 0.25 ML. To increase the oxygen coverage further, NO₂ was used as the oxidant.

Full text of DOI:10.1016/S0039-6028(00)00599-9

Surface Science 462 (2000) 151–168
www.elsevier.nl/locate/susc
The oxidation of Pd(111)
G. Zheng, E.I. Altman *
Department of Chemical Engineering, Yale University, New Haven, CT 06520, USA
Received 21 January 2000; accepted for publication 8 May 2000
Abstract
The oxidation of Pd(111) was characterized using scanning tunneling microscopy (STM), temperature-programmed
desorption (TPD), and low energy electron diffraction (LEED). Initial exposure of Pd(111) to O at temperatures
2
between 300 and 575 K resulted in a (2×2) structure observable by both LEED and STM. The maximum coverage
achieved by O exposure at 300 K was 0.25 ML. By increasing the temperature above 500 K, the oxygen coverage
2
could be increased to 0.37 ML. To increase the oxygen coverage further, NO was used as the oxidant. On Pd(111),
2
NO dissociatively adsorbs, with NO going to the gas phase below 500 K, leaving oxygen on the surface. Above
2
5
00 K, initial exposure of NO to Pd(111) also produced the (2×2) structure. Increasing the oxygen coverage to
2
between 1.0 and 2.2 ML resulted in a complicated LEED pattern. This pattern could be explained as the superposition
of three equivalent domains of two surface structures: one with a square surface lattice rotated 15° with respect to
Pd[110], the other with a rectangular surface lattice with the short sides of the rectangles parallel to Pd[110]. In STM
movies, ad-islands and peninsulas were observed to nucleate and grow as the oxygen coverage reached this regime.
The rectangular structure was observed on the original Pd(111) terraces, and the square structure on the islands and
peninsulas. This suggested that when the oxygen coverage exceeds 0.25 ML, oxygen atoms penetrate the surface
creating a rectangular structure with a lower Pd atom density than the clean surface; the liberated Pd atoms, along
with oxygen, then form the islands and peninsulas. The lattice constants obtained from the STM images were
0
.679± 0.012 nm for the square structure and 0.394± 0.008 nm and 0.638± 0.022 nm for the rectangular structure,
both consistent with LEED observations. Neither of these structures can be simply related to any crystal orientation
of Pd or PdO, indicating that there are states intermediate between Pd and PdO. After further increasing the oxygen
coverage, the complicated LEED patterns became faint and a low temperature shoulder attributed to PdO
decomposition developed in TPD traces. The results indicate that Pd(111) oxidation proceeds through three stages
involving four distinct surface states. © 2000 Elsevier Science B.V. All rights reserved.
Keywords: Adsorption kinetics; Low energy electron diffraction (LEED); Nitrogen oxides; Oxidation; Oxygen; Palladium; Scanning
tunneling microscopy; Thermal desorption spectroscopy
1
. Introduction
and the catalytic combustion of methane. In cata-
lytic converters, the key reactions include the com-
plete oxidation of any hydrocarbon in the exhaust
gas stream as well as the simultaneous oxidation
of CO and reduction of NO. Compared with the
current Pt/Rh catalyst, Pd is more economical and
can operate at higher temperatures, closer to the
exhaust manifold [1]. Palladium is also the most
Palladium has been identified as a promising
catalyst for applications involving oxidation reac-
tions, including automotive catalytic converters
*
Corresponding author. Fax: +1 203 4327232.
E-mail address: eric.altman@yale.edu (E.I. Altman)
0
039-6028/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S0039-6028 ( 00 ) 00599-9

1
52
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
promising catalyst for the complete combustion of
methane in advanced gas turbines where ultralow
into the Pd(111) surface [14], a state intermediate
between Pd and PdO on Pd(100) [20], PdO
[9,13,17,20,21], and the formation of a very stable
oxide [12,14]. The latter appears to be due to
oxidation of trace amounts of Si segregating to
the surface region from bulk Pd [20,21], while the
other states remain poorly characterized.
NO emissions are desired [2–5].
x
Unusual phenomena in Pd-catalyzed oxidation
reactions, however, have created a number of chal-
lenges that must be overcome before Pd-based
catalysts can be implemented. These include oscilla-
tions in the CO oxidation rate at high O to CO
ratios. It is believed that changes in the nature of
The catalytic activity of the different oxygen
species on Pd surfaces is also unclear. For example,
2
the oxygen species on the Pd surface cause the
oscillations; however, the specific changes remain
poorly understood. Early studies suggested that the
oscillations were due to periodic filling and deple-
tion of subsurface oxygen [6,7] but this idea was
later challenged [8]. More recently, it was found
that PdO could exist under conditions that lead to
oscillations suggesting that the oscillations may be
due to the formation and decomposition of PdO
for CH combustion the oxidation of Pd to form
PdO results in changes in reactivity but dis-
agreement remains as to whether Pd or PdO is
more active [2–5,11,22–25]. It has been suggested
that the more labile oxygen in PdO makes it more
active than Pd [11]. Molecular beam studies, how-
4
ever, indicate that CH oxidation proceeds through
4
dissociative adsorption of CH followed by oxida-
4
tion of the fragments, and that oxygen blocks sites
[
9]. Oscillations have also been observed during
for dissociative adsorption [23,24]. Quantum cal-
culations also indicate that Pd is more active than
Pd-catalyzed CH oxidation when the O /CH ratio
is near stoichiometric [2–4,10]. In addition, it is
well documented that the activity of Pd catalysts
4
2
4
the oxide for CH dissociation [25]. Recent TEM
4
studies have suggested that Pd and PdO coexist
for CH combustion is extremely sensitive to the
catalyst history [2–5,10,11]. This sensitivity is gen-
when the catalyst is most active [3].
4
Because of the uncertainties outlined above, we
have begun to characterize the oxidation and
reduction of Pd surfaces on the atomic scale using
scanning tunneling microscopy (STM) along with
temperature-programmed desorption (TPD) and
low energy electron diffraction (LEED); in this
paper the oxidation of Pd(111) is described. It will
be shown that Pd(111) oxidation proceeds through
three stages, involving four distinct oxygen states.
These include an interesting mixed palladium–
oxygen layer that is observed following saturation
of the (2×2) chemisorbed oxygen layer, but prior
to the formation of bulk PdO. As this mixed
surface layer forms, the Pd density in the surface
decreases, liberating Pd atoms. Along with oxygen,
the liberated Pd nucleates and forms islands and
peninsulas with a structure that is distinct from
both the original terraces and PdO. The implica-
tions of these results on Pd-catalyzed oxidation
reactions are discussed.
erally ascribed to changes in the catalyst surface on
oxidation and reduction; however, the nature of
these changes and how they influence the catalytic
activity remains the subject of debate [2–5,10,11].
The results outlined above indicate that the
oxidation of Pd surfaces plays a key role in deter-
mining the catalytic behavior of Pd. The inter-
action of Pd with oxygen, however, remains poorly
characterized. Previous studies have shown that
initial exposure of Pd surfaces to O at or below
2
room temperature leads to dissociation and the
formation of a chemisorbed layer of oxygen atoms.
On Pd(111), the adsorbed oxygen atoms form a
(
0
2×2) structure that saturates at a coverage of
.25 ML (ML=adsorbed atoms/surface metal
atom) [12–15]. Higher oxygen coverages could be
obtained at elevated temperatures and higher O
pressures [13,15], or by using stronger oxidants
2
such as NO or NO [16–18]. At these higher
2
coverages, it has been suggested that oxygen can
exist on and beneath the Pd surface in several
different forms, depending on the oxidation condi-
tions and the crystallographic orientation. These
different forms include reports of subsurface
oxygen on Pd(110) [9,19], incorporation of oxygen
2. Experimental
The experiments were performed using a UHV
system equipped with a quadrupole mass spec-

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
153
trometer (UTI 100C-12) for TPD, a double pass
cylindrical mirror analyzer for Auger electron
spectroscopy (AES), an ion gun for sputtering,
LEED optics, and a high-speed variable-temper-
ature STM. The pumping is provided by a
was used to determine when the surface was free
of C; a C-free surface exhibits no CO or CO
2
desorption when exposed to O near room temper-
2
ature [27]. During all TPD runs, CO and CO
2
were monitored in addition to O ; if either CO or
2
3
00 l s−1 ion pump, resulting in a base pressure of
CO desorption was observed, the data were
discarded.
2
less than 1×10−10 Torr. A small turbo pump is
employed to remove sputter gases and to evacuate
a side chamber used for sample and tip introduc-
tion. The side chamber can also be used to heat
samples above 700 K at pressures above 10 Torr.
The custom-designed microscope is capable of
recording STM movies at speeds exceeding 1 image
per second in the temperature range of 300–900 K
3. Results
3.1. TPD of O from Pd(111)
2
Initial exposure of Pd(111) to O results in a
2
single O TPD peak at 850 K. As shown in Fig. 1,
2
[
26], and allows two dosers to be aimed at the
tunnel junction during scanning.
the desorption peak temperature shifts from 850 K
The oxidants were led into the chamber through
two variable leak valves, one for TPD and LEED
experiments, and the other for in situ STM experi-
ments. A stainless steel tube guided the gases from
each leak valve to the sample. For TPD and LEED
experiments, the sample was positioned within a
few mm of the end of the tube, leading to effective
pressures at the sample surface an estimated 45
times the background pressure. This factor was
obtained by comparing the time required to
achieve a certain oxygen coverage, determined by
the TPD peak integral, with the sample located in
front of the tube, with the time required with the
sample turned away from the tube. In the micro-
scope, the sample was further from the doser, and
the factor was estimated to be 5. For all the data
presented in this paper, the gas exposures and the
dosing pressures have taken these factors into
account.
at low coverage to 800 K after exposure to 30 L;
consistent with second-order desorption due to
recombination of adsorbed oxygen atoms. The
uptake curves in Fig. 2 show that at room temper-
ature the oxygen coverage saturates after exposure
The Pd(111) sample was heated via conduction
from resistively heated Ta wires spot-welded to a
0.0127 mm thick Ta foil, which was in turn
clamped to the back of the crystals by ceramic
posts. The temperature was measured using a
chromel–alumel thermocouple attached to the
front of the crystal. The samples were subjected to
repeated sputter–anneal cycles (sputtering with
5
00 eV Ar+ with the surface at 900 K and annea-
Fig. 1. Series of TPD curves for O desorption from Pd(111)
ling at 1000 K) until no impurities were detected
by AES. Since the KLL signal of C overlaps the
MNN signal of Pd, it is difficult to detect trace
amounts of C by AES. Instead, oxygen titration
2
exposed to O at temperatures between 500 and 550 K. The
2
background pressure for each curve was essentially zero; the
data have been offset so the traces for the different doses do
not overlap. A heating rate of 5 K s−1 was used.

1
54
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
of the (2×2) layer, this is the first time on Pd(111)
that an O thermal desorption signal other than
2
that due to desorption of the (2×2) layer has been
reported following O exposure.
2
The oxygen coverage could not be significantly
increased by raising the oxygen pressure above
10 Torr while maintaining the crystal at 500 K.
This difficulty may stem from the fact that high
pressure O can stimulate the replacement reaction
2
2
between O and the stainless steel walls of the
UHV chamber to produce CO [28], which in turn
can remove surface oxygen from Pd through CO
formation. As shown in Fig. 2, even at 500 K the
2
O sticking coefficient is very low above 30 L, and
2
so trace amounts of CO can easily lead to CO
2
formation rates that equal the oxygen adsorption
rate. To circumvent this problem, nitrogen dioxide,
which is a much stronger oxidant than O , was
2
used to access higher oxygen coverages. On Pd
surfaces, NO dissociatively adsorbs below room
2
temperature, yielding NO and O [17]. Above
500 K, the NO desorbs leaving atomic O on the
surface [18]. Fig. 3 shows the O TPD traces
2
following NO exposure at 500 K; the results are
2
Fig. 2. Comparison of oxygen uptake for exposure to NO and
2
O . The data were obtained by integrating the oxygen TPD
2
very similar to those of Banse et al. [17]. For
curves. The dashed line shows the maximum coverage obtained
reference, the bottom TPD curve was obtained
for O exposure at room temperature.
following 30 L O exposure and corresponds to an
2
2
oxygen coverage of 0.25 ML. Comparison of the
to 30 L. Both LEED and STM results indicate
that oxygen forms a (2×2) adsorbate structure
under these conditions. This structure saturates at
a coverage of 0.25 ML and so the TPD peak
curves for NO and O exposure reveals that an
oxygen coverage roughly an order of magnitude
2
2
higher can be obtained with NO . The data also
2
reveal that the 725 K peak seen in Fig. 1 eventually
integral following saturation O doses at room
temperature was used to calibrate the oxygen
coverage for surface treatments that led to higher
coverages.
It was found that the oxygen coverage could be
increased beyond 0.25 ML by exposing the surface
completely replaces the 800 K peak and finally
saturates at a dose of 30 L NO . At higher NO
2
2
2
exposures, an additional desorption feature that
started at around 590 K was observed. In a previ-
ous study of the oxidation of Pd(110), a similar
desorption peak appeared simultaneously with an
XPS signature indicative of bulk PdO and thus
the feature was associated with PdO decomposition
[9]. Since the rate of PdO decomposition is
expected to be independent of the orientation of
the underlying Pd, the same assignment is sug-
gested for Pd(111).
to O at temperatures above 500 K. As shown in
2
Fig. 1, a lower temperature shoulder began to
appear at a dose of 6500 L. As the coverage was
increased, a peak at ~725 K developed that grad-
ually began to replace the 800 K peak. The highest
exposures were obtained by dosing
O
at
2
1
0
0−3 Torr which resulted in oxygen coverages of
Comparison of the results to the simulated TPD
curves for PdO decomposition shown in Fig. 4 are
consistent with the assignment of the desorption
feature starting at 590 K to PdO decomposition.
.37 ML. Although previous photoemission
spectroscopy [14] and ellipsometry studies [15]
suggested that O could adsorb beyond saturation
2

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
155
Fig. 4. Simulated O TPD curves due to PdO decomposition
Fig. 3. Series of TPD curves for O desorption from Pd(111)
2
2
for a heating rate of 5 K s−1. The curves correspond to initial
exposed to NO at a temperature of 500 K. The bottom curve,
2
PdO coverages of 0.5, 1, 2, and 3 ML respectively.
denoted by the asterisk, was obtained following a 30 L O dose
2
which corresponds to a coverage of 0.25 ML. The heating rate
was 5 K s−1.
uptake as a function of NO and O exposure.
2
2
The initial sticking coefficient for NO dissociative
2
The simulated curves were obtained by calculating
adsorption is on the order of unity. In contrast,
the rate of O desorption from the equilibrium
dissociation pressure of PdO, assuming the oxide
the initial sticking coefficient for O is over a factor
of 20 lower and drops dramatically when the
2
2
layer was uniformly dispersed across the surface,
and that the background oxygen pressure was so
low that the oxygen flux to the surface could be
neglected. The dissociation pressure was estimated
by extrapolating the data of Bayer et al. [29] to
lower temperatures; Voogt et al. [15] have used a
similar approach to estimate the stability of PdO
(2×2) layer saturates, suggesting that the O
adsorption mechanism changes at 0.25 ML. The
2
oxygen uptake due to NO exposure rolls off at
2
much higher coverages; at about 2.0 ML as the
725 K desorption peak approaches saturation. This
roll-off is not due to slow decomposition of PdO
at the 500 K adsorption temperature. At 500 K,
the PdO dissociation pressure was estimated to be
10−10 Torr and so for 1 ML of PdO the oxygen
coverage would drop by only 0.01 ML in 4 min,
while the TPD curves were obtained seconds after
at low pressure. Fig. 4 shows that O desorption
2
due to PdO decomposition peaks at 600 K at 1 ML
and shifts to higher temperature as the coverage
increases. The predicted peak position occurs in
the same range as that of the lowest temperature
desorption peak seen in Fig. 3, again suggesting
that the desorption feature is due to PdO
decomposition.
the NO exposure was stopped. Thus the fall in
2
the sticking coefficient when the coverage
approaches 2.0 ML suggests a change in the ability
of the surface to dissociate NO as the 725 K peak
2
Fig. 2 provides a comparison between oxygen
saturates. In Section 3.3 it will be shown that this

1
56
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
change can be associated with a structural change
in the surface.
visible when the energy was changed. This is
illustrated in Fig. 6e, which shows the LEED
pattern obtained at 89 eV after an NO dose of
2
3.2. LEED results
8.1 L, the same exposure as in Fig. 6d. The sche-
matic diagram in Fig. 6f shows the positions of all
the observed spots. While the pattern is rather
complicated, there are several distinctive features:
an inner ring of 18 spots, a second ring of 12
spots, and then groups of three spots. In the
following section, it will be shown with the aid of
STM that this pattern is due to three equivalent
domains of two different structures.
Exposure of Pd(111) to O at temperatures
2
between 300 and 575 K resulted in (2×2) patterns
as shown in Fig. 5. The (2×2) pattern was
observed starting at a dose of 8 L corresponding
to an oxygen coverage of 0.15 ML, and the pat-
terns became more intense and sharper as the
exposure was increased to 30 L. Increasing the
O exposure beyond 30 L produced no new LEED
2
patterns.
3.3. STM study of Pd(111) oxidation
Initial exposure to NO at 500–575 K also
2
resulted in a (2×2) LEED pattern as shown in
Scanning tunneling microscopy images obtained
Fig. 6a. Increasing the NO exposure slightly pro-
duced new diffraction spots in addition to the
after O exposure at room temperature were con-
sistent with LEED observations. The images in
2
2
spots associated with the (2×2) pattern. As the
Fig. 7 reveal hexagonal patterns of bright protru-
sions; it is unclear whether the bright features are
due to the adsorbed O atoms or the exposed Pd
atoms in the ordered adlayer. The nearest neighbor
spacing between the protrusions is 0.549±
0.005 nm, or within experimental error of twice
the measured 0.277± 0.004 nm spacing of the Pd
atoms on the (111) surface; the actual spacing of
the Pd atoms on the (111) surface is 0.275 nm.
Further, atomic resolution images of the Pd sub-
strate indicate that the close-packed directions of
the adlayer run parallel to Pdꢀ110ꢁ directions.
Thus the images indicate that the adsorbed oxygen
atoms form a (2×2) structure. Also, note that the
images in Fig. 7 were obtained at subsaturation
coverage, consistent with LEED observations of a
NO exposure was increased, the intensity of these
2
new spots became stronger, while the (2×2) spots
faded, as shown in Fig. 6b–d. The oxygen coverage
associated with these changes was 0.25–2.2 ML,
the coverage regime where the 725 K TPD peak
replaces the 800 K desorption peak. Increasing the
NO exposure into the regime where the thermal
2
desorption feature associated with bulk PdO
decomposition was observed caused the LEED
pattern to fade and the diffuse background to
increase.
The LEED patterns in Fig. 6b–d were all
obtained at 62 or 63 eV. The spot intensities of
these complicated LEED patterns were strongly
dependent on energy, and additional spots became
Fig. 5. LEED patterns for (a) clean Pd(111) surface and (b) Pd(111) exposed to 30 L O at 300 K.
2

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
157
Fig. 6. LEED patterns illustrating structural changes corresponding to the following doses and exposure temperatures: (a) 0.68 L at
00 K, (b) 0.79 L at 575 K, (c) 1.35 L at 575 K, (d) and (e) 8.1 L at 575 K. The sample was rotated between experiments accounting
5
for the different orientations in (a) and (e) compared with (b)–(d). The diffraction spots are shown schematically in (f ), where the
black solid spots illustrate those due to the Pd(111) substrate.
(
2×2) pattern at low coverage. The adsorbate
Increasing the oxygen coverage beyond 0.25 ML
caused dramatic changes in the surface. As shown
in Fig. 8, STM images obtained after exposure of
islanding also suggests sufficient mobility of
adsorbed oxygen atoms at room temperature to
form ordered islands, and through surface attrac-
tive interactions between adsorbed oxygen atoms
that are strong enough to stabilize the islands so
that they can be imaged with STM.
Pd(111) to 8.1 L NO at 575 K, a dose that results
2
in the LEED pattern shown in Fig. 6f and an
intense 725 K desorption peak, reveal the growth
of islands and peninsulas on the surface. Close

1
58
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
Fig. 7. STM images of Pd(111) exposed to O at 300 K, illustrating the expected periodicity and orientation for a (2×2) adsorbed
2
oxygen layer. Image (a) was obtained at an essentially bare area and the periodicity corresponds to that of Pd(111); the dark areas
are attributed to O. The image was obtained at a sample bias of 0.25 V and a tunnel current of 0.18 nA. Images (b) and (c) were
obtained at a sample bias of 0.1 V and a tunnel current of 0.22 nA; under these conditions the (2×2) patches appear raised above
the bare terraces.
Fig. 8. Room temperature STM images of Pd(111) exposed to 8.1 L NO at a temperature of 575 K. The images were obtained at
2
the following sample biases and tunnel currents: (a) −1 V and 0.55 nA, (b) −1 V and 0.55 nA, (c) −1 V and 0.75 nA, (d) −0.75 V
and 0.70 nA, (e) −0.5 V and 0.70 nA. The boxes in (b) highlight the areas where the images in (c), (d), and (e) were obtained.
inspection of the terraces in Fig. 8c reveals stripes
rotated 0, 120, and 240° with respect to Pd [110].
As shown in Fig. 8d, higher resolution images of
the terraces showed that the stripes are due to a
rectangular structure with lattice constants equal
to 0.394± 0.008 and 0.638± 0.022 nm; the short
sides of the rectangles run parallel to a Pdꢀ110ꢁ
direction. In contrast, the islands and peninsulas

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
159
display a distinctly different structure. As shown
in Fig. 8e, high-resolution STM images of the
islands and peninsulas revealed a square structure
with a lattice constant of 0.679± 0.012 nm; the
squares are rotated 15° from a Pdꢀ110ꢁ direction.
The STM images in Fig. 8 can explain the
LEED patterns in Fig. 6d–f. Three equivalent
domains of the rectangular structure with lattice
constants a=0.39 nm and b=0.64 nm and a: paral-
lel to Pdꢀ110ꢁ directions produces the LEED
pattern in Fig. 9a. Fig. 9b shows the LEED pattern
due to three equivalent domains of the square
structure with lattice constant 0.68 nm and each
domain rotated 15° with respect to a Pdꢀ110ꢁ
direction in real space. Note that the (2,2) diffrac-
tion spots due to the squares overlap the first-
order spots due to the Pd(111) substrate. Such a
pattern has been reported previously and it was
suggested that it was due to three equivalent
domains of a square structure with a lattice con-
stant of 0.75± 0.05 nm [15]. This discrepancy in
the lattice constant is due to the failure in the
earlier work to account for the non-orthogonality
of the lattice vectors of the hexagonal surface unit
cell which leads to a factor of 2ǰ3/3 that must be
considered when the reciprocal lattice is inverted.
It should also be noted that if the (2,2) spots due
to the squares perfectly coincide with the Pd(111)
Closer inspection of the STM images in Fig. 8
provides further insight into how oxidation
changes the Pd surface. First, the islands and
peninsulas are the same height, 0.22± 0.01 nm, as
a step on the Pd(111) surface, independent of the
imaging bias. This indicates that the islands are
Pd and that the square structure is due to an
oxygen-induced reconstruction of a Pd(111) sur-
face. Further, since the measured island height is
independent of tip bias, the square structure must
be electronically similar to the rectangular struc-
ture observed on the original terraces. Also, the
islands are only seen on relatively wide terraces
while the peninsulas are observed on narrower
terraces. These observations are consistent with a
nucleation and growth mechanism in which Pd
atoms are liberated on the terraces and then diffuse
across the surface. When the terraces are narrow,
the diffusing atoms attach to pre-existing steps
and, together with oxygen, form reconstructed
peninsulas, while on wider terraces the diffusing
Pd atoms collide with one another, nucleating
islands. The Pd atoms are liberated from the
original surface indicating that the rectangular
structure is due to oxygen penetration into the
surface, creating a rectangular structure with a
lower Pd atom density. This mechanism is consis-
tent with the sequence of LEED patterns observed
as the O coverage was increased. As shown in
Fig. 6, the intensity of the spots due to both the
square and rectangular structures increase simulta-
neously as the (2×2) spots fade.
Oxygen penetration into the surface layer has
been reported for other metals, and it has been
suggested that a nearly coplanar metal–oxygen
layer can be formed on Al(100) [30,31]. A similar
mechanism has also been observed for S on
Ni(111) [32]. In this case, S atoms displace Ni
atoms from the surface leading to a missing row
reconstruction on the original terraces and Ni
islands created from the ejected Ni atoms.While
the observed rectangular unit cell is also suggestive
of a missing row structure for O on Pd(111), the
lattice constants do not match any simple multiple
of the Pd(111) surface lattice constant and so it is
difficult to assign the structure to a surface with
periodic missing rows.
(
1,0) spots, then the lattice constant of the square
structure would be 0.674 nm, which is within the
experimental uncertainty of the STM measure-
ments. Superimposing the patterns in Fig. 9a and
b results in the pattern in Fig. 9c, which still does
not completely reproduce the observed LEED
pattern. First-order multiple scattering between
the Pd(111) substrate and the square structure,
however, leads to the additional spots highlighted
in Fig. 9d, and now the LEED pattern derived
from the STM images (Fig. 9d) matches the
observed pattern (Fig. 6f ). Therefore, both STM
and LEED indicate that replacement of the desorp-
tion peak associated with desorption of oxygen
chemisorbed in a (2×2) layer by a lower temper-
ature peak is due to replacement of the (2×2)
structure by two distinct structures: (1) a rectangu-
lar structure on the original terraces; and (2) a
square structure that grows on top of the origi-
nal surface.
To learn more about island nucleation and

1
60
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
Fig. 9. LEED schematics for: (a) three equivalent domains of the square structure (light grey circles) with a lattice constant of
.68 nm; (b) three equivalent domains of the rectangular structure (dark grey circles) with lattice constant 0.39 and 0.64 nm, respec-
tively; (c) the superposition of (a) and (b); and (d) the superposition of (a) and (b) plus spots due to first order multiple scattering
open circles) between the square structure and the substrate. In (a) and (b), the squares and rectangles highlight the unit cells of
the three equivalent domains. The black solid spots represent spots due to the Pd(111) substrate.
0
(
growth, STM movies were recorded while the
middle of the large terrace only 17 s after the leak
sample was exposed to NO at elevated temper-
atures. The images in Fig. 10 were taken from a
movie recorded while the surface was exposed to
valve was opened, before any peninsulas are
observed. The islands appear without any notice-
able change in the step position, and through the
entire sequence of images there is no apparent
retreat of the step edge, indicating that the steps
2
2
.5×10−8 Torr NO at 575 K. As shown in
2
Fig. 10a, small islands begin to appear in the

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
161
Fig. 10. Frames taken from an STM movie recorded during in situ exposure to 2.5×10−8 Torr NO at a temperature of 575 K. The
2
sample was biased at −0.15 V and the tunnel current was 0.60 nA. The box in (i) highlights the area shown in Fig. 13.
are not the source of Pd atoms for the islands.
Further, there is no evidence of a higher island
density near the step edges suggesting that the
steps are not significantly more reactive than the
terraces for oxygen incorporation. The sequence
of images also show that the islands and peninsulas
exhibit preferred growth directions; in particular,
the two peninsulas at the upper part of the center
of the images grow much more rapidly perpendicu-
lar to the step than parallel to the step. Also notice
that the island at the bottom left of the image
grows towards the step until it gets very close to
it, at which point growth in that direction stops.
During the movie, several islands were observed
to coalesce with no obvious slowdown in growth
prior to coalescence, while others avoided one

1
62
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
surface, 0.275 nm. While the edge positions vary
with time, the minimum observed gap between
clusters that do not coalesce is 1.1± 0.16 nm. Since
the islands have a different structure than the
original terraces, the islands always avoid pre-
existing steps. The observation of islands before
peninsulas and the faster growth of the peninsulas
away from rather than parallel to the step edges
also suggest a barrier for adatoms to attach to
pre-existing steps. This is not unexpected since
attachment to pre-existing steps again creates
domain boundaries. A result of the barrier is the
development of series of voids at the step edges,
as can be seen in Figs. 8a and 10i. The movie also
shows that the island and peninsula growth rate
remains relatively constant then slows after ~300 s
and completely stops after about 400 s with
roughly 32% of the surface covered by islands and
peninsulas, as illustrated in Fig. 11. These observa-
tions are consistent with Fig. 2, and suggest that
Fig. 11. Plot of the percentage of the surface covered by the
islands and peninsulas versus time for the movie shown in
Fig. 10.
NO dissociation proceeds through a precursor
2
mechanism with the (2×2) structure containing
the more reactive dissociation sites. This would
lead to a constant growth rate until the (2×2)
structure is nearly completely removed from the
surface.
The movie frames highlighted in Fig. 12 directly
show the correlation between the structural change
in the original terraces and island and peninsula
growth. In Fig. 12a, irregular sections of the origi-
nal terraces are raised 0.06 nm above the remainder
another in a manner similar to the island and the
step. This variation can be understood in terms of
domain boundary formation. When the islands are
misaligned so that coalescence creates a domain
boundary, the islands avoid one another, while if
the islands are in the same registry with the sub-
strate they easily coalesce. The minimum gap
widths are consistent with this explanation. The
domain boundary width must be greater than the
of the terraces. As the NO exposure increases,
2
0.68 nm lattice constant of the square structure by
these areas decrease in size while the islands and
at least the spacing of the Pd atoms on the (111)
peninsulas grow as shown in Fig. 12b and c. Thus,
Fig. 12. Additional frames taken from the same movie shown in Fig. 10. The frames highlight an intermediate height area of the
original terraces attributed to the (2×2) structure that shrinks in size as the islands and peninsulas grow.

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
163
Fig. 13. Higher resolution images of the area highlighted in Fig. 10i after growth of the islands and peninsulas stopped. The sample
was biased at (a) −0.50 V and (b) −0.25 V, respectively, with a tunnel current of 0.60 nA. A model is shown in (c) that explains
the observed moir e´ patterns. In the model, the solid circles represent substrate Pd atoms and the partially transparent circles the
position of the maxima in STM images due to the square structure. The solid lines highlight the positions where the maxima of the
two lattices coincide, while the dashed lines highlight rows of maxima that are nearly coincident.
the images show that increasing oxygen coverage
simultaneously leads to a transformation in the
terraces and growth of islands and peninsulas with
a different structure. While it was not possible to
image the raised areas with atomic resolution,
since they disappear with increasing exposure and
because the terraces are ultimately completely cov-
ered by the rectangular structure, they are attrib-
uted to Pd(111) covered by a (2×2) layer of
oxygen.
After growth stopped, higher resolution images
of the area highlighted in Fig. 10i were obtained.
As shown in Fig. 13a, images of the peninsulas
revealed a new striped pattern. The stripes appear
as pairs with the distances between the pairs
alternating between 1.4 and 1.9 nm. The stripe
corrugation is 6.0± 0.3 pm. After a tip-change was
experienced, the square structure was observed in
addition to the stripes as shown in Fig. 13b. In
this image, the corrugation of the stripes is again
maxima along Pd[110] lie nearly on top of the Pd
atoms. This would lead to a pair of raised stripes
in STM images. If we move two more rows of
maxima along Pd[110], for a total of 1.44 nm, the
maxima are again nearly on top of the Pd atoms,
which would also lead to a stripe. The next row
of maxima, 1.92 nm from where we started, lie
almost directly on top Pd atoms creating another
stripe. Comparing Fig. 13a and c shows that this
model reproduces the experimentally observed
periodicities. Thus the results indicate that the
square structure grows on top of an unre-
constructed Pd surface, consistent with the
multiple scattering observed between the Pd(111)
substrate and the square structure with LEED. It
should be mentioned that the reason the moire´
pattern was not observed in Fig. 8e is due to the
quality of the image; when 6.0 pm high stripes
were added to the image in Fig. 8e, the stripes did
not rise above the noise level in the image.
6.0± 0.3 pm while the corrugation of the squares
The results described above indicate that at
is 15.1± 0.3 pm; the lattice constant of the square
structure is the same as observed in Fig. 8. These
observations indicate that the square structure is
superimposed on a longer-range periodicity due to
the lattice mismatch with the underlying layers.
Fig. 13c illustrates what happens if we place the
square structure on top of an unreconstructed
Pd(111) surface. Starting with the row of maxima
due to the square structure laying directly on top
of the Pd atoms, we see that the next row of
575 K and an NO pressure in the 10−8 Torr range,
2
the oxidation of the Pd surface stops when the
rectangular and square structures completely cover
the surface. This is not surprising since Fig. 4
indicates that PdO decomposes at 575 K.
Therefore, PdO formation was studied by dropping
the temperature to 500 K while recording STM
movies during NO exposure. As shown in
2
Fig. 14b, after exposure to 61 L NO (32.5 min at
2
3.5×10−8 Torr), clusters were observed on the

1
64
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
Fig. 14. STM images recorded in the bulk oxide formation regime. Images (a) and (b) show the same area of the surface (a) before
and (b) after in situ exposure to 61 L NO at a temperature of 500 K. Image (c) was obtained after Pd(111) was dosed with 68 L
2
NO at 575 K, followed by same amount of NO at 500 K. The sample bias and the tunnel currents for these images are as follows:
2
2
(
a) −0.1 V and 0.55 nA, (b) −0.1 V and 0.70 nA, (c) −0.3 V and 0.55 nA.
surface. As expected for a nucleation and growth
process, there is a higher density of smaller clusters
on the surface at the lower temperature. The total
coverage of the clusters was also higher than that
of the islands and peninsulas observed at 575 K.
Zooming in on the individual clusters revealed no
evidence of atomic scale features. After the STM
images were recorded, LEED showed only a (1×1)
pattern and some faint spots. TPD from this
surface revealed the low temperature desorption
feature associated with PdO.
The failure to observe either an ordered LEED
pattern or atomic-scale features in STM images
once the oxide forms could be due to the small
size of the clusters in Fig. 14b which would severely
limit the size of the domains on the original
terraces and the clusters. To investigate this possi-
bility, a surface with large islands and peninsulas
structures give way to a surface with no discernible
crystalline order.
4. Discussion
The STM, LEED, and TPD results indicate
that Pd(111) oxidation proceeds through a three-
step mechanism: (1) rapid adsorption that stops
at 0.25 ML of atomic oxygen; (2) oxygen penetra-
tion into the surface accompanied by island growth
leading to the formation of two phases that are
intermediate between Pd and PdO; (3) formation
of bulk PdO. The driving forces for this mechanism
can be understood as follows. Assuming second-
order desorption and a typical pre-exponential of
the order 1013 s−1, the O desorption peak temper-
2
ature at coverages less than 0.25 ML suggests a
was prepared by NO exposure at 575 K, and then
further oxidized at 500 K. Oxidation of this surface
heat of adsorption of roughly −50.2 kcal mol−1
[33], which is significantly higher than the
−25.65 kcal mol−1 (at 1050 K) heat of formation
of PdO [29]. Thus, as expected, initial exposure of
either NO or O to Pd(111) results in oxygen
adsorption. Increasing the coverage of the
adsorbed layer beyond 0.25 ML requires compres-
sion of the (2×2) layer that results in repulsive
adsorbate–adsorbate interactions that reduce the
heat of adsorption below the heat of formation of
the oxide. Therefore, at higher coverages, oxide
formation is favored. The PdMPd nearest neighbor
distance in PdO is 0.305 nm versus 0.275 nm in
Pd, so oxidation requires expansion of the Pd
2
caused the sharp LEED pattern due to the square
and rectangular structures to fade. An STM image
of a surface prepared in this manner is shown in
Fig 14c. As expected, the island and peninsula
sizes are now characteristic of initial oxidation at
2
2
575 K. Zooming in on these larger islands, how-
ever, again failed to reveal atomic-scale features,
thus eliminating the possibility that the small
domain size precluded the observation of atomic-
scale periodicities following bulk oxidation at
500 K. Thus, both STM and LEED indicate that
as bulk PdO forms, the square and rectangular

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
165
surface; the resulting compression is relieved by
ejecting Pd atoms from the terraces. At this stage,
the moir e´ pattern seen in the STM images, and
the multiple scattering seen in the LEED patterns,
indicate that single oxidic, or alternatively recon-
structed, layers form on top of an unreconstructed
Pd(111) surface. As a result, the oxygen atoms
and the Pd atoms in these layers interact with
underlying Pd atoms that are bound to no other
oxygen atoms. Therefore, it is not surprising that
the oxygen atoms are more strongly bound in
these layers than in the bulk oxide. Further, since
the oxidic layers are so thin, interactions with the
substrate and surface effects determine the struc-
ture, and so the layers exhibit structures that are
distinct from either Pd or PdO. As the oxygen
coverage increases, bulk effects become important
and the OMPd bond strength becomes characteris-
tic of the bulk oxide. It should be noted that these
arguments assume that the shifts in the desorption
peak temperatures are dominated by changes in
the desorption activation energy and not the pre-
exponential.
While the general oxidation mechanism can be
readily understood, a number of interesting ques-
tions regarding Pd(111) oxidation remain. The
first centers on why the islands and peninsulas
exhibit different structures than the original ter-
races. Since the structures are different, at least
one of the structures must be due to a metastable
phase, and kinetic limitations must play a role in
determining the structure. It can be argued that
the islands encounter little or no barrier to expan-
sion because they have unbounded edges, while,
in contrast, expansion within the terraces requires
compressing the surface and ultimately ejecting Pd
atoms. Along the same lines, a lower barrier is
expected for restructuring a small cluster into a
lower energy configuration than a terrace tens
of nm wide. Both arguments suggest that the
square structure observed on the islands is the
lower energy structure. In prior work on S on
Ni(111), which exhibits a similar oxidation mecha-
nism, it was found, however, that the islands and
terraces displayed different structures at room tem-
perature, and that annealing caused the islands to
restructure into the same structure seen on the
terraces [32,34]. It was therefore concluded that
diffusional limitations prevented the islands from
adopting the lower energy structure. Similar annea-
ling experiments could not be performed for O on
Pd(111) because O begins to desorb from Pd(111)
only ~50 K above the 575 K reaction temperature
required to form domains large enough to be
imaged reliably with STM.
The second question regards the specific atomic
arrangements associated with each observed phase.
Previous studies of chalcogen adsorption on Ni
[
32,34–37] and Pt(111) [38–40] at room temper-
ature and below have also revealed (2×2) struc-
tures. A variety of techniques including surface
extended X-ray absorption fine structure [35,36],
photoelectron diffraction [41], LEED spot inten-
sity analysis [36,39], and low temperature STM
[
40], all indicate that the chalcogen atoms sit
above the fcc-type threefold hollows on the surface.
Therefore, the (2×2)-O structure on Pd(111) is
attributed to a simple chemisorbed layer in which
the O atoms sit atop on an essentially unre-
constructed surface. In prior work on the oxidation
of supported Pd catalysts, it was suggested that
initial oxidation created amorphous PdO [42]. The
fading of the LEED pattern and the lack of any
discernible order in STM images when evidence of
bulk PdO was observed with TPD are consistent
with the formation of an amorphous oxide.
Determining the atomic positions of the inter-
mediate states from the STM and LEED data
remains a challenge. It is clear that the square and
rectangular structures do not correspond to simple
adsorbed layers in primitive structures. If this were
the case, the lattice constants of the rectangular
and square structures would suggest coverages of
only 0.26 and 0.14 ML respectively, well below the
measured coverage of 2.2 ML. It is also clear that
the observed periodicities cannot be obtained by
simple manipulation of either a Pd or PdO lattice.
Palladium oxide crystallizes in a tetragonal struc-
ture with lattice constants a=b=0.305 nm and c=
0.534 nm at room temperature; Pd atoms are
located at (0, 0, 0) and (0.5, 0.5, 0.5), and oxygen
atoms at O at (0.5, 0, 0.25) and (0.5, 0, 0.75).
Palladium is an fcc metal with a lattice constant
of 0.389 nm. Although the lattice constant of the
square structure is roughly ǰ6 times the nearest

1
66
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
the square structure is rotated 15° with respect to
the substrate, and so this model suggests an 11.6°
rotation between PdO[100] and Pd[110], which
yields no obvious epitaxial relationship between
the oxide and the metal. Thus the results suggest
that the intermediate states form structures that
are distinct from both the metal and the bulk oxide.
Regardless of the details of the structure of the
intermediate phases, the results have important
implications for Pd-catalyzed oxidation reactions.
In the catalysis literature, variations in the catalytic
activity of Pd oxidation catalysts are generally
attributed to differences in activity between PdO
and Pd [2–5,42]. Since the intermediate states
exhibit distinct structures and distinct PdMO bond
strengths, these phases should also exhibit distinct
catalytic behavior and therefore must be consid-
ered in modeling Pd-catalyzed oxidation reactions.
The adsorption and oxidation mechanism can also
impact catalysis. The displacement of the Pd atoms
from the (111) terraces prior to bulk oxidation
results in massive restructuring of the surface at
very modest temperatures (500 K). For sustained
catalytic reactions, the oxidation–reduction cycle
will lead to catalyst restructuring and sintering at
temperatures far lower than those normally associ-
ated with these phenomena. This facile restructur-
ing mechanism may play a key role in the observed
sensitivity of Pd oxidation catalysts to catalyst
history and preparation conditions.
Fig. 15. Schematic diagram of the oxygen-terminated PdO(001)
surface with 20% oxygen atoms on the top layer removed. This
produces a (ǰ5×ǰ5)R26.6° structure with a lattice constant
of 0.682 nm.
neighbor spacing in Pd(111), as shown in Fig. 13c,
the different symmetries of the surface and the
underlying substrate dictate that few of the atoms
in the square structure lie in the same registry with
the substrate. Along similar lines, the
5. Summary
(
ǰ5×ǰ5)R26.6° ordered vacancy structure on
PdO(001) shown in Fig. 15 yields a square surface
with a lattice constant of 0.682 nm, within experi-
mental error of the measured 0.679 nm lattice
constant of the square structure. There are, how-
ever, difficulties with this model. If the layer were
essentially PdO, a step height closer to the
The oxidation of Pd(111) proceeds through
three stages involving four distinct surface states.
The first stage is chemisorption of atomic oxygen
in a simple (2×2) overlayer. Desorption of the
chemisorbed layer is a second-order process with
a desorption peak temperature of 800 K at satura-
0
0
.267 nm PdO(001) step height rather than the
.226 nm Pd step height would be expected. Also,
tion. When Pd(111) is exposed to O at 300 K,
oxygen uptake does not proceed beyond this point.
2
PdO(001) is a polar surface and so an ordered
vacancy structure is not unexpected; however, the
Increasing the exposure temperature to 500 K and
using stronger oxidants allows the second and
third stages to be accessed. In the second stage,
oxygen penetrates the surface creating a rectangu-
lar structure with lattice constants equal to
0.394± 0.008 and 0.638± 0.022 nm; the short sides
(
ǰ5×ǰ5)R26.6° structure corresponds to remov-
ing 20% of the oxygen atoms, while neutrality
would dictate that 50% of the oxygen atoms be
removed. Finally, STM and LEED indicate that

G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
167
of the rectangles run parallel to Pdꢀ110ꢁ direc-
References
tions. Formation of the rectangular structure
results in a decrease in the Pd atom density in the
surface. The excess Pd atoms are ejected onto the
terraces where they diffuse, collide, and ultimately
nucleate and grow islands and peninsulas. The
islands and peninsulas exhibit a square structure
with a lattice constant of 0.679± 0.012 nm; the
[1] S.M. Vesecky, D.R. Rainer, D.W. Goodman, J. Vac. Sci.
Technol. A 14 (1996) 1457.
[
[
2] M. Lyubovsky, L. Pfefferle, Catal. Today 47 (1999) 29.
3] M. Lyubovsky, L. Pfefferle, A. Datye, J. Bravo, T. Nelson,
J. Catal. 187 (1999) 275.
[
4] M. Lyubovsky, L.D. Pfefferle, Appl. Catal. A: Gen 173
(1998) 107.
ꢀ
10ꢁ directions of the squares are rotated 15° with
[5] R.J. Farrauto, M.C. Hobson, T. Kenelly, E.M. Waterman,
Appl. Catal. A: Gen 81 (1992) 227.
respect to Pdꢀ110ꢁ directions. Neither the square
nor rectangular structures could be referenced to
any orientation of either Pd or PdO. At the end
of this stage, oxygen desorbs in a single peak at
[
[
[
6] S. Ladas, R. Imbihl, G. Ertl, Surf. Sci. 219 (1989) 88.
7] M.R. Bassett, R. Imbihl, J. Chem. Phys. 93 (1990) 811.
8] V. Bondzie, P. Kleban, D.A. Browne, J. Vac. Sci. Technol
A 11 (1993) 1946.
7
25 K; lower than the 800 K peak temperature
[9] V.A. Bondzie, P. Kleban, D.J. Dwyer, Surf. Sci. 347
1996) 319.
10] T. Furuya, K. Sasaki, Y. Hanakata, K. Mituyasu, M.
Yamada, T. Tsuchiya, Y. Furuse, in: Proceedings of the
Second International Workshop on Catalytic Combustion,
Tokyo, Japan, April 18–20 (1994) 162.
(
observed for chemisorption but higher than the
peak temperature expected for bulk PdO dissoci-
ation, indicating a PdMO bond strength intermedi-
ate between chemisorbed oxygen and bulk PdO.
Thus the second stage involves the formation of
two intermediate phases that are distinctly different
from both the simple chemisorbed oxygen layer
and the bulk oxide. The second stage stops at
[
[11] J.G. McCarty, V.L. Wong, Y.F. Chang, Scr. Metal. Mater.
1 (1994) 1115.
12] H. Conrad, G. Ertl, J. K u¨ ppers, E.E. Latta, Surf. Sci. 65
1977) 245.
3
[
[
(
13] P. L e´ gar e´ , L. Hilaire, G. Maire, G. Krill, A. Amamou,
Surf. Sci. 107 (1981) 533.
2.2 ML, increasing the oxygen coverage further
leads to the third stage: bulk PdO formation.
Oxidation at 500 K produces an oxide that displays
no evidence of crystalline order in either LEED or
STM measurements. Both the formation of four
distinct states and the surface restructuring that
occurs during oxidation are expected to play key
roles in Pd-catalyzed oxidation reactions.
[14] D.L. Weissman-Wenocur, M.L. Shek, P.M. Stefan, I.
Lindau, W.E. Spicer, Surf. Sci. 127 (1983) 513.
[
15] E.H. Voogt, A.J.M. Mens, O.L.J. Gijzeman, J.W. Geus,
Surf. Sci. 373 (1997) 210.
[
16] H. Conrad, G. Ertl, J. K u¨ ppers, E.E. Latta, Surf. Sci. 65
(
1977) 235.
[17] B.A. Banse, B.E. Koel, Surf. Sci. 232 (1990) 275.
[18] D.T. Wickham, B.A. Banse, B.E. Koel, Surf. Sci. 243
(
1991) 83.
[
[
19] J.-W. He, P.R. Norton, Surf. Sci. 204 (1988) 26.
20] S.D. Bader, L. Richter, T.W. Orent, Surf. Sci. 115
(
1982) 501.
[
[
21] T.W. Orent, S.D. Bader, Surf. Sci. 115 (1982) 323.
22] K. Sekizawa, M. Machida, K. Eguchi, H. Arai, J. Catal.
Acknowledgements
1
42 (1993) 655.
[
[
[
23] M. Valden, J. Pere, M. Hirismaki, S. Suhonen, M. Pessa,
Surf. Sci. 377 (1997) 605.
24] M. Valden, J. Pere, N. Xiang, M. Pessa, Chem. Phys. Lett.
The authors thank M. Lyubovsky and L.D.
Pfefferle for stimulating discussions on unusual
effects in Pd catalyzed oxidation reactions. The
authors also acknowledge the assistance of H.L.
Chan, P. Meethunkij, C.Y. Nakakura, M. Piech,
and R.E. Tanner in carrying out this work. This
research is supported by the Petroleum Research
Foundation of the American Chemical Society
under grant number ACS-PRF-34181-AC5. GZ
also acknowledges the support of the Department
of Energy through grant number 96ER14679.
2
57 (1996) 289.
25] E. Broclawik, R. Yamauchi, A. Endou, M. Kubo, A. Miya-
mota, Int. J. Quant. Chem. 61 (1997) 673.
[26] C.Y. Nakakura, V.M. Phanse, G. Zheng, G. Bannon, E.I.
Altman, K.P. Lee, Rev. Sci. Instrum. 69 (1998) 3251.
[
27] R.D. Ramsier, K.-W. Lee, J.T. Yates Jr., J. Vac. Sci. Tech-
nol. A 13 (1995) 188.
28] T. Matsushima, Surf. Sci. 157 (1985) 297.
29] G. Bayer, H.G. Wiedemann, Thermochim. Acta 11
(1975) 79.
[
[

1
68
G. Zheng, E.I. Altman / Surface Science 462 (2000) 151–168
[
[
30] L.L. Lauderback, S.A. Larson, Surf. Sci. 233 (1990) 276.
31] H. Krakauer, M. Posternak, A.J. Freeman, D.D. Koelling,
Phys. Rev. B 23 (1981) 3859.
32] L. Ruan, I. Stensgaard, F. Besenbacher, E. Lægsgaard,
Phys. Rev. Lett. 71 (1993) 2963.
33] D.A. King, Surf. Sci. 47 (1975) 384.
34] L. Ruan, I. Stensgaard, F. Besenbacher, E. Lægsgaard,
J. Vac. Sci. Technol. B 12 (1994) 1772.
[37] E. Schmidtke, C. Schwennicke, H. Pfnur, Surf. Sci. 312
(1994) 301.
[38] B.C. Stipe, M.A. Rezaei, W. Ho, Phys. Rev. Lett. 78
(1997) 4410.
[39] C. Puglia, A. Nilsson, B. Hernnas, O. Karis, P. Bennich,
N. Martensson, Surf. Sci. 342 (1995) 119.
[40] N. Materer, U. Starke, A. Barbieri, R. Doll, K. Heinz,
M.A. van Hove, G.A. Somorjai, Surf. Sci. 325 (1995) 207.
[41] V. Fernandez, K.M. Schindler, O. Schaff, Ph. Hofmann,
A. Theobald, A.M. Bradshaw, V. Fritzsche, R. Davis, D.P.
Woodruff, Surf. Sci. 351 (1996) 1.
[
[
[
[
[
35] J. Haase, B. Hillert, L. Becker, M. Pedio, Surf. Sci. 262
(
1992) 8.
36] M. Pedio, L. Becker, B. Hillert, S. D’Addato, J. Haase,
Phys. Rev. B 41 (1990) 7462.
[42] S.C. Su, J.N. Carstens, A.T. Bell, J. Catal. 176 (1998) 125.