The local structure of SO2 and SO3 on Ni(1 1 1)

Article abstract of DOI:10.1016/j.susc.2004.12.023

The normal incidence X-ray standing wave (NIXSW) technique, supported by X-ray photoelectron spectroscopy and near-edge X-ray absorption fine structure (NEXAFS), has been used to determine the local adsorption geometry of SO ₂ and SO₃ on Ni(1 1 1). Chemical-state specific NIXSW data for coadsorbed SO₃ and S, formed by the disproportionation of adsorbed SO₂ after heating from 140 K to 270 K, were obtained using S 1s photoemission detection. For adsorbed SO₂ at 140 K the new results confirm those of an earlier study [Jackson et al., Surf. Sci. 389 (1997) 223] that the molecule is located above hollow sites with its molecular plane parallel to the surface and the S and O atoms in off-atop sites; corrections to account for the non-dipole effects in the interpretation of the NIXSW monitored by S 1s and O 1s photoemission, not included in the earlier work, remove the need for any significant adsorption-induced distortion of the SO₂ in this structure. SO₃, not previously investigated, is found to occupy an off-bridge site with the C_3v axis slightly tilted relative to the surface normal and with one O atom in an off-atop site and the other two O atoms roughly between bridge and hollow sites. The O atoms are approximately 0.87 ? closer to the surface than the S atom. This general bonding orientation for SO₃ is similar to that found on Cu(1 1 1) and Cu(1 0 0) both experimentally and theoretically, although the detailed adsorption sites differ.

Full text of DOI:10.1016/j.susc.2004.12.023

Surface Science 577 (2005) 31–41
www.elsevier.com/locate/susc
The local structure of SO and SO on Ni(111)
2
3
a
G.J. Jackson , D.P. Woodruff , A.S.Y. Chan , Robert G. Jones ,
a,
b
b
*
c
B.C.C. Cowie
a
Physics Department, University of Warwick, Coventry CV4 7AL, UK
b
School of Chemistry, University of Nottingham, Nottingham NG7 2RD, UK
c
ESRF, BP 220, 33043 Grenoble Cedex, France
Received 14 October 2004; accepted for publication 20 December 2004
Available online 6 January 2005
Abstract
The normal incidence X-ray standing wave (NIXSW) technique, supported by X-ray photoelectron spectroscopy
and near-edge X-ray absorption fine structure (NEXAFS), has been used to determine the local adsorption geometry
of SO
tionation of adsorbed SO after heating from 140 K to 270 K, were obtained using S1s photoemission detection. For
2 3 3
and SO on Ni(111). Chemical-state specific NIXSW data for coadsorbed SO and S, formed by the dispropor-
2
2
adsorbed SO at 140 K the new results confirm those of an earlier study [Jackson et al., Surf. Sci. 389 (1997) 223] that
the molecule is located above hollow sites with its molecular plane parallel to the surface and the S and O atoms in off-
atop sites; corrections to account for the non-dipole effects in the interpretation of the NIXSW monitored by S1s and
O1s photoemission, not included in the earlier work, remove the need for any significant adsorption-induced distortion
2 3
of the SO in this structure. SO , not previously investigated, is found to occupy an off-bridge site with the C3v axis
slightly tilted relative to the surface normal and with one O atom in an off-atop site and the other two O atoms roughly
˚
between bridge and hollow sites. The O atoms are approximately 0.87 A closer to the surface than the S atom. This
general bonding orientation for SO
retically, although the detailed adsorption sites differ.
2005 Elsevier B.V. All rights reserved.
3
is similar to that found on Cu(111) and Cu(100) both experimentally and theo-
ꢀ
Keywords: X-ray standing waves, NEXAFS, catalysis; Chemisorption; Surface structure, nickel, sulphur dioxide; Low index single
crystal surfaces
1. Introduction
*
Corresponding author. Tel.: +44 2476523378; fax: +44
Sulphur dioxide, SO , is an important air pollu-
2
2
476692016.
E-mail address: d.p.woodruff@warwick.ac.uk (D.P.
tant created by the burning of sulphur-contami-
nated fossil fuels, yet also has beneficial effects in
Woodruff).
0
039-6028/$ - see front matter ꢀ 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2004.12.023

32
G.J. Jackson et al. / Surface Science 577 (2005) 31–41
that it has also long been used as a sterilising
agent, and there have been several investigations
of its surface chemistry. It is known to adsorb in-
tact on noble and transition metal surfaces at low
temperatures, but chemical interaction occurs on
all these surfaces other than Ag [1] as the temper-
ature is raised, leading to dissociation or dispro-
portionation and the presence of coadsorbed SOx
species in which x has been variously proposed
on different surfaces to be 1, 3 or 4 e.g. [2–7]. In
addition to quite a number of studies to characte-
rise these surface chemical reactions, there have
been a small number of structural studies on Ni
plane positions. By measuring such layer spacings
for two or more sets of (non-parallel) scatterer
planes, the absolute location of the absorber can
be established by triangulation. If the X-ray
absorption is monitored by the photoemission sig-
nal from the absorbing (photoionised) atom, the
Ôchemical shiftsÕ in the photoelectron binding
energy, widely exploited in conventional X-ray
photoelectron spectroscopy (XPS) allows one to
separate the absorption profiles for atoms of the
same element in different chemical states [19–21].
In the case of SO interactions with surfaces, this
2
means that shifts in the S1s photoelectron binding
energy can be used to obtain separate NIXSW
structural data from coadsorbed SO species [14].
[7–10] and Cu [11–14] surfaces aimed at establish-
ing the local adsorption geometry, principally of
x
the intact SO species; a short review of much of
2
Photoemission detection does, however, require
special precautions in the data analysis; non-dipole
effects in the photoemission create a forward/back-
ward asymmetry in the angular dependence of the
photoemission relative to the direction of photon
propagation which means that the photoemission
from the incident and reflected X-rays are not de-
tected in an equivalent fashion by a detector hav-
ing a finite acceptance angle [22]. However, this
now seems to be well understood [23], and the
method of data analysis can be modified to ac-
count for this effect [24].
this work to 1996 is available [15]. One of the dif-
ficulties involved in studying the local bonding
geometry of the reaction products is that these
usually show no long-range order and the presence
of coadsorbed species, each containing S, renders
even element-specific local structural probes such
as surface extended X-ray absorption fine struc-
ture (SEXAFS) inadequate for separating out the
individual components, although there have been
several studies using this technique [12,13].
An alternative local structural probe which of-
fers both elemental and chemical-state specificity
is X-ray standing waves, notably performed near
normal incidence to the Bragg scatterer planes
In the case of SO interaction with Ni(111), a
2
NIXSW investigation of the intact adsorbed SO2
at 150 K was conducted by our group several years
ago, but at that time (using a bending magnet
source of synchrotron radiation on the SRS
(synchrotron radiation source) at Daresbury Lab-
oratory, we were unable to perform the chemical-
state specific study necessary to investigate the
surface reaction products [10]. This earlier study
also predated our current understanding of non-
dipole effects in photoemission-detected NIXSW,
so some small systematic errors are present in the
analysed data. Here we present the results of a
newer investigation exploiting the high brilliance
undulator beamline (ID32) on the third-generation
source of the ESRF (European synchrotron
radiation facility) in Grenoble which allows chem-
ical-state specific NIXSW experiments to be per-
formed. We also consider the implications of the
revised methods of data analysis which take ac-
count of non-dipole effects in photoemission.
(
NIXSW) [16] which allows investigation of imper-
fect crystalline substrates (such a typical metal sin-
gle crystals) [17,16]. This technique is a variant of
the general XSW method [18] and exploits the fact
that under conditions of a X-ray Bragg reflection
in a crystal, the interference of the incident and
scattered waves gives rise to a standing wave in,
and outside, the crystal which has a periodicity
in intensity equal to the scatterer plane spacing
associated with the Bragg condition. As one scans
through the Bragg condition in incident angle or
photon energy, there is a finite range over which
the standing wave exists; within this range, the
standing wave is systematically displaced in phase
relative to the scatterer planes. The profile of the
X-ray absorption at a particular atom as one scans
through this range is characteristic of the spacing
of this absorber relative to the extended scatterer

G.J. Jackson et al. / Surface Science 577 (2005) 31–41
33
2
. Experimental details and surface characterisation
quent re-cooling to minimise the effects of thermal
vibrations.
The interaction of the SO with the Ni(111)
The Ni(111) sample, oriented by Laue X-ray
2
diffraction, was mechanically polished and then
cleaned in situ in the UHV end-station of the
ESRF beamline by cycles of 2 keV argon ion bom-
bardment and annealing at 630 ꢁC. Low energy
electron diffraction (LEED) was used to confirm
the presence of a (1 · 1) ordered surface whilst
the quality of the near-surface sample order and
cleanness could be also assessed by the substrate
NIXSW profiles and by electron-stimulated Auger
electron spectroscopy. The beamline was fitted
with an in-vacuum double-crystal monochromator
using Si(111) crystals. Photoelectron and Auger
electron emission spectra were recorded using
the Physical Electronics concentric hemispherical
dispersive analyser (fitted with multi-channel
detection) installed at 45ꢁ to the incident X-radia-
tion the (horizontal) plane of polarisation. SO2
dosing was effected by introducing the gas
surface was first characterised by XPS using a
nominal photon energy of 3060 eV (chosen to be
well away from the standing wave excitation en-
ergy range), mainly in the S1s photoelectron
energy range. Following SO dosing onto the sam-
2
ple at low temperature the sample was briefly
heated to higher and higher temperatures, taking
XP spectra after each heating. No absolute cali-
bration of the photoelectron binding energies
was undertaken, so all such energies are nominal
although, of course, the relative energies of the dif-
ferent S1s chemically shifted components should
be reliable. The results are summarised in Fig. 1.
Fig. 1(b) shows the set of S1s spectra obtained in
this way which are compared with similar data col-
lected in the earlier investigation using the SRS
[10] in Fig. 1(a). Fig. 1(c) shows on a larger scale
S1s spectra recorded at 140 K and after heating
to 270 K taken during NIXSW measurements
including some fitting of the components (see be-
low). The overall trend of these spectra is as previ-
ously discussed [10]. At the lowest temperature the
main peak at a nominal binding energy of approx-
imately 2472 eV is assigned to the intact adsorbed
(
99.99% manufactured purity) into the main
experimental chamber with the sample at low tem-
perature (120–140 K), checking the purity in the
UHV vessel with a mass spectrometer. Nominal
ꢀ6
exposures of about 10 · 10 mbar s were found
to produce an essentially saturated layer of the
adsorbate; to investigate some of the surface reac-
tion products brief heating to specified tempera-
tures were used, but NIXSW data were always
recorded at the lowest temperature after subse-
SO . There is a clear shoulder on the high bind-
2
ing energy (low kinetic energy) side which has
also been observed by Yokoyama et al. [8], and, in-
deed a similar effect has been seen in conventional
S 1s photoemission spectra
a)
(b)
(
140 K
00 K
20 K
(
c)
1
40 K
1
50 K
70 K
2
1
SO2
SO3
S
2
2
20 K
250 K
270 K
320 K
370 K
2
70 K
270 K
S
300 K
SO3
2490 2480 2470 2460 2490 2480 2470 2460 2480
2470
nominal binding energy (eV)
2
Fig. 1. S1s XP spectra recorded at a photon energy of 3060 eV from SO adsorbed on Ni(111) at low temperature followed by heating
to successively higher temperatures. The spectra in panel (b) are from the present study while those in (a) are from an earlier study [10]
and are included for comparison. (c) shows two of the spectra from (b) on an expanded scale with the three component peaks labelled.

34
G.J. Jackson et al. / Surface Science 577 (2005) 31–41
(
laboratory-based) XPS on the S2p photoemission
S K-edge NEXAFS
S KLL Auger spectrum
140 K
peaks be Zebisch et al. [4]. It is not clear whether
this feature is due to partial occupation of a second
SO adsorbed layer or to an intrinsic satellite of
1
40 K
o
o
o
90
55
15
2
the main chemisorbed SO peak, although the fact
2
that it seems to be present at all temperatures at
which the SO peak is seen rather suggests that
2
it is an intrinsic feature of this one state. Heat-
ing causes this SO2 feature to be replaced by
two new peaks with nominal binding energies of
approximately 2474 eV and 2469 eV. At the high-
est temperature only the lowest binding energy
peak survives, and this is clearly assigned to chem-
isorbed atomic S, the energy being identical to that
seen from a Ni(111)(2 · 2)–S surface [10]. The
identification of the 2474 eV peak is somewhat
more controversial; the general consensus is that
2
70 K
2
70 K
2460
2480
2500
photon energy (eV)
2100 2105 2110 2115 2120
kinetic energy (eV)
it can be assigned to SO although on Cu(100)
3
2
Fig. 2. S K-edge NEXAFS from Ni(111) exposed to SO at
low temperature, and after heating briefly to 270 K, at different
grazing incidence angles as labelled. Also shown are the S KLL
Auger spectra used to monitor the X-ray absorption.
Polcik et al. have argued that it should be attrib-
uted to SO [5]. However, on Cu there are compel-
ling arguments on the basis of vibrational
spectroscopy [3], S1s XPS and NEXAFS [12]
and S2p and O1s XPS [25], that the intermediate
For the low temperature state, attributable (al-
most) entirely to SO , the spectra show the behav-
is SO , produced by the disproportionation reac-
3
2
tion (3SO ! 2SO + S). The fact that the relative
iour already shown to be characteristic of this
surface [8,10]. There is a strong polarisation
dependence of the edge feature at a nominal pho-
ton energy of 2473 eV which is assigned to a tran-
sition to a final state of p-symmetry. The fact that
this peak is intense at grazing incidence and van-
ishes at normal incidence clearly indicates that
2
3
binding energies of these features on Cu and Ni
surfaces are the same, and that the NEXAFS of
this intermediate state on Ni(111) (see below) is
also essentially identical to that assigned to SO3
on Cu(111), leads us to the clear conclusion that
this intermediate in our present study on Ni(111)
can also be assigned to this sulphite species.
SO adsorbs with its molecular plane parallel to
2
To provide further characterisation of these dif-
ferent surface species S K-edge NEXAFS spectra
were recorded from both the as-deposited SO2
adsorbate at 140 K and from the same surface
after heating to 270 K at different grazing inci-
dence angles to the surface. Because the plane of
incidence is the same horizontal plane as that of
the polarisation of the incident synchrotron radia-
tion, these grazing incidence angles also corre-
spond to the angle of the polarisation vector
relative to the surface normal. NEXAFS data were
recorded by measuring the intensity of the S KLL
the surface as previously established. In the case
of the data taken after heating to 270 K, the
XPS data of Fig. 1 show clearly that in this case
one has coadsorbed S and SO . However, the large
3
chemical shift in the S KLL Auger peak associated
with this species appears to help considerably in
separating these components. The NEXAFS spec-
tra for this surface show a pronounced peak
at much lower photon energies (approximately
2467 eV) attributable to the near-edge Ôwhite lineÕ
of the chemisorbed atomic S, but the edge jump
associated with this component is quite small, indi-
cating that the two features around 2477 and
Auger peak at approximately 2108 eV (SO ) and
2
2
111 eV (SO ) at each photon energy in 0.2 eV
3
2479 eV can be assigned entirely to the SO . As re-
3
steps. The results are shown in Fig. 2, together
with the associated S KLL Auger electron spectra.
marked above, these features, and their polarisa-
tion angle dependence, are essentially identical to

G.J. Jackson et al. / Surface Science 577 (2005) 31–41
35
the behaviour seen for the reaction product ob-
tained by heating SO adsorbed at low tempera-
or contamination. We should, perhaps, stress that
the use of chemical-state resolved NIXSW means
that the structure determinations are not sensitive
to the exact surface preparation conditions (tem-
perature, exposure) which may give rise to differ-
ent relative surface coverages of the different
coadsorbed species. The one caveat to this state-
ment, of course, is that we assume that the local
adsorption site of each species is not significantly
influenced by the relative coverage of each coad-
sorbate; to some extent, the fact that each species
appears to have a single well-defined S1s photo-
electron binding energy, validates this assumption.
2
ture on Cu(100) [12] and Cu(111) [14]. In these
cases the two peaks are attributed respectively to
transitions of e and a symmetry in SO , so the fact
1
3
that the lower energy peak is strongest at normal
incidence indicates that the C3v symmetry axis of
the SO is essentially perpendicular to the surface.
3
In order to obtain quantitative information on
the adsorption geometries, NIXSW absorption
profiles using both the chemical-state specific S1s
and O1s photoemission peaks were recorded for
ꢀ
both the (111) and ð111Þ Bragg reflections. To
achieve this, photoelectron energy distribution
curves (EDCs) were measured around the S1s
and O1s peaks for a succession of photon energies
in 0.2 eV steps in a range of 14 eV centred on the
Bragg condition. The EDCs were recorded using
an analyser pass energy of 30 eV in 0.25 eV steps
in kinetic energy over a range of 18 eV for the
S1s and 12 eV for the O1s spectra. The overall
spectral resolution was 0.3 eV. To extract NIXSW
absorption profiles the component peaks of these
EDCs were fitted by Doniach-Sunjic line-shapes,
and the resulting peak areas were plotted as a
function of photon energy after normalising to
the beam current monitor to correct for the time-
dependent decay of the photon flux. XP spectra
were recorded before after each NIXSW measure-
ment to check that there was no significant change
in the adsorbate spectra due to radiation damage
3. NIXSW results
The NIXSW absorption profiles from the chem-
ically shifted S1s and O1s photoemission signals
from the SO and SO adsorbate species are sum-
2
3
marised in Fig. 3; the experimental data are shown
as individual data points and the best-fit theoreti-
cal curves are shown as continuous lines. To
achieve these fits we first determine the key non-
structural parameters of the absolute energy and
the energy broadening of the monochromated inci-
dent X-rays by fitting to the bulk absorption pro-
file extracted from the inelastically-scattered
electron background signal below the photoemis-
sion peaks. A further check was to fit similar
absorption profiles extracted from monitoring
_
_
(
111)
S 1s
SO3
SO2
(111)
(111)
S 1s
(111)
1
.4
1.0
1
.4
1
.0
1
1
.4
.0
O 1s
O 1s
1
.4
.0
1
3
045 3050 3055 3060
3045 3050 3055 3060
3045 3050 3055 3060
3045 3050 3055 3060
photon energy (eV)
photon energy (eV)
Fig. 3. Experimental NIXSW absorption profiles for the S and O atoms in SO
2 3
and SO on Ni(111) (data points) compared with the
best-fit theoretical curves (full lines). The structural parameters found for these fits are given in Tables 1 and 2.

36
G.J. Jackson et al. / Surface Science 577 (2005) 31–41
the Ni2p3/2 photoemission signal through the
XSW range, although in this case an approximate
correction for the forward–backward asymmetry
of the photoemission had to be applied (a rigorous
treatment is not currently available for emission
from p-states [23,24]). Using these values of the
non-structural parameters derived from the sub-
strate absorption profile the adsorbate S and O1s
photoemission signals for each reflection were
fitted to two structural parameters, the coherent
lation is simplest when a single high symmetry site
is occupied. In particular, on an fcc(111) surface,
there are three such high symmetry sites, atop a
surface layer substrate atom, in a ÔhcpÕ hollow atop
a second layer atom, and in a ÔfccÕ hollow atop a
third layer atom. If the substrate scattering plane
spacing is D(111), then for these three geometries
the expected value of d
d(111)/3, (d(111) + D(111))/3 and (d
ꢀ
are, respectively,
+ 2D(111))/3.
ð111Þ
(111)
Taking the values specific to the S atom in SO2
on Ni(111) the predicted d values are thus
position, dhkl, and the coherent fraction, fhkl
,
ð 1ꢀ 11Þ
0.67 A (atop), 1.75 A (hcp hollow) and 2.03 A
˚
˚
˚
including full corrections for non-dipolar effects
in the angular dependence of the photoemission
˚
(fcc hollow). The experimental value of 0.71 A is
[
resulting fits is good.
Table 1 shows the structural parameter values
obtained in this way for the SO2 adsorbed at
24]. Fig. 3 shows clearly that the quality of the
therefore in reasonable agreement with the atop
site and clearly quite different from the other pre-
dicted values. Notice, incidentally, that the value
of d
and fcc hollow sites (and also for occupation of
ꢀ
expected for equal occupation of the hcp
ð111Þ
1
40 K. Also included (in square brackets) are the
values obtained in an earlier investigation in which
no correction was included for non-dipole effects in
the photoemission [10]. The two experimental data
sets yielded almost identical absorption profiles,
and the difference in structural parameter values
is almost entirely due to these non-dipolar correc-
tions. By far the largest difference in these values
the lower symmetry bridging sites) is (d(111)
˚
+
1.5D(111))/3 of 1.88 A, also far from the experimen-
tal value. The same arguments show that the pre-
dicted d
site of 0.65 A is also much the closest to the exper-
ꢀ
value for the O atoms for the atop
ð111Þ
˚
˚
imental value of 0.74 A, although the discrepancy
is a little larger.
ꢀ
is in the coherent fraction for the ð111Þ reflection,
While simple triangulation of the coherent posi-
tions thus favours occupation of atop sites for
both the S and O atoms in adsorbed SO , it is clear
f
ꢀ
, for the O absorbers; the non-dipole correction
ð111Þ
reduces this value from 0.70 to 0.29. These non-
dipolar corrections most commonly lead to (mod-
est) reductions in the coherent fraction, although
for certain combinations of parameter values there
can be, as in this case, much larger changes in the
coherent fraction and for some parameter combi-
nations even substantial changes in the coherent
positions. It transpires (see below) that this reduc-
2
that the f
ꢀ
coherent fraction values of 0.3 for
ð111Þ
both S and O are small, indicating that this simple
identification of high symmetry sites is not valid.
For low symmetry adsorption sites a real NIXSW
measurement must average over all symmetrically
equivalent positions on the surface, and for a set
of Bragg scatterer planes not parallel to the surface
ꢀ
tion in f
ð 1ꢀ 11Þ
for the O atoms improves the fit to
the structural model proposed earlier [10].
such as ð111Þ (at 70.5ꢁ to the surface) this means
that the measurement averages over several differ-
ent values of the spacing of the absorber site from
the scattering planes, and this summation leads to
Interpreting these pairs of coherent position
values to determine the adsorption site by triangu-
Table 1
Best-fit structural parameter values for NIXSW from the SO
2
species adsorbed at 140 K on Ni(111)
˚
˚
Absorber
d
(111)/A
f
(111)
d ꢀ /A
f ꢀ
ð111Þ
ð111Þ
S
2.02(5) [1.95]
1.94(5) [2.03]
0.90(5) [0.90]
0.74(5) [0.75]
0.71(5) [0.66]
0.74(5) [0.66]
0.30(5) [0.25]
0.29(5) [0.70]
O
The numbers in round brackets are the estimated precision in the last significant figure. The numbers in square brackets are the values
obtained in an earlier study [10] which did not include any correction for non-dipole effects in the photoemission.

G.J. Jackson et al. / Surface Science 577 (2005) 31–41
37
a coherent position which is generally a weighted
Ni
Ni
average of these spacings, while the coherent frac-
tion falls from its ideal value of unity for a single
spacing. A simple procedure for performing this
summation, which can be visualised as a sum of
vectors in an Argand diagram, the amplitude and
phase of the vectors being determined by relative
occupation and layer spacing respectively, has
been described previously [26]. Using this ap-
proach one can map the values of the two struc-
coherent
position
tural parameters, d
ð 1ꢀ 11Þ
and f
ð 1ꢀ 11Þ
as a function of
the lateral position on an fcc(111) surface for a
particular value of d(111) [10,16]. Fig. 4 shows the
result of this exercise in the form of grey-scale
maps with associated contours for each parameter
over a rectangular region with the area of a prim-
itive unit mesh on this surface. For this figure,
however, an additional assumption has been made
that the adsorbate does not distinguish between
the two inequivalent (hcp and fcc) hollow sites,
effectively introducing a mirror plane which is
present in the outermost layer alone but not in
the underlying substrate. Fig. 5, which we will dis-
Ni
0.0
0
.9 0.6
.8 0.4
0.7
0
0.3
0.4
O
O
coherent
fraction
S
cuss later in the context of the SO data, shows the
3
same coherent fraction map over a larger area of
the surface and without the grey scale, allowing
the contours to be seen more clearly. We have dis-
cussed previously the need for the added mirror
Fig. 4. Grey-scale maps (with superimposed contours) of the
ꢀ
ð111Þ coherent positions and coherent fractions for an adsor-
bate atom located at different lateral positions on a Ni(111)
surface. The calculation sums over all symmetrically equivalent
positions for low symmetry sites, and also assumes that the
adsorbate does not distinguish between the two inequivalent
(hcp and fcc) hollow sites, thus imposing an additional mirror
symmetry plane. In the coherent position map, only two
possible values are allowed; one of these corresponds to that
expected for atop adsorption and is represented by black. The
other value is that expected for mixed hollow site occupation
and is shown as white. The lateral positions of outermost layer
Ni atoms are labelled. In the coherent fraction map, contours
are shown at spacings of 0.1 and a few of these contours are
labelled with the values. The white rings superimposed on the
map correspond to those regions in which the coherent position
is that for an atop site but the coherent fraction is 0.30 ± 0.05,
plane in interpreting data for the Ni(111)/SO sys-
2
tem [10] and the equivalent map obtained without
this additional assumption has been presented else-
where [16]. In particular, without this additional
symmetry our NIXSW structural parameters can-
not be reconciled with any position of the SO2
molecule on the Ni(111) surface.
Fig. 4 shows a rather surprising result of these
simple calculations, namely that only two values
of d
ꢀ
can occur, one corresponding to atop tri-
ð111Þ
angulation, the other to the mixture of hcp and
fcc hollows (or equivalently, bridge sites), for any
position of the adsorbing atom on the surface. This
as found for both the S and O atoms for adsorbed SO . A local
2
is why the nominal grey-scale map of d
4
ð 1ꢀ 11Þ
in Fig.
has only two different levels, black (correspond-
2
adsorption geometry for SO consistent with these constraints is
superimposed on the figure.
ing to the atop value) and white (corresponding to
the value for the mixed hollows). As one displaces
an adsorbate from an exact atop site across the
Eventually, f
placement causes d
low value and f
sudden switch of the d
ð 1ꢀ 11Þ
falls to zero and a further dis-
to switch to the mixed hol-
to rise again. Notice that this
ð 1ꢀ 11Þ
surface the atop value of d
ð 1ꢀ 11Þ
remains constant
, falls steadily.
ð 1ꢀ 11Þ
but the coherent fraction, f
ð 1ꢀ 11Þ
ð 1ꢀ 11Þ
value occurs when the

38
G.J. Jackson et al. / Surface Science 577 (2005) 31–41
to be compared with the value for SO in the gas
2
Ni
Ni
0
.8
phase of 119.5ꢁ. Clearly some modest distortion
0
.6
of this kind is not unreasonable as a result of the
chemisorption. This same basic model and analysis
was used to interpret the results of the earlier
NIXSW study, but in this case the apparently
0.4
0
.2
0
.2
0
.4
O
O
much higher value of 0.70 for f
ꢀ
of the O atoms
ð111Þ
implied that the O atoms were significantly closer
to the true atop sites, requiring a larger distortion
of the O–S–O bond angle and a substantial expan-
sion of the S–O bondlength. Using the non-dipole
corrections, the true value of this coherent fraction
is found to be much smaller, consistent with the O
atoms being displaced further from the atop sites
and thus a much smaller implied distortion of
the adsorbed SO2.
Ni
S
O
Ni
Fig. 3 also shows the NIXSW adsorption pro-
files and best-fit calculations for the S and O ab-
sorber atoms in the adsorbed SO species, while
3
Ni
Ni
Table 2 shows the resulting structural parameter
values. Notice that in this case the value of d(111)
for the S atom has had one (111) substrate inter-
layer spacing added to give a value which may
be consistent with realistic S–Ni nearest neighbour
bondlengths. An intrinsic feature of the XSW tech-
nique is that the true adsorbate–substrate layer
spacing may differ from the coherent position by
any integral number of substrate interlayer spac-
ings, the correct number to be used being readily
identified by simply considering the plausibility
of adsorbate–substrate bondlengths. In this partic-
ular case the value of the coherent position used in
ꢀ
Fig. 5. ð111Þ coherent fraction contour map for different
positions on Ni(111) as in the lower panel of Fig. 4 but with the
grey scale omitted and the map repeated over several unit
ꢀ
meshes. The black ring corresponds to a ð111Þ coherent
position consistent with mixed hollow sites and a coherent
fraction of 0.15 ± 0.08 as found for the S atom in adsorbed SO
The projection of an adsorbed SO species in a geometry
described in the text is superimposed on the map.
3
.
3
associated coherent fraction is zero and so the
value of the coherent position at this location on
the surface is meaningless. It is now straightfor-
ward to identify the regions on these maps which
correspond to our measured coherent position
and fraction values. Specifically, the coherent posi-
tions indicate atop triangulation and thus imply
positions of both the S and O atoms within the
area defined by the black levels in the coherent po-
sition map. Within this area we must place the
atoms on the coherent fraction contour corre-
sponding to 0.3, but within a range defined by
the estimated precision. The white arcs superim-
posed on the coherent fraction map correspond
to these regions, and superimposed on this map
are possible positions of the S and O atoms within
˚
the fitting was 0.65 A, but it is clearly not possible
for this to represent a true S–Ni interlayer spacing
(even in the highest-coordination hollow sites this
would infer a S–Ni nearest neighbour distance of
˚
1.58 A which is unrealistically short); adding one
interlayer spacing overcomes this problem (and
Table 2
Best-fit structural parameter values for NIXSW from the SO3
species adsorbed Ni(111) after briefly heating an SO
surface from 140 K to 270 K
2
-covered
˚
˚
Absorber
d
(111)/A
f
(111)
dð ꢀ1 11Þ/A
fð ꢀ1 11Þ
an adsorbed SO molecule. The S–O bondlength
2
˚
used in this diagram is 1.48 A, the value obtained
S
2.69(5)
1.82(5)
0.80(5)
0.62(10)
2.08(5)
1.04(5)
0.15(8)
0.08(8)
O
experimentally by SEXAFS on this surface [8].
The O–S–O bond angle in this diagram is 110ꢁ,
The numbers in round brackets are the estimated precision in
the last significant figure.

G.J. Jackson et al. / Surface Science 577 (2005) 31–41
39
adding two such spacings is incompatible with any
plausible combination of SO –Ni bondlengths).
Turning to the O NIXSW coherent position
values, a similar prediction of the d values to
be expected for single high-symmetry adsorption
3
ð 1ꢀ 11Þ
The value of d
fied. Despite these corrections the d(111) value for
ð 1ꢀ 11Þ
for S has been similarly modi-
˚
site occupation gives values of 0.61 A (atop),
˚
˚
the S absorbers of 2.69 A is significantly longer
˚
1.29 A (hcp hollow), 1.97 A (fcc hollow) and
˚
1.63 A (mixed hollows), none of which is reason-
˚
ably close to the experimental value of 1.04 A. In
than one might expect for a Ni–S chemisorption
bond, whereas the equivalent value for the O
absorbers is much shorter, but has a value which
could be more readily reconciled with a Ni–O
chemical bond, particularly as the nearest neigh-
bour bondlength must be larger than d(111) for
any adsorption site other than atop. We therefore
truth, this is not surprising. This procedure is for-
mally only consistent for occupation of these high
symmetry sites, although for a single low symme-
try adsorption site (and all its symmetrically equiv-
alent positions), the coherent position map of Fig.
4 shows that (with the added mirror symmetry)
this type of triangulation must return one of two
values, that for atop or that for mixed hollows.
However, as is clear from Fig. 5, if the S atoms
infer that the SO bonds to the surface through the
3
O atoms with the S atoms higher above the sur-
face, as has been found for SO on Cu surfaces
3
[
12,14].
As may be seen from Table 2, the f
ꢀ
values
of SO occupy a low symmetry site, and the C
3 3v
ð111Þ
for both the S and O atoms are extremely low,
clearly indicating that these atoms are not in single
high-symmetry adsorption sites, but we first con-
sider the possibility of simple triangulation of the
two coherent position values for each atom. In
the case of the S, the measured d(111) value can
molecular axis is perpendicular to the surface (as
implied by the NEXAFS data), then the three O
atoms cannot occupy equivalent sites, and must
actually occupy at least two inequivalent low sym-
metry sites. This situation is not covered by the
contour maps of Figs. 4 and 5 and further model-
ling is required.
be used to calculate the expected d
the different high-symmetry adsorption sites. The
ꢀ
values for
ð111Þ
For this purpose our starting point is the fact
that the d(111) values for the S and O absorbers dif-
˚
˚
values thus derived are 0.90 A (atop), 1.58 A
˚
hcp hollow), 0.22 A (fcc hollow) none of which
˚
fer by 0.87 A. If we assume that this represents the
projection of the S–O bonds onto the C3v axis, and
˚
that the S–O bondlengths are 1.50 A (as in
(
˚
is close to the measured value of 2.08 A. However,
the value expected for mixed hcp + fcc hollows (or
bridge sites) is 1.92 which is much closer to the
experimental value. In this regard we note that,
as shown in Fig. 4, for adsorption on (single)
low symmetry sites on Ni(111) in which the hcp
and fcc sites are not distinguished, this value of
the coherent position is the one alternative to the
atop site value. While the agreement between the
Na SO ), then the angle of the S–O bonds to this
2
3
symmetry axis is 55ꢁ and the length of the S–O
bond projection onto the plane perpendicular to
˚
the symmetry axis is 1.22 A. We now note that
the d
ð 1ꢀ 11Þ
value for O is between that expected for
atop sites and for the hollow sites, so some combi-
nation of off-atop and off-hollow sites may give a
better fit to the experiment. The model of SO3
superimposed in the coherent fraction contour
map of Fig. 5 has these properties. Note that the
O atom in the off-atop site will lie higher above
the surface than those O atoms in the sites roughly
between the hollow and bridge sites. Taking a
plausible value for the Ni–O nearest neighbour
expected and measured values of d
ꢀ
is therefore
not very good, adsorption of SO with the S atom
ð111Þ
3
in this part of the surface seems the most reason-
able interpretation. Fig. 5 shows the same f
ꢀ
ð111Þ
coherent fraction map of the surface as in Fig. 4
with the grey shading omitted and drawn over a
wider area of the surface. Superimposed on this
map is a black ring corresponding to the region
of the surface around one of the Ni surface atoms
for which both the coherent fraction and the
coherent position correspond to the values found
experimentally for the S atom.
˚
bondlength of 2.0 A, these two layer spacings are
˚
estimated to be 1.98 A and 1.48 A. This would
˚
actually imply a tilt of the C3v axis of the SO away
3
from the surface normal of 14ꢁ, small enough to be
consistent with the NEXAFS conclusion that this

40
G.J. Jackson et al. / Surface Science 577 (2005) 31–41
p
axis is perpendicular to the surface (typical preci-
sion estimates for this assignment in NEXAFS
are ±10–20ꢁ). More importantly, this height differ-
ence for the three O atoms above the surface is cal-
culated to cause a 25% reduction in f(111), almost
exactly the difference between the measured values
of this parameter for the S atoms (in identical sites)
and the O atoms (in the two different sites). The
f
ꢀ
is zero [27]. The simple Ni(111)ð 3 ꢁ
ð111Þ
p
ꢂ
3ÞR30 –S phase, which should have a single
well-defined S high-symmetry adsorption site and
for which NIXSW triangulation indicated fcc site
occupation, gave a low value of f
ꢀ
of only 0.50
ð111Þ
[27], and is this was attributed to partial coverage
of the reconstructed phase. A very similar effect
could account for the results for the atomic chem-
isorbed S in the present experiments.
model defined by the superimposed SO on Fig.
3
ꢀ
5
also allows us to determine the expected ð111Þ
coherent fraction and coherent position values
for the individual O atoms which may be com-
bined to give the values expected experimentally;
4. Conclusions
a simple calculation yields values of 0.13 and
˚
This new NIXSW study of the interaction of
SO2 with Ni(111), reinforced by measurements
using XPS and NEXAFS, has refined the structure
1
.01 A respectively, both falling well within the
estimated precision limits of the experimental val-
ues. This model is therefore clearly consistent with
all of the experimental data; we have been unable
to identify any other such model.
found for the intact SO species adsorbed at low
2
temperature in an earlier such study [10] and has
determined the local geometry of the SO species
3
Finally, we should note that our NIXSW
measurements of the surface annealed to 270 K
also provided information on the local adsorption
geometry of the chemisorbed atomic S. The coher-
formed by the disproportionation reaction which
occurs on heating. For adsorbed SO the new data
2
are essentially identical to the old, and the only
real difference is in corrections applied to the data
analysis to account for non-dipole effects in the
photoemission which cause some modification of
the derived structural parameter values, most
˚
ent positions obtained were 1.59 A for (111) and
˚
ꢀ
1
.73 A for ð111Þ. On the basis of this (111) value
ꢀ
˚
the predicted values for ð111Þ are 0.53 A (atop),
˚
˚
˚
.21 A (hcp), 1.89 A (fcc) and 1.55 A (mixed hol-
1
notably the f
ð 1ꢀ 11Þ
coherent fraction value for the
O atoms. As a result of these corrections the orig-
low or bridge). The closest fit is therefore to the
fcc hollow site, although the measured value may
be rationalised by partial co-occupation of the
hcp hollow site. However, while the measured
inally-proposed model is confirmed. SO lies above
2
hollow sites with its molecular plane parallel to the
surface and the S and O atoms in off-atop sites.
The corrected NIXSW structural parameters,
however, greatly reduce the chemisorption-in-
duced distortion which was first thought to occur.
Specifically, with the revised analysis the S–O
bondlengths used are those found in an earlier
experimental SEXAFS study [8], while only the
O–S–O bond angle is modified by some 10ꢁ rela-
tive to its value in the gas phase. Note that this
adsorption site is quite different from that pro-
posed on the basis of the SEXAFS study of
Yokoyama et al. [8] which concluded that the S
occupied bridge sites, although these authors only
considered high-symmetry locations for the S
atoms. Our data are clearly not compatible with
S atoms in bridge sites. This S-bridging model
has been supported by total energy cluster calcula-
tions [28], but these calculations do not appear to
(
111) coherent fraction was high (0.90) that for
ꢀ
the ð111Þ reflection was low (0.30) clearly indicat-
ing that the true situation must be more complex.
In this regard, we may note that equal occupation
of hcp and fcc hollow sites leads to a highest possi-
ble f
ꢀ
value of 0.50, so if there is some co-occu-
ð111Þ
pation of the hcp hollow, some reduction of this
coherent fraction is to be expected; however, this
cannot entirely account for the low value measured
here. The origin of this effect is most probably par-
tial local reconstruction of the surface. Atomic S
chemisorbed on Ni(111) forms several ordered
phases, two simple overlayers (2 · 2) and ð 3ꢁ
p
p
p
ꢂ
3
ÞR30 and one reconstructed phase, ð5 3 ꢁ 2Þ-
rect. In the reconstructed phase, with equal
co-occupation of the three rotationally-equivalent
domains, NIXSW experiments show the value of

G.J. Jackson et al. / Surface Science 577 (2005) 31–41
41
have considered our structure in clusters with more
than four Ni atoms.
For adsorbed SO our data are shown to be
compatible with a model in which the C3v symme-
try axis is slightly tilted (but by only about 14ꢁ) rel-
ative to the surface normal, the S–O bonds bent
out of plane such that the O atoms are closer to
[6] K. Wilson, C. Hardacre, C.J. Baddeley, J. L u¨ decke, D.P.
Woodruff, R.M. Lambert, Surf. Sci. 372 (1997) 279.
[7] L. Wilde, M. Polcik, J. Haase, B. Brena, D. Cocco, G.
3
Comelli, G. Paolucci, Surf. Sci. 405 (1998) 215.
[8] T. Yokoyama, S. Terada, S. Yagi, A. Imanishi, S.
Takenaka, Y. Kitajima, T. Ohta, Surf. Sci. 324 (1995)
25.
9] S. Terada, A. Imanishi, T. Yokoyama, S. Takenaka, Y.
[
Kitajima, T. Ohta, Surf. Sci. 336 (1995) 55.
˚
the surface by some 0.87 A, implying the adsor-
[
10] G.J. Jackson, J. L u¨ decke, S.M. Driver, D.P. Woodruff,
R.G. Jones, A. Chan, B.C.C. Cowie, Surf. Sci. 389 (1997)
223.
bate–surface bonding is through these O atoms.
The molecular species occupies an off-bridge site
such that one O atom is quite close to atop a sur-
face Ni atom and the other two O atoms are
roughly between hollow and bridge sites. There
appears to be no previous determination of the
[11] N. Pangher, L. Wilde, M. Polcik, J. Haase, Surf. Sci. 372
1997) 211.
(
12] T. Nakahashi, S. Terada, T. Yokoyama, H. Hamamatsu,
[
Y. Kitajima, M. Sakano, F. Matsui, T. Ohta, Surf. Sci.
73 (1997) 1.
3
structure of SO on a Ni surface. However, previ-
3
[13] M. Polcik, L. Wilde, J. Haase, Phys. Rev. B 57 (1998)
1868.
ous investigations of this species on Cu(111) by
NIXSW [14], and on Cu(100) by SEXAFS [12]
have both concluded that the C3v axis is essentially
perpendicular to the surface and that the O atoms
are closer to the surface and thus involved in the
adsorbate substrate bonding. For Cu(100) this
bonding geometry is also favoured by total energy
calculations [29]. This similarity of behaviour for
[
14] G.J. Jackson, S.M. Driver, D.P. Woodruff, N. Abrams,
R.G. Jones, M. Butterfield, M.D. Crapper, B.C.C. Cowie,
V. Formoso, Surf. Sci. 459 (2000) 231.
[15] J. Haase, J. Phys.: Condens. Matter 9 (1997) 3647.
[16] D.P. Woodruff, Prog. Surf. Sci. 57 (1998) 1.
[17] D.P. Woodruff, D.L. Seymour, C.F. McConville, C.E.
Riley, M.D. Crapper, N.P. Prince, R.G. Jones, Phys. Rev.
Lett. 58 (1987) 1460.
[18] J. Zegenhagen, Surf. Sci. Rep. 18 (1993) 199.
SO on Cu and Ni surface is in contrast to the dif-
3
[19] M. Sugiyama, S. Maeyama, S. Heun, M. Oshima, Phys.
Rev. B 50 (1995) 14778.
ferences seen for SO , which adsorbs with its
2
molecular plane parallel to Ni surfaces but perpen-
dicular to Cu surfaces.
[20] G.J. Jackson, D.P. Woodruff, R.G. Jones, N.K. Singh,
A.S.Y. Chan, B.C.C. Cowie, V. Formoso, Phys. Rev. Lett.
8
4 (2000) 119.
[
21] G.J. Jackson, J. L u¨ decke, D.P. Woodruff, A.S.Y. Chan,
N.K. Singh, J. McCombie, R.G. Jones, B.C.C. Cowie, V.
Formoso, Surf. Sci. 441 (1999) 515.
Acknowledgement
[22] C.J. Fisher, R. Ithin, R.G. Jones, G.J. Jackson, D.P.
Woodruff, B.C.C. Cowie, J. Phys.: Condens. Matter 10
The authors acknowledge the benefit of access
to the ESRF and the financial support of the Engi-
neering and Physical Sciences Research Council.
(
1998) L623.
23] I.A. Vartanyants, J. Zegenhagen, Solid State Commun.
13 (1999) 299.
[
[
[
1
24] J. Lee, C. Fisher, D.P. Woodruff, M.G. Roper, R.G.
Jones, B.C.C. Cowie, Surf. Sci. 494 (2001) 166.
25] H. Lu, E. Janin, M.E. Davila, C.M. Pradier, M. G o¨ thelid,
Vacuum 49 (1998) 171.
References
[
1] D.A. Outka, R.J. Madix, G.B. Fisher, C. Dimaggio,
Langmuir 2 (1986) 406.
[26] D.P. Woodruff, B.C.C. Cowie, A.R.H.F. Ettema, J. Phys.:
Condens. Matter 6 (1994) 10633.
[
[
2] M.L. Burke, R.J. Madix, Surf. Sci. 194 (1988) 223.
3] K.T. Leung, X.S. Zhang, D.A. Shirley, J. Phys. Chem. 93
[27] J. L u¨ decke, A.R.H.F. Ettema, S.M. Driver, G. Scragg, M.
Kerkar, D.P. Woodruff, B.C.C. Cowie, R.G. Jones, S.
Bastow, Surf. Sci. 366 (1996) 260.
(
1989) 6164.
4] P. Zebisch, M. Weinelt, H.-P. Steinr u¨ ck, Surf. Sci. 295
1993) 295.
[
[
[28] Y. Sakai, M. Koyanagi, K. Mogi, E. Miyoshi, Surf. Sci.
513 (2002) 272.
(
5] M. Polcik, L. Wilde, J. Haase, B. Brena, D. Cocco, G.
Comelli, G. Paolucci, Phys. Rev. B 53 (1996) 13720.
[29] J.A. Rodriguez, J.M. Ricart, A. Clotet, F. Illas, J. Chem.
Phys. 115 (2001) 454.