Characterization of dielectric properties ofultrathin SiO2 film formed on Si substrate

K. Hirosea,*, H. Kitaharab, T. Hattorib

aInstitute of Space and Astronautical Science, 3-1-1 Yoshinodai, Sagamihara, Kanagawa 229-8510, JapanbMusashi Institute of Technology, 1-28-1 Tamazutumi, Setagaya, Tokyo 158-8557, Japan

Abstract

The dielectric constant of ultrathin (0.55–7.96 nm) SiO2 films formed on Si(0 0 1) substrates was characterized in terms of the

modified Auger parameter, a0. The a0 values for Si atoms were found to shift by about 0.7 eV for ultrathin SiO2 films compared

with thick SiO2 films. This shift is apparently caused only by a change in the electrostatic screening energy originating from the

dielectric discontinuity between the bulk dielectric constants of SiO2 and Si at the SiO2/Si interface. This indicates that the bulk

dielectric constant also holds for ultrathin SiO2 films.

# 2003 Elsevier Science B.V. All rights reserved.

Keywords: SiO2; Si; Dielectric constant; XPS; AES; AES parameter

1. Introduction

The atomic and/or electronic structures at SiO2/Si

interfaces, such as intermediate oxidation states and

valence-band offsets, have attracted much interest

from both the scientific and technological points of

view [1–3]. However, little attention has been paid to

those in the ultrathin SiO2 film formed on the Si

substrate. We previously revealed that ultrathin

(�1 nm thick) SiO2 film is strained compressively

due to the lattice mismatch between the SiO2 and

the Si substrate [4].

Another aspect attracting interest is the dielectric

constant of ultrathin SiO2 film. For such film, the use

of the bulk dielectric constant is open for question. We

have evaluated the dielectric constant of strained

ultrathin SiO2 film in terms of the modified Auger

parameter of Si atoms, a0, which is defined as the sum

of the binding energy of the photoelectrons and the

kinetic energy of the Auger electrons for the core

levels of Si atoms [5]. A shift in the value of a0 is twice

that in the relaxation energy associated with the core-

hole for Si atoms, so it reflects the change in the

dielectric constant of the SiO2 [6–9].

2. Experiment

Device-quality SiO2 films with thicknesses ranging

from 0.55 to 7.96 nm were formed on either 6 in. n-

type Si(1 0 0) wafers (resistivity ¼ 10–20 O cm) or

6 in. p-type Si(1 0 0) wafers (resistivity ¼ 10–

20 O cm) by oxidizing the substrate at 800 8C under

dry oxygen conditions at 1 atm. The thicknesses of the

oxide films were determined by high-resolution XPS

(VG ESCALAB220i-XL). We used the metrology

Applied Surface Science 216 (2003) 351–355

* Corresponding author. Tel.: þ81-427-59-8326;

fax: þ81-427-59-8463.

E-mail address: hirose@pub.isas.ac.jp (K. Hirose).

0169-4332/03/$ – see front matter # 2003 Elsevier Science B.V. All rights reserved.

doi:10.1016/S0169-4332(03)00426-4

proposed by Lu et al. [10]. The crucial conditions are

the largest available acceptance angle of the analyzer

to reduce possible elastic scattering effects of photo-

electron and high take-off angles to eliminate photo-

electron refraction effects.

The samples were examined using XPS. The Si 2p

photoelectron spectrum and Si KLL Auger spectrum

were measured at a photoelectron take-off angle of 308by using the bremsstrahlung (braking radiation) from

an Al Ka X-ray source [11]. The a0 values in the films

were determined by summing up the binding energy of

the Si 2p peak and the kinetic energy of the Si KLL

peak.

Although the modified Auger parameter is indepen-

dent of the charge-up effect and the referencing method,

we must pay special attention to another factor, the

carrier trapping phenomena, which affects the energy

levels of SiO2/Si samples [12–14]. We thus measured

the core level peak and the Auger line for Si atoms as

quickly as possible (in less than 6 min) to minimize

errors due to the carrier trapping phenomena. We

obtained a0 from the average of three measurements.

The reproducibility is shown by the error bars in Fig. 3.

The modified Auger parameters for the SiO2 films

were calculated using a first-principles molecular

orbital method—the Hartree–Fock–Slater method

using the discrete variational (DV)-Xa code [15].

The molecular orbitals were constructed by linearly

combining the atomic orbitals, which were generated

numerically. The basis functions of Si, O, and H were

1s–3d, 1s–3d, and 1s–2p, respectively. The Auger and

photoemission energy levels of the peaks were calcu-

lated using the Slater transition state procedure, which

is suitable for studying the binding energy in XPS

spectra and the kinetic energy of Auger peaks. We

used Si5O16H12 clusters, illustrated in Fig. 1, to repre-

sent the SiO2 structures. The 12 hydrogen atoms were

arranged so as to terminate the dangling bonds of the

surrounding 12 silicon atoms and to make the cluster

representative of bulk SiO2 without the surface effects.

We used constant interatomic distances of 0.162 nm

for the Si and O and 0.092 nm for the O and H, which

are consistent with those used by Edwards [16]. We

used a bond angle of 144–1358 for the Si–O–Si bond

angle to investigate the effect of strain on the modified

Auger parameter. The validity of the parameter values

used in this calculation was previously confirmed by

comparing the calculated energy levels for the SiO2

cluster with the experimental XPS spectrum for bulk

SiO2 and ultrathin SiO2 film [4,17].

3. Results and discussion

Fig. 2 shows the wide-region spectrum of the Si core

levels for 0.86 nm-thick SiO2 film formed on a Si

Fig. 1. Si5O16H12 cluster with Cs point symmetry.

352 K. Hirose et al. / Applied Surface Science 216 (2003) 351–355

substrate, measured using the bremsstrahlung from an

Al Ka X-ray source. It shows the Si 2p photoelectron

peaks and the Si KLL Auger peaks from the SiO2 film

(Siox 2p and Siox KLL) and the Si substrate underneath

(Siel 2p and Siel KLL). The a0 values were determined

precisely from a much narrower region spectrum

including Siox 2p and Siox KLL.

Table 1 shows the a0 for 0.55–7.96 nm-thick SiO2

film. The value for 7.96 nm-thick SiO2 film was

1711.48 eV, which agrees well with the value reported

for bulk SiO2, 1711.5 eV [18]. This excellent agree-

ment is due to the fact that the modified Auger

parameter is unaffected by changes in the surface

Fermi level due to calibration of the spectrometer

and/or to the charge-up effect. The modified Auger

parameter shifts, Da0, were determined with respect to

that for 7.96 nm-thick SiO2 film or to that of bulk

SiO2. The Da0 value was 0.24 eV for SiO2 film

1.60 nm thick and as high as 0.7 eV for SiO2 film

0.68 nm thick.

The dependence ofDa0 on film thickness is plotted in

Fig. 3. It increases monotonically with decreasing

thickness. This observed shift is in contrast with pre-

vious reports. Wagner et al. did not observe a shift in the

modified Auger parameter for oxide thicknesses ran-

ging from 2 to 7 nm, even though they expected one

near the SiO2/Si interface [19]. Iqbal et al. measured the

Auger parameter for Si and found that there is no

difference between thin SiO2 film (0.8–20 nm) formed

on a Si substrate and bulk SiO2 [20]. However, careful

analysis of their data reveals a shift of about 0.3 eV for

1.2–1.3 nm-thick film, but not for 0.8 nm-thick film.

This exception is probably be due to inhomogeneity of

the film thickness or errors in measurement. In contrast

to these studies, we used device-quality SiO2/Si sam-

ples and paid special attention to the carrier trapping

Fig. 2. Wide-region spectrum for Si core levels of 0.86 nm-thick SiO2 film formed on Si substrate measured using bremsstrahlung from an Al

Ka X-ray source.

Table 1

Modified Auger parameter, a0, in thermal SiO2 formed on Si(0 0 1)

substrate

Substrate

type

SiO2 film

thickness (nm)

Modified Auger parameter (eV)

a0 Da0

p 0.55 1712.02 0.64

p 0.68 1712.04 0.70

n 0.86 1711.94 0.51

p 0.95 1711.93 0.52

p 1.04 1711.99 0.55

p 1.60 1711.72 0.24

p 1.72 1711.81 0.34

p 2.13 1711.77 0.29

n 4.10 1711.55 0.07

n 6.26 1711.50 0.02

n 7.96 1711.48 0.00

Shifts in a0 were obtained with respect to those of 7.96 nm-thick

SiO2 film or of bulk SiO2 and are shown as Da0.

K. Hirose et al. / Applied Surface Science 216 (2003) 351–355 353

phenomena, which was ignored in previous studies, to

obtain precise Da0 values.

Now we consider the reason for the observed shift in

Da0 in conjunction with the relaxation energy depen-

dence on the SiO2 film thickness. The shift in the

modified Auger parameter, Da0, is twice the shift in the

relaxation energy, DR, associated with the core-hole

accompanying the photoionization of the SiO2 film

[7]:

Da0 ¼ 2DR: (1)

The relaxation energy, R, of the SiO2 film formed on

the Si substrate is determined by the polarization of the

neighboring SiO2 molecules, Epol, and by the electro-

static interactions on the phase borders, Ez [21]:

R ¼ Epol þ Ez: (2)

We can then estimate Da0 using

Da0 ¼ 2ðDEpol þ DEzÞ; (3)

where DEpol and DEz are the shifts in Epol and Ez with

respect to bulk SiO2. Strong spatial variation in the

relaxation energy caused by DEz is limited to the thin

interface layer, which has a thickness on the order of

the characteristic screening length, �0.1 nm, for the

SiO2/Si system [22]. Beyond the interface layer, DEz

is expressed as the electrostatic screening energy or as

the classical image potential of the core-hole [23,24].

First, we estimated DEz, which depends on the

distance from the SiO2/Si interface or on the SiO2

film thickness. If Da0 is twice DEz, DEpol should be

zero, meaning that the dielectric constant of ultrathin

SiO2 film is identical to that of bulk SiO2 since the

change in the dielectric constant can be calculated,

using the Born equation [21], from the DEpol. We

estimated DEz values by using the theory proposed

by Pasquarello et al. [24]. They obtained the shifts in

the relaxation energy of SiO2 film formed on a Si

substrate by calculating the change in electrostatic

screening energy caused by dielectric discontinuity at

the vacuum/SiO2/Si interface. They calculated the

average shift for various SiO2 film thicknesses using

the bulk dielectric constants, e0 ¼ 1, eSiO2¼ 2:1, and

eSi ¼ 12. They used two factors: the Z dependence

resulting from the electrostatic model and a weighting

function characterized by the electron escape length,

lSiO2, 2.6 or 0.71 nm.

Next, we compared the doubled DEz given by

Pasquarello et al. with the Da0 in this study. The

doubled DEz values are shown in Fig. 3 by the solid

(lSiO2of 0.71 nm) and dotted (lSiO2

of 2.6 nm) lines.

The escape depth under these analysis conditions is

estimated to be 1.3 nm, which is between the two

theoretical curves. Our results agree well with both

curves. The good agreement between the observedDa0

values and the doubled DEz values indicates that DEpol

is zero and that the bulk dielectric constant is applic-

able to ultrathin SiO2 film.

Finally, to reconcile this finding with our previous

finding that ultrathin (�1 nm) SiO2 film is compres-

sively strained, we investigated the effect of strain on

Da0 for SiO2 film. We calculated the a0 values for an

SiO2 cluster as a function of the Si–O–Si bond angle and

obtained the Da0 values from the shifts with respect to

Da0 for an Si–O–Si bond angle of 1448, which is the

most probable average value for bulk SiO2 [25]. The

calculated Da0 value was as small as 50 meV for an Si–

O–Si bond angle of 1358, which we assumed to be the

angle for strained ultrathin SiO2 film on Si substrates.

This finding agrees well with our experimental finding

that DEpol should be zero for ultrathin SiO2 films,

including a strained layer (0.55–1.04 nm thick).

4. Summary

In summary, we measured the modified Auger

parameters for SiO2/Si(0 0 1) interfaces by using

Fig. 3. Shift in Auger parameter, Da0, in SiO2 film with respect to

those in bulk SiO2. Solid and dotted curves are from theoretical

study by Pasquarello taking into account the electrostatic screening

effect caused by dielectric discontinuity at SiO2/Si interface.

354 K. Hirose et al. / Applied Surface Science 216 (2003) 351–355

X-ray photoelectron spectroscopy and Auger electron

spectroscopy. We found that they differed by about

0.7 eV between thick SiO2 films and ultrathin SiO2

films. The observed shift in the parameters can be

explained only by a change in the electrostatic screen-

ing energy originating from dielectric discontinuity

between the bulk dielectric constants of SiO2 and Si at

the SiO2/Si interface. The bulk dielectric constant is

thus applicable to ultrathin SiO2 films.

Acknowledgements

We are grateful to E. Hasegawa and Y. Miura of

NEC Corporation for supplying the device-grade sam-

ples. We also thank H. Adachi and M. Uda for

providing the DV-Xa.

References

[1] T. Hattori, Crit. Rev. Solid State Mater. Sci. 20 (1995) 339.

[2] K. Hirose, K. Sakano, H. Nohira, T. Hattori, Phys. Rev. B64

(2001) 155325.

[3] K. Takahashi, M.B. Seman, K. Hirose, T. Hattori, Jpn. J.

Appl. Phys. 41 (2002) L223.

[4] K. Hirose, H. Nohira, T. Koike, K. Sakano, T. Hattori, Phys.

Rev. B59 (1999) 5617.

[5] C.D. Wagner, L.H. Gale, R.H. Raymond, Anal. Chem. 51

(1979) 466.

[6] R.H. West, J.E. Castle, Surf. Interf. Anal. 4 (1982) 68.

[7] W.F. Egelhoff Jr., Surf. Sci. Rep. 6 (1987) 253.

[8] S. Kohiki, S. Ozaki, T. Hamada, K. Taniguchi, Appl. Surf.

Sci. 28 (1987) 103.

[9] P. Streubel, R. Franke, Th. Chasse, R. Fellenberg, R. Szargan,

J. Electr. Spectrosc. Relat. Phenom. 57 (1991) 1.

[10] Z.H. Lu, J.P. McCafarrey, B. Brar, G.D. Wilk, R.M. Wallace,

L.C. Feldman, S.P. Tay, Appl. Phys. Lett. 71 (1997) 2764.

[11] R.J.E. Castle, R.H. West, J. Electr. Spectrosc. Relat. Phenom.

18 (1980) 355.

[12] Y. Hagimoto, T. Fujita, K. Ono, H. Fujioka, M. Oshima, K.

Hirose, M. Tajima, Appl. Phys. Lett. 75 (1999) 2011.

[13] Y. Hagimoto, H. Fujioka, M. Oshima, K. Hirose, Appl. Phys.

Lett. 77 (2000) 4175.

[14] K. Hirose, K. Sakano, K. Takahashi, T. Hattori, Surf. Sci.

507–510 (2002) 906.

[15] I. Tanaka, J. Kawai, H. Adachi, Phys. Rev. B52 (1995) 11733.

[16] A.H. Edwards, Phys. Rev. Lett. 71 (1993) 3190.

[17] K. Hirose, H. Nohira, K. Sakano, T. Hattori, Appl. Surf. Sci.

166 (2000) 455.

[18] L. Lozzi, M. Passacantando, P. Picozzi, S. Santucci, J. Electr.

Spectrosc. Relat. Phenom. 72 (1995) 97.

[19] C.D. Wagner, A. Joshi, L. Gulbrandsen, B.E. Deal, J. Vac.

Sci. Technol. B2 (1984) 107.

[20] A. Iqbal, C.W. Bates, J.W. Allen, Appl. Phys. Lett. 47 (1985)

1064.

[21] V.I. Nefedov, J. Electr. Spectrosc. Relat. Phenom. 63 (1993)

355.

[22] F. Bechstedt, R. Enderlein, D. Reichardt, Phys. Stat. Sol. 118

(1983) 327.

[23] R. Browning, Phys. Rev. B38 (1988) 13407.

[24] A. Pasquarello, M.S. Hybersten, R. Car, Phys. Rev. B53

(1996) 10942.

[25] R.L. Mozzi, B.E. Warren, J. Appl. Crystallogr. 2 (1969) 164.

K. Hirose et al. / Applied Surface Science 216 (2003) 351–355 355