Steam reforming of methanol on PdZn near-surface alloys on Pd(1 1 1) and Pd foil studied by in-situ XPS, LEIS and PM-IRAS

Article abstract of DOI:10.1016/j.jcat.2010.09.006

The CO₂ selectivity in methanol steam reforming was investigated for a multilayer PdZn 1:1 surface alloy (thickness of ～1.3 nm) and for a subsurface-Zn diluted monolayer Pd-Zn surface alloy, both exhibiting a 1:1 composition in the surface layer. Despite having almost the same surface layer stoichiometry, these two types of near-surface alloys exhibit different corrugations and electronic structures. The CO₂-selective multilayer alloy features a lowered density of states close to the Fermi edge and surface ensembles of PdZn exhibiting a Zn-up/Pd-down corrugation, acting as bifunctional active sites both for reversible water activation as ZnOH and for reaction of methanol (via formaldehyde + ZnOH) toward CO₂. The thermochemical stability limit of the multilayer alloy at around 573 K was determined in-situ at elevated pressures of water, methanol and CO, applying in-situ XPS, PM-IRAS spectroscopy, LEIS and AES. Above 573 K, the coordination of the surface 1:1 PdZn layer with subsurface-Zn gradually decreased by bulk diffusion of Zn escaping from the second and deeper layers, resulting in a transition from the CO₂-selective PdZn multilayer state to the unselective monolayer state, which only catalyzes methanol dehydrogenation to CO.

Full text of DOI:10.1016/j.jcat.2010.09.006

Journal of Catalysis 276 (2010) 101–113
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Steam reforming of methanol on PdZn near-surface alloys on Pd(1 1 1) and Pd
foil studied by in-situ XPS, LEIS and PM-IRAS
Christoph Rameshan a,b, Christian Weilach c, Werner Stadlmayr a, Simon Penner a, Harald Lorenz a,
Michael Hävecker b, Raoul Blume b, Tulio Rocha b, Detre Teschner b, Axel Knop-Gericke b, Robert Schlögl b,
Dmitry Zemlyanov d, Norbert Memmel a, Günther Rupprechter c, Bernhard Klötzer a,
⇑
a Institute of Physical Chemistry, University of Innsbruck, Innrain 52a, A-6020 Innsbruck, Austria
b Department of Inorganic Chemistry, Fritz-Haber-Institute of the Max-Planck-Society, Faradayweg 4–6, D-14195 Berlin, Germany
c Institute of Materials Chemistry, Vienna University of Technology, Veterinärplatz 1, A-1210 Vienna, Austria
d Purdue University, Birck Nanotechnology Center, 1205 West State Street, West Lafayette, IN 47907-2057, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 7 June 2010
Revised 1 September 2010
Accepted 4 September 2010
Available online 14 October 2010
The CO2 selectivity in methanol steam reforming was investigated for a ‘‘multilayer” PdZn 1:1 surface
alloy (thickness of ꢀ1.3 nm) and for a subsurface-Zn diluted ‘‘monolayer” Pd–Zn surface alloy, both
exhibiting a 1:1 composition in the surface layer. Despite having almost the same surface layer stoichi-
ometry, these two types of near-surface alloys exhibit different corrugations and electronic structures.
The CO2-selective multilayer alloy features a lowered density of states close to the Fermi edge and surface
ensembles of PdZn exhibiting a ‘‘Zn-up/Pd-down” corrugation, acting as bifunctional active sites both for
reversible water activation as ZnOH and for reaction of methanol (via formaldehyde + ZnOH) toward CO2.
The thermochemical stability limit of the multilayer alloy at around 573 K was determined in-situ at
elevated pressures of water, methanol and CO, applying in-situ XPS, PM-IRAS spectroscopy, LEIS and
AES. Above 573 K, the coordination of the surface 1:1 PdZn layer with subsurface-Zn gradually decreased
by bulk diffusion of Zn ‘‘escaping” from the second and deeper layers, resulting in a transition from the
CO2-selective PdZn ‘‘multilayer” state to the unselective ‘‘monolayer” state, which only catalyzes meth-
anol dehydrogenation to CO.
Keywords:
PdZn near-surface alloy
Pd(1 1 1)
Pd foil
Methanol dehydrogenation
Methanol steam reforming
Water activation
High-pressure XPS
In-situ polarization-modulation IR
Ó 2010 Elsevier Inc. All rights reserved.
1. Introduction
‘‘isolated” bimetallic surfaces (i.e. without contact to the respective
oxidic support under realistic catalytic conditions). Most impor-
Complementing recent advances in Cu/ZnO-based methanol
steam reforming ‘‘MSR” catalysis, Iwasa et al. have identified palla-
dium nanoparticles supported on ZnO, Ga2O3 and In2O3 as highly
selective for MSR [1]. These catalysts have in common that bime-
tallic Pd–M nanoparticles are formed upon H2 treatment. Espe-
cially 1:1 PdZn nanoparticles supported on ZnO offer the
advantage of higher thermal and chemical stability [1,2] than their
Cu counterpart, giving rise to better long-term stability and less
deactivation with time-on-stream [3].
Empirically, the presence of intermetallic particles of a specific
composition, corresponding e.g. to the stable Pd1Zn1 or Pd2Ga bulk
phases, has been shown to be highly beneficial for the desired
selectivity of the MSR process CH3OH + H2O ? CO2 + 3H2 [1] and
for suppression of the methanol dehydrogenation reaction
CH3OH ? CO + 2H2. This result is very interesting, but also raises
several questions regarding details of the action of the respective
tantly, the ‘‘working catalyst” surface may be substantially affected
by chemical and/or thermal modification, or segregation effects.
Under realistic steady-state reaction conditions, a specific dynamic
surface state of PdZn can be expected, directly adapting to the ther-
modynamic limitations resulting from the chosen gas phase.
In this work, we present a detailed model catalyst study of near-
surface PdZn alloys on both Pd(1 1 1) and polycrystalline Pd sam-
ples. The model approach using extended bimetallic surfaces under
close-to-real conditions, in combination with advanced in-situ
spectroscopic techniques (X-ray photoelectron spectroscopy
(XPS) and polarization-modulation infrared absorption spectros-
copy (PM-IRAS) in the mbar range), allows to elucidate the correla-
tion between catalytic selectivity, details of the bimetallic
coordination on and near the surface (‘‘ligand effect”) and surface
composition (‘‘ensemble effect”). Furthermore, the alloy stability
under ambient-pressure reaction conditions can be examined in
a direct manner.
Concerning the already existing knowledge, a number of studies
focused on ZnO-supported PdZn catalysts and Pd–Zn intermetallic
⇑
Corresponding author. Fax: +43 512 507 2925.
E-mail address: Bernhard.Kloetzer@uibk.ac.at (B. Klötzer).
0021-9517/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2010.09.006

102
C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
systems. Reduction of Pd supported on ZnO in H2 at T ꢀ 523 K leads
to the formation of particularly stable 1:1 PdZn nanoparticles crys-
tallizing in the tetragonal AuCu-type lattice [1,2,4]. Both experi-
mental and theoretical studies have already addressed the
important issue of surface composition and segregation trends of
the Pd–Zn intermetallic system (near-surface alloy on a single
crystal and 1:1 PdZn nanoparticles on ZnO [5–11]) as a function
of the Zn:Pd ratio (Zn coverage) and the temperature. For the spe-
cific 1:1 stoichiometry, DFT calculations identified a 1:1 ratio of Pd
and Zn as the most stable configuration also in the surface layer, at
least from the viewpoint of surface/bulk energetics that were not
affected by high gas pressures or elevated temperature [8,10]. Re-
lated experimental studies of the Zn/Pd(1 1 1) system using XPS re-
ported different stages of the alloying process as a function of
temperature, among them the formation of the theoretically pre-
dicted 1:1 surface composition [5,11,12]. Studies performed by
some of us revealed tetragonal AuCu-type PdZn nanoparticles [2]
and a tetragonal PdZn near-surface alloy formed on Pd(1 1 1) at
elevated temperatures and high Zn coverages beyond 5 ML, which
closely resembled the active tetragonal bulk alloy state [6]. Since
the electronic structure of the 1:1 PdZn alloy is similar to that of
metallic copper, a phenomenological interpretation of the ob-
served analogous catalytic performance of both Cu and PdZn was
suggested [5,11–15]. A general concept involving the cooperative
action of the Cu-like electronic structure of an extended PdZn
‘‘multilayer” surface alloy, in combination with a corrugated 1:1
Pd–Zn surface layer, being capable of efficient water activation,
was presented recently by Rameshan et al. [14]. In the current
work, we focus on a more detailed characterization of the most ac-
tive/selective PdZn bimetallic surface state. Details of its high-pres-
sure thermochemical stability and its mechanistic role in the
selective MSR process toward CO2, involving especially the influ-
ence of surface Zn–OH species originating from water activation,
will be discussed.
As highlighted in recent theoretical work [11,16], the presence
of the – with respect to clean Pd – electronically altered Pd–Zn sur-
face is certainly crucial for steering the individual dehydrogenation
steps from methanol to CO. The enhanced stability and thus pref-
erential formation of formaldehyde on the PdZn(1 1 1) surface
and the suppression of its dehydrogenation to CO were already
predicted in [16]. However, the mechanistic link between bimetal-
lic CH2O stabilization and the preferential formation of CO2 in MSR
is still unclear. The latter obviously requires water activation as a
‘‘source of oxygen” at or near the bimetallic surface, in order to
add an additional oxygen atom into a dehydrogenated methanol
species (e.g. CH2O(ads)), to create a CO2 precursor such as formic
acid at the surface. Theory examined the likely forms of formalde-
hyde and water at the Pd–Zn surface, which are then supposed to
interact toward CO2 [11,16,17]. We may ask the questions where
and how water is activated, e.g. at the bimetallic Pd–Zn surface it-
self, at local ‘‘ionic” Zn sites arising from chemically induced segre-
gation, or at the phase boundary PdZn/ZnO, as suggested in [15],
and where and how these oxygen-containing species, e.g. –OH
groups, interact with which specific methanol-derived C1-oxygen-
ates to form the optimum CO2 precursor species. For this reason,
the propensity of Cu or other metal surfaces [18] and of ZnO sur-
faces [19–23] for water dissociation, being an important aspect
for all reforming reactions, has already been studied extensively.
However, pure ZnO is unselective with respect to CO2 formation
and rather favors CO production [24].
dence for a low-temperature reaction channel for the oxidation
of methanol-derived species to CO2 on Pd/ZnO(0 0 0 1), in which
oxygen was supplied by the ZnO(0 0 0 1) support. The preferential
decomposition of methanol toward CO and hydrogen was observed
both on Pd and PdZn deposited on ZnO(0 0 0 1), when the influence
of the oxide-metal phase boundary was suppressed. From this
study, it appears likely that also water activation and the oxidation
processes leading to CO2 involve reactions at the bimetal-oxide
phase boundary. On PdZn, in the absence of water, the formation
of CO and H2 both from formaldehyde and methanol [25] is a mat-
ter of fact, but Rösch and coworkers rather attributed this process
to structural defects present on the 1:1 PdZn system [16].
On the other hand, a complementary study is available which
highlights the exclusive role of bimetallic PdZn particles prepared
on an inert support for highly selective MSR. Suwa et al. showed
that PdZn alloy particles supported on carbon-black led to en-
hanced CO2 selectivity [26]. Again, the unknown oxidation state(s)
of the involved Zn surface species occurring under ‘‘real” MSR con-
ditions hamper a clear experimental proof for the exclusive PdZn
bimetallic action. Nevertheless, this study provided some bias for
the water-activating role of Pd–Zn surface ensembles, or, at least,
local ionic Zn species within a bimetallic surface environment,
e.g. formed under reaction conditions by partial Zn segregation.
The model catalyst approach applied in this work not only al-
lows us to differentiate the catalytic properties of the pure bimetal-
lic systems from those involving an oxide-metal phase boundary,
but – most importantly – to identify the Zn oxidation state(s)
and the electronic state of Pd in-situ on the working catalyst
surface.
2. Experimental
The preparation and characterization of the PdZn model cata-
lysts were performed under ultrahigh vacuum (UHV) and mbar
gas pressure, using three different experimental setups. The first
setup (for high-pressure kinetic studies) is located at Innsbruck
University [27], the second one (for high-pressure in-situ XPS) at
the HZB/BESSY II synchrotron facility (Berlin, Germany) [28], and
the third one (for in-situ PM-IRAS) at Vienna University of Technol-
ogy [29]. A detailed description of all three systems is given in the
supporting information section of [14].
In order to compare the results in an unambiguous manner, the
sample preparation was monitored in all setups by XPS and LEED,
and the same Pd foil sample was used for the in-situ XPS and high-
pressure catalysis studies, using identical preparation/reaction
conditions.
2.1. Innsbruck experimental setup
The UHV system with attached high-pressure reaction cell [27]
is designed for catalytic studies up to 1 bar on a larger piece of
2 cm ꢁ 2 cm polycrystalline Pd foil, allowing us to detect reaction
products and even minor intermediates with high sensitivity,
either by discontinuous sample injection into the gas chromatogra-
phy–mass spectrometry (GC–MS) setup (HP G1800A) or by direct
online MS analysis of the reaction mixture via a capillary leak into
the GC/MS detector. The system consists of an UHV chamber with a
long-travel Z-manipulator and a small-volume Pyrex glass reactor
(52 ml, no hot metal components) attached to the outside of the
UHV chamber and accessible via a sample transfer port. The UHV
chamber is equipped with an XPS/Auger/ISS spectrometer (Thermo
Electron Alpha 110) and a standard double Mg/Al anode X-ray gun
(XR 50, SPECS), an Omicron ISE 100 ion gun to provide the focused
1 keV He+ ions for ISS, an electron beam heater, an ion sputter gun
and a mass spectrometer (Balzers). All the ISS experiments were
The ideal, ‘‘theory-compatible” case of a bifunctional 1:1 PdZn
alloy surface being highly selective, without the need of a phase
boundary to the ZnO support, is opposed by the classical picture
of a mechanistic synergism between PdZn and ZnO. That ZnO
may still be important to enhance CO2 formation has been clearly
demonstrated by Bera and Vohs [15]. These authors found evi-

C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
103
performed at an angle of beam incidence
W
= 45° and a scattering
using high precision piezoresistive pressure gauges, which allow
for a very accurate determination of both the ratio and the total
amount of reactant molecules initially supplied to the reactor –
thus exact control of the ‘‘starting amount” of reactants is possible.
Then, the reaction mixture is completed to 1 bar by a He–Ar mix-
ture of defined composition. The reaction can be started at room
temperature (as it was the case for the present work, in order to
monitor temperature-dependent selectivity effects) using a linear
ramp generator to increase the reactor temperature up to the final
isothermal section, or alternatively be started isothermally already
at the final reaction temperature (gas admission after T-ramp). In
any case, the partial pressure changes vs. time are monitored using
the MSD of our HP GCMS system directly via a capillary leak, which
will withdraw less than 1% of the total reaction mixture during
ꢀ1 h of reaction, so this change can be basically neglected. In the
same turn, the reactor is heated to approach the final temperature,
which increases the total pressure in the system by ꢀ20%. To com-
pensate for the total resulting pressure effect, we always use a
fixed partial pressure of Argon added to the initial reaction mixture
and use the Ar MSD signal change to renormalize all pressure
changes back to the basis of the initial conditions.
In our MSR- and methanol dehydrogenation experiments on ac-
tive Pd–Zn surfaces typically ꢀ50% of the initially admitted meth-
anol is reacted off within 30 min, depending on the catalyst’s
activity. The (already pressure-change corrected) product mass
traces increase from a base level (corresponding to no product in
the beginning) to a final value. The precision of the whole rate-
determination procedure now depends on how precisely the final
reactant pressures can be determined. We use an internal standard
method, i.e. the addition of a well-defined molar amount of the
respective product will create a step in the MSD trace which then
can be used to determine the individual product amounts at a gi-
ven time t with high accuracy. Now we can normalize the MSD
traces to (corrected) pressure change, whereby the limiting (pre/
post) values are exactly known. On the basis of the precisely
known reactor volume, we can now recalculate how many mole-
cules of reactants were consumed/how many product molecules
were formed as a function of time. The first derivative of these
curves will yield molecules consumed/formed per second.
angle of # ¼ 90ꢂ, and after correction for the different cross-sec-
tions, intensity normalization of the Pd and Zn signals was per-
formed relative to the total backscattering yield, i.e.
IZn(normalized) = IZn/(IPd + IZn) and IPd(normalized) = IPd/(IPd + IZn).
The Pd and Zn scattering cross-sections valid for our specific setup
were determined by measurement of the clean Pd foil and a suffi-
ciently thick pure Zn metal surface layer covering all Pd under
identical experimental conditions. For controlled Zn deposition, a
home-built Zn evaporator was attached, which consists of a resis-
tively-heated glass effusive cell filled with Zn shot (99.99%) and a
water-cooled quartz-crystal microbalance to control the amount
of deposited Zn. Samples are transferred by means of a magneti-
cally coupled transfer rod at first to the UHV sample holder and fi-
nally to a Pyrex glass sample holder used inside the reaction cell.
With this all-glass setup, no wires or thermocouples are connected
to the sample during catalytic measurement, and therefore, back-
ground (blind) activity of the reaction cell is negligible. The main
chamber is pumped by a turbomolecular pump, an ion getter pump
and a titanium sublimation pump to a base pressure in the low
10ꢃ10 mbar range. High purity gases (5.0 H2, O2, Ar) were used as
supplied from Messer-Griesheim and dosed via UHV leak valves.
The high-pressure cell is evacuated sequentially by a rotary pump
(via LN2 cooled zeolite trap) and then via the main chamber down
to UHV base pressure and can be heated from outside to 723 K with
an oven covering the cell. For better mixing of the reactants, the
high-pressure cell is operated in circulating batch mode. By using
an uncoated GC capillary attached to the high-pressure cell, the
reaction mixture in the close vicinity of the sample is analyzed
continuously by the mass spectrometer of the GC/MS system. MS
signals of methanol, CO2, CO, H2 and CH2O were externally cali-
brated and corrected for fragmentation (that is, CO and CH2O frag-
ments for methanol, CO fragment for CO2).
A polycrystalline palladium foil (Goodfellow, purity 99.999%,
0.125 mm thick, size 3.5 cm2) was cleaned on both sides by succes-
sive cycles of Ar+ ion bombardment (6.0 ꢁ 10ꢃ5 mbar Ar, 503 eV,
1
l
A sample current), oxidation (5.0 ꢁ 10ꢃ7 mbar O2, T = 1000 K),
and annealing in hydrogen (5.0 ꢁ 10ꢃ7 mbar H2, T = 700 K) and in
vacuum (T = 1000 K) until no impurities were detected by AES
and XPS. Details of the preparation of the PdZn multilayer alloy will
be given in Section 3.1. Methanol and methanol/water mixtures
were degassed by repeated freeze-and-thaw cycles. All the MSR
reactions were conducted with methanol/water mixtures of a
1:10 composition of the liquid phase. This corresponds to a room
temperature partial pressure ratio of methanol/water in the equi-
librium gas phase of 1:2.
To determine TOF values, some additional assumptions with re-
spect to the number of active sites must be made. Since the geo-
metric surface area of our foil sample is precisely known, we
usually assume a 1:1 mixture of (1 1 1) and (1 0 0)-oriented facets,
and thus the average Pd atom density of our foil is assumed at
1.4 ꢁ 1015 atoms/cm2, just half way between the values for the
respective single crystals. On the 1:1 Pd–Zn surface alloys, we
may additionally assume 50% of this number.
The catalytic experiments were performed in a temperature-
programmed manner, i.e. the reaction cell was heated at a constant
linear rate of ꢀ8 K/min to the final temperature of 623 K, and then
kept isothermally at this temperature for ꢀ20 min. Experimental
details will be given in context with the individual reaction runs,
see Sections 3.4 and 3.5. The advantage of the TPR (temperature-
programmed reaction) runs is that pronounced selectivity changes
can be monitored via the partial pressure changes as a function of
the reaction temperature, yielding useful qualitative information
about changes of the reaction mechanism and the catalyst state.
The main strength of the all-glass high-pressure cell, as de-
scribed in detail in [27], is that any heated metal parts coming into
contact with the respective reaction gas mixture are avoided,
which is of paramount importance when measuring reaction rates
on the model catalyst surfaces that exhibit only a small surface
area of a few cm2. Therefore, also blind reaction runs under the ex-
actly identical reaction conditions are regularly performed to en-
sure the ‘‘intrinsic inactivity” of our setup. The reactor volume,
including recirculation pump and gas lines, is fixed, and the initial
reactant mixture is first premixed and then admitted to the reactor
2.2. HZB/BESSY II experimental setup
The HZB/BESSY II system [28] (at beamlines U49/2-PGM1 and
later at ISISS) allowed us to perform in-situ photoelectron spectros-
copy up to 1 mbar total reactant pressures. The data obtained at
these – with respect to the HP cell described in Section 2.1. – rel-
atively low pressures complement the kinetic experiments per-
formed in Innsbruck, which were in part also performed in the
sub-mbar pressure range, in order to study a possible pressure
gap effect between the Innsbruck and the HZB measurements.
The sensitivity of the simultaneous MS detection of the reaction
products at HZB/BESSY II was not sufficient to extract reliable reac-
tion rate and selectivity data for CH2O/CO2, mainly because of an
unfavorable ratio of the large total reactant flow through the XPS
high-pressure cell (which is generally operated in constant flow
mode) relative to the minor amounts of CO2 and CH2O formed on
the low surface area catalyst (only ꢀ0.5 cm2 PdZn alloy on Pd foil
or Pd(1 1 1)). However, experiments using the same reaction

104
C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
conditions with respect to initial reactant pressures, PdZn prepara-
tion setup and reaction temperature range, provided a reliable
‘‘connection” between the data obtained in either experimental
setup.
ported in [33], the microbalance was cleaned by Ar+ sputtering
and cooled to below 270 K during evaporation.
To calibrate the amount of Zn required for one monolayer, Zn
was deposited on the clean Pd foil at ꢀ300 K sample temperature
at an evaporation rate of 12 Hz/min. Zn was successively deposited
in 3 Hz intervals, and Auger spectra of the Zn–L3M45M45 (1G) (ki-
netic energy 992 eV) and the Pd–M5N45N45 (1G) (commonly
MVV) peaks (kinetic energy 327.8 eV) were recorded after each
evaporation/deposition step. After differentiation, the peak-to-
peak height of the Zn signal was plotted versus the Zn exposure
(Fig. 1). In a further experiment, an analogous calibration proce-
dure was performed, but modified by an additional 503 K thermal
annealing step after each deposition. After each step, Auger spectra
were again collected, and the results are depicted in Fig. 1.
In order to interpret the data of Fig. 1, additional information
about the growth mode and morphology of Zn on Pd, both at
ꢀ300 K and 503 K, is required. Some – partly controversial – infor-
mation is reported in the literature. For Zn deposited at 105–150 K
simultaneous multilayer (Frank-van-der-Merwe), growth of Zn on-
top of Pd, as reported in [5,34], can be safely assumed both on
Pd(1 1 1) and polycrystalline Pd. At ꢀ300 K, the degree of PdZn al-
loy formation is still a matter of debate. According to Stadlmayr
et al. [34], alloying on Pd(1 1 1) starts just above 300 K, in good
agreement with an XPS study by Bayer et al. [5], who also observed
the very onset of the transformation from the Pd–Zn(ads) interface
to the PdZn alloy state around 300 K. A recent STM study by Wei-
rum et al. [32] reported spontaneous formation of PdZn bilayer al-
loy islands at ꢀ300 K, as concluded from the local occurrence of
ordered p(2 ꢁ 1) patches and the height of the islands. Considering
the combined information, it appears rather unlikely that temper-
atures below 300 K are sufficient to induce complete alloying of
1 ML of Zn toward a full ‘‘bilayer alloy” situation on Pd(1 1 1). Con-
cluding, only a minor degree of alloying might appear on the
Pd(1 1 1) surface e.g. at 293 K.
The HZB/BESSY II setup is equipped with differentially pumped
electrostatic lenses and a SPECS hemispherical analyzer. The sam-
ple is positioned inside the high-pressure/analysis chamber
ꢀ2 mm away from a 1-mm aperture, which is the entrance to
the lens system separating gas molecules from photoelectrons.
Binding energies (BE) were generally referred to the Fermi edge re-
corded after each core level measurement. Samples were mounted
on a transferable sapphire holder. The temperature was measured
by a K-type Ni/NiCr thermocouple spot-welded to the side of the
sample, and temperature-programmed heating was done by an
IR laser from the rear. Sample cleaning procedures consisted of re-
peated cycles of Ar+ sputtering at room and elevated temperatures,
annealing up to 950 K in UHV, and exposure to O2, followed by
flashing at 950 K for 60 s in UHV. The sample cleanliness was
checked by XPS. Two kinds of samples were used: a smaller piece
of the identical Pd foil sample used in Innsbruck and, in addition, a
(1 1 1)-oriented Pd single crystal.
2.3. Vienna experimental setup
Polarization-modulation infrared reflection absorption spec-
troscopy (PM-IRAS) provides surface vibrational spectra of ad-
sorbed species and can be applied from UHV to ambient
pressure, based on the accurate subtraction of gas phase contribu-
tions [29–31]. As described by the metal surface selection rule [32],
the effective surface intensity of s-polarized infrared light on a me-
tal surface is basically zero and no surface absorption occurs. Thus,
IR spectra acquired in s-polarization are identical to IR gas phase
spectra. In contrast, IR spectra acquired in p-polarization contain
absorption contributions of both surface and gas phase species.
Consequently, (p–s) spectra represent the vibrational signature of
the surface species. Polarization modulation is performed by a
photoelastic modulator (Hinds-PEM-90; ZnSe) as described in
more detail in [29,31]. In order to carry out ambient-pressure
PM-IRAS spectroscopy on single-crystal-based model catalysts, a
UHV-compatible high-pressure cell, coupled to a UHV surface
preparation/analysis system, has been used [31]. After sample
preparation/characterization in UHV, the model catalyst was trans-
ferred, with the help of a xyzh manipulator and still under UHV, to
the spectroscopy/reaction cell, where experiments in the range
from UHV to 1 bar can be performed (such as PM-IRAS using a
BRUKER IFS66v/S spectrometer). The preparation of Pd(1 1 1) and
PdZn multilayer near-surface alloys was performed in analogy to
the Innsbruck and HZB/BESSY experiments and confirmed by LEED
and XPS. For PM-IRAS, CO (purity >99.997%) was passed over a car-
bonyl absorber cartridge and then introduced via a cold trap filled
with liquid nitrogen in order to remove potential Ni- and Fe-car-
bonyl impurities.
In contrast, Zn deposition at temperatures around 300 K on the
Pd foil sample – exhibiting a more structural defects (steps, grain
boundaries) than well-annealed Pd(1 1 1) – induces a much higher
degree of alloying. In the Pd3d spectra taken directly after deposi-
tion (but before heating), the presence of a pronounced Pd3d com-
ponent at ꢀ336.3 eV provided clear evidence.
3. Results and discussion
3.1. The PdZn ‘‘multilayer” alloy preparation verified by AES and LEIS
Prior to catalytic measurements, the preparation of a suitable
‘‘multilayer” PdZn alloy both on the surface of Pd foil and
Pd(1 1 1) was examined, building upon knowledge obtained by
previous structural investigations of Zn deposition and alloy for-
mation on Pd(1 1 1), by Bayer et al. [5], Gabasch et al. [6], Weirum
et al. [32], Kratzer et al. [33] and Stadlmayr et al. [12,34]. To avoid
sticking coefficients of Zn smaller than 1 on the quartz microbal-
ance, which would lead to underestimation of Zn thickness as re-
Fig. 1. Peak-to-peak intensities of the differentiated Zn–L3M45M45 (1G) Auger peak
and the Pd–M5N45N45 (1G) peak (usually denoted as MVV) as a function of the Zn
deposition at 300 K and after annealing at 503 K.

C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
105
From this viewpoint, we interpret the ꢀ300 K AES calibration
data in Fig. 1 by assuming sequential deposition of Zn monolayers
partly in the alloy state and partly on-top of the (already) alloyed
Pd foil surface: the first Zn ML is completed at ꢀ24 Hz, likely yield-
ing the PdZn ‘‘bilayer” structure suggested by Weirum et al. [32],
and the second ML of Zn is completed at ꢀ48 Hz, corresponding
to a pure Zn adlayer on-top of the more or less complete PdZn bi-
layer structure. Finally, ‘‘Zn on Zn” layers are deposited above
48 Hz. This sequence is supported by the obvious changes of the
slope, which can be explained by superimposing the relative AES
intensity contributions of the first, second, third, and all additional
alloy/metal layers.
After completion of the first two alloy layers (the full bilayer al-
loy) at 1 ML of Zn, the (inter)layer below is still clean Pd and does
therefore not contribute to the Zn intensity; after the second ML of
Zn, both the PdZn bilayer and the Zn adlayer on-top of it contrib-
ute, but not the then fourth (inter)layer of Pd, and so on. Essen-
tially, the Zn intensity should level off at the value of pure Zn,
once enough layers were deposited to fully screen the Pd substrate.
In order to cross-check our coverage assignment independently,
the convergence of the Pd and Zn AES signals at ꢀ300 K with
increasing Zn exposure was modeled by a simple calculation of
the exponential intensity attenuation with probe depth.
amount to ꢀ4.0 ML of Zn, as compared to the 4.8 ML derived from
Fig. 1, reasonably confirming our calibration in Fig. 1.
In any case, according to Bayer and coworkers [5,32,34], when
>3 ML Zn deposited on-top of Pd(1 1 1) are annealed to 503 K, com-
plete Zn alloying with Pd is obtained. In Ref. [5], the maximum al-
loy-induced binding energy (BE) of the Pd3d component was
ꢀ336.1 eV and was reached upon annealing at ꢀ500 K. This can
be interpreted in terms of a Zn-coordination number of Pd atoms
of up to 6 (1:1 bulk alloy), which can be regarded as optimum ini-
tial state for catalytic experiments, because an optimized (modi-
fied) surface electronic structure together with a defined (1:1)
surface composition were already verified [12]. Bayer et al. [5] also
distinguished between a ‘‘2D-alloy”, obtained by annealing one
single ML of Zn, and a thicker ‘‘3D” alloy state, obtained by thermal
annealing of a 300 K 3ML Zn deposit or by directly depositing 3ML
at an elevated sample temperature of 503 K. The latter two proce-
dures then give rise to the typical multilayer PdZn alloy state with
a Pd3d component around 336.1 eV. Using only 1 ML of Zn, the
maximum BE of the Pd3d ‘‘2D-alloy” component did not exceed
335.6 eV at the optimum annealing temperature around 503–
550 K, corresponding to complete formation of a bilayered ‘‘2D-
alloy” over the whole surface area, provided that a perfect 1:1 PdZn
ratio is present in the first two metal layers after ‘‘quantitative”
alloying.
On this basis may we assign the pronounced break of the 503 K
calibration curve in Fig. 1, observed at 24 Hz = 1 ML (as derived
from the ꢀ300 K curve), to the completion of the ‘‘2D
alloy = bilayer” state. We therefore have to assume that the first
two alloy layers contribute most of the Zn AES intensity and that
formation of additional alloy layers ‘‘deeper than two” by the extra
Zn coverage >1 ML gives rise to the reduced slope of the 503 K
curve beyond 1 ML, due to the discussed attenuation effect. The
steep and close-to-linear increase in the Zn AES 503 K intensity be-
tween 0 and 22 Hz can be interpreted in terms of an increase in the
relative bilayer alloy island surface area between 0% and 100% be-
tween 0 and 1 ML Zn coverage, in accordance with the growth
model proposed by Weirum et al. [32].
ꢀ
ꢁ
0
1
1
ꢃt
auger
L
B
C
A
328eV
IPd ¼ I0;Pd exp
@
The maximum attainable Zn-LMM signal at a given Zn (multi-)layer
thickness can be calculated by:
ꢀ
ꢁ
0
1
1
ꢃt
auger
L
B
C
A
992eV
IZn ¼ I0;Zn 1 ꢃ exp
@
Here, t is the thickness of the overlayer in Å; Lauger are the electron
attenuation lengths (or Inelastic Mean Free Path (IMFP)), which are
a function of kinetic energies of the Auger electrons; I0 are the
intensities of the respective signals at t ? 1.
In order to determine the catalytically most relevant top layer
composition of the PdZn alloy model catalyst as a function of Zn
From Fig. 1 we derive that the MVV signal of Pd at 114 Hz is
attenuated to ꢀ18% of the value of pure Pd (ratio IPd/I0 of ꢀ0.18).
Vice versa, based on comparison with Zn layers sufficiently
thick to fully screen the Pd underneath, we derived that the Zn-
LMM intensity at 114 Hz corresponds to ꢀ60% of the maximum
attainable intensity (i.e. of pure Zn metal), i.e. at 114 Hz we found
IZn/I0 ꢀ 0.6. L992eV in Zinc is ꢀ12 Å, and L328eV ꢀ 5 Å. Thus, t at
114 Hz can be calculated both from the Pd and Zn signal attenua-
tion by the Zn overlayers:
ꢀ
ꢁ
ꢂ
ꢃ
IPd
I0
auger
328eV
t ¼ ln
ꢃL
¼ 8:6 Å
and alternatively:
ꢀ
ꢁ
ꢂ
ꢃ
IZn
I0
auger
992eV
t ¼ ln 1 ꢃ
ꢃL
¼ 12:8 Å
The discrepancy between the two numbers can be explained if one
accounts for the partial bilayer alloy formation above 300 K. Partial
alloying will cause incorporation of Pd atoms in layers closer to the
top layer, causing less Pd–MVV attenuation than for a simple on-top
layer-by-layer growth scenario without alloying. Vice versa, addi-
tional Zn layers will be required to obtain saturation of the Zn signal
at the value of pure Zn, when compared to the simple on-top depo-
sition case.
The Zn inter-atomic distance in clean Zn metal is 2.7 Å, i.e. using
this calculation we estimate the Zn layer thickness at 114 Hz to
Fig. 2. LEIS-derived Pd:Zn surface composition on Pd foil after ꢀ300 K Zn
deposition and subsequent thermal annealing at 503 K.

106
C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
amount and annealing temperature, additional ꢀ300 K Zn-deposi-
tion + 300–800 K annealing experiments were examined by LEIS.
As can be seen in Fig. 2, the LEIS intensities of Pd and Zn approach
the expected ꢀ1:1 ratio already at a Zn deposition of 1 ML, fol-
lowed by annealing at 503 K (42:58 = Zn:Pd). Beyond 1 ML, up to
2 ML, the ratio gradually changed to 53:47 in favor of Zn (which
can be assigned to an effective 1:1 ratio within the error limit of
our measurements), and above 2 ML this ratio remained almost
constant with further Zn exposure.
corroborated by the LEIS data of Fig. 2. In full analogy to
Pd(1 1 1), diffusion of subsurface-Zn into the deeper Pd bulk sets
in above ꢀ573 K, as evidenced by the strong decrease in the AES
intensity above 573 K. Note that this decrease in AES is associated
with only a small change in the LEIS-detected 1:1 Pd–Zn surface
composition.
The annealing temperature of 503 K, chosen for the ‘‘Zn on Pd
foil” model catalyst preparation, is therefore located just in the
center of the plateau region. It is highly relevant for the catalytic
properties of the surface alloy discussed below, that the surface
composition did not change abruptly above 573 K (i.e. only from
Zn:Pd = 52:48 to 48:52 between 573 and 640 K), whereas the Zn
Auger intensity decreased much faster. Obviously, the top layer
of 1:1 PdZn exhibits a higher thermal stability than the alloy layers
below, resulting in a diffusional loss of the multilayer alloy Zn to
deeper layers before the top layer changes considerably. The tran-
sition to the ‘‘monolayer” alloy state above 573 K can be explained
by a bulk diffusion induced lowering of the Zn coordination of the
topmost alloy layer, thus reducing the overall number of Zn neigh-
bors adjacent to the Pd atoms in the surface layer.
Some of the consequences for XPS spectroscopy of this process
on Pd(1 1 1), for the Pd3d and Zn2p core levels, were already de-
scribed in [5]. In [12], XPS using tunable photon energies at HZB
was used to study the related changes in the VB region in greater
detail, in particular because changes of the VB electronic structure,
especially of the DOS (density of states) near the Fermi edge, are of
paramount importance for the catalytic performance [11,14].
Therefore, we also measured the corresponding series of Pd3d,
Zn3d and valence band XPS spectra for thermal annealing of
3 ML Zn on Pd foil at HZB/BESSY II, which closely resembled the
one obtained on Pd(1 1 1) [12] after the respective equivalent
treatment (thus not shown here). We explicitly state that the ob-
served thermal changes of the PdZn coordination chemistry are
virtually the same on Pd(1 1 1) and Pd foil.
The valence band region shown in [12], thus also valid for the
Pd foil model system, is particularly interesting in view of the
aforementioned ‘‘pseudo-copper” hypothesis for selective MSR
[11,12], suggesting a phenomenological link between the high
selectivities of Cu and PdZn catalysts resulting from their similar
electronic structure. The trend observed in the VB region (BE of 0
to ꢀ6 eV) is in good qualitative agreement with this idea. For clean
Pd and the highly Zn-lean 780 K alloy state, the maximum of the Pd
4d related intensity is below 1 eV, corresponding to a high DOS
close to the Fermi edge. The higher the Zn coordination of the Pd
near-surface layers becomes (with decreasing annealing tempera-
ture), the more this intensity maximum shifts toward higher BE.
For clean copper, we expected a position of the 3d-related intensity
maximum at ꢀ2.3 eV, confirmed by Cu spectra obtained at 650 eV
photon energy in the same apparatus. The high BE shift of the Pd4d
maximum for our ‘‘optimum” PdZn alloy obtained by annealing
3ML Zn at 503 K is very close to this value and does not change
upon annealing to e.g. 540 K, i.e. within the stability plateau of
the multilayer PdZn alloy state. From this viewpoint, we may antic-
ipate that this state is in fact the most ‘‘Cu-like” which can be
prepared.
According to Weirum et al. [32], thermal annealing of >1 ML Zn
offers two different pathways for Zn: desorption back to the gas
phase or alloying to form a thicker alloy which exceeds the
‘‘bilayer = 2D” state. Apparently, the relative rates of desorption
and alloying depend both on the total amount of extra Zn > 1 ML
and on the time available for alloying without fast simultaneous
desorption. The TPD series presented in [32] suggests that at
503 K, the b1-state, assigned to the first pure Zn ML deposited
on-top of the already completed ‘‘bilayer = 2D” PdZn alloy state,
desorbed only at a low rate (peak maximum at ꢀ530 K), whereas
all further Zn layers (i.e. Zn on-top of Zn) desorbed quickly well be-
low 503 K. Consequently, upon holding the sample isothermally at
503 K, in total P2 ML of Zn should contribute to the final alloy
layer, meaning that the multilayer alloy state should consist of four
or more layers of (1:1) PdZn alloy. Also, the differences between
the Pd–Zn Pd3d ‘‘2D-alloy” and ‘‘multilayer” alloy components
(335.6 vs. 336.1 eV, respectively), observed by Bayer et al. after
annealing 1 ML and 3 ML Zn at 503 K [5], should be interpreted
on this basis, suggesting for the multilayer state an increased coor-
dination number of Pd by Zn residing in deeper alloy layers.
On this basis, the region between 1 and 2 ML Zn exposure in
Fig. 2 is interpreted in terms of the transition from the ‘‘2D-alloy”
to the ‘‘multilayer” alloy state, by consumption of the second ML of
Zn to form a multilayered surface alloy consisting of ꢀ4 layers of
PdZn. It appears possible that this is accompanied by a further in-
crease in the Zn:Pd surface ratio toward the eventually expected
stable (1:1) composition of the PdZn bulk alloy state. Another
explanation comes from a recent LEIS study by Stadlmayr et al.
on Zn/Pd(1 1 1), revealing a Zn ‘‘outward” relaxation observed for
the multilayer alloy case, thus improving the ‘‘visibility” of Zn for
ion scattering [12].
The standard preparation procedure, finally chosen for the PdZn
model catalyst (both on Pd(1 1 1) and Pd foil, and irrespective
which UHV system was used), thus involved deposition of 3 ML
Zn at ꢀ300 K (in order to exceed the 2 ML ‘‘multilayer alloy” limit),
followed by thermal annealing at 503 K for 5 min. This procedure
yields the multilayer PdZn alloy state first described by Bayer
et al. [5] and characterized in more detail by Stadlmayr et al. with
respect to its specific surface corrugation by LEIS [12,14] and with
respect to the homogeneous distribution of catalytically active
sites by PM-IRAS [12,14].
3.2. Thermal stability of the Pd–Zn surface alloy on Pd foil determined
by AES and LEIS
The temperature-dependent surface composition and, most
importantly, the thermal stability of the 3ML PdZn multilayer alloy
state both on Pd foil and Pd(1 1 1) in vacuum between ꢀ300 K and
800 K has already been discussed in Ref. [12,14], revealing a de-
fined stability region of the multilayer alloy between 500 and
570 K.
Whereas the rationale in [12] was to study purely the thermal
stability of the multilayer PdZn alloy on Pd(1 1 1) under UHV con-
ditions (i.e. in the absence of reactive gases), here we focus on the
thermochemical stability under realistic catalytic high-pressure
conditions.
In this region, also a well-ordered p(2 ꢁ 1) LEED pattern has
been observed on Pd(1 1 1) by various authors [5,6,32,34], and a
1:1 PdZn composition in the surface and subsurface layers has
been verified by LEIS [12,14]. Hence, we conclude that in this tem-
perature region the ‘‘Pd foil equivalent” of the well-ordered multi-
layer 1:1 terminated PdZn alloy is present, as also strongly
3.3. PdZn multilayer surface alloy: thermochemical stability under
high-pressure conditions
In order to assess the – for the ‘‘real” catalytic world more rel-
evant – thermochemical stability of the multilayer Pd–Zn surface

C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
107
alloy state, we performed thermal annealing experiments exposing
the multilayer alloy to elevated pressures of water (0.24 mbar),
methanol (0.12 mbar), as well as CO (5 mbar).
ure, which occur to a major extent within a ꢀ80 K temperature
range (from ꢀ550 K to 630 K), but also continue at higher
temperatures.
Fig. 3 shows the Pd 3d, Zn 3d and the VB spectral regions during
heating from 473 K to 573 K in water (0.24 mbar; i.e. the pressure
later on used for MSR). The most remarkable observation in the
middle panel of Fig. 3 is a high BE shoulder of the Zn 3d peak at
ꢀ10.5 eV, which increased approximately two times throughout
the chosen T-range. This component was not observed in vacuum
(UHV) annealing experiments (compare with Zn 3d region in
Fig. 3 of [12]). XPS depth profiling by varying the photon energy
demonstrated the exclusive surface nature of the 10.5 eV species
(spectra not shown), being characteristic of an oxidized Zn species
directly at the surface, supposedly ZnO(H). The Pd 3d signal shifts
toward the BE position of metallic Pd upon annealing (left hand pa-
nel). In the VB region, a strong change in the DOS close to the Fermi
edge is observed between 473 K and 573 K, leading to a much more
‘‘Pd-like” electronic structure at 573 K (see Fig. 3, right panel).
In Fig. 4, the binding energies at the peak maxima of the Pd 3d5/
As CO is a (potential) by-product of MSR and a frequently used
probe molecule, we have also examined the thermal stability of the
Zn/Pd(1 1 1) system in a CO environment using PM-IRAS (Fig. 5).
After room temperature Zn deposition on Pd(1 1 1) in UHV, the
sample was annealed (10 min) to the indicated temperatures in a
background pressure of 10ꢃ6 mbar CO (Fig. 5a) or 5 mbar CO
(Fig. 5b). After each annealing step, the Zn/Pd(1 1 1) system was
cooled back to 150 K (10ꢃ6 mbar) or 300 K (5 mbar), and PM-IRAS
spectra were again acquired. This was intended to characterize the
resulting Pd–Zn surface structure under conditions close to UHV
annealing and close to reactive conditions.
The 1:1 Pd–Zn surface alloy is characterized by a single sharp
peak at ꢄ2070 cmꢃ1 (FWHM of 9 cmꢃ1), originating from CO ad-
sorbed on-top of individual Pd atoms in the Pd rows of the
PdZn(1 1 1) surface alloy (cf. Fig. 10). This vibrational frequency
agrees well with the on-top CO peak reported for PdZn–ZnO–
Al2O3 powder catalysts, after hydrogen reduction/activation of
the catalyst [35]. After annealing in 10ꢃ6 mbar CO to 473 K, which
in the range of the alloy stability window, the PdZn alloy was
clearly established. This proves the homogeneity of the entire mul-
tilayer alloy surface, because in case of remaining Pd patches
and Zn 3d core levels are plotted as a function of the annealing
2
temperature, highlighting the stability ranges of the multilayer al-
loy state in water (crossed circles), in clean methanol vapor (open
circles) and in a 1:2 methanol/water steam reforming mixture
(filled circles). For the respective vacuum annealing experiment
compare Ref. [12]. The major spectral changes occur in all gaseous
environments above ꢀ550 K and are largely complete at ꢀ630 K. It
appears that the chemical environment exerts no pronounced
influence on the stability of the multilayer alloy state. Rather, the
changes observed under ambient-pressure conditions are in good
agreement with those observed under UHV conditions [12]. There-
in, e.g. the Pd 3d5/2 peak position analogously shifted from
ꢀ335.9 eV at 500 K to ꢀ335.2 eV at 630 K. We conclude that the
dominant effect is thermal decomposition of the multilayer alloy
by Zn bulk diffusion. Even in clean water vapor, the alloy decompo-
sition is only to a minor extent a surface oxidation process forming
oxidized ZnO(H) species. The related XPS spectra for methanol (not
shown) showed one main difference to the water series of Fig. 3,
namely the absence of the 10.5 eV oxidic ZnO(H) component in
methanol, because clean methanol vapor represents a strongly
reducing chemical environment.
bridge-bonded CO (at ꢄ1955 cmꢃ1
)
and on-top CO (at
ꢄ2085 cmꢃ1) would additionally occur at these p/T conditions
[36,37,31] (note that the diameter of the IR beam is several mm).
Taking into account thermal desorption data (Tdes of CO from PdZn
is ꢀ210 K, hsat = 0.5 ML CO) [38], each Pd atom is occupied by a CO
molecule.
Fig. 5 also shows the effect of higher annealing temperatures,
until the PdZn alloy started to decompose at 573 K < T < 623 K.
The integrated area of the ꢀ2070 cmꢃ1 on-top CO peak is used as
indicator and its temperature dependence is plotted in Fig. 5c. In
agreement with the LEIS/AES/XPS measurements, there is a stabil-
ity window up to ꢀ600 K. Interestingly, the intensity of the on-top
CO species is highest for the corrugated ‘‘Zn-out/Pd-in” surface.
Annealing to 623 and 673 K induced
a decrease of the
ꢀ2070 cmꢃ1 species, which was successively replaced by CO reso-
nances characteristic of Zn-lean alloy (broad peak around
1930 cmꢃ1) and Pd(1 1 1), the latter with bridge and on-top
bonded CO at ꢄ1955 cmꢃ1 and ꢄ2085 cmꢃ1, respectively, and a
shoulder at 1890 cmꢃ1 of hollow-bonded CO [36,37,31].
At this point, it becomes clear that the transition from the mul-
tilayer alloy state to the subsurface-Zn-lean ‘‘monolayer” alloy is
by no means a discontinuous process, as already shown in [12].
Changes of the Zn-coordination number of Pd are of gradual nat-
Fig. 3. Pd 3d5/2, Zn 3d and VB regions measured in-situ during thermal annealing of the 503-K annealed multilayer PdZn alloy on Pd(1 1 1) in 0.24 mbar water pressure
between 473 K and 573 K. Photon energy was 650 eV.

108
C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
Fig. 4. Temperature-dependent peak position of the Pd 3d5/2 and Zn 3d XPS peaks during annealing in water (black filled circles), methanol (open circles) and a methanol/
water steam reforming mixture (1:2, crossed circles).
Fig. 5. (a) PM-IRAS spectroscopy of 3 ML Zn deposited at 300 K on Pd(1 1 1). Spectra were acquired at 150 K in 10–6 mbar CO (a), and at 300 K in 5 mbar CO (b), after
successive annealing to the indicated temperatures. The IR intensity in the frequency range of multiple-coordinated CO (below 2000 cmꢃ1) was multiplied by three in some
spectra. (c) Integrated area of the ꢀ2070 cmꢃ1 on-top CO peak as function of T(anneal).
Annealing the Zn deposit in 5 mbar CO led to a remarkable dif-
ference in the formation of the Pd–Zn alloy (Fig. 5b and c).
Although one may expect that CO ‘‘pulls out” Pd (due to the larger
binding energy of CO on Pd when compared to Zn), it was observed
that elevated CO pressure shifted the onset of alloying to consider-
ably higher temperature. However, once formed, the onset of PdZn
decomposition was again found around 600 K (excluding a simple
deviation in temperature measurement as being responsible). The
origin of this effect is currently not understood but may be related
with the CO-poisoning of surface sites (e.g. steps) that are required
for PdZn formation via (partial) diffusion of the Zn overlayer from
the surface to subsurface regions.
experiment yielded a measurable positive rate of CH2O formation
around 565 K, but no CO2 was detected in the stability range of
the multilayer alloy. Beyond about 565 K, CO started to be formed,
due to switching of the surface to the more Pd-like ‘‘monolayer” al-
loy state. According to related theory work [11,16], the Cu-like
electronic structure of the PdZn valence band is important for
the selective control of the entire series of dehydrogenation steps,
leading from methanol to CO. In particular, the PdZn 1:1 alloy was
predicted to stabilize formaldehyde and suppress its further dehy-
drogenation toward formyl –CHO and finally CO (on Pd, CHx spe-
cies induce
a similar stabilization [30,31]). Our experiment
strongly supports this concept, since formaldehyde formation is
predominant in the stability range of the multilayer Pd–Zn surface
alloy. Once this state decays, the selectivity shifts toward CO. From
the negative formation rate of CH2O (consumption of intermediate
gas phase formaldehyde in the closed batch reactor system) above
ꢀ600 K, we can even deduce that CH2O(gas) is the key intermedi-
ate toward CO formation (note the coincidence of CH2O consump-
tion rate maximum (with negative sign) and (positive) CO
3.4. PdZn multilayer surface alloy: performance in methanol
dehydrogenation without water
The temperature-programmed methanol dehydrogenation run
shown in Fig. 6 was performed in the absence of water on a
503 K annealed multilayer Pd–Zn surface alloy on Pd foil. This

C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
109
formation rate maximum after 29 min). We conclude that the ob-
served selectivity shift results from the change of the combined
electronic and geometric alloy structure above 565 K (typical
methanol conversion ꢀ50%).
Nevertheless, in the absence of water (in contrast to the MSR
experiments of Fig. 8 shown below), the reaction pathway toward
CO2 is not available, which is reasonable since the activation of
water is required to provide a ‘‘source of oxygen” for the further
oxidation of any of the surface C1-oxygenates to CO2. In (pure)
methanol, the oxidized Zn(OH) component in the Zn3d spectra
(ꢀ11.5 eV) was also absent, even when the spectra were acquired
at a photon energy of 130 eV to increase the surface sensitivity
(not shown).
Methanol dehydrogenation on PdZn multilayers was also exam-
ined by PM-IRAS. Fig. 7 (upper panel) shows spectra acquired at
room temperature in 5 mbar CH3OH, comparing Pd(1 1 1) and
PdZn(1 1 1)/Pd(1 1 1). On the Pd surface, methanol was in part
decomposed via CH2O to CO, producing hollow-bonded CO (char-
acterized by the peak at 1873 cmꢃ1, indicative of ꢀ0.4 ML cover-
age). As reported in [30,31,39], this leads, together with the
parallel formation of surface CHx species, to a self-poisoning of
the process at low temperature. On PdZn, no methanol dehydroge-
nation occurred at 300 K.
When the reaction temperature was increased (Fig. 7 lower pa-
nel), methanol was in part dehydrogenated to CO. However, due to
the low binding energy of CO to PdZn, adsorbed CO could not be
detected during the reaction (note that for Pd surfaces, CO result-
ing from methanol decomposition can be easily observed at
423 K [40]). However, the product CO could be detected after cool-
ing the model catalysts to room temperature in the gas. The ob-
served CO frequency was then characteristic of the surface state
of PdZn, i.e. up to reaction temperatures <600 K, the typical
2070 cmꢃ1 peak characterized the PdZn multilayer; whereas upon
reaction at >600 K, the PdZn multilayer decomposed and CO reso-
nances typical of Zn-lean Pd–Zn or of developing Pd patches/is-
lands were observed after cooling. For example, after methanol
dehydrogenation at 623 K, cooling to 300 K in the gas yielded a
spectrum with peaks at 2070 cmꢃ1 (still characteristic of Pd–Zn)
Fig. 6. Temperature-programmed methanol dehydrogenation (initial pressure
60 mbar MeOH) without water addition on the multilayer Pd–Zn surface alloy on
Pd foil. The temperature program involved a linear ramp (6.9 K/min) up to 623 K
followed by isothermal reaction for 21 min.
Fig. 8. MSR temperature-programmed reaction on the multilayer PdZn alloy on Pd
foil (3 ML Zn annealed at 503 K for 5 min). The reaction conditions for the high-
pressure reaction were 12 mbar methanol + 24 mbar water, and for the low
pressure reaction 0.12 mbar + 0.24 mbar, in both cases diluted in 1 bar He inert
gas to ensure reliable gas recirculation. The temperature program involved a linear
ramp (9.3 K/min) up to 623 K followed by isothermal reaction for 24 min
(viewgraph was extracted from supporting information of Ref. [14]).
Fig. 7. PM-IRAS spectra of surface and gas phase species, acquired during and after
methanol reaction on the PdZn multilayer alloy on Pd(1 1 1).

110
C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
Fig. 9. Pd 3d5/2 core level spectra (left), Zn3d spectra (middle) and VB spectra (right) obtained in-situ during methanol steam reforming (0.12 mbar MeOH + 0.24 mbar H2O)
on the multilayer surface alloy (3ML Zn on Pd(1 1 1) annealed to 503 K in vacuum). Pd 3d core level spectra were recorded with 650 eV photon energy, the VB and Zn3d region
with 120 eV in order to enhance the surface sensitivity.
and at 1918 cmꢃ1 (characteristic of bridging/hollow CO on
Pd(1 1 1) [31] and/or of Zn-lean PdZn [41]). Intermediate species
of methanol dehydrogenation such as CH2O could not be detected
(the peak at 1034 cmꢃ1 is due to the C–O stretching vibration of gas
phase methanol), due to a presumably too small steady-state con-
centration below 565 K and its decomposition above 565 K. In
addition, the Pd(1 1 1) single crystal used for PM-IRAS was
<1 cm2 so that the expected amount of CH2O is much smaller than
for Pd foil in the all-glass reactor.
with hardly any activity in CO2-selective MSR. The strong catalyst
deactivation observed after reaching the final temperature of 623 K
after ꢀ15 min could be rationalized by ex-situ AES and XPS carbon
spectra recorded after the MSR experiment, showing increasing
amounts of carbon deposits progressively blocking the active metal
surface with increasing reaction time [39]. Moreover, the experi-
ments were all performed in a re-circulating batch reactor, with
the consequence of decreasing reactant pressures over time, which
also contributes to the isothermal rate decrease in product forma-
tion with time.
Finally, temperature-programmed MSR experiments were car-
ried out on the most promising ‘‘multilayer” alloy state prepared
by thermal annealing of 3 ML Zn on Pd foil at 503 K for 5 min
(Fig. 8). The reason for the comparison of the ‘‘low- and high-pres-
sure” experiments in Fig. 8 was to connect the batch reactor stud-
ies (36 mbar) with the experimental conditions chosen for the
corresponding in-situ XPS experiment at HZB/BESSY II (see
Fig. 9), which was limited in total pressure to about 0.36 mbar.
The analogous 0.36 and 36 mbar experiments in the all-glass HP
cell in Innsbruck allowed us to establish an unambiguous correla-
tion between the catalytic and spectroscopic measurements per-
formed in the two different systems. In fact, Fig. 8 shows no
substantial ‘‘pressure gap” effect between 0.36 mbar and 36 mbar
total reactant pressure, when using the same MeOH/water pres-
sure ratio.
3.5. MSR on clean Pd vs. ‘‘monolayer” vs. ‘‘multilayer” Pd–Zn surface
alloy
In a recent publication [14], the MSR catalytic performance of
the 630 K annealed subsurface-Zn-lean ‘‘monolayer” surface alloy
was compared with that of clean Pd foil under otherwise identical
experimental conditions.
Both samples were characterized before reaction by LEIS con-
firming the ꢀ1:1 Pd:Zn ratio of the 2D alloy surface (after anneal-
ing 3 ML Zn deposit to 623 K; cf. Fig. 2), and the absence of Zn on
clean Pd. On both types of surfaces, CO was the only reaction prod-
uct observed during the temperature-programmed reaction,
whereas CH2O remained below the detection limit throughout
the whole temperature range. Despite the presence of 0.24 mbar
water in the gas phase, CO2 was also not detected. LEIS performed
after the reaction showed an only slightly changed Pd:Zn ratio of
56:44, i.e. the ‘‘monolayer” alloy exhibited high thermochemical
stability. Moreover, the in-situ Zn3d + VB spectra did not change
during MSR (and are thus not shown) and did also not indicate
any water-induced Zn(OH) component at ꢀ10.5 eV. This clearly
shows that the ‘‘monolayer” alloy is apparently not capable of acti-
vating water as efficiently as the ‘‘multilayer” surface alloy state,
even though the ‘‘multilayer-to-monolayer” change in the Pd:Zn
surface ratio is small in between ꢀ573 and 623 K. Possibly, water
activation is suppressed by the more Pd-like electronic structure
of the ‘‘monolayer” alloy surface, despite the presence of suitable
Pd–Zn surface ensembles (but this awaits theoretical
confirmation).
The absence of the final product CO2 can therefore be rational-
ized by the lack of the required ‘‘source of oxygen” both on PdZn
monolayer and Pd. For the same reason, clean Pd metal is well
known to be an excellent methanol dehydrogenation catalyst but
Fig. 10. Left: p(2 ꢁ 1) surface structure of the 1:1 multilayer PdZn alloy on
Pd(1 1 1). Right: side view of the multilayer PdZn alloy with likely surface
intermediates reacting toward CO2 (viewgraph was extracted from supporting
information of Ref. [14]).

C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
111
For the MSR reaction mixture, the temperature region of CH2O
formation and of the selectivity change from CH2O toward CO is
well comparable to the experiment of Fig. 6 with (pure) methanol.
The main difference is the much higher CO2 fraction in the prod-
ucts, likely because the reaction channel from C1-oxygenates to-
ward CO2 was now available due to the activation of water.
Whereas CH2O was formed and then again consumed by transfor-
mation to CO (negative reaction rate of CH2O coincided again with
positive CO formation rate at ꢀ22 min), revealing its intermediate
nature, CO2 was a final product. Again, deactivation of the catalyst
in the isothermal part of the experiment occurred via progressive
carbon deposition at 630 K, as verified by the decreasing CO forma-
tion rate and post-reaction carbon AES spectra.
supported catalysts, due to much stronger contribution of ZnO sur-
face or metal–support interface sites, and 50% of the original Pd
surface atoms (on the clean foil) are available for reaction).
In contrast to the supported catalysts, at ꢀ500 K our model cat-
alyst surface is hardly active for MSR, possibly because of facili-
tated water activation on the supported catalysts, and the
maximum MSR activity at the highest selectivity for CO2 is ap-
proached between 560 and 570 K. Moreover, it seems that most
MSR experiments on supported catalysts were done below
ꢀ523 K in order to avoid selectivity losses due to the increasing
influence of the inverse water gas shift reaction at higher temper-
atures, thus no direct comparison at the equivalent temperatures
was possible.
The results of Fig. 8 imply that formaldehyde is the key interme-
diate toward CO2 at least up to ꢀ573 K, but starts to act as an inter-
mediate toward CO once the transition from the multi- to the
monolayer alloy has started, in the sense that intermediate CH2O
already desorbed to the gas phase becomes later-on dehydrogenat-
ed to CO by re-adsorption and reaction at higher temperatures, i.e.
on the more Pd-like ‘‘monolayer”-alloy state of the surface.
A mechanistic model of the MSR reaction in the stability range
of the multilayer alloy up to 573 K should therefore not only in-
volve the suppression of CH2O dehydrogenation to CO, resulting
from the particular electronic surface structure of the PdZn multi-
layer, but should also account for efficient water activation on this
surface, yielding e.g. OH(ads) at Zn sites. The resulting ‘‘oxygen
donor” surface species, e.g. ZnOH, is then capable of fully oxidizing
at least a part of the surface-adsorbed CH2O. As some form of sur-
face-bonded formaldehyde may either become quickly oxidized by
interaction with the observed Zn–O(H) species or simply thermally
desorbed to the gas phase, the relative probabilities of thermal
desorption as competing ‘‘chemical reaction” to total oxidation will
determine the product partial pressure ratio. If the overall rate-
limiting step is located within the dehydrogenation steps prior to
formaldehyde, both the subsequent formaldehyde desorption and
total oxidation pathways should be much faster, which can explain
the observed simultaneous formation of CO2 and CH2O up to 573 K.
This scenario is supported by Ranganathan et al. [35], who reported
a faster conversion of formaldehyde than methanol toward CO2 un-
der otherwise identical reaction conditions. Moreover, the reaction
probability toward CO2 appears to be much higher on the sup-
ported PdZn/ZnO systems, which may be attributed to the much
higher abundance of ZnO surface or ZnO/Pd–Zn interface sites,
i.e. the surface chemistry of the support and/or of the metal–sup-
port interface (especially for efficient water activation) must be
considered for PdZn/ZnO, whereas these sites must be absent on
our model surface. Why these sites are more reactive for formalde-
hyde conversion than the ‘‘isolated” PdZn bimetal surface sites is
an open question.
Another difference is important, namely the total reactant pres-
sure, which was 36 mbar at maximum in our MSR experiments,
but much higher (around 1 bar) in the supported catalyst studies
cited above. This is certainly another main reason for the some-
what lower TOF values derived for our model catalyst, but even
so we are not far off the ‘‘lower end” of the usual values.
We also attempted to estimate an apparent activation energy
for the initial increase in the H2CO and CO2 formation rates, with
the assumptions of a stable surface and hardly any change of meth-
anol concentration up to ꢀ573 K. We calculated an apparent Ea of
around 1 eV for the initial CO2 formation between 510 and 560 K
(with a large error bar of at least 20%), which may be compared
to recent DFT work e.g. by Huang et al. [35], reporting an absolute
value of 0.91 eV for the most likely rate determining methoxy for-
mation from methanol on a bilayered PdZn 1:1 surface alloy. Of
course, as always with absolute energies from DFT, e.g. the zero
point energy corrections and other computational assumptions
make a big difference, so one has to be cautious with such
comparisons.
The XPS spectroscopic fingerprints associated with MSR on the
(initially present) 503 K multilayer Pd–Zn surface alloy on
Pd(1 1 1) at 0.36 mbar total reactant pressure are shown in
Fig. 9. The decomposition of the PdZn multilayer alloy occurs
above 573 K, as expected from the thermal and thermochemical
stability measurements, which again manifests itself by the shift
of Pd 3d5/2 toward lower BEs (Fig. 9, left panel) and by the trans-
formation of the VB from the ‘‘Cu-like” to the ‘‘Pd-like” density of
states. According to the related kinetic measurements of Fig. 8, at
this temperature CO formation sets in. Despite the formation of
the surface-limited Zn–OH species, the Zn-rich alloy remained
stable up to 573 K. The oxidized state of Zn, which can be seen
as a shoulder at ꢀ10.5 eV of the main Zn 3d peak (Fig. 9, middle
panel) is present already at the lowest temperature, but only sta-
ble up to ꢀ573 K, i.e. once the stability limit of the multilayer al-
loy was exceeded, the related Zn–OH species also became
unstable.
On the supported PdZn/ZnO catalyst, formaldehyde was also
suggested as central intermediate, but it could not be detected
due to its small steady-state concentration [1]. It was, however, de-
tected indirectly via follow-up products of formaldehyde, such as
methyl formate HCOOCH3 (methanol condensation product, high
selectivity only in the absence of water at T ꢀ 473 K).
These spectral series illustrates the cooperative, bifunctional ac-
tion of the lowered DOS near the Fermi edge and the water activa-
tion by the Pd–Zn surface ensembles. At the maximum of CH2O/
CO2 selectivity, the effects of suppressing the ‘‘CH2O to CO path-
way” by the still prevailing multilayer alloy electronic structure
and the oxidized Zn species, likely the source of oxygen for the
reaction of C1-oxygenates toward CO2, combine in a unique man-
ner. Experimentally, the mechanistic role of the Zn–OH species
was additionally corroborated by water treatment of the surface
(see Fig. 3) followed by testing of the reducibility of the Zn–OH
species in different gases (H2, methanol, CO). At around 560–
570 K, there are different ways to suppress the ZnOH intensity in
the Zn 3d region: exposure to methanol or hydrogen after water
oxidation, or also switching off the water partial pressure during
MSR. The Zn–OH species can in turn be reversibly populated by
switching from clean methanol to the MSR mixture or to clean
water.
Also a comparison of ‘‘benchmark” turnover rates from the lit-
erature on different catalyst systems with the values of Fig. 8
may be useful at this point. Iwasa et al. reported a TOF of 0.497/
3 = 0.166 on supported Pd/ZnO (division by three because turnover
is given with respect to H2 formation) at 493 K [1], whereas Ranga-
nathan et al. find 0.8 at 503 K, again on Pd/ZnO [35] and Conant
et al. report 0.39 at 523 K, on a comparable supported catalyst
[35]. These values are somewhat higher than those reported in this
work: we estimate a TOF of 0.01ꢅ2ꢅ2 = 0.04 at ꢀ540 K and a maxi-
mum TOF of 0.031ꢅ2ꢅ2 = 0.144 at 570 K (assumptions for factor 4:
all formaldehyde would become quickly converted toward CO2 on

112
C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
The Pd 3d and VB results shown in Fig. 9 are complemented
Metallic surfaces generally favor decarboxylation of formates,
which would finally result in CO2 (gas) and additional H(ads), the
latter finally desorbing associatively to H2(gas). Nevertheless, re-
cent attempts to verify this mechanistic picture by DRIFTS spec-
troscopy on Cu catalysts were unsuccessful [46]. On Pd–Zn this
mechanism is not supported by spectroscopic information, too,
and has to be considered as speculative at present.
well by the related C 1s spectra of Ref. [14] (supporting informa-
tion). The presence of CH2O or related oxygenates on the surface
was observed therein in the temperature range up to 505 K.
Around 573 K, the surface CH2O was replaced by CO. It was con-
cluded that gas phase CH2O is formed just in between 505 and
573 K by desorption of the related surface species, whereas forma-
tion and desorption of CO takes place above 573 K, once the
‘‘monolayer” alloy state starts to develop. PM-IRAS spectra of
MSR did not yield further information yet.
4. Conclusions
A thermal/thermochemical stability limit of up to 573 K has
been verified for the MSR (CO2)-selective PdZn multilayer alloy
phase on Pd(1 1 1) and Pd foil via thermal annealing experiments
in vacuum, mbar pressure of water, methanol, CO, and under
MSR reaction conditions. Up to this limit, the PdZn multilayer alloy
model catalyst maintained the predicted Cu-like electronic struc-
ture of the valence band even under realistic MSR conditions. Con-
sequently, the dissociation barrier between CH2O and –CHO was
higher than for pure Pd, thus leading to increased formation of
formaldehyde in the gas phase, which we were able to verify
experimentally as the central intermediate. The theory-predicted
control of dehydrogenation steps of CH2O could not only be con-
firmed, but moreover the importance of effective total oxidation
of CH2O by activated water at oxidized Zn sites on the PdZn cata-
lyst surface could be proven experimentally. The formation of
CH2O and CO2 at reactive bifunctional Pd–Zn surface ensembles
exhibiting a Cu-like local density of states can be explained on
the basis of formaldehyde as common surface intermediate either
desorbing to the gas phase or becoming oxidized by Zn–OH
species.
3.6. Surface chemistry on PdZn during MSR
At this stage, it becomes clear that previous ideas for selective
MSR, involving full methanol dehydrogenation followed by CO
conversion to CO2 via the water gas shift reaction, are not applica-
ble to the PdZn alloy system. On the basis of kinetic studies in mi-
cro-channel reactors, Pfeifer et al. [42] have shown that methyl
formate is unlikely the main sequential intermediate toward CO2.
It rather represents a by-product resulting from the reaction of for-
mate species with a sufficient amount of coadsorbed methanol, i.e.
when little water is present, as also concluded from [1].
The central role of formaldehyde is apparent from the present
experiments, which leads us to the question of the local structural
and molecular chemistry responsible for selective CO2 formation.
Fig. 10 shows a schematic drawing of the 1:1 Pd:Zn surface alloy
with its well established p(2 ꢁ 1) surface structure [5,6,32,34],
along with the most likely intermediates toward CO2.
In analogy to Cu surfaces, Iwasa et al. suggested that aldehydes
on the PdZn alloy are preferentially adsorbed and stabilized in the
g
1(O)-structure (bonding via oxygen, right side), while they rather
exist as g
2(C,O)-structure (left side) on metallic Pd and Pt. On clean
Acknowledgments
Pd these ‘‘C–O bond weakened” aldehyde species more easily un-
dergo transformation to carbonaceous deposits via C–O bond scis-
sion, but also decarbonylation to carbon monoxide and hydrogen
[1,11,31]. Both forms of formaldehyde have been detected by Jer-
oro et al. using HREELS [43], coexisting after formaldehyde adsorp-
tion on the Pd–Zn surface in the temperature region 150–200 K,
This work was financially supported by the Austrian Science
Fund through Grant P208920-N19 and by TU Vienna via IP 2008
VSFG. Ch. Rameshan acknowledges a PhD scholarship granted by
the Max-Planck-Society. Support for the measurements at HZB/
BESSY II was granted through EU program RII3-CT-2004-506008,
Proposal No. 2008_2_80336. The authors thank the BESSY staff
for their support of the in-situ XPS measurements.
whereby a trend toward the g
1(O)-structure was observed at high-
er Zn loadings of Pd(1 1 1). Nevertheless, a clear assessment of
higher reactivity to either species was not possible. DFT calcula-
tions favored a top–bridge–top (tbt) configuration with the C and
References
O atoms each bridging two substrate atoms, i.e. a g
2(C,O)-structure
[1] (a) A. Szizybalski, F. Girgsdies, A. Rabis, Y. Wang, M. Niederberger, T. Ressler, J.
Catal. 233 (2005) 297;
(b) N. Iwasa, N. Takezawa, Top. Catal. 22 (2006) 215.
[2] S. Penner, B. Jenewein, H. Gabasch, B. Klötzer, D. Wang, A. Knop-Gericke, R.
Schlögl, K. Hayek, J. Catal. 241 (2006) 14.
[3] J.D. Holladay, Y. Wang, E. Jones, Chem. Rev. 104 (2004) 4767.
[4] M. Lenarda, E. Moretti, L. Storaro, P. Patrono, F. Pinzari, E. Rodriguez-Castellon,
A. Jimenez-Lopez, G. Busca, E. Finocchio, T. Montanari, R. Frattini, Appl. Catal. A
312 (2006) 220.
[5] A. Bayer, K. Flechtner, R. Denecke, H.-P. Steinrück, K.H. Neyman, N. Rösch, Surf.
Sci. 600 (2005) 78.
[6] H. Gabasch, S. Penner, B. Jenewein, B. Klötzer, A. Knop-Gericke, R. Schlögl, K.
Hayek, J. Phys. Chem. B 110 (23) (2006) 11391.
[7] J.A. Rodriguez, Prog. Surf. Sci. 81 (2006) 141.
[8] Z. Chen, K.M. Neyman, N. Rösch, Surf. Sci. 548 (2004) 291.
[9] K.M. Neyman, R. Sahnoun, C. Inntam, S. Hengrasmee, N. Rösch, J. Phys. Chem. B
108 (2004) 5424.
[10] Z. Chen, K.M. Neyman, A.B. Gordienko, N. Rösch, Phys. Rev. B 68 (2003)
075417.
[11] K.M. Neyman, K.H. Lim, Z.-X. Chen, L.V. Moskaleva, A. Bayer, A. Reindl, D.
Borgmann, R. Denecke, H.-P. Steinrück, N. Rösch, PCCP 9 (27) (2007) 3470.
[12] W. Stadlmayr, Ch. Rameshan, Ch. Weilach, H. Lorenz, M. Hävecker, R. Blume, T.
Rocha, D. Teschner, A. Knop-Gericke, D. Zemlyanov, S. Penner, R. Schlögl, G.
Rupprechter, B. Klötzer, N. Memmel, J. Phys. Chem. C 114 (2010) 10850.
[13] A.-P. Tsai, S. Kameoka, Y. Ishii, J. Phys. Soc. Jpn. 73 (2004) 3270.
[14] Ch. Rameshan, W. Stadlmayr, C. Weilach, S. Penner, H. Lorenz, M. Hävecker, R.
Blume, T. Rocha, D. Teschner, A. Knop-Gericke, R. Schlögl, N. Memmel, D.
Zemlyanov, G. Rupprechter, B. Klötzer, Angew. Chem. Int. Ed. 49 (2010) 3224.
[15] P. Bera, J.M. Vohs, J. Phys. Chem. C 111 (19) (2007) 7049.
[16] K.H. Lim, Z.X. Chen, K.M. Neyman, N. Rösch, J. Phys. Chem. B 110 (2006) 14890.
of rather weakly adsorbed formaldehyde. The elementary step of
formaldehyde dehydrogenation starts from this species and is sim-
ilar on Pd(1 1 1), Cu(1 1 1) and also PdZn(1 1 1), whereby the bar-
riers on Pd–Zn and Cu are significantly higher [11,16].
A
thermodynamic analysis of coupled water dissociation equilibria
involving H(ads), OH(ads) and O(ads), based on DFT-derived free
surface energies, was presented in [11,17] as well. Accordingly, at
a given water pressure of ꢀ1 atm at ꢀ500 K, combined with hydro-
gen pressures in the range 10ꢃ6–1 atm (being realistic H2 back-
ground pressures during MSR), the dominant oxygen-containing
surface species on the PdZn alloy was OH(ads) rather than O(ads).
At present, open mechanistic questions are rather related to the
elementary reaction steps, starting out from coadsorbed formalde-
hyde and water species, and finally ending with the desorption of
CO2 and H2. Assuming the reactive species being the dominant
OH(ads) at 3-fold hollow sites (with a larger coordination of Zn
atoms; as calculated in [17]) and g
2 formaldehyde in tbt configura-
tion, a mechanism already qualitatively suggested by Takezawa
and Iwasa [44] and Takahashi et al. [45] for a Cu/SiO2 catalyst could
possibly proceed via a dioxomethylene species originating from
the addition of surface hydroxyls to adsorbed formaldehyde. Diox-
omethylene, also the intermediate of the reverse methanol synthe-
sis reaction, is then supposed to dehydrogenate to a formate group.

C. Rameshan et al. / Journal of Catalysis 276 (2010) 101–113
113
[17] K.H. Lim, L. Moskaleva, N. Rösch, ChemPhysChem 7 (2006) 1802.
[18] N. Schumacher, A. Boisen, S. Dahl, A.A. Gokhale, S. Kandoi, L.C. Grabow, J.A.
Dumesic, M. Mavrikakis, I. Chorkendorff, J. Catal. 229 (2005) 265.
[19] O. Dulub, B. Meyer, U. Diebold, Phys. Rev. Lett. 95 (2005) 136101.
[20] M.A. Henderson, Surf. Sci. Rep. 46 (2002) 1.
[21] C.T. Au, J. Breza, M.W. Roberts, Chem. Phys. Lett. 66 (1979) 340.
[22] B.J. Hinch, L.H. Dubois, J. Chem. Phys. 96 (1992) 3262.
[23] A.F. Carley, P.R. Davies, M.W. Roberts, N. Shukla, Y. Song, K.K. Thomas, Appl.
Surf. Sci. 81 (1994) 265.
[32] G. Weirum, M. Kratzer, H.-P. Koch, A. Tamtögl, J. Killmann, I. Bako, A. Winkler,
S. Surnev, F.P. Netzer, R. Schennach, J. Phys. Chem. C 113 (22) (2009) 9788.
[33] M. Kratzer, A. Tamtögl, J. Killmann, R. Schennach, A. Winkler, Appl. Surf. Sci.
255 (11) (2009) 5755.
[34] W. Stadlmayr, S. Penner, B. Klötzer, N. Memmel, Surf. Sci. 603 (1) (2009) 251.
[35] T. Conant, A.M. Karim, V. Lebarbier, Y. Wang, F. Girgsdies, R. Schlögl, A. Datye, J.
Catal. 257 (2008) 64;
E.S. Ranganathan, S.K. Bej, L.T. Thompson, Appl. Catal. A 289 (2005) 153;
Y. Huang, Z.X. Chen, Langmuir 26 (13) (2010) 10796.
[24] A. Neramittagapong, S. Hoshino, T. Mori, J. Kubo, Y. Morikawa, Chem. Lett. 11
(2002) 1078.
[25] J. Pantförder, S. Pöllmann, J.F. Zhu, D. Borgmann, R. Denecke, H.-P. Steinrück,
Rev. Sci. Instrum. 76 (2005) 014102.
[26] Y. Suwa, S. Ito, S. Kameoka, K. Tomishige, K. Kunimori, Appl. Catal. A 267
(2004) 9.
[36] H. Unterhalt, G. Rupprechter, H.-J. Freund, J. Phys. Chem. B. 106 (2002) 356.
[37] G. Rupprechter, H. Unterhalt, M. Morkel, P. Galletto, L. Hu, H.-J. Freund, Surf.
Sci. 502–503 (2002) 109.
[38] A. Tamtögl, M. Kratzer, J. Killman, A. Winkler, J. Chem. Phys. 129 (2008) 224706.
[39] M. Morkel, V.V. Kaichev, G. Rupprechter, H.-J. Freund, I.P. Prosvirin, V.I.
Bukhtiyarov, J. Phys. Chem. B 108 (2004) 12955.
[27] W. Reichl, G. Rosina, G. Rupprechter, C. Zimmermann, K. Hayek, Rev. Sci.
Instrum. 71 (3) (2000) 1495.
[40] M. Bäumer, J. Libuda, K.M. Neyman, N. Rösch, G. Rupprechter, H.-J. Freund,
Phys. Chem. Chem. Phys. 9 (2007) 3541.
[28] H. Bluhm, M. Hävecker, A. Knop-Gericke, E. Kleimenov, R. Schlögl, D. Teschner,
V.I. Bukhtiyarov, D.F. Ogletree, M. Salmeron, J. Phys. Chem. B 108 (2004)
14340.
[29] G. Rupprechter, C. Weilach, J. Phys. Condens. Matter 20 (2008) 184020.
[30] M. Borasio, O. Rodríguez de la Fuente, G. Rupprechter, H.-J. Freund, J. Phys.
Chem. B Lett. 109 (2005) 17791.
[41] I. Bako, R. Schennach, G. Palinkas, J. Phys.: Conf. Ser. 100 (2008) 052067.
[42] P. Pfeifer, A. Kölbl, K. Schubert, Catal. Today 110 (2005) 76.
[43] E. Jeroro, J.M. Vohs, J. Am. Chem. Soc. 130 (2008) 10199.
[44] N. Takezawa, N. Iwasa, Catal. Today 36 (1997) 45.
[45] K. Takahashi, N. Takezawa, H. Kobayashi, Appl. Catal. 2 (6) (1982) 363.
[46] B. Frank, F.C. Jentoft, H. Soerijanto, J. Kröhnert, R. Schlögl, R. Schomäcker,
J. Catal. 246 (2007) 177.
[31] G. Rupprechter, Adv. Catal. 51 (2007) 133.