Surface Science 555 (2004) 94–100

www.elsevier.com/locate/susc

Structure of clean and H-saturated epitaxial two-dimensionalEr silicide on Si(1 1 1) studied by SEXAFS

M.-H. Tuilier, C. Pirri, D. Berling, D. Bolmont, G. Gewinner, P. Wetzel *

Laboratoire de Physique et de Spectroscopie Electronique––UMR CNRS 7014, Facult�e des Sciences et Techniques,

4 rue des Fr�eres Lumi�ere, F-68093 Mulhouse cedex, France

Received 14 January 2004; accepted for publication 19 February 2004

Abstract

The atomic structure of H-saturated epitaxial two-dimensional (2D) Er silicide on Si(1 1 1) has been studied by

means of surface-extended X-ray absorption fine structure. This structure consists of a single hexagonal erbium plane

intercalated between the substrate and a bulk-like Si top bilayer, and oriented in the same way as the substrate double

layers (A-type orientation). The Er atoms are positioned on T4 sites of the Si substrate. The interlayer spacings between

the Er plane and the upper and lower Si top bilayer planes are 3.14± 0.03 and 2.24± 0.03 �A, respectively, and theinterlayer spacings between the Er plane and the first and second Si substrate planes are 2.12± 0.03 and 2.95 ± 0.03 �A,respectively. These results clearly indicate that H adsorption induces a remarkable switch of the Si top layer buckling

from B-type (clean ErSi2) to A-type orientation. In addition, a strong outward relaxation of the Si top bilayer atoms

with respect to their positions in clean ErSi2 is observed upon H dosing.

� 2004 Elsevier B.V. All rights reserved.

Keywords: Surface-extended X-ray absorption fine structure (SEXAFS); Lanthanides; Silicon; Hydrogen atom; Silicides; Epitaxy

1. Introduction

Rare-earth (RE) silicides epitaxially grown on

Si(1 1 1) substrates have been subjected to numer-

ous experimental and theoretical studies [1] due to

their attractive technological properties for micro-

electronic applications. For instance, these mate-

rials present very low Schottky-barrier heightson n-type Si (0.3–0.4 eV). In addition, highly per-

fect epitaxial two-dimensional (2D) p(1 · 1) RESi2

* Corresponding author. Tel.: +33-03-89-33-60-08; fax: +33-

03-89-33-60-83.

E-mail address: p.wetzel@uha.fr (P. Wetzel).

0039-6028/$ - see front matter � 2004 Elsevier B.V. All rights reserv

doi:10.1016/j.susc.2004.02.023

phases have been successfully grown on Si(1 1 1)

for a number of trivalent RE such as Er [2–11],

Ho [11], Dy [12,13], which are characterized by

Schottky-barrier heights below 0.1 eV [4,7,8].

These 2D materials are obtained in a simple

way by deposition of the RE at room temperature

(RT) followed by annealing at temperature around

450 �C. A lot of work has been dedicated to thestudy of the structure of all these 2D phases (see

for example [2–13] and references therein). It

consists of a single hexagonal monolayer of RE

atoms located on threefold T4 sites of the Si(1 1 1)

substrate and covered by a buckled Si top layer

rotated by 180� around the surface normal with

respect to the relevant double layers of the

ed.

[111]

(b)

Si2Si1

[121]

(1)

(3)

(2)

(1)

(3)

(2)

ddown1 dup1

dup2 ddown2

ddown1 dup1

dup2 ddown2

SiEr

(a) Er

Si4 Si3

Si1Si2

Er

Si4 Si3

Fig. 1. Side view of the (a) clean 2D ErSi2 and (b) 2D H–ErSi2structural models as derived from Refs. [4,15], respectively.

Hydrogen atoms are not shown in (b) since they cannot be seen

directly in the present EXAFS study.

M.-H. Tuilier et al. / Surface Science 555 (2004) 94–100 95

substrate (B-type orientation). The atomic

arrangement is shown schematically in Fig. 1a.

More recently, experimental studies of the

structural properties of 2D RE germanides, have

been reported. Spence et al. [14] using medium-

energy ion scattering (MEIS) and Bonet et al. [13]

using quantitative low energy-electron diffraction(LEED) I–V analysis, found the growth of a 2D

dysprosium germanide DyGe2 on the Ge(1 1 1)

surface adopting an atomic structure similar to

that of 2D rare-earth silicides.

In this paper we report a structural analysis on

both clean 2D ErSi2 and hydrogen-saturated 2D

ErSi2 (2D H–ErSi2) film on Si(1 1 1) performed by

surface-extended X-ray absorption fine structure(SEXAFS). Previous Auger electron diffraction

(AED) [15], angle-resolved photoemission (ARP)

[16–18], high-resolution electron energy loss spec-

troscopy (HREELS) [17,19,20], and thermal

desorption spectroscopy (TDS) [20] experiments

have been employed in the study of the chemi-

sorption of atomic hydrogen on 2D ErSi2 silicide.

Drastic modifications in ErSi2 atomic and elec-tronic structures were found upon hydrogen dos-

ing. The origin for these changes was attributed to

atomic H adsorbed at two inequivalent sites: H

adsorbs on the Si dangling bonds present at the

ErSi2 surface and makes a monohydride (1 · 1)phase similar to the Si(1 1 1)-(1 · 1) H surface, as

well as on Er in the interstitial voids of the Er

hexagonal plane below Si species of the outermost

atomic plane of the buckled Si top layer [17–19].

From the structural point of view, the Si top bi-layer is found to switch from B-type to A-type

orientation upon H dosing, i.e., the H–ErSi2exhibits now a buckled Si top layer oriented in the

same direction as the substrate double layers (Fig.

1b) [15]. Furthermore, this surface rearrangement

is accompanied by a large vertical outward relax-

ation by �0.3 �A of the Si top bilayer atoms with

respect to positions in the clean 2D ErSi2. As forthe interfacial geometry, crystallographic order

and interlayer distances (ddown2 and dup2) have notyet been determined experimentally for H–ErSi2.

Finally, we observe besides the change in the

geometry, a strong change of the electronic struc-

ture. Whereas the clean 2D ErSi2 silicide is semi-

metallic, the H–ErSi2 silicide is semiconducting

[17–19]. As pointed out in Ref. [17], the presence oftwo different adsorption sites provides a simple

explanation of the observed semimetal-to-semi-

conductor transition in terms of chemical bonding

and electron counting arguments. Actually, the

atomic rearrangement inferred from the experi-

mental studies mentioned before, was further

supported by theoretical works [18].

Recently, Spence et al. [14] showed that in thecase of 2D DyGe2 germanide grown on the Ge-

(1 1 1) surface, the buckling of the top layer re-

versed upon adsorption of H in a same way as for

2D H–ErSi2. Similar results were also found by

Kitayama et al. in studying 2D HoSi2 silicide after

H dosing by LEED technique [21].

In the present work we used the SEXAFS

technique which yields information on the localstructure of thin layers. The atom specific nature of

the X-ray processes makes them ideal to charac-

terize the local structure around an Er adsorber

and allows, in particular to probe the bonds to the

neighbours located both above (surface) and below

(interface) the Er plane. Hence, such local char-

acterization is a good complement of the crystal-

lographic order information obtained by AEDbecause SEXAFS provides the possibility of access

to interfacial information. Assuming the clean

96 M.-H. Tuilier et al. / Surface Science 555 (2004) 94–100

ErSi2 and H–ErSi2 top layer atomic structures

previously determined by AED [4,15], quantitative

structural analysis was performed by EXAFS.

Actually, we also performed simulations of the

EXAFS spectra using the FEFF6 code, which

calculates the absorption cross-section in a multi-ple scattering formalism.

2. Experimental

SEXAFS measurements were performed at the

beamline DW21 of the DCI storage ring at LURE

(Laboratoire pour l’Utilisation du RayonnementElectromagn�etique) in Orsay (France). The inci-

dent X-ray beam was monochromatized by a

Ge(2 2 0) flat two-crystal spectrometer. The spectra

were acquired in the region of Er L3 absorption

edge (8300–8850 eV) using the fluorescence yield

mode with a seven element Ge detector. All mea-

surements were done at 80 K. It is interesting to

note that the best signal to noise ratio is obtainedfor the smallest angle between the X-rays and the

substrate surface (12 · 12 mm2). Indeed, record-

ing the spectra in very low incidence requires a

narrowing of the vertical slits to less than 2 mm,

instead of 10 mm in normal incidence. Conse-

quently, the intensity scattered by the Si atoms of

the substrate is considerably reduced in grazing

incidence. Though the acquisition time were di-vided by two, raw data of very good quality were

obtained in grazing incidence.

Both clean and H-saturated 2D ErSi2 were

prepared on Si(1 1 1) surface. In this way we are

able to directly compare the relevant structures

and determine the remarkable structural changes

induced by H chemisorption on 2D ErSi2. The

Si(1 1 1) surfaces were cleaned by cycles of argonion sputtering followed by annealing to 850 �C to

achieve a sharp 7 · 7 reconstruction. The 2D ErSi2silicide was grown by deposition of one Er

monolayer (ML) onto Si(1 1 1) at RT and anneal-

ing at �450 �C. This procedure gives a highly

ordered structure, as evidenced by a bright well-

contrasted p(1 · 1) LEED pattern with marked

threefold rotational symmetry. Er was evaporatedfrom an electron bombardment cell, at a rate of

0.5 ML/min. During the Er deposition the base

pressure in the chamber did not exceed 2 · 10�10Torr.

Hydrogen saturation was achieved by exposing

the silicide surface to about 1000 Langmuirs (L) (1

L¼ 1 · 10�6 Torr s) of atomic H produced by

thermal dissociation of H2 at a tungsten filamentplaced close to the sample surface. This hydroge-

nation procedure yields well saturated H surfaces

as demonstrated in our previous angle-resolved

photoemission and high-resolution electron en-

ergy-loss spectroscopy studies [17]. Note that, the

effective atomic H exposure is a critical function of

sample-filament-ionization gauge geometry as well

as filament temperature.After H saturation, the silicide (1 · 1) periodic-

ity is preserved but beam intensities versus in-plane

reciprocal lattice vector are inverted with respect

to the clean ErSi2 one [15]. This inverted LEED

pattern agrees well with the rotation by 180� of thebuckled Si top layer observed by AED [15] after H

dosing. A similar reversal of the threefold sym-

metry of the p(1 · 1) LEED pattern induced by Hadsorption has also been reported for the Si(1 1 1)–

Ho [21] and for the Ge(1 1 1)–Dy [14] systems.

3. Results and discussion

SEXAFS data were reduced using the conven-

tional procedure [22]. The spectra were normalizedto the intensity of the non-oscillatory background

which was approximated by fitting the post-edge

region between 8357 and 8850 eV by a five-degree

polynomial (the fluorescence yield is close to zero

below the Er L3 edge). Yet, keeping in mind the

silicide thickness (�5 �A), we analyzed preferen-

tially the data recorded in grazing incidence (W ¼10�), i.e., with polarization of the X-ray almostperpendicular to the surface plane. So, this angle

of incidence was chosen for measuring all the

spectra. In Fig. 2 we compare the modulus of

Fourier transforms (FT) of the k-weighted vðkÞdata from clean 2D ErSi2 (circles) and H–ErSi2(full line) silicides recorded at grazing angle. Tak-

ing into account the quality of the raw data (not

shown here), the coordination shells visible in Fig.2 can be considered as significant at least up to 6�A. As can be seen from the figure, the FT are

0

0.05

0.1

0.15

0.2

0 1 2 3 4 5 6R )

°°° ErSi2/Si(111)H-ErSi2/Si(111)

bcb' d

a'

a

2

Mod

ulus

of F

ourie

rtra

nsfo

rm

2

(A

Fig. 2. Comparison of the FT for clean 2D ErSi2 (circles) and

H-saturated 2D ErSi2 (full lines) recorded at grazing incidence

(W ¼ 10�).

M.-H. Tuilier et al. / Surface Science 555 (2004) 94–100 97

dominated by a main single feature at about 2.6 �Adue to backscattering from the first shell of Si

atoms. It is apparent that the Si nearest-neighbour(NN) environment of Er in H–ErSi2 is shifted to-

wards higher coordination radii compared to clean

ErSi2 indicating that the NN Er–Si bond length

distances are slightly larger. This relaxation is not

surprising and can be straightforwardly related to

the large expansion of the Er–Si top bilayer spac-

ing (�0.30 �A) induced after hydrogenation, as re-ported by AED [15]. The lengthening of the Er–Sibond can be explained by the chemical bond that

the H has formed with the silicide which weakens

considerably the Er–Si interaction. Above 3 �A,additional qualitative information can be obtained

from simple examination of the FT. As can be

seen, the FT of clean ErSi2 and H–ErSi2 exhibit

major differences in the 3–4 �A region where the

peak labelled ‘b’ is shifted towards lower coordi-nation radii with respect to peak b. On the other

hand, peak locations are very similar in position in

the 4–6 �A region (peaks labelled c and d). The

major contribution to these distant neighbour

shells is made of Si atoms belonging to deep Si

planes of the substrate. Qualitatively, according to

these observations, the hydrogenation induces

changes in Er–Si distances for Si atoms of the topmost layers (i.e., above the Er plane), whereas the

interfacial geometry below the single Er atomic

plane should be very similar in both ErSi2 and H–

ErSi2. It has been established [4,9–11] that among

the possible bonding configurations at the ErSi2/

Si(1 1 1) interface, namely S (substitutional), T

(top), and hollows H3 and T4, Er occupies theeclipsed threefold hollow T4 site in clean 2D ErSi2.

This geometry results in a marked anisotropy of

the Er environment that is qualitatively in line

with the observed polarization dependence of the

SEXAFS data. The amplitude of the main oscil-

lation (Si NN) is strongly enhanced in grazing

incidence with respect to normal incidence for

both clean and H–ErSi2 silicides. So, for the sim-ulation we have assumed Er atoms in the T4 sites

for both clean ErSi2 and H–ErSi2. It is noteworthy

that this bonding configuration is quite consistent

with recent work on the electronic structure of H–

ErSi2 [18].

The quantitative interpretation of the spectra

requires an assignment of the various features

to one or another single scattering (SS) or multi-ple scattering (MS) paths involving neighbours

belonging to the Si top bilayer, in-plane Er

monolayer or Si substrate. An efficient way to do

that is the use of the ab initio code FEFF6 [23],

which computes the absorption spectra from the

values of atomic positions of the various scatterers

surrounding the absorbing atom and takes into

account the polarization dependence of the inci-dent radiation.

For clean 2D ErSi2/Si(1 1 1), an initial cluster

was built using the atomic structure determined by

AED [4], which consists of an Er monolayer lo-

cated underneath a bulk-like Si bilayer (1) of B-

type (Fig. 1a). The vertical distances between the

Er and Si planes, called ddown1 and dup1, were takento be 1.92 and 2.70 �A, respectively. For the sim-ulation of the Si top bilayer (1) in H–ErSi2, the

cluster was modified by reversing the buckling to

an A-type orientation (Fig. 1b) and increasing

the vertical distances between the Er plane and Si

top bilayer planes ddown1 and dup1 to 2.24 and 3.00�A, respectively, according to the results of AED[15]. Finally, two Si substrate bilayers, labelled (2)

and (3), were introduced below the Er plane byassuming a T4 geometry for both clean ErSi2 and

H–ErSi2. The interplanar distances between the Er

0

0.05

0.1

0.2

0.2

0.3

0 1 2 3 4 5 6

Mod

ulus

of F

ourie

r tra

nsfo

rm

++ Experiment–Theory

2D ErSi2/Si(111)

1

2 3

2

R ( )A

Fig. 3. Comparison between the FT of experimental spectrum

(crosses) recorded in grazing incidence from clean 2D ErSi2 and

theoretical partial RDF (radial distribution function) (full line).

The structural parameters used for the calculations are sum-

marized in Table 1.

0

0.05

0.1

0.15

0.2

0 1 2 3 4 5 6

Mod

ulus

of F

ourie

r tra

nsfo

rm

ExperimentTheory

H-ErSi /Si(111 )2

R ( )A

Fig. 4. Comparison between the FT of experimental spectrum

(crosses) recorded in grazing incidence from H-saturated 2D

ErSi2 and theoretical partial RDF (full line). The structural

parameters used for the calculations are summarized in Table 1.

98 M.-H. Tuilier et al. / Surface Science 555 (2004) 94–100

plane and the first and second substrate planes,

dup2 and ddown2 were taken to be 2.16 and 3.09 �A,respectively, as reported from the previous SEX-

AFS measurements [4]. The former value could

not be accurately determined previously [4], Er

atoms were found to be coordinated to four Sineighbours at a mean distance of 3.09 �A, three ofthem belonging to the upper substrate Si plane

(dup2 ¼ 2:16 �A) and one just below Er in the secondsubstrate Si plane. Yet, with the present good

quality data recorded at low temperature (80 K)

which gives information about the deeper Si dou-

ble layer (3), we expect to improve the precision of

the ddown2 distance. Below the Si double layer (2),bilayer (3) and beyond was assumed to continue in

an undistorted Si bulk lattice. The in-plane Er

neighbours were taken into account in the calcu-

lations.

On the other hand, electronic transitions from p

initial to both s and d final states symmetry are

involved in Er L3 edge. Now, FEFF6 neglects the

former, as the best value of the amplitude ratiofrom s to d final sates is known to be c ¼ 0:2.Consequently, the interference between final states

of s and d symmetry arising from anisotropic L2;3edge spectra is not taken into account. Fortu-

nately, it has been demonstrated that neglecting

this interference in the calculation of polarization

dependent EXAFS spectra only gives rise to very

small additional errors (of the order of magni-tude of 0.01 �A) in the determination of bond

lengths [24]. Thus, the accuracy seems to be suffi-

cient to determine safely the atomic structure

by using FEFF6, although some errors in abso-

lute amplitudes are expected to occur from these

approximations. Nevertheless, the validity of the

calculations was directly tested on bulk ErSi1:7grown epitaxially on Si(1 1 1), whose crystallo-graphic structure is now well known and both L1and L3 edge spectra have been studied in details

previously [25]. This study also allowed us to ad-

just the Debye temperature for the initial calcula-

tions on the thin films and the origin E0 of theEXAFS spectra for the conversion into the k-space.

For both clean ErSi2 and H–ErSi2 the ddowni anddupi (i ¼ 1–3) interlayer spacings were varied. The

best fits obtained in this way, are shown compared

to the experimental data in Figs. 3 and 4, respec-

tively. We see that agreement between experiment

and theory is good. The values of the best fitting

parameters used in the calculations are summa-

rized in Table 1.

Table 1

Interlayer spacings between the Er plane and the upper and lower Si top bilayer planes and the first and second Si substrate planes for

clean and H-saturated 2D ErSi2 surfaces, as determined from the SEXAFS measurements

Vertical

distance (�A)

Clean ErSi2 H–ErSi2

This work Previous works This work Previous work

SEXAFS

AED [4]

SXRD [10] MEIS [10] MEIS [11] AED [15]

dup1 2.68± 0.03 2.70± 0.05 2.60± 0.08 2.62±0.05 2.69± 0.03 3.14± 0.03 3.00± 0.05

ddown1 1.90± 0.03 1.92± 0.05 1.78± 0.08 1.82±0.06 1.77± 0.03 2.24± 0.03 2.24± 0.05

dup2 2.12± 0.03 2.12± 0.03

ddown2 2.95± 0.03 3.09± 0.04 3.10± 0.06 2.95± 0.03

The values from AED [4,15], SXRD [10] and MEIS [10,11] studies are shown for comparison.

M.-H. Tuilier et al. / Surface Science 555 (2004) 94–100 99

In clean 2D ErSi2 an Si2–Er distance of 2.92 �Ais found corresponding to the three Si NN locatedin the lower plane of the buckled top layer (1)

whereas a bond length of 3.48 �A is found for Si

atoms belonging to the upper surface plane. The

relevant interlayer spacings ddown1 ¼ 1:90 �A and

dup1 ¼ 2:68 �A are in good agreement with that

previously reported [4,10].

In H–ErSi2, the distance between the Er and the

Si NN belonging to the lower plane of the buckledtop layer (1) increases up to 3.15 �A. This bondlength correspond to an interlayer spacing of 2.24�A. Moreover, the same distance of 3.84 �A is found

between the Er and the three Si belonging to the

upper plane of the buckled top layer and the in-

plane Er atoms. The relevant interlayer spacing is

dup1 ¼ 3:14 �A. The Er–Er distance is in good

agreement with the value expected for strainedepitaxial silicide on Si(1 1 1).

In our determination of the interfacial struc-

tures, we find within experimental error, identical

parameters for both clean and H–ErSi2 silicides.

The Er atoms have three Si NN at a distance of

3.07 �A (Si3–Er) in the upper atomic plane of the

substrate bilayer (2), and a fourth neighbour at a

distance of 2.95 �A (Si4–Er) in the second substratelayer located directly underneath Er absorbers.

The corresponding interlayer spacings, dup2 andddown2 are found to be 2.12 and 2.95 �A, respec-tively. These results are in line with observations

reported in Ref. [18]. In this study the 2D band

dispersions of 2D H–ErSi2 were determined by

ARP and compared with band structure calcula-

tions. It was found that an increase of 0.15 �A in the

interlayer distance dup2 does not markedly affectthe calculated energy bands and density of states

(DOS), and the H adsorption energies.

The structural parameters of clean ErSi2 are in

good overall agreement with the results reported in

the literature as can be seen in Table 1. With re-

gard to H–ErSi2, the buckling of the topmost layer

has slightly higher value 0.9 �A than in previous

AED study [15] where an essentially Si(1 1 1) bulk-like (0.76 �A) value was found. Finally, the sameddown1 value has been determined in both AED and

SEXAFS studies.

4. Conclusion

Wehave used SEXAFS in order to determine theatomic structure of 2D H–ErSi2 silicide. It consists

of amonolayer of Er atoms covered by a bulk-like Si

top bilayer oriented in the same way as the substrate

double layers. Whereas the interfacial geometry

remains similar to that of clean 2DErSi2 (Er adopts

T4 site), H produces an expansion in the direction

perpendicular to the interface of the topmost Si

bilayer with respect to clean 2D ErSi2. This surfacerelaxation is assigned to the chemical bond that H

forms with the surface silicide. Indeed, as expected

from simple chemical arguments, the formation of

H–Si and H–Er chemical bonds weakens consider-

ably the Er–Si bonds in the silicide. A similar

relaxation has been observed uponH adsorption on

the DyGe2/Ge(1 1 1) system, using MEIS [14].

100 M.-H. Tuilier et al. / Surface Science 555 (2004) 94–100

Acknowledgements

The authors would like to acknowledge the

LURE staff for technical help.

References

[1] F.P. Netzer, J. Phys.: Condens. Matter 7 (1995) 991.

[2] O. Sakho, F. Sirotti, M. De Santis, M. Sacchi, G. Rossi,

Appl. Surf. Sci. 56–58 (1992) 568.

[3] P. Wetzel, C. Pirri, P. Paki, D. Bolmont, G. Gewinner,

Phys. Rev. B 47 (1993) 3677.

[4] M.-H. Tuilier, P. Wetzel, C. Pirri, D. Bolmont, G.

Gewinner, Phys. Rev. B 50 (1994) 2333.

[5] J.Y. Veuillen, D.B.B. Lollman, T.A. Nguyen, L. Magaud,

Appl. Surf. Sci. 65/66 (1993) 712.

[6] L. Ming, L. Grill, M.G. Ramsey, F.P. Netzer, J.A.D.

Matthew, Surf. Sci. 375 (1997) 24.

[7] M.-H. Tuilier, G. Gewinner, C. Pirri, P. Wetzel, D.

Bolmont, O. Heckmann, J. Phys. IV 4 (1994) 187.

[8] S. Vandre, T. Kalka, C. Preinesberger, M. D€ahne-Prietsch,

Phys. Lett. 82 (1999) 1927.

[9] L. Stauffer, A. Mharchi, P. Wetzel, C. Pirri, D. Bolmont,

G. Gewinner, Phys. Rev. B 47 (1993) 10555.

[10] M. Lohmeier, W.J. Huisman, G. ter Horst, P.M. Zagwijn,

E. Vlieg, C.L. Nicklin, T.S. Turner, Phys. Rev. B 54 (1996)

2004.

[11] D.J. Spence, S.P. Tear, T.C.Q. Noakes, P. Bailey, Phys.

Rev. B 61 (2000) 5707.

[12] S. Vandre, T. Kalka, C. Preinesberger, I. Manke, H. Eisele,

M. D€ahne-Prietsch, R. Meier, E. Weschke, G. Kaindl,

Appl. Surf. Sci. 123–124 (1998) 100.

[13] C. Bonet, D.J. Spence, S.P. Tear, Surf. Sci. 504 (2002) 183.

[14] D.J. Spence, T.C.Q. Noakes, P. Bailey, S.P. Tear, Phys.

Rev. B 62 (2000) 5016.

[15] P. Wetzel, C. Pirri, G. Gewinner, Europhys. Lett. 38 (1997)

359.

[16] S. Saintenoy, P. Wetzel, C. Pirri, D. Bolmont, G. Gewin-

ner, Surf. Sci. 349 (1996) 145.

[17] P. Wetzel, T. Angot, C. Pirri, G. Gewinner, Surf. Sci. 383

(1997) 340.

[18] P. Sonnet, L. Stauffer, S. Saintenoy, C. Pirri, P. Wetzel, G.

Gewinner, C. Minot, Phys. Rev. B 56 (1997) 15171.

[19] T. Angot, J.J. Koulmann, G. Gewinner, Europhys. Lett. 35

(1996) 215.

[20] P. Louis, T. Angot, D. Bolmont, G. Gewinner, Surf. Sci.

422 (1999) 65.

[21] H. Kitamaya, S.P. Tear, D.J. Spence, T. Urano, Surf. Sci.

482–485 (2001) 1481.

[22] A. Michalovicz, J. Phys. III C2-235 (1997).

[23] S.I. Zabynsky, J.J. Rehr, A. Ankudinov, R.C. Albers, M.J.

Eller, Phys Rev. B 52 (1995) 2995–3009.

[24] P.H. Citrin, Phys. Rev. B 31 (1985) 700, and references

cited therein.

[25] M.-H. Tuilier, C. Pirri, P. Wetzel, G. Gewinner, J.Y.

Venillen, T.A. Nguyen Tan, Surf. Sci. 307 (1994) 710.