Adsorption geometry of CN on Cu(1 1 1) and Cu(1 1 1)/O

Article abstract of DOI:10.1016/j.susc.2004.06.167

The adsorption geometry of CN on Cu(111), both with and without predosing with oxygen, has been investigated using N K-edge near-edge X-ray absorption fine structure (NEXAFS) and C 1s and N 1s scanned-energy mode photoelectron diffraction (PhD). The NEXAFS shows clearly that adsorbed onto clean Cu(111) the C-N axis is closely parallel to the surface, but in the presence of coadsorbed oxygen the average orientation has the axis tilted by 25° away from the surface; this confirms a much earlier report of an oxygen-induced reorientation of CN on this surface based on vibrational spectroscopy. The PhD data show very weak modulations which are rather insensitive to the emission geometry, clearly implying a high degree of disorder or a local adsorption site well-removed from any position of high point group symmetry. The best-fit structure corresponds to the CN lying slightly displaced from the three-fold coordinated hollow sites but with the C and N atoms having single Cu atom nearest neighbours at distances of 1.98±0.05 and 2.00±0.05 A? respectively.

Full text of DOI:10.1016/j.susc.2004.06.167

Surface Science 563 (2004) 159–168
www.elsevier.com/locate/susc
Adsorption geometry of CN on Cu(1 1 1) and Cu(1 1 1)/O
a
M. Polcik , M. Kittel , J.T. Hoeft , R. Terborg , R.L. Toomes ,
a
a
a
b
b,
*
D.P. Woodruff
a
Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14195 Berlin, Germany
b
Department of Physics, University of Warwick, Coventry CV4 7AL, UK
Received 26 May 2004; accepted for publication 11 June 2004
Available online 25 June 2004
Abstract
The adsorption geometry of CN on Cu(1 1 1), both with and without predosing with oxygen, has been investigated
using N K-edge near-edge X-ray absorption fine structure (NEXAFS) and C 1s and N 1s scanned-energy mode
photoelectron diffraction (PhD). The NEXAFS shows clearly that adsorbed onto clean Cu(1 1 1) the C–N axis is closely
parallel to the surface, but in the presence of coadsorbed oxygen the average orientation has the axis tilted by 25ꢀ away
from the surface; this confirms a much earlier report of an oxygen-induced reorientation of CN on this surface based on
vibrational spectroscopy. The PhD data show very weak modulations which are rather insensitive to the emission
geometry, clearly implying a high degree of disorder or a local adsorption site well-removed from any position of high
point group symmetry. The best-fit structure corresponds to the CN lying slightly displaced from the three-fold
coordinated hollow sites but with the C and N atoms having single Cu atom nearest neighbours at distances of
ꢀ
1.98 0.05 and 2.00 0.05 A respectively.
ꢁ 2004 Elsevier B.V. All rights reserved.
Keywords: Photoelectron diffraction; Near edge extended X-ray absorption fine structure (NEXAFS); Chemisorption; Surface
structure, morphology, roughness, and topography; Copper; Cyanogen; Oxygen; Low index single crystal surfaces
1. Introduction
tion of amino acids (e.g. [1]). At these solid–liquid
interfaces there is a general consensus that the CN
adsorbs with is molecular axis essentially perpen-
dicular to the surface bonding through the C end
(e.g. [2–5]), although in some cases there may be
some tilting [6] and the extent to which this
bonding is predominantly ionic or significantly
covalent has been a matter of debate [5]. Adsorbed
CN has also been investigated in ultra-high vac-
uum studies of gas-surface reactions, mainly by
dissociative adsorption of cyanogen, C2N2; of
particular practical importance in this case is the
identification of a CN-containing surface inter-
mediate in the reaction of CO with N2 and NO
Studies of the adsorption of the cyanide species,
CN, at metal–solution and electrode–electrolyte
interfaces, using various vibrational spectroscop-
ies, have been motivated in part by its relevance to
electroplating, although this species has also been
identified as a surface intermediate in the oxida-
* Corresponding author. Tel.:+44-2476523378; fax: +44-
2476692016.
E-mail address: d.p.woodruff@warwick.ac.uk (D.P. Wood-
ruff).
0039-6028/$ - see front matter ꢁ 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2004.06.167

160
M. Polcik et al. / Surface Science 563 (2004) 159–168
[7–9], raising the spectre of poisonous HCN pro-
duction in three-way automotive exhaust catalysts.
By contrast to the solid–liquid interface studies,
investigations of adsorbed CN at the solid–vac-
uum interface show the molecular axis to be
essentially parallel to the surface.
note that ab initio calculations of CN on Ni(1 0 0)
have found that while there is almost no energetic
difference between perpendicular and parallel ori-
entations of the C–N axis on the clean surface, in
the presence of coadsorbed O atoms the end-on
perpendicular geometry is strongly favoured [27].
Here we report on the application of quantita-
tive structural methods to investigate the behav-
iour of CN on Cu(1 1 1). In particular, N K-edge
NEXAFS has been used to gain quantitative
information on the average orientation of the
surface CN species on Cu(1 1 1) with and without
preadsorbed oxygen. We have also measured
extensive C 1s and N 1s PhD spectra in order to
try to establish the local adsorption sites.
The most complete structural studies are for
CN adsorbed in ordered c(2 · 2) overlayer phases
on the fcc(1 1 0) surfaces of Ni [10–12], Rh [13] and
Pd [14], variously based on one or more of the
techniques of scanned-energy mode photoelectron
diffraction (PhD) [15], near-edge X-ray absorption
fine structure (NEXAFS) and quantitative low
energy electron diffraction (LEED). In all of these
systems the CN lies above second layer substrate
atoms with its molecular axis in the [0 0 1] azimuth
such that it straggles a pair of close-packed sub-
strate atoms rows. The C atom lies slightly lower
on the surface such that this atom is three-fold
coordinated to two outermost layer and one sec-
ond layer substrate atoms, whereas the N atom is
two-fold coordinated to adjacent outermost layer
atoms. This leads to a small tilt of the C–N axis
out of the surface which falls in the range 18–25ꢀ
for these three different substrates. The polarisa-
tion-angle dependence of NEXAFS has also been
used to determine that the orientation of the CN
molecular axis on Pd(1 1 1) is tilted by 14ꢀ away
from the surface [16,17]. In addition, spectroscopic
data (mainly high-resolution electron energy loss
spectroscopy––HREELS) has been interpreted as
indicating that CN adsorbs with its molecular axis
essentially parallel to the surface on Pd(1 0 0)
[18,19], on Ru(0 0 0 1) [20] and on Cu(1 1 1) [21].
HREELS has also been used to study the
interaction of cyanogen with Cu surfaces predosed
with oxygen [21–23], motivated in part by interest
in the formation of the surface cyanate species,
NCO [24–26]. A particularly interest conclusion of
this work [23] is that on Cu(1 1 1) predosing of
oxygen to an estimated coverage of 0.1 ML did not
lead to the formation of a surface cyanate, but a
strong enhancement of the intensity of the C–N
stretching band in the HREELS was taken to infer
that the molecular axis of the adsorbed CN was
now much more nearly perpendicular to the sur-
face. While there appear to be no relevant calcu-
lations for the Cu(1 1 1) surface, it is interesting to
2. Experimental details
The experiments were conducted using syn-
chrotron radiation from the HE-TGM mono-
chromator [28] on the BESSY electron storage ring
in Berlin, in a surface science end-station equipped
with the usual facilities for in situ sample prepa-
ration and characterisation. The Cu(1 1 1) sample
was prepared in situ by argon ion bombardment
and annealing to produce a clean and well-ordered
surface as judged by soft X-ray core level photo-
electron spectra and the low energy electron dif-
fraction (LEED) pattern. The CN-covered surface
was produced by a nominal saturation exposure
(200–400 · 10ꢀ6 mbar s) of cyanogen gas produced
by heating carefully-outgassed AgCN to a nominal
temperature of 360 ꢀC. The coadsorption experi-
ments were performed by first exposing the Cu
surface to 60–100 · 10ꢀ6 mbar s oxygen followed
by similar 150–300 · 10ꢀ6 mbar s exposures of
cyanogen; these conditions were chosen to repro-
duce those used in the earlier HREELS study of
this system [21–23]. Photoemission spectra were
recorded using a 152 mm mean radius spherical-
sector electrostatic analyser (VG Scientific)
equipped with three-channeltron parallel detec-
tion. This analyser was mounted at 60ꢀ relative to
the incident synchrotron radiation, and has an
acceptance angle of 4.5ꢀ. We have found that this
acceptance angle appears to be around optimal for
PhD determinations of local adsorbate structure.

M. Polcik et al. / Surface Science 563 (2004) 159–168
161
Much smaller acceptance angles greatly reduce the
fcc hollow site
measured intensity and make the computational
simulations very time-consuming (see below),
whereas much larger acceptance angles ‘wash out’
the PhD modulations. By comparison with pho-
toemission spectra taken previously with the same
instrumentation for the 0.5 ML coverage phases of
O on N on Cu(1 0 0), we estimate that the adsor-
bate coverages studied in the present experiments
were: pure CN, 0.28 ML; coadsorbed O and CN,
O coverage 0.12 ML, CN 0.42 ML. The precision
of these estimates is probably about 10–20%.
[211]
[110]
[121]
hcp hollow site
Fig. 1. Schematic plan view of the Cu(1 1 1) surface defining
azimuthal directions and identifying the two inequivalent three-
fold coordinated hollow sites available on this surface.
N K-edge NEXAFS spectra were recorded by
monitoring the intensity of the N KLL Auger
emission for a range of different grazing incidence
angles, normalised to the sample drain current as a
measure of the incident beam intensity variations
with photon energy. Similar measurements were
made from the clean Cu(1 1 1) surface and the
adsorbate spectra were normalised by dividing by
these reference spectra. The PhD technique [15]
exploits the coherent interference which occurs
between the directly emitted component of
the photoelectron wavefield resulting from the
photoionisation of an adsorbate core level, and the
components of this same wavefield which are
elastically scattered by the surrounding atoms. The
photoelectron kinetic energy, and thus the associ-
ated photoelectron wavelength, is varied by
changing the photon energy, causing different
scattering pathways to switch in and out of phase.
This leads to modulations in the photoemission
intensity as a function of energy in any specific
direction. Using relatively low photoelectron
energies (<500 eV), at which elastic backscattering
cross-sections are reasonably large, these data
provide quantitative information on the adsor-
bate-substrate registry. Notice that the same pho-
toelectron diffraction effects can be monitored by
measuring the intensity as a function of polar and
azimuthal emission angles at fixed energies [15],
but we have found the use of energy scans in fixed
directions to be a more convenient way of probing
the information and our experiment is designed for
this purpose. In the present case PhD spectra were
obtained by measuring 80 eV wide photoelectron
energy distributions centred around the C and N
1s emission peaks at photoelectron energies in the
range 80–350 eV at 2 eV steps. Each of these peaks
was fitted by a Gaussian peak, a Gaussian-
broadened step, and a template background
function to obtain a graph of the integrated 1s
peak area as a function of photoelectron kinetic
energy, IðEÞ. Normalisation of this gave the PhD
modulation spectrum vðEÞ ¼ ðIðEÞ ꢀ I0ðEÞÞ=I0ðEÞ
where I0ðEÞ is the non-diffracted intensity which is
taken to be a smooth stiff spline through the
experimental data. PhD modulation spectra of this
type were obtained in most of the directions within
the range of polar emission angles from 0ꢀ to 60ꢀ
in 10ꢀ steps and in each of the three symmetrically
inequivalent high symmetry azimuths, namely one
of the Æ1 1 0æ azimuths and two inequivalent Æ1 1 2æ
azimuths at +30ꢀ and )30ꢀ relative to Æ1 1 0æ la-
ꢁ
ꢁ ꢁ
belled here as [2 1 1] and [1 2 1] (at +30ꢀ and )30ꢀ
ꢁ
to [1 1 0]––see Fig. 1).
3. Results and data analysis
3.1. NEXAFS
Fig. 2 shows a summary of the N K-edge
NEXAFS spectra from the pure CN and CN/O
coadsorption surfaces for several different inci-
dence angles; in the standard NEXAFS nomen-
clature the angles shown are the grazing incidence
angles (i.e. normal incidence corresponds to 90ꢀ)
but this angle also corresponds to the angle, hE,
between the electric vector of the incident plane-
polarised radiation and the surface normal.

162
M. Polcik et al. / Surface Science 563 (2004) 159–168
nance in which the two contributions are unre-
N K- edge NEXAFS
solved. In this case, for perfectly plane-polarised
incident radiation, after averaging over all equiv-
alent azimuths on the three-fold symmetric sur-
face, the NEXAFS intensity variation is given by
I / sin2 hEð1 þ cos2 bÞ þ 2 sin2 b cos2 hE.
CN/Cu(111)
θE
By simple fitting of the NEXAFS p-resonance
peak to a Gaussian profile one can extract the
polarisation-angle dependence of the amplitude of
this feature in a more quantitative fashion (Fig. 3)
and compare this with the predicted dependence
for different tilt angles. Fig. 3 shows the best-fit
theory curves obtained for values of the tilt angle,
b, of 90ꢀ for the pure CN layer and 65ꢀ for the CN
with preadsorbed oxygen. These theoretical curves
assumed that the incident radiation was 90%
horizontally polarised and 10% vertically polar-
ised. Other measurements of the polarisation of
the output flux from the HE-TGM beamline
indicate a polarisation as high as 96%, but using
this figure gave essentially identical (within less
than 1ꢀ) optimum tilt angles. We therefore con-
clude that for the pure CN layer the C–N axis is
exactly parallel to the surface (to within the esti-
90°
60°
40°
20°
O+CN/Cu(111)
80°
60°
40°
20°
400
404
408
412
Photon energy (eV)
Fig. 2. N K-edge NEXAFS spectra for CN on Cu(1 1 1) with
and without coadsorbed O recorded at several different grazing
incidence angles, hE.
Clearly there is a significant difference induced by
the preadsorbed oxygen. For the pure CN layer
the single NEXAFS peak close to threshold,
associated with excitation into the unoccupied p-
resonance, increases in intensity by about a factor
of two as the grazing incidence angle is decreased.
By contrast, with preadsorbed oxygen this varia-
tion in intensity is much weaker. This intensity
variation provides a basis for determining the
average tilt angle of the molecular axis away from
the surface normal, b. To this end we exploit the
dependence of the NEXAFS intensity on the an-
gle, d, between the electric vector of the incident
radiation and the final state p-orbital (i.e. the
direction perpendicular to the molecular axis in the
plane in which the p-orbital is symmetric) which is
given by cos2 d [29]. For a molecular axis tilted
relative to the surface normal there are formally
two non-degenerate p-orbitals, one lying within
the plane of tilt and the other perpendicular to this
plane, and we here assume that these two final
states contribute equally to the observed p-reso-
CN/Cu(111)
O+CN/Cu(111)
20
30
40
50
60
70
80
90
Grazing incidence angle, θE (°)
Fig. 3. Dependence of the N K-edge NEXAFS p-resonance
peak intensity from CN on Cu(1 1 1) as a function of the
grazing incidence (and polarisation) angle, hE, extracted from
the spectra of Fig. 2. The full lines are theoretical fits corre-
sponding to tilt angle of the C–N axis relative to the surface
normal of 90ꢀ for the pure CN layer and 65ꢀ with preadsorbed
O.

M. Polcik et al. / Surface Science 563 (2004) 159–168
163
mated precision of 15ꢀ) and with preadsorbed
oxygen the average tilt angle of the C–N axis is 25ꢀ
from parallel to the surface. We should stress that
because the oxygen precoverage is relatively low
(0.12 ML) and the range of interaction of the CN
and O is unknown, it is possible that on the oxygen
precovered surfaces the NEXAFS measurements
represent an average over some CN molecules
which are parallel to the surface which is locally
devoid of oxygen while others, influenced by the
preadsorbed oxygen, are more strongly tilted to
the surface.
we have found that for adsorption in a high sym-
metry site the PhD modulations seen in a direction
corresponding to a near-180ꢀ nearest-neighbour
backscattering direction are typically of order
30–40%, with values as large as 80% in a few
cases. By contrast, weak modulation can result
from the emitter atoms being well-removed from
such high symmetry sites, because the experiment
then averages (sums incoherently) over all the
symmetrically equivalent local domains of the
adsorbate on the surface; in this case, a favoured
180ꢀ backscattering direction for one domain
(which should give strong modulations) will not
correspond to such a favourable geometry for the
many other symmetrically equivalent domains
which will contribute only weak modulations of
different periodicities. A similar situation, of
course, can also arise if multiple (symmetrically
inequivalent) adsorption sites are co-occupied. A
further feature of the data which indicates low
symmetry or multiple sites are occupied is the
generally weak dependence of the PhD spectra on
the azimuth of measurement, clearly seen in Fig. 4.
Our standard procedure for extracting the
structure from a PhD data set contains two
ingredients. In the first stage we generally apply
3.2. PhD
Fig. 4 shows an overview of the C 1s and N 1s
PhD modulation spectra obtained from the pure
CN layer on Cu(1 1 1). The most striking feature of
these data is that the modulations are in all cases
very weak. A few spectra show consistent modu-
lations in excess of 10%, but in many cases the
modulations are only of order 5% or less. A nec-
essary consequence of this is that the signal-to-
noise ratio is also rather poor. Based on previous
studies of a very large number of atomic and
molecular adsorbate systems (many on Cu(1 1 1))
N 1s
C 1s
[211]
60°
[121]
[110]
60°
50°
40°
40°
20°
0°
30°
20°
0°
0.2
0.0
0.0
-0.2
100
200
300
400
100
200
300
400
Photoelectron kinetic energy (eV)
Fig. 4. Overview of the N 1s and C 1s PhD spectra recorded from the pure CN on Cu(1 1 1) at different polar emission angles and in
different azimuthal planes (see Fig. 1). The shaded regions correspond to the modulation range 10%, providing a clear indication of
the weakness of the modulations.

164
M. Polcik et al. / Surface Science 563 (2004) 159–168
estimate the errors associated with the individual
structural parameters we use an approach based
on that of Pendry which was derived for LEED
[36]. This involves defining a variance in the min-
the ‘projection method’ of direct data inversion
[30,31] to obtain an approximate ‘image’ of the
scattering atoms around the emitter. This method,
however, relies on the data set containing strong
modulations in directions corresponding to near-
180ꢀ nearest neighbour backscattering directions
as is obtained from highly symmetric adsorption
sites. For emitter sites well-removed from such
high symmetry locations these clear emitter-
neighbour directions cannot be identified due to
domain averaging and the method fails to identify
near-neighbour scatterer locations in a convincing
fashion [32]. In the present case we must therefore
rely entirely on what is normally the second stage
of trial-and-error modelling, simulating the
experimental data with multiple scattering calcu-
lations for different trial structures. These calcu-
lations are performed with codes developed by
Fritzsche [33–35] which are based on the expan-
sion of the final state wave-function into a sum
over all scattering pathways which the electron can
take from the emitter atom to the detector outside
the sample. A magnetic quantum number expan-
sion of the free electron propagator is used to
calculate the scattering contribution of an indi-
vidual scattering path. Double and higher order
scattering events are treated by means of the re-
duced angular momentum expansion (RAME).
The finite energy resolution and angular accep-
tance of the electron energy analyser are accounted
for analytically; both of these instrumental effects
reduce the importance of more distant scatterers
and thus allow convergence of the calculations
with less scattering paths. Anisotropic vibrations
for the emitter atom and isotropic vibrations for
the scattering atoms are also taken into account.
The comparison between theoretical and experi-
mental modulation amplitudes, vth and vex, is
quantified by the use of a reliability factor
imum of the R-factor, Rmin as
pffiffiffiffiffiffiffiffiffi
VarðRminÞ ¼ Rmin 2=N
where N is the number of ‘independent pieces of
structural information’ contained in the set of
modulation functions used in the analysis. All
parameter values giving structures with R-factors
less than Rmin þ VarðRminÞ are regarded as falling
within one standard deviation of the ‘best-fit’
structure. More details of this approach, in par-
ticular on the definition of N, can be found else-
where [37].
A particular problem in establishing the correct
structure in the present case arises from the very
weak PhD modulations and the associated impli-
cation that both the C and N emitter atoms are
well-removed from any high symmetry adsorption
site. Weak modulations imply poor signal-to-noise
ratios which ensure that even for a correct struc-
ture the R-factor will be relatively large, making
this parameter less sensitive to subtle changes in
the adsorption geometry. Domain averaging over
symmetrically equivalent low symmetry adsorp-
tion sites also leads to a relatively weak depen-
dence of the theoretical PhD modulation spectra
on the exact geometry. We may therefore antici-
pate that the precision with which we can locate
the adsorption geometry will be relatively low.
Initial attempts to determine the local adsorption
sites of the C and N atoms independently, using six
or more selected PhD spectra from each emitter
confirmed the fact that no adsorption geometries
close to high symmetry sites yielded even plausible
fits to the experimental spectra. Moreover,
including experimental PhD spectra with very
weak modulations was counter-productive, be-
cause the R-factors for these individual spectra
were always large and simply lowered the sensi-
tivity of the global (multi-spectral) R-factor to
structural changes. The final structural optimisa-
tion was therefore performed by simultaneously
optimising the fit to a smaller subset of both C 1s
and N 1s PhD spectra which showed the strongest
modulations, applying the constraint that the C–N
.
X
X
2
Rm ¼
ðvth ꢀ vexÞ
ðv2th þ v2exÞ
where a value of 0 corresponds to perfect agree-
ment, a value of 1 to uncorrelated data, and a
value of 2 to anti-correlated data. The search in
parameter space to locate the structure having the
minimum R-factor was performed with the help of
an adapted Newton–Gauss algorithm. In order to

M. Polcik et al. / Surface Science 563 (2004) 159–168
165
ꢀ
distance was always 1.09 A and that the C–N axis
is closely parallel to the surface as indicated by the
NEXAFS results. Additional tests allowing the C–
N axis to tilt slightly yielded no improvement in
the fit to the PhD spectra, although similar checks
on the influence of the assumed C–N distance
found that the lowest R-factor was found with a
C
N
ꢀ
bondlength of only 1.00 A. While the true value of
this intramolecular bondlength is not known, in
Fig. 6. Schematic plan view of the Cu(1 1 1) surface (the Cu
atoms shown as large lightly-shaded spheres) with CN adsorbed
in the geometry leading to the best-fit to the experimental PhD
spectra (Fig. 5).
ꢀ
molecules the CBN bond is typically 1.16 A [38]
ꢀ
while adsorbed on Ni(1 1 0) values of 1.25 0.12 A
ꢀ
[10] and 1.17 0.08 A [12] have been found. We
ꢀ
therefore regard the value of 1.00 A as improbably
short and give the results for the somewhat larger
ꢀ
value of 1.09 A. Notice, though, that the optimal
coordinates of the N and C atoms found using C–
N bondlengths constrained in the range 1.00–1.20
ꢀ
A all fell within the estimated precision of these
mum achievable R-factor is less than 0.3, and in-
deed in some cases less than 0.2, but in the present
case of weak modulations the value of 0.33 is
actually rather encouraging. Fig. 6 shows a sche-
matic diagram of the geometry corresponding to
these best-fit PhD spectra. The C and N atoms are
in sites offset from atop towards bridging sites on
the Cu(1 1 1) surface such that one Cu atom is very
significantly closer to both the C and N atom
ꢀ
values for the value of 1.09 A.
The best-fit simulated spectra are compared
with the corresponding experimental spectra in
Fig. 5. The overall R-factor value is 0.33 and re-
flects the moderately good agreement in all the
main modulations. For systems showing stronger
modulations we have generally found the mini-
ꢀ
(1.98 0.05 and 2.00 0.05 A respectively) than
the next-nearest neighbour Cu atoms (2.58 and
ꢀ
ꢀ
2.51 A). The molecule is 1.82 0.04 A above the
outermost Cu layer of a bulk-terminated substrate;
no significant improvement in the fit resulted from
allowing relaxation of the substrate. The precision
in the lateral positions of the atoms on the surface
N 1s
C 1s
60° [110]
40° [121]
0.1
0.0
-0.1
ꢀ
was found to be 0.09 A, although the optimised
fit yielded an anomalously large optimum value of
the root-mean-square vibrational amplitude of the
30° [211]
30° [121]
30° [211]
0.1
0.0
-0.1
ꢀ
C atoms (0.37(+0.60/)0.25) A) which may reflect
significant static disorder in this position. Notice
that the centre of the molecule occupies a site
roughly midway between atop and a three-fold
coordinated hollow. There are, of course, two in-
equivalent hollow sites on a fcc(1 1 1) surface, one
above a second layer substrate atom (‘hcp hollow’)
and one above a third layer substrate (‘fcc hollow’)
(Fig. 1). Our calculations indicate that these two
sites are occupied with equal probability.
0.2
0.0
30° [121]
40° [211]
0.1
0.0
theory
experiment
-0.1
100
200
300
100
200
300
In the case of CN adsorbed onto the Cu(1 1 1)
surface predosed with oxygen, Fig. 7 shows a
comparison of C 1s and N 1s PhD spectra re-
corded from the surface compared with compa-
rable data from the pure CN layer. Clearly the
PhD spectra from the two phases are very similar;
Photoelectron kinetic energy (eV)
Fig. 5. Comparison of a subset of the N 1s and C 1s PhD
experimental modulation spectra (Fig. 3) recorded from the
pure CN layer on Cu(1 1 1) with the results of the theoretical
multiple scattering simulations for the best-fit adsorption
geometry (see Fig. 6).

166
M. Polcik et al. / Surface Science 563 (2004) 159–168
[121]
CN/Cu(111)
O+CN/Cu(111)
N 1s
60°
40°
30°
[211]
[110]
40°
60°
20°
30°
20°
20°
0°
0.2
0.0
-0.2
100
200
300
100
200
300
100
200
300
60°
C 1s
60°
30°
0°
30°
20°
0.2
0.0
100
200
300
100
200
300
100
200
300
Photoelectron kinetic energy (eV)
Fig. 7. Overview of the N 1s and C 1s PhD spectra recorded from CN adsorbed on Cu(1 1 1) with and without preadsorbed O.
there are certainly detailed differences, but these
are not very systematic. For example, for the N 1s
spectra recorded at 30ꢀ and 40ꢀ polar emission
clear. Chemisorbed oxygen on Cu(1 1 1) does not
form any long-range ordered structures, but at
high coverage there is clear evidence from several
different techniques for restructuring of the out-
ermost Cu atom layer(s), possibly to form a local
pseudo-(1 0 0) reconstruction (see [39] and refer-
ences therein). At lower coverages even less is
known about the structure of O on Cu(1 1 1), but it
is certainly possible that similar reconstruction
occurs locally. In this case it may be inappropriate
to think of the CN adopting a slightly modified
geometry on the Cu(1 1 1) surface, because the
reconstruction may produce a far more atomically
rough surface on which a tilted molecule might
still bond to the surface Cu atoms through both
the C and N atoms (as has been found for the
similarly weakly tilted CN species on several
fcc(1 1 0) surfaces [10–14]). On the other hand, if
ꢁ
angle in the [2 1 1] azimuth the modulations are
somewhat stronger with the preadsorbed O than
without, but the contrary is true in the equivalent
ꢁ ꢁ
spectra in the [1 2 1] azimuth. Similar effects are
seen in the C 1s PhD spectra. As the NEXAFS
clearly indicates a change in the orientation of the
C–N axis away from parallel to the surface, one
might have expected that the molecule would now
be bonded to the surface through only one end and
that the atom at the other end, no longer directly
bonding to the Cu and thus more distant from the
Cu scatterers, would lead to significantly weaker
PhD modulations. No such systematic behaviour
can be discerned for either emitter atom. Of
course, the role of the coadsorbed O is far from

M. Polcik et al. / Surface Science 563 (2004) 159–168
167
the interaction between the adsorbed O and CN
species is highly local, it is possible that some
fraction of the CN molecules are far more strongly
titled out of the surface plane through interaction
with the O atoms, while the remainder (perhaps
the majority) retain the geometry on the clean
surface.
a research grant and a Senior Research Fellowship
for DPW.
References
ꢂ
ꢂ
[1] F. Huerta, E. Morallon, F. Cases, A. Rodes, J.L. Vazquez,
A. Aldaz, J. Electroanal. Chem. 431 (1997) 269.
[2] R.E. Benner, K.U. von Raben, R. Dornhaus, R.K. Chang,
B.L. Laube, F.A. Otter, Surf. Sci. 102 (1981) 7.
[3] K. Kunimatsu, H. Seki, W.G. Golden, J.G. Gordon II,
M.R. Philpott, Surf. Sci. 158 (1985) 596.
4. Conclusions
N K-edge NEXAFS studies of CN and CN/O
coadsorption on Cu(1 1 1) confirm earlier qualita-
tive results from HREELS that the oxygen coad-
sorption causes the C–N axis to tilt up out of the
surface plane, and provides some quantification of
these conclusions. Specifically, in the absence of
coadsorbed O the molecular axis of the adsorbed
CN is parallel to the surface to within the limits of
precision ( 15ꢀ) whereas in the presence of coad-
sorbed O the axis is tilted by an average angle of
25 15ꢀ away from parallel to the surface. Note
that we estimate the coverages of O and CN in the
coadsorption phase to be 0.12 and 0.42 ML
respectively, so it is possible that this tilt angle
represents an average over two distinct values with
CN unaffected by O remaining parallel to the
surface while the second species associated with
the O is more nearly perpendicular to the surface.
C 1s and N 1s scanned-energy mode photoelectron
diffraction from the pure CN layer show very weak
modulations characteristic of a low symmetry
adsorption site or multiple sites. Constraining the
C–N bond length and bond orientation (the latter
from the NEXAFS results) we find the best-fit
structure corresponds to the molecule occupying
an off-atop geometry with both C and N atoms
having a single nearest-neighbour Cu atom at
[4] D.S. Corrigan, P. Gao, L.-W.H. Leung, M.J. Weaver,
Langmuir 2 (1986) 744.
€
[5] W. Daum, F. Dederichs, J.E. Muller, Phys. Rev. Lett. 80
(1998) 766 (erratum Phys. Rev. Lett. 85 (2000) 2655).
[6] B. Ren, X.-Q. Li, D.-Y. Wu, J.-L. Yao, Y. Xie, Z.-Q. Tian,
Chem. Phys. Lett. 322 (2000) 561.
[7] P. Jakob, Chem. Phys. Lett. 263 (1996) 607.
[8] D.R. Mullins, Lj. Kundakovic, S.H. Overbury, J. Catal.
195 (2000) 169.
[9] J.H. Miners, A.M. Bradshaw, P. Gardner, Phys. Chem.
Chem. Phys. 1 (1999) 4909.
[10] N.A. Booth et al., Surf. Sci. 416 (1998) 448.
[11] D. Brown, D.P. Woodruff, T.C.Q. Noakes, P. Bailey, Surf.
Sci 476 (2001) L241.
[12] C. Bittencourt, E.A. Soares, D.P. Woodruff, Surf. Sci. 526
(2003) 33.
[13] F. Bondino, A. Baraldi, H. Over, G. Comelli, P. Lacovig,
S. Lizzit, G. Paolucci, R. Rosei, Phys. Rev. B 64 (2001)
085422.
[14] F. Bondino, E. Vesselli, A. Baraldi, G. Comelli, A. Verdini,
A. Cossaro, L. Floreano, A. Morgante, J. Chem. Phys. 118
(2003) 10735.
[15] D.P. Woodruff, A.M. Bradshaw, Rep. Prog. Phys. 57
(1994) 1029.
[16] J. Somers, M.E. Kordesch, T. Lindner, H. Conrad, A.M.
Bradshaw, G.P. Williams, Surf. Sci. 188 (1987) L693.
[17] M.E. Kordesch, T. Lindner, J. Somers, W. Stenzel, H.
Conrad, A.M. Bradshaw, G.P. Williams, Spectrochim.
Acta 43A (1987) 1561.
[18] K. Besenthal, G. Chiarello, M.E. Kordesch, H. Conrad,
Surf. Sci. 178 (1986) 667.
[19] M.E. Kordesch, W. Stenzel, H. Conrad, Surf. Sci. 186
(1987) 601.
ꢀ
distances of 1.98 0.05 and 2.00 0.05 A respec-
tively.
[20] W.H. Wienberg, D.F. Johnson, Y.-Q. Wang, J.E. Parme-
ter, M.M. Hills, Surf. Sci. 235 (1990) L299.
[21] M.E. Kordesch, W. Feng, W. Stenzel, M. Weaver, H.
Conrad, J. Electron. Spectrosc. Relat. Phenom. 44 (1987)
149.
Acknowledgements
[22] W. Feng, W. Stenzel, H. Conrad, M.E. Kordesch, Surf.
Sci. 211/212 (1989) 1044.
This work was supported by the European
Community through Large Scale Facilities sup-
port to BESSY, and by the Physical Sciences and
Engineering Research Council (UK) in the form of
[23] M.E. Kordesch, W. Stenzel, H. Conrad, M.J. Weaver, J.
Am. Chem. Soc. 109 (1987) 1878.
[24] D.A. Outka, S.W. Jorgensen, C.M. Friend, R.J. Madix, J.
Mol. Catal. 21 (1983) 375.

168
M. Polcik et al. / Surface Science 563 (2004) 159–168
€
[25] H. Celio, K. Mudalige, P. Mills, M. Trenary, Surf. Sci. 394
(1997) L168.
[32] D.P. Woodruff, P. Baumgartel, J.T. Hoeft, M. Kittel, M.
Polcik, J. Phys.: Condens. Matter 13 (2001) 10625.
[33] V. Fritzsche, J. Phys.: Condens. Matter 2 (1990) 1413.
[34] V. Fritzsche, Surf. Sci. 265 (1992) 187.
[35] V. Fritzsche, Surf. Sci. 213 (1989) 648.
[36] J.B. Pendry, J. Phys. C: Solid State Phys. 13 (1980) 937.
[37] N.A. Booth et al., Surf. Sci. 387 (1997) 152.
[38] R.C. Weast (Ed.), CRC Handbook of Chemistry and
Physics, 58th edition, CRC Press, West Palm Beach, USA,
1977.
[26] A.F. Carley, M. Chinn, C.R. Parkinson, Surf. Sci. 537
(2003) 64.
[27] H. Yang, J.L. Whitten, J. Chem. Phys. 107 (1997)
8518.
[28] E. Dietz, W. Braun, A.M. Bradshaw, R.L. Johnson, Nucl.
Instrum. Methods Phys. Res. A 239 (1985) 359.
€
[29] J. Stohr, R. Jaeger, Phys. Rev. B 26 (1982) 4111.
[30] Ph. Hofmann, K.-M. Schindler, Phys. Rev. B 47 (1993)
13941.
[39] R.L. Toomes, D.P. Woodruff, M. Polcik, S. Bao, Ph.
Hofmann, K.-M. Schindler, A.M. Bradshaw, Surf. Sci. 445
(2000) 300.
[31] Ph. Hofmann, K.-M. Schindler, S. Bao, A.M. Bradshaw,
D.P. Woodruff, Nature 368 (1994) 131.