Si interfaces

Applied Surface Science 56-58 (ITS)2)408-415 NnUh-Holland

surface

science

Schottky barriers at epitaxial silicide/Si interfaces Hideaki Fujltani Ftzjltslr Labotarorie~ Ltd.. I0-1 Morint~ato-|t~lkami)'a. At.~ngt-adli,Kanagawa-,~ot 24.?-01. J(ll~ltl

and Setsuro Asano ht~toute of Ph)'.m:s. Colhx'e of Arts aJM Svu.a¢¢~. Uni, t,r.iay of Tokyo, 3-8-1 Komaba, Mcgum.kll, Tokyo 153, Jalmn

Received 6 May Iqgl; accepted for publicalion 5 Augusl Iqgl

we studied Ih¢ elcccro~ic structure of the YSi:/Si[I I I ) and NiSi2/Si(0Pl ) interfaces using Ihe linear muffin :in orhitals in the atomic sphere approximation (LMTO-ASAI based on the local density approximation (LDAL Together with the previous r~suhs,n the CoSiffSi( 11 l) and NiSi:/Si(I I 1) interfaces, we showed that LMTO-ASA calculations with a large supercell give an adequate Schonky barrier height (SBH: E F E v) P~r real silicide/Si interfaces although the LDA depresses the hand gap of hulk Si to almost hall (ff the experimental value. AP ¢ightfotd IqiSi2/Sil001) interface showed almost the same SBtl as ~ho type A NiSi,/sitlll) intcrfa¢~¢. From Ih¢ relation between Ihe interface slructur¢ and the calculated SBH. we speculate that the btmd angle at the interface affecls Ihc SBH.

I. Introduction

The atomic structure of a metal-semlconductor interface is difficult to clarify because the interface is buried by an overlayer and its structure depends on the materials and conditions during formation. This kept the electronic structure of the real metal-semiconductor interface from being clarified for a long time. To understand how the Sehottky barrier is formed at such an interface, we must be able to clarify the electronic structure of a well-defined interface. T h e interface of metal silicide and Si can be a good example for this. The epitaxial NiSi2/Si(I l 1) interface has type A and type B structure in which the interface Ni atoms are sevenfold coordinated, l y p e A and type B indicate two interfere structures with the overlayer rotated 180~ around the Si(111) axis [I]. In 198, T u n g discovered that the Schottky barrier heights (SBH's) in these N i S i z / S i ( l 11) interfaces

differ [2]. This conflicts with the conventional understanding of the Schottky barrier [3,4]. and it is much disputed [5,6]. We performed a firstprinciple calculations based on the local density approximation (LDA) and obtained different SBH's for the two structural types, which is consistent with T u n g ' s discovery [7]. Until now, some groups repc,rted calculations on the NiSi 2/Sit I 11) interfaces, and obtained differences in the SBH's between the two types of interfaces [8-10]. But the reason the two types have different SBH's has not yet been resolved [10]. The S B H differences between the two types of interfaces is only 0.14 eV, which is one percent of the valence band width. It is difficult to understand the formation mechanism of the Sehottky barrier only by exami n i n g t h e e a l c u l a t i o n a l r e s u l t s o n the NiSi2/Si(I f 1) interfaces. It is well known that the L D A depresses the bulk band gap to almost half of the experimental value. So, it i: questionable how accurately the

0169-4332/92/$U5110 © 19q2 .- Elsevier Science Publisaers B.V. All righls reserved

H. FujtlanL S. Asano / Schottky barricra at epitaxial sdictde / Si imel faees

calculation with the LDA describes the electronic structure of the Si interfaces with different materials. The supercell size is an another problem. Since the supereell size affects the calculated SBH, the supereell must be large enough to get a reliable result [7]. With large supereells v,e obtained SBH's (in this paper, SBH always means Fermi level minus the top of the Si valence band: E F - E~) of the two types of NiSi J S i ( I 11 ) interfaces. The calculated SBH's were about 0.1 eV smaller than the experimental values, probably because of the error in the LDA. Rare-earth silieides are unique in that they have SBH's larger than half of the Si band gap [11]. It is interesting to know what SBH can be obtained by calculations with the LDA. We studied the YSi~/Si(I I 1) interface assuming a atomic structure experimentally suggested [12]. Recently, Tung et al. reported a new SBH measurement on a single-crystal, uniform, planar NiSi,/Si(001) interface formed by it novel technique in which stoichiomctric NiSi 2 is co-deposited on St(001) at low temperatures and annealed at high temperatures ( > 700°C) [13]. This interface exhibits a large SBH of about 0 72 eV. They say that together with the two types of (111 ) interface the SBH at the NiSi2/Si interface changes about 0.4 eV depending rjn the interface atomic structure. To clarify this fact, we examined the NiSiJSi(001) interface.

2. Calculation We made scalar relativistic calculations using the linear muffin-tin orbital method in the atomic sphere api~oximation (LMTO-ASA) [14]. Exchange and correlation are determined by the LDA with the parameterization of Janak, Morazzi, and Williams [15]. A nearly orthogonal representation was used for the muffin-thin orbitals and the combined correction was not included. We adopt s-, p-, and d-orbitals for St, Co, Ni, Ca, F, and empty spheres. As for Y we include f-orbitals [16]. The atomic sphere radii are determined for the bulk band dispersion of each material to agree with that obtained by the full potential linearized augmented plane wave

409

method. The radii are 1.337 ,~. for St, 1.222 ,~ for Ni and Co, 1.991 A for Y, 1.466 A for Ca, and 1.240 ,~ for F. The empty sphere radii of the diamond and fluorite structures are determined for the total volume of the snheres being the unit cell volume. YSi 2 has the C32 crystal structure and does not contain an empty sphere. The empty sphere radii at the interface were chosen to decrease the sphere overlap. We used an Si lattice constant of 5.429 A and did not include lattice relaxation except for the T 4 site structure of the C a F 2 / S i ( l l l ) interface in which the distance between the interface layers is taken to be the experimental value [17]. The supereells we used have space-group symmetries of P3ml (D{.j) for all the S t ( l I D interfaces and Cmmm ( D ~ ) fro" the NiSiz/Si(001) interface. To get reliable results, the supercell must be large enough because the supercell size affects the calculated SBH through the valence band width of each silicide and Si layer [7]. We used supercells which contain 9-11 Si z layers and 8-11 silicide layers.

3. Results 3.1. Y S i 2 / S i ( I I I )

YSi2 has a layered structure. Within each Si and Y layer, the atoms are arranged in a planar mesh with sixfold symmetry. Baptist et al. observed that the YSi 2 surface is Si terminated with an upward displacement (0.8 ,~) of one Si atom out of two so that it exhibits the same geometry as a S t ( I l l ) (1 × 1) surface. So, it is natural to assume that at the interface the top Si layer of YSi 2 continues to the Si substrate, so the Y atom at the interface resides at the H a site (fig. l). In this structure the interface bonds are only slightly bent and there will be no dangling bond at the interface. The distance between the interface Y and the Si layers (d in fig. l) is assumed to be the same as the distance between the Si and Y layers of bulk YSi 2. Fig..7. is a two-dimensional projected hand structure of the Y S i z / S i ( l l l ) interface, which was obtained by examining the wavefunetion

H I'itjilatlk & Aseno / S~lumky I~lrrh't~at #patL~udsili~tde/ Si mtvr~we~

410

YSi2tSi(111)

Niai=/Si(001)

C)

Y

• Si

0 Ni

Fig I. AlomicslruCture~:it the YSiz/Si(I I I ) and NiSi2/Si(Olll ) interface.

weights of the energy eigenvalues of the supcrccll in each atomic sphere of the Si and YSi z ia,'ers. In the projected band, there are many region', in which bulk YSi 2 states do not appear. This comes partly from the layered crystal structure of bulk YSi 2. For instance the energy dispersion of the bulk s-band at the bottom of the valence band is

larger in the layered plane than in the direction perpeudicular to the layers. Bold lines in fig. 2 are interface states whose wavefunctlons are localized near the interface. The interface states in the Si band gap are formed mainly of interface Sip-orbitals and Y d-orbitals, and they maintain a bonding and anfibonding

4

~

-4

uJ

YSI,

8

1 F

M K Wave vector

F

Fig. 2. Two-dimensional prujected band structure of the YSi~/Si(Il I } interface obtained from the supercell calculation.The zero energ~point is the Fermi level.The bold lines indlcate the interface states which are l~allzed near the interface.

H. Fujllant, S, Asftno / Schottk)' btlrrier~ at epitaxial siltcidc~St hm'r]'acex

Table 1 Seht)llky barrier height, (E r - E~) of the silicide/Si(lllt interface in eV Material YSi.~ cosi_~

Structure Calculation Experiment Ha u.sfi 11.73 T4 a,4a 8A a,37 81~ a,25 n,4 I).b NiSi, 7A 1134 (I.47 " 7o a,lg a.33 The calculated values are obtained from tile ciscnvalucs tlf the supercells which conlail) 9 Si2 luyer~ and 8-10 silicide layers. Experimenlal values ate from rcf~ 12,1t25],

character, because the atomic structure changes little at the interface. The interface state which overlaps the bulk YSi z states is inclined to extend from the interface. Near - 8 eV there is the interface state which is formed mainly of interface Si s-orhitals. The calculated SBH is 0.56 cV. It was obtained from the Fermi energy and eigeovalue of the supercelL This is 0.17 eV smaller t~mn the experimental value (table 1). Since we think of the po~ibility that the distance between the interface Si and Y layers (d in fig. 1) may be smaller than the bulk, we performed calculations with the interface Y - S t distance 0.1 ,~ smaller than the bulk, In this structure the SBH was 0.66 eV. However the conduction bottom of the Si layer was 0.02 eV lower than the Fermi energy of the supercell. Anyway, the contraction of the interface Y - S t distance makes the SBH higher, although, at the N l z l i 2 / S i ( l l l ) interface, the contraction of the interface St-St bond length makes the S B H slightly lower [S]. 3.2. N i S i e / S i ( O 0 1 )

N i S i J S i ( 0 0 1 ) is considered to have an interface structure as given in fig. 1. It is questionable whether the interface Si atoms (circled in fig. 1) exist or not. If these atoms do not exist, the interface Ni atoms are sixfold coordinated. On the other hand if they exist the interface Ni atoms are eightfold coordinated. We examined both structures.

411

The calculated SBH at the NiSiz/Si(O01) was 0.36 cV for the eightfold model with a supercell containing 7 Si 2 and 7 NiSi 2 layers, and it was --(I.02 e V for the sixfold model with a supercell containing 6 Si 2 and 7 NiSi:: layers. The Fermi energy of the sixfold model is lower than the E v of the Si layer. This is an un':easonable SBH. We previously reported that a wrong flvefold model of the C o S i , / S i ( l l l ) interface gives a negative SBH although a correct eightfold model gives a positive and reasonable SBH [18,19]. So, we conelude that the sixfold model does not represent the NiSi J S i ( 0 0 1 ) interface. Since the calculated SBH depends on the su. pereell size, we performed calculations with 11 Si z and 11 NiSi~ layers for the eighffold model to compare the SBH's between the (001) and ( l i d NiSi2/Si interfaces. Since we did not include spin-orbit interaction, the E~ of bulk Si has three-fold degeneracy. At the (111) interface, E~ of the Si layer was doubly degenerate and was easily distinguishable by the wavelianction weights of each eigenvalue of the supercell [7]. At thc C001) interface all elgenstates at the F point are single-valued because the interface structure lowers the space symmetr~ of the superecll, so it is difficult t.., determine the Ev of the Si layer from the eig,:nvalues of the supercell. To obtain the SBH of the NiSi2/Si(001) interface, we used the frozen potential method in which the one-electron potentials of the NiSi 2 and Si 2 layers farthest from the interface are cut from the self-consistent potential obtained by supercell calculations, and exported to the bulk band calculations, which yield the Ni$i z Fermi energy and Si valence band maximum. We obtained 0.35 eV for the eightfold (001) interface. The S B H of the N i S i 2 / S i ( l 11) interface obtained by the supercells with 12 Si 2 and 11 NiSi 2 layers was 0.36 eV for type A and 0,19 eV for type B [7]. The eightfold N i S i J S i ( 0 0 1 ) interface has almost the same SBH as the type A (111) interface. Although these supereells may he still too small to remove the cell s i z e dependence from the calculated SBH the final SBH values probably stay within 0.04 eV from these values. Fig. 3 is the local density of states ( L D O S ) of

412

IL l'~tjtmm, S. ztxlulo / St hollky tmrrit'r~ at eptnt~tal slhl kh, / Si Ptletfiu'e~

(111) Type A

(OOl)

a~l

.A[

- ' ' 4;

.~

0i

"~

[

i

i oL

"~, " ~ 1

~

r .....

0k

o

,,

1i

Energy

Energy (eV)

(eV)

Fig. 3. Local dcn~ily o[ ~talCs( [ DOSI of the eightfold NiSi:/Si(l~)l ) interface anti of the type A NiSi:/Si( 111) inlerhlce. From top hi hlHIllm, ate ~ho~n Ihu sixlh NiSi z layur, Ih¢ first NiSi~ layer, Ihe [irsl Si z layer, avid Ih¢ Si 2 farlhesl from the interface. The d~lted line~ are hulk density of ~talus of Si and NiSi_,, The dolted regit~n~indlcale interface slates. The zero energ~ poinl is the Fermi energy o[ Ihe supercclls.

_

the eightfold (ll01) and ( l i d type A N i ~ i 2 / S i interl'accs. T h e L D O S is very different for t h e (0{)1) and (I I 1) interfaces. A t the (I 11) interface, the large peak of d-clcctron~ of t h e interface Ni a t o m is shifted to higher eftergy, however at t h e (O(}l) interface it stays at almost t h e same energy. A t the ( I l 1) interface, the interface state originates in the interface Ni d-orbitaL At the (001) interface, interface states originate in the two dangling bonds of the interface Si atom, and they are formed near t h e Fermi energy and at t h e bottom of the valence band. Roughly speaking, t h e et~ctron distribution is mainly d e t e r m i n e d by t h e atomic structure at t h e silicide/Si interface. Since a t o m s are more crowded at the eightfold f 0 0 D interface than at t h e (111) type A interface, electrons overt'low at t h e (OO1) interface and decrease at the ( I l l )

3I 2 ,~. ~

i ~ee-

o

0

_

m

-e-(O01) •-C'- (111) A

j.

-~ -1 ~. -2 -a -4

16 Si spheres

Fig. 4. Potenlial at Si h~utiong. The potential is calculaled

using Madelung constants and charges in olher atomic spheres. The lines arc shifted for the righl-h;md side to become zero The center llne indicalcs the inlerface. The abscissa does nat mean real distance.

H. Fail/ant, S. Asano / Schottlo, barriers a¢ ep,a~ial sflicide / Si imerfaces

interface, In fig. 4, we plotted the intersphere electrostatic potential on each Si site at the NiSi2/Si interfaces. For convenience of comparison we set the value of both series on the righthand side to be zero. Near the interl~tee the potential is quite different for the (001) and (I I I) interfaces according to the electron distributions. This large difference is dielectrically screened nut in two or three layers on both Si and NiSi, sides, and the potential difference between the Si and NiSi~ layers goes to the same level as for the ((gH) and ( l i d interface. The (001) inte.Caee has a much lower potential at the interface. This is the reason why there exist interface states lower than the valence hand.

413

this structure, a large peak appears in the Si band gap near the interface and the Fermi level is pinned by this interface state. When this interface is annealed at a high temperature, F atoms are dissociated from the interface and the T 4 structure remains [17,21,22]. The large peak in the LDOS disappeal,',, and separates into a bonding and antihonding interface states. At the cightfold NiSiJSi(0Ol) interface the interface Si atom has two °'dangling bonds", which form the interface states both in the Si band gap and at the bottom of the valence hand, similarly to the eightfold C o S i 2 / S i ( l l l ) interfaces where the inter|ace Si atom has one "dangling bond" [18.23]. The interface states originated in the "dangling bond" of the Si atoms differ quite from the silicide/Si interfaces and the C a F 2 / S i ( I l I ) interface, which is an insulator--semiconductor interface. At the NiSi2/Si(00I) interface formed by the conventional template technique, there are many ( I l l ) facets which have the (111) type A structure. So, the observed SBH of 0.47 cV at the NiSi:/Si(O01 ) interface was attributed to the (1 l l)

4. Discussiun According to the experiment by Zegenhagen ana Patti [20], C a F : / S i ( l l t ) interfaces formed al low temperatures have the sevenfold type B structure (fig. 5). In the local density of states of

.~"

7B

==

pea -15

-10

-5

0

5

,F .St ~

T~

nI -~5

-10

-5

O

5

Energy (eV) Fig. 5. Local density of Slates in the Si: layer nearest IO the C a F J S i ( l l 1) interface and the interface atomic structures f'lr Ihe se;'en fold and T 4 type n CaF:/Si{ I 11 ) interfaces, The dtmed lines are Ihe denshy of states for bulk St. The dotted rcgitms indicate the interface slates. Zero energy point is the Fcrmi energy of the supcrccll.

414

/L Fto?tauL S, A~ano / Xcho:ri,v hamer~ a~ t'pittLriwlsiltcide /.SJ Hiterfaccs

facets, ltowevcr, our calculation shows that the eightfold NiSi2/Si(O01) interface has the same SBH as the ( l i d type A interface. We suppose that the observed SBH at the (001) interface is attriboted not only to the (111) facets with type A structure but also to the eightfold (0fill structure. Although the NiSi,/Si(001) interface was ennsidered to be a sixfold structure from transmission electron microscopy (TEM), the (001) interface is not clearly observed in comparison with the (I 11 ) facet (fig. 3 in ref. [24]). Since there are two "'dangling bonds" at the eightfold NiSi.~/Si(001) interface, we suppose that during the formation, of the interface the "Mangling bonds" make bonding states with each other or with another atom so that other atomic structures partly exist in the (001) interface. At a C o S i 2 / S i ( l l l ) interface with a low density of dislocations. Tung reported that the S B H varies from ~ 0 . 4 to ~ 0 . 6 [25]. Using the T 4 structure of the CaF2/Si(I 11) interface we performed calculations for the C o S i 2 / S i ( l l 1) interface and obtained a S B H of 0.49 cV which is 0.24 e V larger than the SBH of the eighifold type B 18B) interface (table 1). Although the T.t structure at the C o S i . , / S i ( l l l ) interface is not verified by experiment, this result shows which characteristics Of the interface structure affects the SBH. Type B has a lower S B H than type A at the CoSi_,/Si(l 11) and NiSi2/Si; 111 ) interfaces. T h e "I"4 CoSi2/Si(I I 1 ) and H 3 Y S i 2 / S i ( I 1 I) interfaces have larger SBH's than the type A interfaces. There are common features among the above interfaces, i.e. the twisted interface lowers the SBH and the interface bond bending raises the SBH. At the NiSi2/Si(001) interface formed by Tung el al. there is a "'chain" of recq~n~'r::,.;.~d ~et, inus. where I}t¢, 1,~'~-' b~,,umg probably occurs. We sp,~culate that the reconstructed regions are concerned with the large experimental S B H of 0.7 e V at the NiSiz/Si(001) interface formed by the novel technique [29]. The Y S i J S i ( l l l ) interface naturally takes the H 3 structure as explained above. T h e C o S t 2 / St( I l 1) interface prefers the eightfold str,aeture to the sevenfold structure because the Co atom has one d-electron less than the Ni atom [26,27]. The interface properly is intimately eonnectcd with

YSi z ~

2

. . . . . . .

o:

~6 a

- -

NISi~

~a m °_15

-10

-5

0

5

Energy (eV) Fig. 6. Density of states (DOS) of bulk NiSi-. CoSt2. afld YSi,. Solid lines ate partial DOS of the d-mbitals of the mehd atoms. The dotted lines are the tolal DOS. The zero energy is the Fermi enetgy. The positions o[ the large peak, of the d-orbital arc indiea:ed by arrows.

the bulk property of the silicides. Fig. 6 shows the bulk density of states of each silicide, T h e dotted lines are the total density of states ( D O S ) and the solid lines are the partial D O S of the d-orbitals of the metal atoms. The positions of the large peak of the d-orbitals (indicated by arrows) move to higher energy in the order of NiSi,. CoS!a, and YSi 2. If compared wi*~ :.h~ same types, the NiS; /~;;(il i,~ interface has lower S B H than the C o S i 2 / S i ( l 11 ) interface. This corresponds to the position of the d-orbitals, although the energy scale differs by an order of fifty between the d-position and SBH's [28]. In summary. L M T O - A S A calculations with large supercells give adequate SBH's for real silicide/Si interfaces although the L D A depresses the Si band gap to almost half of the experimental value. The elghtfold NiSi2/Si(001) interface shows almost the same S B H as the type A N i S i 2 / S i ( l l l ) interface. This fact suggests in principle that the SBH changes depending on

tL Fl~jilatli, S. Asuno / Schottkg barriers at epitaxial silicide / S i interfaces specific c h a r a c t e r i s t i c s of t h e interface a t o m i c structure.

Ac "lmowledgments W e t h a n k R.T. T u n g for c o m m u n i c a t i n g his results prior to publication. T h i s work is p a r t l y s u p p o r t e d by a G r a n t - i n A i d for Scientific R e s e a r c h (03216205) from t h e M i n i s t r y o f E d u c a tion, S c i e n c e , a n d C u l t u r e , J a p a n .

References [I] R,T. Tung, J,M. Gibson and J.M. Poate, Phys. Roy. Lea. 511( 19831 429, [26 RT. Tung. Phys. Roy. l.etL 52 11984) 461r t3] C. To[odor. F. Flutes and E. Louis, J. Phy~rC 01 (19771 2163. 14] J. Tersuff. Phys, Roy. Left. 52 (19841 465; phys. Rev. 32 (1985} 6t)68. 156 M. Liehr. P.E, Schmidt, F.K. LeGoue~ and P.S. H(I. Phys. Rev Letl, 54 11985) 2139. [g] R.T. Tullg, K.K, Ng, J.M. Gib~nn and F.J. Levi. Phys. Rev. B 33 (1986) 71)77; R.J. Hauenstein. T.E. Schlesinger, T.C. McGill. B.D. Hunt and LJ. Sunowalter, J. Vac. Sci, Technol. A 4 ( 1986} 860; M. O~pelt. J. Henz, L. Flepp and tl. ','on Kiinch Appl. Phys. LcIL 52 11988) 227. [7] 11. Fujitani and S. Asano. J, Phys. Soc. Jpn. 57 (19931 2253: Phys. Rev. B 42 119901 Ifi96 [g] G.P. D~:. P. BI;~chl.N.E. Christensen and O,K. Andcrs~'n. ;~1: Mclallizalion and Metal Semiconductor Interfaces. Ed. I.P. Batra (Plenum, New York. 1988); G.P. Das, P. Bliichl. O.K. Andersen. N.E. Chrislensen and O. Gunnarsson, Phys. Rev. Len. 63 11989} 1168. [q] M.D. Stiles and D.R. Hamann, Phys. Roy. B 40 11989) 1349. El0] S, Ossicilli. O. Bisi and C.M. 13erloni. Phys. Rev. B 42 (1990) 5735.

415

[11] K.N. Tu, R.D. Thompson and B.Y, Tsaur. Appl. Phys. Lett. 38 (19el) 626. [12] R. Bapllst, S. Fetter, G, Gnenet and H.C. Pupa, Phys. Roy. Lgtt. 64 (19901 311, [13] R.T. Tung, A.F,J, Levi, J.P. Sullivan ard F. Scllr~:y,Phys, RCV. LOll. (~h( ]991 ) 72. 614] O.K. Andersen, Phys. Rev. B 12 (1973! 3060; O.K. Andersen, O, .lepsen and D. Gl/31zcl, in: Highlighls uf Condensed Mallet Theory, Eds. F. gassanl, F. Fumi, and M,Pr Tosi (North-Holland, Amslerdam, 1985) p, 59; O.K. Andersen, O, Jgpsgn and M, Sob, in: Electronic Band Structure and its Applications, Ed. I~, Yussuaff (Springer, Heidelberg, 1087). [15] J,F. Janak, V.L. Moru~:zl and A.R, Williams, Phys. Roy. B 12119751 1257. [16] The SBH of YSiz/St(Ill1 is smaller in tel, [23] than in this paper, because we did nol include an t-orbilal for Y in ref. [23], It seems mere accurale that a large atomic spllere for Y includes f-orbitals. [17] R.M. Tromp and M.C. Router, Phys. Roy. Lctt. 61 11988) 1756. [18] H, Fujdani and S. Asano, Appl. Surf. Sci, 41/42 (1999) 164. [196 N.V. Rues and C.C. Matthai, J, Phys. C (Solid State Phys.) 21 1198811.981. [20] L :~egenhi~gen and J.R, PateL Phys. Roy. B 41 119901 5315, [211 L L Batstone. LM. Phillips and E.C. Itunke, Phys. Roy. Leu. 60 (1988) 1394. [22] We call the T4 site model TR model in H. Fujilani and S. Asano, Phys. Roy. B 40 (19891835% I~1 H. Fujitani and S. Asano, Mater. Res. Soc. Proc. 193 ( 19901 77. 124l D. Cherns, C.J.D. Hctherington and CJ. tlumphreys, Phil. Mag. A 49 (19841 165. [25] R.T. Tung, J. Vac. Sci. T¢chnol. A 7 (1989) .598. [26] DR. Hamann, Phys, Roy. Leu, 60 (19881 313. [271 P.J. van deP Hook, W. Ravenek and EJ. Baerends, Phys. Roy. Lctt. 00 119881 1743; Surf. Sci. 205 119881 549. [28] When we artificially shifted all Ni d-orbitals to a 0.136 eV higher energy for type A NJSi2/St( I 11) inlerface, we gel a 0.04 ¢V higher SBH. [29] K. Hirose. K. Akimotu, L Hirosawa, L Mizuki. T. Mizutani and J. Matsui. Phys. Roy. B 43 119911 4538.