Surface Science 583 (2005) 80–87

www.elsevier.com/locate/susc

Band bending at the Si(100)–Si3N4 interfacestudied by photoreflectance spectroscopy

Kapil Dev, E.G. Seebauer *

Department of Chemical Engineering, University of Illinois, 600 S. Mathews, Urbana, IL 61801, United States

Received 21 December 2004; accepted for publication 16 March 2005

Available online 7 April 2005

Abstract

Photoreflectance spectroscopy has been used to measure the band bending at the p-Si(100)–Si3N4 interface subjected

to annealing and ion implantation. Upon annealing, unimplanted interfaces exhibit a constant band bending of about

0.77 eV, even though the spectral amplitude changes due to variations in the way minority carriers are annihilated at the

interface. Implantation reduces the band banding, although subsequent annealing in stages up to 900 �C progressively

restores the bending to its original value through pathways exhibiting a wide range of activation barriers.

� 2005 Elsevier B.V. All rights reserved.

Keywords: Silicon; Silicon nitride; Interface; Photoreflectance spectroscopy; Band bending; Defects

1. Introduction

Interfaces of silicon with dielectrics appear

widely in microelectronic devices and thereforehave attracted considerable study. Electrically ac-

tive defects at such interfaces—particularly Si–

SiO2—have garnered particular attention because

of the role such defects play in degrading device

behavior. Si–SiO2 interfaces can be fabricated with

very few defects, but processing steps such as ion

implantation or X-ray lithography can induce sub-

0039-6028/$ - see front matter � 2005 Elsevier B.V. All rights reserv

doi:10.1016/j.susc.2005.03.026

* Corresponding author. Tel.: +1 2173334402; fax: +1

2173335052.

E-mail address: eseebaue@uiuc.edu (E.G. Seebauer).

stantial defect formation. The formation and dis-

appearance of such defects [1–6] has been studied

in considerable detail, with most attention focusing

on the so-called ‘‘Pb center’’ [7,8]. Other kinds ofdefects can form, however, particularly in response

to ion implantation. In contrast with Pb centers,

whose concentrations evolve quickly at tempera-

tures around 600 �C, some defects require temper-

atures up to 1000 �C to evolve at similar rates.

This laboratory has examined the behavior of

such defects formed by sub-keV Ar+ bombardment

at Si(100)–SiO2 [9] and the Si(111)–SiO2 interfaces[10]. The bombardment led to a band bending

�0.5 eV at both the interfaces, and upon annealing,

both interfaces exhibited two kinetic regimes for

ed.

K. Dev, E.G. Seebauer / Surface Science 583 (2005) 80–87 81

the evolution of band bending due to defect heal-

ing. Such defect-induced band bending was shown

in subsequent simulations of pn junction formation

to persist throughout typical short-time annealing

steps after implantation and to have significanteffects on junction depth [11]. The atomistic mech-

anism responsible for defect healing remained

unclear, though unusually low activation energies

(less than 1 eV) for band bending evolution sug-

gested that a distribution of energies was involved.

The present work represents an attempt to better

understand the mechanism for defect healing by

examining an analogous interface of Si–Si3N4.There already exists some literature documenting

the behavior of Si–Si3N4. For example, interface

trap densities have been obtained by capacitance–

voltage measurements for nitride grown by photo-

chemical vapor deposition [12] and jet vapor

deposition [13,14]. A charge pumping technique

has been used to determine the trap densities at

the same interface where thermally grown nitridefilms were used as gate insulators of MISFET tran-

sistors [15]. As in earlier reports [9,10] for Si–SiO2,

the present work employs optical technique of

photoreflectance to examine the evolution of band

bending in response to low-energy ion implantation.

2. Experiment

Photoreflectance (PR) is one of a class of

modulation spectroscopies in which a semiconduc-

tor is periodically perturbed, and the resulting

change in dielectric constant is detected by reflec-

tance [16,17]. PR accomplishes the modulation

with a chopped laser beam having hm greater than

the fundamental bandgap energy Eg. Photogener-ated minority carriers migrate to the interface

and recombine with charge stored there. The

resulting change in built-in field affects the surface

reflectance R in narrow regions of wavelength

corresponding to optical transitions of the sub-

strate material. The small reflectance change DR/Rexhibits a spectral dependence that is monitored

with a weaker, independent probe beam usingphase sensitive detection. The presence of a non-

zero PR spectrum demonstrates unequivocally

the existence of surface band bending, and experi-

ments as a function of temperature and pump

intensity can yield useful estimates of the degree

of this band bending [18].

Experiments were performed in a turbomolecu-

larly pumped ultrahigh vacuum chamber set up inconjunction with optics for PR as described else-

where for a similar system [17]. Base pressures in

the low 10�9 Torr range were regularly achieved.

The chamber was equipped with a variable energy

ion gun (up to 2.0 keV) for ion implantation and

Auger electron spectroscopy (AES) for surface

characterization. Samples of dimensions 1.7 cm ·0.7 cm were cut from boron doped Si(100) waferswith resistivity of 0.014 X cm corresponding to a

doping level of 1 · 1018 cm�3. Resistive heating

of samples was employed, with temperature mon-

itored by a chromel–alumel thermocouple. A

He–Ne laser operating at 632 nm served as the

pump beam. Spectra were collected at 302 K (un-

less otherwise specified) in the vicinity of the nearly

degenerate E1 and E00 [19] optical transitions of Si,

which lie near 3.4 eV.

Clean Si(100) surfaces were obtained by re-

moving native oxide with aqueous HF, quickly

transferring the sample to the vacuum chamber,

baking to achieve ultrahigh vacuum, and heating

the surface at 850 �C for 5 min. AES revealed less

than 3% contamination with C and O after this

treatment. Nitrided surfaces were prepared byexposure to 3 · 10�6 Torr of ammonia for 10 min

at 800 �C. AES scans showed that this procedure

resulted in the formation of about 1.2 monolayers

of silicon nitride. After nitridation, samples were

implanted with 1.0 keV Ar ions. We used Ar to

induce interface bond breakage without intention-

ally affecting the doping level of the underlying Si,

which affects both the amplitude and the lineshapeof the PR spectra and therefore complicates data

interpretation. An ion fluence of 1 · 1015/cm2 was

used in all experiments reported here.

3. Results

3.1. Band bending after thermal nitridation

Fig. 1 shows room temperature raw PR spectra

at various illumination intensities for thermally

Pump Beam Intensity0.0 0.4 0.8 1.2

C (A

mpl

itude

)

0

2e-5

4e-5

6e-5

8e-5

302 K

310 K

317 K

Fig. 2. Variation of the photoreflectance amplitude factor C

82 K. Dev, E.G. Seebauer / Surface Science 583 (2005) 80–87

nitrided samples. The nonzero spectral amplitude

in the figure demonstrates the existence of band

bending within the Si at the interface. To quantify

the magnitude of the band bending, we used a

procedure detailed elsewhere [18]. Briefly, the PRspectrum was fitted to the classic third-derivative

functional form expected for electromodulation

spectroscopies of this type at optical transitions

far from the fundamental bandgap [20]

DR=R ¼ Re Cei/ðE � Ecrit þ iCÞ�n� �; ð1Þ

where C denotes an amplitude factor, / a phase

factor, C a broadening parameter, and Ecrit and

n the energy and dimension of the critical point

associated with the transition. As mentioned ear-lier, there are actually two optical transitions in

the region of the measurements: the E1 and E00.

These transitions of Si are nearly degenerate, how-

ever, being separated by less than 0.1 eV [19]. The

resulting spectra can be adequately fit by a single

lineshape having the form of Eq. (1) with the

parameter n chosen phenomenologically to be 3

[18]. The remaining parameters can then be ex-tracted from the experimental spectra according

to the methods outlined in Ref. [20]. It can then

be shown [18] that C obeys

C ¼ A1 ln½A2I expðV s=kT Þ þ 1�; ð2Þwhere k denotes Boltzmann�s constant, T the tem-perature, I the pump beam intensity, and Vs the

surface potential referenced to flat band. A1 and

Fig. 1. Photoreflectance spectra from the unimplanted Si(100)–

Si3N4 interface for different pump beam intensities. Spectra here

and in the other figures were collected at 302 K unless otherwise

specified.

A2 represent constants describing optical proper-

ties of the substrate. For the silicon–nitride inter-

face, this logarithmic dependence of C on inten-

sity is presented in Fig. 2 at different temperatures.

The solid lines in the figure represent least squarefits through the data based on Eq. (2), which yields

the composite parameter A2exp(Vs/kT). An Arrhe-

nius plot of A2exp(Vs/kT) at different temperatures

yields Vs, which for the silicon–nitride interface

appears in Fig. 3. This procedure yielded a value

of 0.48 eV for Vs just after nitridation.

The nitrided films were subsequently subjected

to isochronal annealing steps (5 min) in vacuumat temperatures ranging from 200 �C up to

900 �C. The resulting PR spectra taken at room

temperature after each annealing step are shown

in Fig. 4. Spectra were taken at room temperature

with illumination intensity for the spectra in Fig. 1. Solid lines

represent logarithmic fits according to Eq. (2).

1/kT (eV)36.0 36.5 37.0 37.5 38.0 38.5

Ln(A

2exp

(Vs/k

T))

-0.4

0.0

0.4

0.8

1.2

Vs = 0.48 eV

Fig. 3. Arrhenius plot of the quantity A2exp(Vs/kT) taken from

the data of Fig. 2. Slope of the plot gives Vs.

Pump Beam Intensity0.0 0.4 0.8 1.2

C (A

mpl

itude

)

0.0

4.0e-5

8.0e-5

1.2e-4

1.6e-4

297 K

305 K

313 K

Fig. 5. Variation of the photoreflectance amplitude factor C

with illumination intensity for the unimplanted Si(100)–Si3N4

interface annealed at 900 �C in vacuum for 5 min.

1/kT (eV)36.5 37.5 38.5 39.5

Ln(A

2exp

(Vs/k

T))

-2.4

-1.6

-0.8

0.0

Vs = 0.77 eV

Fig. 6. Arrhenius plot of quantity A2exp(Vs/kT) taken from

data of Fig. 5. Slope of the plot gives Vs.

Fig. 4. Photoreflectance spectra from the interface of Fig. 1

annealed for 5 min at various temperatures in vacuum.

K. Dev, E.G. Seebauer / Surface Science 583 (2005) 80–87 83

because the PR spectral amplitude decreases rap-

idly with increasing temperature due to thermal

free carrier generation; no signal can be measured

at all above roughly 200 �C. Fig. 4 shows that no

change in the PR spectral lineshape results from

annealing regardless of temperature. The ampli-

tude does change, however. Over the range 200–

500 �C, the amplitude increases progressively toroughly three times its original value. Intensity

studies using Eq. (2) at 200 �C and 400 �C yielded

values of roughly 0.8 eV for Vs. At higher temper-

atures (600–900 �C), the spectral amplitude pro-

gressively decreases again and ultimately settles

to a constant value around 1.7 times the original

value. Intensity studies using Eq. (2) yielded the

data of Figs. 5 and 6, with a value of 0.77 for Vs.

3.2. Band bending after implantation and annealing

The nitrided films were implanted with Ar+ as

described above and subsequently subjected to iso-

chronal annealing steps (5 min) in vacuum at tem-

peratures ranging from 200 �C up to 900 �C. Theresulting PR spectra taken at room temperatureafter each annealing step are shown in Fig. 7.

Immediately after implantation (i.e., before

annealing), the specimens exhibited no PR spec-

trum whatsoever (zero amplitude), implying that

no band bending existed at the interface. Subse-

quent annealing induced three different regimes

of spectral behavior. At 200 �C, the PR signal

reappeared, although the lineshape differed fromthat before ion bombardment. Further heating

up to 400 �C led to a progressive increase in ampli-

tude with no change in lineshape. Fig. 7a shows

the behavior in this first regime. From 500 to

600 �C, the PR spectra decreased again in magni-

tude while retaining the same lineshape. Fig. 7b

shows the behavior in this second regime. Heating

in the range 700–900 �C led to an increase in thespectral magnitude as well as a gradual change in

the lineshape. Ultimately the original lineshape

before bombardment was recovered. Fig. 7c shows

the behavior in this third regime.

Fig. 8 shows example Arrhenius plots deriving

from intensity studies at 400 �C and 900 �C. Fig.9 shows how Vs measured in such studies varied

with annealing temperature, increasing progres-sively from 0.50 eV at 300 �C to 0.79 eV at 900 �C.

Fig. 7. Series of raw PR spectra from the Ar+-implanted Si(100)–Si3N4 interface annealed for 5 min in vacuum in the temperature

range of (a) 200–400 �C, (b) 400–600 �C and (c) 600–900 �C.

84 K. Dev, E.G. Seebauer / Surface Science 583 (2005) 80–87

4. Discussion

For unimplanted material, the p-Si(100)–Si3N4

interface exhibits significant band bending regard-

less of the annealing temperature, indicating a sub-

1/kT(eV)36 37 38 39 40

Ln(A

2exp

(Vs/k

T))

-1

0

1

2

Vs = 0.57 eV

Vs = 0.79 eV

400oC

900oC

Fig. 8. Example Arrhenius plots of the quantity A2exp(Vs/kT)

for Ar+-implanted Si(100)–Si3N4 annealed at two different

temperatures in vacuum for 5 min. Slopes of the plots give Vs.

stantial number of electrically active defects at the

interface. This finding accords with the literature

for this interface [13,21,22], and contrasts withthe Si–SiO2 interface that can be prepared free

of such defects. The band bending of 0.77 eV

Temperature (oC)

300 500 700 900

V s (e

V)

0.5

0.6

0.7

0.8

Fig. 9. Variation of band bending Vs with annealing temper-

ature for the Ar+-implanted Si(100)–Si3N4 interface. Line is

guide to the eye.

K. Dev, E.G. Seebauer / Surface Science 583 (2005) 80–87 85

obtained after high-temperature annealing agrees

well with Si 2p core-level measurements made

by Stober et al. [23], who measured 0.75 ± 0.10 eV

for p-Si(100)–Si3N4. Note that photoreflectance

measures only the magnitude of the band bending,and not its direction. However, since the Fermi en-

ergy in the bulk Si lies near the valence band, band

bending of this magnitude can occur only when the

Fermi energy at the interface lies in the upper half

of the bandgap. Thus, the Si at the interface is

effectively n-type.

During the annealing steps between 200 and

900 �C, progressive changes occur over nearly theentire temperature range. Fig. 4 shows continual

changes in the PR spectra, and intensity studies

also show an increase in band bending from

0.48 eV immediately after nitridation to 0.77–

0.8 eV after annealing to any temperature between

200 and 900 �C. Note that the band bending in-

creases by less than a factor of two, however,

whereas the spectral magnitude first increases bya factor of three and then decreases. This behavior

contrasts markedly with the behavior observed for

Si–SiO2, where the spectral amplitude scales line-

arly in Vs.

To explain the variations in amplitude not

caused by Vs, we must examine the equations gov-

erning the PR signal. The PR amplitude factor C

scales linearly in DVs [24,25]. In turn, DVs obeysthe following relation [26,27]:

DV s ¼gkTq

lnJpc

qJ 0

þ 1

� �. ð3Þ

In this expression, J0 denotes the dark current

density to the surface, and Jpc the corresponding

photocurrent density. The dark current J0 origi-

nates from thermal carrier generation and contains

Vs according to [27,28]

J 0 ¼ A��T 2 expð�V s=kT Þ; ð4Þwhere A** denotes the modified Richardson con-

stant of 3.2 · 105 A/m2 K2 for p-type Si(111) and

11.2 · 105 A/m2 K2 for n-type1 [29]. For constant

1 A more complete treatment multiplies the right side of Eq.

(4) by a term related to surface recombination: (1 + BT3/2)�1.

This term becomes significant only above about 600 K for Si,

and is therefore neglected here.

Vs, there is no surface-dependent parameter within

J0. The photocurrent Jpc is described by [25,27,31]

Jpc ¼qIcð1� RÞ

hm1� e�aW þ aLd

1þ aLd

e�aW

� �; ð5Þ

where R represents the reflectivity (0.4 for Si at

632.8 nm), c the quantum efficiency (0.6 for Si

[30]), h Planck�s constant, m the frequency of thelight, a the absorption coefficient, W the depletion

width, Ld the diffusion length of the minority car-

riers. The primary factor that might be affected by

implantation and annealing is W, which varies as

the square root of Vs. Since Vs does not change sig-

nificantly upon annealing between 200 and 900 �C,however, there can be little corresponding varia-

tion in W. The remaining parameters in Eq. (3)are the electronic charge q, the quantum mechani-

cal ideality factor g, and the area factor q. Theparameter g usually lies near unity [27,31] and does

not depend on the surface. By contrast, q can vary

with the condition of the surface due to the differ-

ing areas where dark current and photocurrent is

nominally discharged. Normally dark current dis-

charges on surface states, which may be a smallfraction of the surface atomic density. However,

photocurrent can sometimes discharge in the entire

illuminated area [27,31]. This analysis implies that

the spectral amplitude changes that occur between

200 and 900 �C originate from variations in the

surface states controlling the area factor q, but

not those that control the degree of band bending

Vs.The Si(100)–Si3N4 interface after ion bombard-

ment showed no band bending, in contrast with

both the Si(100)–SiO2 and Si(111)–SiO2. The

behavior of the nitride interface is surprising be-

cause the unimplanted interface already has elec-

trically active defects that induce band bending,

and ion bombardment should increase that num-

ber. We ascribe the lack of band bending to theproduction of donor defects within the near-

surface Si bulk. If present in sufficient quantity,

such defects could locally convert the p-doped Si

to n-type due to generalized damage or knocked-

in nitrogen. Generalized bombardment damage

is well known to produce donor defects within sil-

icon over a wide range of conditions [32]. Further-

more, about 5% of knocked-in nitrogen resides in

86 K. Dev, E.G. Seebauer / Surface Science 583 (2005) 80–87

substitutional sites, where it acts as a donor [33].

Thus, the near-surface Si is probably n-type. The

Si at the interface is already n-type before bom-

bardment, and probably remains so afterward. It

is therefore possible that after bombardment, theFermi levels at the interface and within the near-

interface bulk coincide, resulting in little or no

electric field at the interface. This lack of field

would reduce the photoreflectance signal to zero.

The annealing behavior of the Si(100)–Si3N4

interface after ion bombardment also contrasts

with the silicon–oxide interface. The photoreflec-

tance lineshape at both the Si(100)–SiO2 andSi(111)–SiO2 interfaces remains invariant upon

annealing [9,10], and changes in amplitude result

solely from changes in band bending. In the pres-

ent case, however, the lineshape varies, and

changes in amplitude result from up to four differ-

ent contributions. First, the total band bending

changes, as evidenced by intensity studies. Second,

the area factor q almost certainly varies, possiblyin a manner akin to that of unimplanted nitride.

Third, the number and spatial distribution of

donor defects from generalized implantation dam-

age changes as the defects heal and diffuse. Since

such damage includes a wide variety of defect

types, such transformations would occur over a

broad temperature range. Thus, the near-interface

electric field also evolves over a broad temperaturerange as observed in the spectra. Fourth, the num-

ber and spatial distribution of donor defects from

knocked-in donor nitrogen changes through diffu-

sion. Implanted nitrogen diffuses to the surface

readily upon heating [33–35] with an apparent dif-

fusion coefficient of 10�2exp(�2.4 eV/kT) cm2/s

[35]. This expression predicts a diffusion length of

roughly 5 nm at 600 �C for a 5-min annealing step.Thus, nitrogen diffusion and the corresponding

change in near-interface electric field should influ-

ence the photoreflectance spectra above this

temperature.

As mentioned earlier photoreflectance lineshape

at both the Si(100)–SiO2 and Si(111)–SiO2 inter-

faces remains invariant upon annealing [9,10]. To

what extent did implantation-induced defects andknocked-in oxygen affect those results? Knocked-

in oxygen initially forms complexes with vacancies,

and subsequent annealing causes precipitation of

amorphous SiO2 [36]. It is unknown whether the

vacancy–oxygen complexes are electrically active,

or whether the formation of buried SiO2 creates

electrically active defects. If few electrically active

complexes or defects form, the effects of knocked-in oxygen on the lineshape would be small. The

lineshape invariance, together with the strong

dependence of band bending kinetics on crystallo-

graphic orientation, suggest that bulk effects on

lineshape due to both generalized damage and

oxygen knock-in are indeed small. By implication,

the complications seen for the nitride interface

probably originate primarily from knocked-innitrogen rather than generalized damage, since

similar kinds of generalized damage probably

occur for both the nitride and oxide interfaces.

Despite these various complications, for the ni-

tride surface it is clear that variations in Vs upon

annealing occur over a broad range of tempera-

tures. This behavior contrasts with that of unim-

planted material, for which Vs remains essentiallyconstant above 200 �C. Implanted Si(100)–SiO2

and Si(111)–SiO2 interfaces do exhibit changes

in Vs over a wide temperature range [9,10], but

two clearly identifiable regimes for annealing exist:

below 500 �C and above 650 �C (with the exact

temperatures depending upon crystallographic ori-

entation). The low activation energies for the evolu-

tion of these changes (below �1 eV) were taken asevidence of distributions of energy barriers for each

of the two regimes. In the present nitride case, the

energy distribution appears to be much broader,

encompassing transformations all the way from

room temperature to 900 �C.

5. Conclusion

This work has sought to better understand the

kinetics and mechanisms of implantation-induced

defect healing at semiconductor–insulator inter-

faces. Photoreflectance studies of band bending

at the Si(100)–Si3N4 interface have been com-

pared to similar work for Si–SiO2 interfaces. For

both nitride and oxide interfaces, band bendingapproaches that for unimplanted material over a

broad range of temperatures extending up to

about 900 �C. However, for nitride the transfor-

K. Dev, E.G. Seebauer / Surface Science 583 (2005) 80–87 87

mation is progressive and continuous over the en-

tire range, while for oxide the transformation takes

place within two distinct temperature regimes.

Changes at the nitride interface are complicated

in part by the evolution of implant-induced defectsin the near-surface bulk, most likely due to

knocked-in nitrogen.

Acknowledgment

This work was partially supported by NSF

(CTS 02-03237).

References

[1] M.A. Jupina, P.M. Lenahan, IEEE Trans. Nucl. Sci. 36

(1989) 1800.

[2] A. Stesmans, Appl. Phys. Lett. 68 (1996) 2076.

[3] A. Stesmans, Physica B 273–274 (1999) 1015.

[4] G.V. Gadiyak, Thin Solid Films 350 (1999) 147.

[5] D.L. Griscom, J. Appl. Phys. 58 (1985) 2524.

[6] S. Kaschieva, P. Danesh, Nucl. Instrum. Methods Phys.

Res., Sect. B 129 (1997) 551.

[7] K.L. Brower, Appl. Phys. Lett. 43 (1983) 1111.

[8] A. Stesmans, Phys. Rev. B 48 (1993) 2418, and references

cited therein.

[9] K. Dev, M.Y.L. Jung, R. Gunawan, R.D. Braatz, E.G.

Seebauer, Phys. Rev. B 68 (2003) 195311.

[10] K. Dev, E.G. Seebauer, Surf. Sci. 550 (2004) 185.

[11] M.Y.L. Jung, R. Gunawan, R.D. Braatz, E.G. Seebauer,

J. Appl. Phys. 95 (2004) 1134.

[12] H. Matsuura, M. Yoshimoto, H. Matsunami, Jpn. J.

Appl. Phys. 35 (1996) 2614.

[13] A. Mallik, X.W. Wang, T.P. Ma, G.J. Cui, T. Tamagawa,

B.L. Haplern, J.J. Schmidt, J. Appl. Phys. 79 (1996) 8507.

[14] T.P. Ma, IEEE Trans. Electron. Dev. 45 (1998) 680.

[15] R.T. Fayfield, J. Chen, M.S. Hagedorn, T.K. Higman,

A.M. Moy, K.Y. Cheng, J. Vac. Sci. Technol. B 13 (1995)

786.

[16] M. Cardona, K.L. Shaklee, F.H. Pollak, Phys. Rev. 154

(1967) 696.

[17] C.R. Carlson, W.F. Buechter, F. Che-Ibrahim, E.G.

Seebauer, J. Chem. Phys. 99 (1993) 7190.

[18] R. Ditchfield, D. Llera-Rodriguez, E.G. Seebauer, Phys.

Rev. B 61 (2000) 13710.

[19] P. Lautenschlager, M. Garriga, L. Vina, M. Cardona,

Phys. Rev. B 36 (1987) 4821.

[20] D.E. Aspnes, Surf. Sci. 37 (1973) 418.

[21] S.V. Hattangady, G.G. Fountain, R.A. Rudder, R.J.

Markunas, J. Vac. Sci. Technol. A 7 (1989) 570.

[22] See, for example, Ultrathin SiO2 and High-K Materials for

ULSI Gate Dielectrics, vol. 567, MRS 1999.

[23] J. Stober, B. Eisenhut, G. Rangelov, Th. Fauster, Surf.

Sci. 321 (1994) 111.

[24] H. Shen, S.H. Pan, Z. Hang, J. Leng, F.H. Pollak, J.M.

Woodall, R.N. Sacks, J. Appl. Phys. 53 (1988) 1080.

[25] T. Kanata, M. Matsunage, H. Takakura, Y. Hamakawa,

T. Nishino, Jpn. J. Appl. Phys. 68 (1990) 5309.

[26] X. Yin, H.M. Chen, F.H. Pollak, Y. Chan, P.A. Montano,

P.D. Kirchner, G.D. Pettit, J.M. Woodall, J. Vac. Sci.

Technol. A10 (1992) 121.

[27] H. Shen, M. Dutta, J. Appl. Phys. 78 (1995) 2151.

[28] E.H. Rhoderick, Metal–semiconductor Contacts, Claren-

don, Oxford, 1978.

[29] J.M. Andrews, M.P. Lepselter, Solid State Electron. 13

(1970) 1011.

[30] K. Kondo, A. Moritani, Phys. Rev. B 14 (1976) 1577.

[31] H. Shen, W. Zhou, J. Pamulapati, F. Ren, Appl. Phys.

Lett. 74 (1999) 1430.

[32] K. Geiwont, S. Ashok, Thin Solid Films 142 (1986) 13.

[33] P.A. Schultz, J.S. Nelson, Appl. Phys. Lett. 78 (2001) 736.

[34] L.S. Adam, M.E. Law, K.S. Jones, O. Dokumaci, C.S.

Murthy, S. Hegde, J. Appl. Phys. 87 (2000) 2282.

[35] L.S. Adam, M.E. Law, O. Dokumaci, C.S. Murthy, S.

Hegde, J. Appl. Phys. 91 (2002) 1894.

[36] T. Ahilea, E. Zolotoyabko, J. Cryst. Growth 198/199

(1999) 414.