chemisorption of co on h-passivated si(1 0 0) surface

6
Chemisorption of Co on H-passivated Si(1 0 0) surface Li Ma * , Jianguang Wang, Qiliang Lu, Guanghou Wang National Laboratory of Solid State Microstructures and Department of Physics, Nanjing University, Hankou Road Number 22, Jiangsu Province, Nanjing 210093, China Received 4 November 2004; in final form 24 January 2005 Abstract Chemisorption of one monolayer Co atoms on a H-passivated Si(1 0 0) surface is studied by using the self-consistent tight-binding linear muffin-tin orbital method. Energies of adsorption systems, the layer projected density of states and charge distributions are calculated. The most stable position is at fourfold hollow for the adsorbed Co atoms, and Co might sit below the H layer. Therefore, a Co–H mixed layer exists at the Co/H-passivated Si(1 0 0) surface. The adsorbed Co atoms cannot sit below the Si surface. The passivated layer of H atoms hinders the intermixing of Co atoms with Si at the interface effectively. Ó 2005 Elsevier B.V. All rights reserved. It is known that the presence of surface impurities is nearly unavoidable in thin film growth. The introduc- tion of the right impurity in the film/substrate interface can greatly improve the morphology of thin films. In re- cent years, the research works have demonstrated that the introduction of certain impurities as adsorbates dur- ing film deposition can improve various film characteris- tics, such as growth mode, interface or surface roughness, etc. An understanding of the impurity effect is eventually important in order to control a given growth process at an atomic level. Hydrogen is the sim- plest impurity and plays an important role in improving the electrical properties of semiconductors such as the termination of dangling bonds and the passivation of deep level impurities [1–4]. In general, changes in the film growth mode for a number of other metal systems (e.g., Ag, In, Cu, Al) have suggested an enhanced metal adatom mobility on H-passivated Si surfaces [5–11]. Cobalt merits special attention since it forms silicides with widespread applications in microelectronic devices [12,13]. This is due to its low electrical resistivity and small lattice mismatch with Si. For industrial applica- tions the Si(1 0 0) surface is more relevant than Si(1 1 1) surface. Palasantzas et al. [14,15] investigated the diffu- sion, nucleation and annealing behavior of Co on H-passivated Si(1 0 0) surfaces by scanning tunneling microscopy (STM). It is found that due to the absence of nucleation sites for silicide formation, the nucleation and growth mode is dominated by the formation of non- epitaxial islands which merge by increasing Co coverage. As the initial stage of the interaction, the adsorption of Co atoms on H-passivated Si(1 0 0) surface is of importance for understanding the properties of the sys- tem. In this Letter, by using the self-consistent tight binding linear muffin-tin orbital (TB-LMTO) method [16–18] with the supercell approach, which based on the density function theory within the local density approximation (LDA) using the Hedin-Lundqvist parameterization of correlation [19], the electronic struc- ture and adsorption properties of a monolayer Co atoms on a H-passivated Si(1 0 0) surface are studied. This ap- proach has been used to deal with the systems of Fe on b-SiC(100) surfaces [20] and Mn on GaAs(100) surface [19] which is in agreement with the experimental results. We have done a test calculation for the comparison of LDA with GGA. The distance of the adatoms from the Si(1 0 0) surface by LDA is only a little less than 0009-2614/$ - see front matter Ó 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.cplett.2005.02.008 * Corresponding author. Fax: +86 258 3595535. E-mail address: [email protected] (L. Ma). Chemical Physics Letters 405 (2005) 208–213 www.elsevier.com/locate/cplett

Upload: li-ma

Post on 26-Jun-2016

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Chemisorption of Co on H-passivated Si(1 0 0) surface

Chemical Physics Letters 405 (2005) 208–213

www.elsevier.com/locate/cplett

Chemisorption of Co on H-passivated Si(100) surface

Li Ma *, Jianguang Wang, Qiliang Lu, Guanghou Wang

National Laboratory of Solid State Microstructures and Department of Physics, Nanjing University, Hankou Road Number 22,

Jiangsu Province, Nanjing 210093, China

Received 4 November 2004; in final form 24 January 2005

Abstract

Chemisorption of one monolayer Co atoms on a H-passivated Si(100) surface is studied by using the self-consistent tight-binding

linear muffin-tin orbital method. Energies of adsorption systems, the layer projected density of states and charge distributions are

calculated. The most stable position is at fourfold hollow for the adsorbed Co atoms, and Co might sit below the H layer. Therefore,

a Co–H mixed layer exists at the Co/H-passivated Si(100) surface. The adsorbed Co atoms cannot sit below the Si surface. The

passivated layer of H atoms hinders the intermixing of Co atoms with Si at the interface effectively.

� 2005 Elsevier B.V. All rights reserved.

It is known that the presence of surface impurities is

nearly unavoidable in thin film growth. The introduc-

tion of the right impurity in the film/substrate interface

can greatly improve the morphology of thin films. In re-

cent years, the research works have demonstrated that

the introduction of certain impurities as adsorbates dur-

ing film deposition can improve various film characteris-

tics, such as growth mode, interface or surfaceroughness, etc. An understanding of the impurity effect

is eventually important in order to control a given

growth process at an atomic level. Hydrogen is the sim-

plest impurity and plays an important role in improving

the electrical properties of semiconductors such as the

termination of dangling bonds and the passivation of

deep level impurities [1–4]. In general, changes in the

film growth mode for a number of other metal systems(e.g., Ag, In, Cu, Al) have suggested an enhanced metal

adatom mobility on H-passivated Si surfaces [5–11].

Cobalt merits special attention since it forms silicides

with widespread applications in microelectronic devices

[12,13]. This is due to its low electrical resistivity and

small lattice mismatch with Si. For industrial applica-

0009-2614/$ - see front matter � 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.cplett.2005.02.008

* Corresponding author. Fax: +86 258 3595535.

E-mail address: [email protected] (L. Ma).

tions the Si(100) surface is more relevant than Si(111)

surface. Palasantzas et al. [14,15] investigated the diffu-

sion, nucleation and annealing behavior of Co on

H-passivated Si(100) surfaces by scanning tunneling

microscopy (STM). It is found that due to the absence

of nucleation sites for silicide formation, the nucleation

and growth mode is dominated by the formation of non-

epitaxial islands which merge by increasing Co coverage.As the initial stage of the interaction, the adsorption

of Co atoms on H-passivated Si(100) surface is of

importance for understanding the properties of the sys-

tem. In this Letter, by using the self-consistent tight

binding linear muffin-tin orbital (TB-LMTO) method

[16–18] with the supercell approach, which based on

the density function theory within the local density

approximation (LDA) using the Hedin-Lundqvistparameterization of correlation [19], the electronic struc-

ture and adsorption properties of a monolayer Co atoms

on a H-passivated Si(100) surface are studied. This ap-

proach has been used to deal with the systems of Fe on

b-SiC(100) surfaces [20] and Mn on GaAs(100) surface

[19] which is in agreement with the experimental results.

We have done a test calculation for the comparison of

LDA with GGA. The distance of the adatoms fromthe Si(100) surface by LDA is only a little less than

Page 2: Chemisorption of Co on H-passivated Si(1 0 0) surface

Fig. 1. Illustration of the present used supercell for Co adsorption on

H-passivated Si(100) surface (a) and the adsorption geometries (b).

The labels A, B1, B2, and C denote the adsorption sites for Co on the

top site, two bridge sites (B1 with a Si atom on the fourth atomic layer

below it and B2, with a Si on the second atomic layer below it) and the

fourfold site, respectively. Open circle, solid circle, downtriangle and

uptriangle represent first, second, third and fourth layer of Si,

respectively.

L. Ma et al. / Chemical Physics Letters 405 (2005) 208–213 209

GGA. They are almost same. So LDA can be used to

deal with the present problem. For clean Si(100) sur-

face, there exists the (2 · 1) reconstruction. The adsorp-

tion of atomic hydrogen onto Si(100) has been

extensively studied using a variety of spectroscopic tools

such as scanning tunneling microscopy [21,22], infraredspectroscopy [23–25], and temperature programmed

desorption (TPD) spectrometry [26,27]. It is reported

[21] that saturation exposure of a clean Si(100)–(2 · 1)

surface to atomic hydrogen at 600 K produces a pure

monohydride phase, HSi–SiH, in which hydrogen termi-

nates the dangling bond of the dimer Si. Saturation

exposures at lower temperatures, 400 and 295 K, lead

to the formation of surfaces which exhibit (3 · 1) and(1 · 1) LEED pattern, respectively [21]. The (3 · 1)

reconstruction was interpreted as being due to alternat-

ing monohydride (SiH) and dihydride (SiH2) units [23].

Traditionally, the (1 · 1) phase has been considered to

have a bulklike arrangement of the Si surface atoms

with the H in a uniform �dihydride� configuration

[21,28]. However, high-resolution infrared data of

Si(100) surface saturated with H at room temperatureshow [23] that the (1 · 1) surface is in fact a phase with

roughly half the H bonded in monohydride and half in

dihydride configurations. Cheng and Yates [29] studied

the H-induced surface structures on Si(100) using tem-

perature programmed desorption mass spectroscopy

and low-energy electron diffraction. They get the same

results. So at the room temperature adsorption the

Si(100)–(2 · 1) surface will change to the (1 · 1) phaseand it contains monohydride. In the calculation the

Si(100) surface will be taken as bulklike for simplicity.

According to above-mentioned experimental results,

this treatment for H-passivated Si(100) surface is rea-

sonable. For metal adsorption on clean Si surface, Si

atoms are expected to move significantly to form a sili-

cide structure. But at present it has a H passivation layer

on Si(100) surface which can make Si surface atomslocate at lattice sites. This system is different from that

of the adsorption on the clean silicon surface. So during

the calculation we fixed the structure of the substrate

and relaxed the adsorb layer according to the total

energy calculations.

When the TB-LMTO method with the atomic-sphere

approximation (ASA) is used to deal with the relatively

open zinc-blende structure, the empty spheres are usu-ally introduced at the tetrahedral interstitial sites for

providing an adequate description of the charge density

and potential in the interstitial region [30]. Lu et al. [20]

have studied the ground-state properties of b-SiC and

have shown that the results agree well with other theo-

retical calculations and experiments. In the case of the

solid-vacuum supercell, the vacuum region is also filled

with empty spheres according to the same structure asthe solid. The adsorption properties of Na, Ca atoms

on the Si(111) surface have been studied with the

solid-vacuum supercell and the results obtained are in

good agreement with experiment [31]. Therefore, it is be-

lieved that the TB-LMTO method with the solid-vac-

uum supercell can be used to deal with the present

problem. The supercell model shown in Fig. 1 is used

to describe the adsorption of a monolayer Co atomson H-passivated Si(100) surface. Fig. 1a is the schematic

diagram of the supercell which consists of five Si atomic

layers with a monolayer of H atoms saturating on each

side of the slab, an adsorbed monolayer of Co atoms is

put on one side of the slab, and four layers of vacuum.

In the region of the slab, some empty spheres are intro-

duced in the usual tetrahedral interstices. Test calcula-

tions of the layer projected density of states (LPDOS)with a thicker vacuum layer (equivalent to eight atomic

layers) and a thicker substrate (equivalent to nine silicon

layers) show that the results are almost the same as the

present model (Fig. 4a, b). The method of using a mono-

layer of H atoms to passivate the dangling bonds of the

slab is effective, which makes it possible to use a smaller

supercell (thinner slab and thinner vacuum) to simulate

the Si(100) surface. The experimental studies showedthat, for the Si(100) surfaces, the number of electroni-

cally active dangling bond states could be dramatically

reduced by the H passivation [32]. This method has been

used to improve the slab calculation. We only need to

put Co atoms on one side of the slab. The other side

can be looked as bulklike. Our calculation shows that

the adsorbed H atoms are more favorable on a bridge

site with a distance 0.056 nm above the Si surface.Therefore, the H atoms are set at those sites on each side

of the slab instead of ideal lattice sites.

Four possible adsorption geometries for a monolayer

of Co atoms on the substrate, namely, the top site (A),

two bridge sites (B1 with a Si atom on the fourth atomic

Page 3: Chemisorption of Co on H-passivated Si(1 0 0) surface

Table 1

The calculated total energy Etot of the supercell versus the vertical

distance D of the atoms from the Si surface

Site A B1 B2 C C1

D (nm) 0.252 0.282 0.232 0.151 0.001

Etot (eV) 1.753 0.918 3.971 0.018 0.000

210 L. Ma et al. / Chemical Physics Letters 405 (2005) 208–213

layer below it, and B2 with a Si atom on the second

atomic layer below it) and fourfold site (C) are consid-

ered (Fig. 1b). The initial distance between the Co

monolayer and the surface is chosen in such a way that

the bond length between the Co atom and its nearest-

neighbor surface atom equal the sum of their covalentradii. In the calculation, the valence electrons in the neu-

tral configurations are 3d74s2 for Co, 1s1 states of H and

3s23p2 for Si. The rest of the occupied levels were frozen.

The Brillionin zone (BZ) integration was performed

by the tetrahedron technique. A grid of 196 k points

in the Brillionin zone is used in the irreducible BZ to

construct the tetrahedrons. As a preliminary study, the

magnetic properties are not considered. We have doneone spin unrestricted test calculation of Co adsorption

on the optimized top site. The calculated spin moment

of the adsorbed Co is 0.82 lB, which is much smaller

as comparedwith the spin moment fora free Co atom

(3 lB).The spin-polarized state is lower energy (per

cell) by only 0.14 eV as compared with the non-spin-

polarized state. Calculating the distance of the adatoms

from the Si surface with spin-polarization and compar-ing withthe non-spin-ploarized, it is almost no effect

on the site of the optimized adsorption. So the effect

of spin-polarization can be ignored.

The variations of the total energy Etot of the supercell

versus vertical distance (D) of the adatoms from the

Si(100) surface are calculated. The minimum values of

Etot corresponding to D with different adsorbed sites

are shown in Table 1. The energies listed in Table 1are relative to the lowest value of the Etot for Co ad-

sorbed on the C1 site (i.e., C site below the H layer in

Fig. 1b). From Table 1, it can be concluded that, among

the adsorption sites considered here, the most stable po-

-0.10 -0.05 0.000.00

0.75

1.50

2.25

3.00

3.75

Distance

Tot

al e

nerg

y/eV

C1

Fig. 2. The total energy Etot (eV) of Co/H-passivated Si(100) system

sition of the Co atom is at the C1 site below the H layer

with a distance of 0.001 nm above the Si surface. This is

almost on the Si surface. Above the H layer, the most

stable adsorption position is a C site with a distance of0.151 nm from Si surface. Fig. 2 gives the curve of the

total energy Etot of the system versus vertical distance

D of the Co atoms from the H layer when Co adsorbed

on the C site. It shows that above and below the H layer,

there exist the stable adsorption sites at distances 0.095

and 0.055 nm. These distances correspond to C and C1

sites, respectively. Near the H layer there a barrier and

a local minimum value lies at 0.04 nm above the H layer.Furthermore, another barrier appears at 0.06 nm above

the H layer. Therefore, a Co–H mixed layer might exist

at Co/H–Si(100) surface. While setting Co below the Si

surface, Fig. 3 gives the total energy versus vertical dis-

tance from the Si surface curve. From Fig. 3, it can be

seen that no stable position exists below Si surface,

which shows that the Co atoms cannot diffuse into the

lattice. The latter excludes the existence of a Co, Simixed layer at the interface. It indicates the H passiv-

ation layer effectively hinders Co, Si intermixing. This

result is in agreement with experiments [14]. When Co

atoms adsorb on an ideal Si(100) surface, both experi-

mental and theoretical investigations showed that the

Co atoms can diffuse into the lattice and form a Co, Si

mixed layer at Co/Si(100) interface [33,34]. The different

0.05 0.10 0.15

/nm

C

versus the vertical distance D of the adatoms from the H layer.

Page 4: Chemisorption of Co on H-passivated Si(1 0 0) surface

-0.10 -0.08 -0.06 -0.04 -0.02 0.00 0.020.00

3.25

6.50

9.75

13.00

Distance/nm

Tot

al e

nerg

y/eV

Fig. 3. The total energy Etot (eV) of Co/H-passivated Si(100) system versus the vertical distance D of the adatoms from the Si surface while Co

adsorbed below Si(100) surface.

L. Ma et al. / Chemical Physics Letters 405 (2005) 208–213 211

adsorptionmorphologies between ideal andH-passivated

Si(100) surface indicate that H can act as an effective

buffer layer. According to the above discussion, it can

be inferred that because the absence of a subsurface stablesite forces the Co into a surface site which is unstable with

respect to island formation. Once the Co has migrated to

islands, some CoSi formation can take place by reactions

at island edges. The path of the reaction is similar to the

interstitial precursor to silicide formation on Si(111)–

(7 · 7) [35]. In the following, only the C, C1 sites adsorp-

tion are considered.

The LPDOS for the clean Si(100) surface, non-adsorbed system, which is moving away the Co layer,

and the Co adsorbed system are shown in Fig. 4. The

density of states (DOS) for the bulk is shown by a dashed

line for comparison. The zero-point energy (marked by a

vertical line) is aligned at the Fermi level. From the fig-

ures, it can be seen that the LPDOS of the middle layer

is bulklike, which indicates that the model of supercell is

reasonable to simulate the Co/H-passivated Si(100) sys-tem. In the case of the clean Si(100) surface (Fig. 4b), it

can be seen that the LPDOS for the surface layer is

greatly different from the bulk one. A high sharp peak

A appears at the Fermi level with a large tail in the en-

ergy gap, which corresponds to the Si dangling band

states contributed mainly by Si 3p state. Another peak

B appears at about 7.5 eV below the Fermi level. In con-

trast to peak A, peak B is resonance state mainly contrib-uted by the Si 3s state. The effect of the H passivation is

noticeable. It can be seen from Fig. 4c that the peak A

damps rapidly after adsorption, peak B decreases and

shifts down. This is the interaction of the passivated H

atoms with surface Si atoms and partial saturation of

the dangling bonds of the surface atoms. Moreover, it

is mainly the bonding of H 1s state with Si 3p state in

the surface layer. From Fig. 4d, e, it is Co adsorbed

above and below H layer, respectively, it can be seen thatthe peaks in the surface layer disappear completely. The

result is attributed to the interaction of Co, H and Si in

the surface layer and mainly come from the Co 3d, H 1s

and Si 3p states.

Usually, chemisorptions are accompanied by charges

transfer between the adsorbates and the substrate.

Table 2 gives the layer effective charges (defined as the

atomic charges relative to the neutral configurations),which are obtained from the sums of the effective

charges of all atoms and the empty sphere on the corre-

sponding layers in the units cell. It is found from Table 2

that, for the non-adsorbed system, the effective surface

charges are negative. This means that the surface Si

atoms transfer some of its electrons to the H-passivated

layer, second layer and empty sphere in the vacuum.

Clearly, H-passivated layer gains 0.23 electrons andthe second layer gains 0.06 electrons. The effective layer

charges after Co adsorption are also given in Table 2.

Comparing with the non-adsorbed system, for the case

of Co adsorption on a C site, the surface layer gains

0.11 electrons and the subsurface gains 0.05 electrons.

At the same time Co loses 0.24 electrons. Therefore,

there are totally about 0.16 electrons are transferred

from the adsorbed Co atom to the substrate. For theCo adsorption on C1 site, the surface layer gains 0.23

electrons and the subsurface gains 0.02 electrons. The

Co layer loses 0.29 electrons. Therefore, the sum of

charges transfer is 0.25 electrons from the Co atom to

the substrate. From Table 2, it can also be seen that

Page 5: Chemisorption of Co on H-passivated Si(1 0 0) surface

0

2

LPD

OS

/ ar

b . u

nits

LPD

OS

/ ar

b . u

nits

LPD

OS

/ ar

b . u

nits

LPD

OS

/ ar

b . u

nits

0

2

0

2B

A

2

0

2

0

2

AB

0

2

0

2

0

2

0

2

H

-12 -8 -4 400

2Middle0

2Subsurface

E/eV-12 -8 -4 40

E/eV

-12 -8 -4 40E/eV

-12 -8 -4 40E/eV

-12 -8 -4 40E/eV

0

2Surface

Middle

Subsurface

Surface

Middle

Subsurface

Surface

Middle

Subsurface

Surface(a)

(d) (e)

(b) (c)

0

2H

Middle

Subsurface

Surface

H0

20

LPD

OS

/ ar

b . u

nits

0

2

0

2

0

2

0

2

0

20Co Co

Fig. 4. The LPDOS for test calculation (a); the clean Si(100) surface (b); non-adsorbed system (c); Co adsorbed on C site (d) and on C1 site (e). The

vertical line indicates the Fermi level and the dashed line is the DOS for the bulk.

Table 2

The layer effective charges (in the unit of electron) in the unit cell for

the non-adsorbed system and Co adsorbed system

Adlayer Passivated

layer

Surface Subsurface Middle

layer

0.23 �0.47 0.06 0.00

C site �0.24 0.28 �0.36 0.11 0.01

C1 site �0.29 0.07 �0.24 0.08 0.02

Here, the affective charges denote the sum of the effective charges of all

atoms and empty spheres inside the layer.

212 L. Ma et al. / Chemical Physics Letters 405 (2005) 208–213

the effective charges on the third layers are almost not

affected by the adsorption of the two cases, showing thatthe thickness of the slab is reasonable.

In summary, the chemisorption of Co adatom on the

H-passivated Si(100) surface is studied by the

TB-LMTO method. An effective method of using a

monolayer of H atoms to passivate the dangling bonds

of the slab is introduced in this study, which makes

possible the use of a smaller supercell to simulate

the Si(100) surface. The total energy calculation and theanalysis of the layer projected density, as well as the

charge distribution show that for the different adsorp-

tion sites, the Co atoms are more favorable on a C site

and might sit below the H layer. Therefore, a Co–H

mixed layer might exist at the Co/H–Si(100) surface.

The adsorbed Co atoms cannot exist below the Si sur-face. So there cannot form a Co, Si mixed layer at the

interface. From above discussion, we can conclude that

the use of a H passivated layer can effectively hinder Co,

Si intermixing at the interface, which is in agreement

with the experimental results.

Acknowledgment

This work was financially supported by the National

Natural Science Foundation of China (Nos. 90206033

and 10274031).

References

[1] S.J. Pearton, J.W. Corbett, M. Stavola (Eds.), Hydrogen in

Crystalline Semiconductors, Springer, New York, 1992.

Page 6: Chemisorption of Co on H-passivated Si(1 0 0) surface

L. Ma et al. / Chemical Physics Letters 405 (2005) 208–213 213

[2] J. Bpland, Surf. Sci. 261 (1992) 17.

[3] J.J Boland, J. Vac. Sci. Technol. A 10 (1992) 2458.

[4] W. Widdra, S.I. Yi, R. Maboudian, G.A.D. Briggs, W.H.

Weinberg, Phys. Rev. Lett. 74 (1995) 2074.

[5] K. Tsubousi, K. Masu, K. Sasaki, Jpn. J. Appl. Phys. 32 (1993)

278.

[6] T.-C. Shen, C. Wang, J.R. Tucker, Surf. Sci. 78 (1997) 1271.

[7] T. Hashizume, S. Heike, M. Lutwyche, S. Watanabe, K. Nakaj-

ima, T. Nishi, Y. Wada, Jpn. J. Appl. Phys. 35 (1996) 1085.

[8] A. Nishiyama, G. ter Horst, P.M. Zagwijn, G.N. van den Hoven,

J.W.M. Frenken, F. Garten, A.R. Schlatmann, J. Vrijmoeth,

Surf. Sci. 350 (1996) 229.

[9] K. sumitomo, T. Kobayashi, F. Shoji, K. Oura, I. Katayama,

Phys. Rev. Lett. 66 (1991) 1193.

[10] F.P. Leisenberger, H. Ofner, M.G. Ramsey, F.P. Netzer, Surf.

Sci. 383 (1997) 25.

[11] T. Yasue, T. Koshikawa, Surf. Sci. 377 (1997) 923.

[12] R.T. Tung, Mater. Chem. Phys. 32 (1992) 107.

[13] V. Scheuch, B. Voigtiander, H.P. Bonzel, Surf. Sci. 372 (1996) 71.

[14] G. Palasantzas, B. Ilge, J. de Nijs, L.J. Geerligs, Surf. Sci. 412

(1998) 509.

[15] B. Ilge, G. Palasantzas, L.J. Geerligs, Appl. Surf. Sci. 144 (1999)

543.

[16] O.K. Andersen, O. Jepsen, Phys. Rev. Lett. 53 (1984) 2571.

[17] O.K. Andersen, Z. Pawlowska, O. Jepsen, Phys. Rev. B 34 (1986)

5253.

[18] H.J. Nowak, O.K. Andersen, T. Fujiwara, O. Jepsen, P. Vargas,

Phys. Rev. B 44 (1991) 3577.

[19] Zongxian Yang, Kaiming Zhang, Sanhuang Ke, Xide Xie, Phys.

Rev. B 56 (1997) 6727.

[20] Wenchang Lu, Kaiming Zhang, Xide Xie, Phys. Rev. B 48 (1993)

18159.

[21] J. Boland, Surf. Sci. 261 (1992) 17.

[22] J.J. Boland, J. Vac. Sci. Technol. A 10 (1992) 2458.

[23] Y.J. Chabal, K. Raghavachair, Phys. Rev. Lett. 54 (1985)

1055.

[24] Y.J. Chabal, J. Vac. Sci. Technol. A 3 (1985) 1448.

[25] Miyako Terashi, Jyun-ko Kuge, Masanori Shinohara, Daisei

Shoji, Miwano Niwano, Appl. Surf. Sci. 130/132 (1998) 260.

[26] S.M. Gates, R.R. Kunz, C.M. Greenlife, Surf. Sci. 207 (1989)

364.

[27] W. Widdra, S.I. Yi, R. Maboudian, G.A.D. Briggs, W.H.

Weinberg, Phys. Rev. Lett. 74 (1995) 2047.

[28] S. Cirici, R. Butz, E.M. Dellig, H. Wagner, Phys. Rev. B 30

(1984) 711.

[29] C.C. Cheng, J.T. Yates Jr., Phys. Rev. B 43 (1991) 4041.

[30] D. Glotzel, B. Segall, O.K. Andersen, Solid State Commun. 36

(1980) 403.

[31] O. Bisi, C. Arcangeli, S. Ossicini, Phys. Rev. B 43 (1991) 9823.

[32] D.R. Lide, H.P.R. Frederikse (Eds.), CRC Handbook of Chem-

istry and Physics, CRC Press, Boca Raton, FL, 1995.

[33] H.L. Meyerheim, U. Dobler, A. Pusechmann, Phys. Rev. B 44

(1991) 5738.

[34] S. Wei, Z. Yang, X. Dai, K. Zhang, Surf. Sci. 37 (2002) 504.

[35] P.A. Bennett, D.G. Cahill, M. Copel, Phys. Rev. Lett. 73 (1994)

452.