Growth of native oxide on a silicon surface

M. Morita, T. Ohmi, E. Hasegawa, M. Kawakami, and M. Ohwada

Citation: Journal of Applied Physics 68, 1272 (1990); doi: 10.1063/1.347181
View online: http://dx.doi.org/10.1063/1.347181
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/68/3?ver=pdfcov
Published by the AIP Publishing

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

Growth of native oxide on a sincon surface

M. Morita, T. Ohmi, E Hasegawa, M. Kawakami,a) and M. Ohwadab )
Department of Electronics, Faculty of Engineering, Tohoku University, Sendai 980, Japan

(Received 19 December 1989; accepted for publication 30 March 1990)

The control factors controlling the growth of native silicon oxide on silicon (SO surfaces have
been identified. The coexistence of oxygen and water or moisture is required for growth of
native oxide both in air and in ultrapure water at room temperature. Layer-by-layer growth of
native oxide films occurs on Si surfaces exposed to air. Growth of native oxides on n-Si in
ultrapure water is described by a parabolic law, while the native oxide film thickness on n + -Si
in ultrapure water saturates at 10 A. The native oxide growth on n-Si in ultrapure water is
continuously accompanied by a dissolution of Si into the water and degrades the atomic
flatness at the oxide-Si interface, prod.ucing a rough oxide surface. A dissolution of Si into the
water has not been observed for the Si wafer having surface covered by the native oxide grown
in air. Native oxides grown in air and in ultrapure de-ionized water have been demonstrated
experimentally to exhibit remarkable differences such as contact angles of ultrapure water
drops and chemical binding energy. These chemical bond structures for native oxide films
grown in air and in ultrapure water are also discussed.

I. INTRODUCTION
The control of native oxide growth rate on Si surfaces is
of great importance in the fabrication of ultralarge-scale integrated devices. Native oxide films on 8i surfaces prevent
the low-temperature growth of high-quality epitaxial Si
films and precise control of the thickness and electrical properties of very-thin-gate oxide films, plus give an increase of
the contact resistance for via-holes of a small area. Consequently, the growth of native oxide films has received an
increasing attention with a decrease of pattern dimension of
integrated circuits. 1-7
The growth of several angstrom (A) thi.ckness native
oxide has been reported to be observed on Si surfaces exposed to air for 10 h. 3 On the contrary, the oxide thickness
formed at room temperature during 10 h is estimated much
below 1 A from the growth rate activation energy determined from temperature dependence data of the thermal
oxidation rate in dry or wet oxygen, i.e., the thermal oxidation data conjecture that the oxide is not grown on Si surface
at room temperature. Therefore, the growth mechanism of
the native oxide is considered entirely different from that of
the thermal oxide. This consideration promotes researches
concerning the origin of the native oxide growth.
Some interesting results for the influence of native oxide
on semiconductor processing have recently been reported, 8. 9
It has been demonstrated that the use of different preoxidation cleaning procedures induding wet chemical cleaning of
silicon surface had a significant effect on the kinetics ofthermal silicon oxidation. 8 This result can be interpreted to demonstrate the influence of native oxides on the thermal oxidation kinetics. It has been pointed out, moreover, that iron

.) Present address: Tokyo Central Research Laboratory, Mitsubishi Gas

Chemical Company Inc., 1-1-6 Niijuku, Katsushika, Tokyo 125, Japan.
b) Present address: Alps Electric Co., Ltd., Furukawa, Miyagi 989-61- Japan.

1272

J. Appl. Phys. 68 (3), 1 August 1990

ions in HNO J solution react with the native silicon oxide to

form an Fe(III)-O complex, but do not react with silicon
itself.9 A drastic degradation of metal-ox ide-semiconductor
device characteristi.cs has been observed by an increase of
iron impurities in native oxides. This indicates that native
oxides serve as the source of iron impurities to diffuse into
the silicon and produce defects in succeeding high-temperature processes. 10- 13 These characteristics have been demonstrated for various metal impurities such as iron, nickel, and
chromium. 10- 12 Thus the existence of native oxide film on
silicon surfaces degrades the control ability of the quality of
device fabrication processings and the performance and reliability of semiconductor devices themselves. Though the existence of native oxides on the silicon surfaces has been demonstrated to give severe influences on the high-quality
processings, however, a method to control the native oxide
growth has not yet been established.
In this paper, we describe the factors controlling the
native oxide growth in air and in ultrapure water at room
temperature and propose methods to control the native oxide growth. I 4-18 We also discuss the structural differences
between native oxides formed in various environments.
II. EXPERIMENT
The wafers used in these experiments are phosphorous
(P) doped n-type Cz (100) wafers with a resistivity of 2-4
n cm (carrier concentration: lO l5 em 3), n -+ -type (1020
em - 3) wafers, and p -+ -type (1020 cm - 3) wafers, where n + and p + -type layers are prepared by chemical vapor deposition of phosphosilicate glass (PSG) and borosilicate glass
(BSG) followed by the thermal diffusion treatments, respectively. AU wafers are chemically cleaned with
(H 2 S04 + H 2 0 2 ),
(NH 4 0H + H 2 0 2 + H 2 0),
and
(HCI + HzO z + H 2 0) solution, etched with (HF + H 2 0)
solution, and then rinsed with ultrapure water with a dissolved oxygen concentration of 0.6 ppm. Most of the wafers
are sllcceedingly dipped in hot ultrapure water (90 ec) and

0021-8979/90/151272-10$03.00

Co) i 990 American Institute of Physics

1272

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

dried with hot ultraclean argon gas blow (70-80 PC). Other
remaining wafers are dried with ultraclean nitrogen gas blow
(23C).
Native oxide film thicknesses are determined by x-ray
photoelectron spectroscopy (XPS) combined with ellipsometry. Native oxide film thickness has been mainly determined by ellipsometry so far, where the thickness has been
derived based on an assumption of a constant oxide refractive index such as 1.46 because an accuracy of the measurement of the refractive index degrades for thin films having a
thickness less than 70 A. But it has been recently reported
that the refractive index for thermal oxides near the oxide-Si
interface is different from that of bulk oxide, i.e., the refractive index of an interlayer having a thickness of 6 Abetween
Si and its thermal oxide is evaluated as 2.8. 19 ,20 It is, therefore, difficult to obtain an accurate film thickness of native
oxides only by using ellipsometry. The native oxide thickness measured by ellipsometry has been demonstrated experimentally thicker than that by XPS. 3
We propose a method to determine the thickness of very
thin oxide films as follows: (1) The thickness of thermal
oxide films (70-140 A) is measured by eHipsometry without
the postulation of the refractive index (see Appendix). (2)
The Si2p spectra of the same oxides are measured by XPS,
( 3) The area ratio of the signal from the oxide (SiO x:
x;;;,0.5) to that from the Si substrate is calibrated with the
thickness measured by ellipsometry. The escape depths of
electrons for the oxide overlayer and the Si substrate are
determined. (4) The thicknesses of the native oxide films are
determined from the area ratio of XPS spectrum of the native oxide and above calibration data. The minimum thickness which can be measured by ellipsometry without the assumption of the refractive index is 70 A. The maximum
thickness from which both the XPS signal from the oxide
overlayer and the Si substrate can be observed for an XPS
system with Al Ka x-ray source is 140 A. 80 70-140-.Athickness oxides are used for the thickness calibration
between ellipsometry and XPS, The native oxide film thickness determined by above~described method is considered to
be correct as long as the atomic density of the native oxide is
equal to that of the thermal oxide, because the peak area of

FIG. 1. Oxide thickness as a function of exposure time of wafers to air at

room temperature.
1273

J. Appl. Phys" Vol. 68, No.3, 1 August 1990

XPS spectrum corresponds to the number of atoms. This

method gives a normalized thickness for native oxides having atomic densities different from that of the thennal oxide,
where the native oxide thickness is normalized by the thermal oxide thickness. The H 2 0 concentration in processing
gases such as oxygen and nitrogen is monitored with a dewpoint meter having a lowest dew point of - 110 "C. The
dissolved oxygen level in ultrapure water is measured by
polarography having a lowest level of 5 ppb. The amount of
the dissolved Si in ultrapure water is measured by atomic
absorption. The roughnesses oftne native oxide surfaces and
the 8i wafer surfaces after removing the native oxide overlayer by (HF + H 2 0) etching are evaluated by light beam
reflection and the scanning tunneling microscopy (STM),
respectively.

iii. RESULTS
A. Growth of native oxide in air

The thicknesses of native oxide films on no, n -'- -, and

p + -type Si wafers are shown as a function of exposure time
to clean :room air at room temperature in Fig. 1, where the
clean room air is characterized by the cleanliness less than
0.000 002 (particle size greater than 0.17 f.lm), an average
temperature of 23.7 'C, and an average humidity of 42%
(where the H 2 0 molecule concentration is 1.2% of total air
molecule concentration)" It is obvious in Fig. 1 that these
plots indicate an approximately successive step functionlike
increase of native oxide. This can be considered to suggest
that layer-by-layer oxide growth occurred on these wafers at
room temperature. 5 For n-Si, it appears in Fig. 1 that the
growth to 5.4 A and the succeeding increa1ie to 7.6 A correspond to two molecular layers of growth followed by one
additional layer of growth of the native oxide. The initial
normalized thickness of native oxide on the n I- -Si wafer is
4.4 A, while the thickness on n- andp + -Si wafer are 1.9 and
2.3 A, respectively. These oxides are mainly grown during
rinsing with ultrapure water and water drying procedure
after a diluted HF treatment. Here, it should be attended
that the Si atoms contributing the initial oxide thickness determined in this measurement includes Si atoms terminated
with hydrogen at 8i surface, Therefore, the actual oxide
thicknesses on each Si surfaces are estimated to be thinner
than above thickness values at the initial stage.
The effect of H 2 0 concentration in air on native oxide
growth is summarized in Table 1. The oxide thickness is 6.7
A after 7 days exposure to the clean room air containing
1.2% HIO. When the moisture level is suppressed less than
0.1 ppm in (N 2 + O 2 ) environment having a concentration
ratio of N2 to O 2 as 4 to 1 and N z environment, the thicknesses of native oxides are 1.7 and 1.9 A, respectively. It is
worthwhile to note in Table I that the native oxide hardly
grows at all even after 7 days exposure to the air when the
Hz 0 concentration in air is suppressed less than 0.1 ppm.
The bare Si surface is significantly oxidized even at room
temperature in the environment of coexistence of oxygen
and moisture.
Morita et al.

1273

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

TABLE 1. Growth of native oxide for H 2 0 concentrations during 7 days.

Ambient

( =

Oxide thickness (A)

~1.2%

6.7

(Humidity: 42%RHJ
<0.1 ppm

1.7

<O.! ppm

1.9

Air

02 /N 2

H 2 0 concentratioll

l)

N2

B. Growth of native oxide In ultrapure water

Figure 2 indicates the oxide thickness on n-Si wafer as a

function of the immersion time in ultrapure water having
different dissolved oxygen concentrations such as 0.04,0.6,
and 9 ppm at room temperature. The dissolved oxygen concentration in ultrapure water is 0.04 ppm (40 ppb), while it
is increased up to 0.6 ppm in overflow rinsing situation, and
up to 9 ppm in equilibrium with air at room temperature. It
is seen in Fig. 2 that the oxide thicknesses increase with increasing immersion time and dissolved oxygen concentration. These curve.." are nearly parabolic; i.e., the native oxide
growth mechanism in the ultrapure water has been confirmed different from that in the air defined by the layer-bylayer growth. This result suggests that the native oxide
growth is suppressed iu ultrapure water by decreasing the
dissolved oxygen concentration.
Figure 3 shows the native oxide thickness on n- and
n +- -Si wafers as a function of the immersion time in ultrapure water with 9-ppm dissolved oxygen concentration. The
native oxide thiockness (3.2 A.) on n + -Si is initially thicker
than that (1.9 A) on n-Si. It is worthwhile to note in Fig. 3
that the native oxide thickness on n -t -Si tends to saturate at
the value of 10 A in spite of its rapid initial growth, while on

16

05

14

(.I)
(.I)

ur 12
Z

:.c

I
IW
Q

x0
w
<:
CJ

a:

10

IN ULTRAPURE WATER
n-Si(100)
o 9ppm D.O.
A

O.6ppm D.O.

//

n-Si, it continuously increases with an increase of the immersion time following parabolic law. This result suggests that
the native oxide growth on n + -Si is considered to come from
a field-assisted mechanism. 21
The native oxide thickness on n + -Si is shown as a function of overflow rinsing time in Fig. 4, where the dissolved
oxygen concentration in ultrapure water is 0.6 ppm and the
wafer surface dry procedure is due to clean N z blow. The
oxide thickness grown during 2-min rinsing after a diluted
HF treatment is 2.6 A, and it increases to 3.3 A with an
increase of the immersion time to 60 min. This result implies
that the initial growth rate of native oxide is very high; i.e.,
the oxide growth on n ~ -Si in ultrapure water can be interpreted by a field assist oxidation model. 21 The oxide thick~
ness grown on n- and n -+- -Si under different rinsing and drying conditions is shown in Table II. The rinsing time is kept
constant at 10 min. The oxide thickness on n + -Si becomes
thin as the dissolved oxygen concentration decreases. The

j1

//
(Y/j

o............. ~,./L:;

101

o~

o O.04ppm D.O.

w
>
<: 2

102
103
TIME (min)

1
...

-o-

_0-

C)'

l--

10

n+-Si(100)

-0

J. Appl. Phys., Vol. 68, No.3, i August 1990

RINSING + N2 BLOW
(O.6ppm D.OJ

:-:

/rJ
104

101

(f)
(/)

10~

FIG. 2. Oxide thicknesses as a function ofitnmersion time of wafers in ultrapure water at room temperature for dilferent dissolved oxygen concentra
tions.

1274

FIG, 3. Oxide thicknesses on n-Si and n I- -Si as a function of immersion

time in ultrapure water with 9-ppm dissolved oxygen concentration.

0
10

TIME (min)

10

101
RINSING TIME (min)

10

~IG. 4. Oxide thickness on n \ -Si as a function of rinsing time by overflow

mg ultrapure water. The dissolved oxygen concentration in ultrapure water
is 0.6 ppm.

Moritaetaf.

1274

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

TABLE II. Growth of native oxide on

pure water rinsing and drying.
Condition

1'1-

and 11" -Si under different ultra-

II-Si( 100)

nt -Sit 100)

1.3

2.9

1.9

3.2

1.9

4.4

Rinsing (0.04 ppm DO)

+ N2 blow
Rinsing (0.6 ppm DO)
+ N2 blow
Rinsing (0.6 ppm DO)
+ hot water dipping
(! min) + hot Ar blow

5000'r---~-~-~-~-~---.--~--r--~---,

02 ANNEALED (49.6A)
---- Nz ANNEALED (39.8A)

-. __ AS GROWN (42.0A)

<Il

t-

:z
::J

0 110 .0

96.0

81NDiNG ENERGY (eV)

FIG. 6. Si,p XPS spectra of native oxides grown in ultrapure water for 46
days and after dry 0, annealing and dry N z annealing at 500 T for 1 h.

oxide thickness on n + -Si increases for hot water dipping and

hot Ar blowing. The oxide thickness (2.3 A) grown on a
p + -Si wafer during rinsing with ultrapure water and a drying procedure after a diluted HF treatment is nearly equal to
that (1.9 A) on n-Si, as seen from Fig. 1. The average bond
strength of Si-Si in p + -Si is weak compared with that in nSi, because p + -Si has many holes. Therefore, this result also
indicates that the native oxide growth in ultrapure water is
dependent on electron concentration, but not the average
strength of Si-Si bonds.
Figure 5 shows the time dependence of the numbers ofSi
atoms in native oxide and of Si atoms dissolved in ultrapure
water, where the dissolved oxygen concentration in ultrapure water is 9 ppm and NSi is defined as the number of Si
atoms per unit area on wafer surface. Si dissolution occurs
simultaneously during native oxide growth. The number of
dissolved 8i atoms is larger than the number of Si atoms in
native oxide by one order of magnitude. The normalized
thickness of the number (1.8 X 10 16 em - 2) of the dissolved
Si atoms to the single-crystal Si( 100) thickness and the na-

17

10

Si IN NATIVE OXIDE
Si

A DISSOL VED
n-Si(100)

..

/A

lOla

Vi

A/

iY"

A/

15r-----------------------------------,
0/
0/

.< .

0"'"
0/
0/

15

10

tive oxide thickness including the number of Si atoms of

9.2X 1O!4 em -2 are 34.9 and 4.1 A for 1315 min (-22 h),
respectively. The number of Si atoms in the native oxide film
increases correspondingly with the number of dissolved Si
atoms. This indicates that native oxide growth is strongly
influenced by the dissolution of Si into ultrapure water.
Figure 6 shows Si 2p XPS spectra of the native oxide
grown in ultrapure water for 46 days at room temperature
and after O 2 annealing and N2 annealing at 500C for 1 h.
The thicknesses of m;-grown oxide, O 2 annealed oxide, and
N2 annealed oxide are 42.0, 49.6, and 39.8 A, respectively.
The oxide thickness increases with O 2 annealing. This indicates that the oxidation reaction of Si at the native oxide/Si
interface proceeds even at 500C. The slight decrease of the
oxide thickness by N z annealing is conjectured to be due to
the deoxidation of Si-~O or Si----OH group, because Si 2p
XPS data indicate that the number of8i atoms bound with 0
or OH decreases by N 2 annealing. The binding energy shifts
for as-grown, O 2 annealed, and N z annealed oxide are 4.49,
4.87, and 4.72 eV, respectively. The binding energy of the
native oxide becomes high by O 2 or N2 annealing. However,
the electrical property of these oxides is stilI leaky.
Figure 7 shows the oxide thickness on n-Si wafer as a
function of the immersion time in a solution ofHl O 2 includ-

__________________________ 0--

100

_0_0 -

80

E
Ul

60

0'4

10'

10

103
TIME (min)

104

a::

::>
t-

tt:
40 llJ
a.

::;e

10~

20

UJ

I-

o~'---L--~10610--~--~2~6o~~--~3~6~o--~~4~b~0~ 0
FIG. 5. Time dependence of the number of5i atoms in native oxide and the
number of dissolved Si atoms in ultrapure water for wafers immersed in
ultrapure water. NSi is defined as the number of Si atoms per unit area on
wafer surface.

1275

TIME (min)

FIG. 7. Oxide thickness on n-Si as a function of immersion time in

(H 2 0, + Pt) solution.
Morita eta/.

J. Appl. Phys., Vol. 68, No.3, 1 August 1990

1275

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211
On: --...
Sun,
2014
15:47:09
'. .T' --;.'

.... ....................
..- ..16
T. Feb
..............
...............................
':'.' .<;.T
..v....... . .. . ....
=._.-......
<;.<;0:0- -~.<;.<;:<:o:.~

v.~.-...

~.-

O;;O. '.' - ......~........<;.-.... ' 0 v.~

7 ; ',....... '..-.-.;0 . . . . . . . . ._._

~;<;....

:OODOl

~~

IN AIR 73days 52A

----- IN WATER 4days 8.7A

(9ppm D.O.l

n-5!(100)

:1

110.0

,/-

:~
BINDING ENERGY (eV)

I
I

96.0

FIG. 8. Si 2P XPS spectra of native oxide on Si in air and in ultrapure water.

(al

iug a mesh ofPt catalyst, where the temperature ofthe solution is also given as a function of the time and the solution is
exchanged for a new solution every 1 h. Pt catalyst is used to
produce oxygen radical from Hz O 2 The oxide thickness saturates at 10 A even on n-Si wafer. This result indicates that
the oxide growth mechanism in (H 2 0 2 + Pt) solution is
different from that in ultrapure water. It is considered that
the oxide growth in (H 2 O 2 + Pt) solution is controlled by a
field assist oxidation even on n-Si having low electron concentration, because a number of oxygen radicals are adsorbed on the Si surface.
C. Native oxide structures

Si2p XPS spectra of the native oxides grown in air and in

ultrapure water are shown in Fig. 8, where the take-off angle
of the photoelectron is 35. The binding energy of SiO~
(x> 1) peak for the native oxide in ultrapure water is higher
than that in air. The intensity of the suboxide (SiO bond) at
the oxide-Si interface for the native oxide in ultrapure water
is smaner than that in air. The angular dependence plots of
the area ratio of the XPS signal from Sio.x to that from Si for
native oxide (15.9 A) in ultrapure water overlapped with the
plots for thermal oxide (17.7 A). It appears that the native
oxide in ultrapure water is not grown in an islandlike configuration, covering the Si substrate. The oxide-Si interface of
the native oxide in ultrapure water is rough at atomic level
compared with that in air, because the number ofSio. bonds
at the interface of the native oxide grown in ultrapure water
is fewer than that in air. The roughness of the native oxide~Si
interface can be observed by STM. Figure 9 shows STM
profile of the Si surface right after removing the native oxide
overlayer grown in ultrapure water by (HF + H 2 0) etching. It can be seen that the roughness of the nati.ve oxide-Si
interface increases as the native oxide grows in ultrapure
water. Observation of the native oxide surface with reflected
white light indicated that the surface of the native oxide in
ultrapure water also becomes rough with the growth of the
oxide. In Fig. 8, the XPS intensity at left side shoulder of the
peak from Si for the native oxide grown in ultrapure water is
larger than that in air. The binding energy of the peak obtained by the subtraction ofthe spectrum for the native oxide
in air from that in ultrapure water is nearly equal to local
electronegativity model 22 for Si=H2 and Si-H bonds. It is
conjectured that the Si-Hz and Si-H bonds exist at the
1276

J. Appl. Phys., Vol. 613, No.3, 1 August 1990

fbI
FIG. 9. STM profiles ofSi surfaces (a; ofdcaned wafer and (b) right aftcr
removing the native oxide grown in ultrapure water with the dissolved oxygen concentration of'} ppm for 20 days.

interface of native oxides grown in ultrapure water. 18

Si 2p XPS spectra of native oxides grown on n-Si by various chemical treatments are shown in Fig. 10. The intensities from SiO bond at the oxide-Si interface of the native
oxides in the NH4 OR solution and in the Hel solution are
smaller than that in the H2 S04 solution. This indicates that
the interface of the native oxide in the NH4 0H solution 23
and in the HeI solution is atomically rough compared with
that in the Hz SO.; solution.

H2S04+H202 (5.5A)
NH40H+H202+H20 (4.1A)
HCl+H202+H20 (5.1A)
n-5;(100)

96.0

BINDING ENERGY (eV)

FIG. 10. Si,l' XPS spectra of native oxides grown on n-Si by various chemical treatments.

Morita et al.

1276

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

TABLE III. Native oxides under various conditions.

cA)

E'

Contact angle

Sampled

(eV)

(")

In air
4duys
In ultrapure
water (9 ppm DO)
2 days
H 2 S04 +H 2 0 1
cleaning
NH.OH + H 2 0? + H 2 0
cleaning

5.6

3.85

35.7

0.14
(9 days)

5.6

4.08

It.4

4.03
(7 days)

5.5

3.83

<10

4.1

3.99

dO

Oxide thickness

Dissolved Si"
weight (ltg/em 2 )

<0.05
(7 days)

II-Si( 100).

bE = E - E(Si}pv,)'
C

Into ultrapure water (9 ppm DO).

The results showing the structural difference between

the native oxides grown under various conditions is summarized in Table III. The binding energy of SiO x (x> 1)
peak for the native oxide in ultrapure water is higher than
others. It is seen from the contact angle data of the ultrapure
water droplet that the surface of the native oxide formed in
air is dose to being hydrophobic. This implies that the many
Si atoms at the native oxide surface are terminated by hydrogen atoms. The amount of Si dissolved upon exposure to
ultrapure water for the native oxides grown under various
environments, with exception of ultrapure water, is little.
This proves that the structure of the native oxide grown in
uitrapure water is different from that of other oxides.

Do Growth of native oxide on HF gas etched silicon

The growth of native oxide films on dean bare silicon
surface obtained in the gas phase etching technology is also
investigated. The bare silicon surface etched by HF gas is
covered by chemically combined fluorine, i.e., fluorine-terminated silicon surface. 24 This chemically combined fluorine give a definite influence on succeeding processes. The
effect of fluorine-terminated silicon surface on succeeding
processes was evaluated by the sputtering epitaxy. 2530 Figure ! 1 shows reflection electron diffraction pattern from Si
epitaxial film grown using rf-dc sputtering technology at a
substrate temperature of 3ooC on three different silicon
surfaces such as fluorine-terminated silicon surfaces. native
oxide surfaces having a thickness on A (SiOx: x > 0.'5), and
wet-cleaned silicon surfaces having 004 A of native oxide
(SiOx:]I. > 0.5). Single-crystalline silicon is obtained only on
the wet-cleaned silicon surface. An amorphous silicon film is
grown even on the native oxide surface having a thickness of
7 A, indicating that the existence of the native oxide degrades the quality of the semiconductor process, in particular, a low-temperature process. The results of this low-temperature sputtering silicon epitaxy are summarized in Table
IV. The sheet resistivity of grown silicon film having a thickness of0.2 ,urn on the wet-cleaned silicon surface is lOOn. 011
the other hand, the grown films are amorphous silicon having the sheet resistivity of 5 X 106 n. on the native oxide surface and amorphous silicon having the sheet resistivity of
1277

J. Appl. Phys., Vol. 68, No.3, i August 1990

3-5 X 106 n on the fluorine-terminated silicon surface. when

aU process parameters are kept at the same conditions. Thus
fluorine-terminated silicon surfaces exhibit similar characteristics to native oxide existence for film depositions.

Cal

Ie)
FIG. 11. Reflection electron diffraction pattern from Si epitaxial film on
(a) l1at.ive oxide surface having a thickness of 7 A (before cleaning), (b)
wet-cleaned silicon surface, and (c) dry-cleaned silicon surface using tfdc
sputtering.
Moritaetal.

1277

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

T ABLE IV. Resistivity of sputtering epitaxial silicon film.

Silicon substrate

Epitaxial silicon
-"---~-~~-.-.---.

HF

Crystal

Sheet
resistivity"

cleaning

SiO x film
thickness CA.)
Cx>0.5)

structure

an

Before cleaning
Wet cleaning
Dry cleaning

7
0.4
0.3

amorphous
single-crystal
amorphous

5XW"

,...

.'"

.-

!!.

!DO
3-5X 10"

...."

"Sheet resistivity: Four-point pmbe method.

694

Figure 12 shows the FIs XPS spectra from the fiuorineterminated surface after I-h thermal treatment up to 930C
in dry N2 at atmospheric pressure. The F" spectra peak
intensity does not change even after heating the substrate up
930 "C. Terminating fluorine can not be eliminated by heating silicon wafer up to 930 cc. F h XPS spectra from a fluorine-terminated silicon surface are shown for the different
bombarding Ar-ion energies in Fig. 13, where the effect of
the thermal treatment in an ultrahigh-vacuum environment
is also given. The peak intensity ofF)s spectra decreases with
an increase in Ar-ion energy, but the elimination of terminating fluorine is not sufficient even when the At-ion energy
is increased up to 10 eV, as shown in the curve.
Figure 14 shows the thicknesses of native oxide films on
HF acid etched and HF gas-etched Si surfaces as a function
of exposure time to clean reom air at room tempemtnre,
where HF gas-etched Si surfaces are prepared by a dry cleaning technology using gas-phase anhydrous hydrogen fluOlide. 24 It can seen from Fig. 14that the growth of the native
oxide on HF gas-etched Si is slightly suspended in initial
stage compared with that on HF acid-etched Si.
Fluorine strongly terminates silicon atoms on a wafer
surface, resulting in the fact that fluorine cannot be sufficiently eliminated by the thermal treatment or the Ar-ion

...

.,<::

692

690
688
686
684
Binding
Energy (eV)

682

680

PIG. 13. Ph XI'S spectra from fluorine-terminated surface by low-energy

AT bomhardment. Bombardment conditions are (a) no bombardment

(dry-cleaned surface), (b) no bombardment (substrate temperature

300 'c, base pressure 5 X 10 _9 Torr), (e) Ar-ion energy of 3 eV, substrate
temperature 300 "c, Ar gas pressure 8 X 10- J Torr. and bombarding AT-ion
energy of 10 e V, substrate temperature 300 C, Ar gas pressure 8 X 10 3
Torr.

bombardment. Thus the fluorine termination has a protection effect of silicon surfaces against heat treatment or Arion bombardment. However, the fluorine termination suffi~
dently cannot protect the silicon surface against the native
oxide growth under the coexistence of oxygen and wafer as
shown in Fig. 14. Therefore, this led us to the conclusion that
the closed system in which wafers are not exposed to air is
necessary in the production line in order to keep the wafer
surface clean.
IV. DISCUSSION

The chemical bond of Si atoms on wet-cleaned wafer

surfaces has been analyzed by some spectroscopic measurements. 31 - 33 Electron energy-loss spectroscopy measure-

------~
~---------

c:

694

692

690
B!nd!ng

688
Energy

686

684

682

600

CeV)

~O~O~~UW1LO'~~~10~2~~~1~O~3~~Wl~O~~~~~1~O~5~
FIG. 12. F,. XPS spectra from fluorine-terminated smfaces heated up to
various temperatures in dry N, gas environment of atmospheric pressure at
(a) room temperatllre, (b) 80 "c for I h, (c) 630 OC for 1 h, and (d) 930 "C
for 1 h.
1278

J. Appl. Phys., Vol. 68, No.3, 1 August 1990

TIME (min)
FIG. 14. Oxide thicknesses on HF acid-etched and HF gas-etched Si surfaces as a function of expose time of wafers to air at room temperature.
Morita et al.

1278

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

ments showed that an Si surface dipped into hydrofluoric

acid is covered mainly by Si-H groups with a small number
ofSi-F bonds. 31 On the other hand, a dip of the wafer in an
ammonia or HeI hydrogen peroxide/water mixture was
shown to create a hydrophilic surface, caused by the formation ofSi-OH groups. Fourier transform infrared spectroscopy measurements proved that the Si surface prepared with
ultraviolet cleaning followed by HF dipping has a hydrogen
monoatomic layer terminating the dangling bonds of 5i. 32
Thennal desorption spectroscopy measurements indicated
that Si surface right after a diluted HF dipping is covered
primarily by H, with some H 2 0 or OH. 33 It can be seen from
above-mentioned research results that most 8i atoms at a
cleaned wafer surface are terminated by hydrogen.
The oxide thickness formed at room temperature after
lO-h exposure to the air is estimated around 10 . 13 A from
the extrapolation of the temperature dependence data of the
oxidation rate of 8i in dry or wet oxygen between 800 and
1050 C shown in Fig. 15. It can be considered, thus, that the
oxidation mechanism of Si in air at room temperature is entirely different from the thermal oxidation mechanism. The
following model is proposed for the oxidation of Si at room
temperature: Si atoms at a cleaned wafer surface are terminated by hydrogen. The oxygen species break Si-Si bonds to
produce Si--O bonds, the remaining Si--H bonds which
cause hydrophobic behavior. The oxygen species are produced at native oxide surfaces by the coexistence of O 2 and
H 0, or the Si-O bonds are produced by the coexistence of
0: (0) andH 2 0(OH) nearSi-Si bonds. TheSi-Si bonds
of the underlayer Si are broken to produce Si-~O bonds after
all of the Si atoms in the overlayer are oxidized; i.e., the
overlayer structure converts to the amorphous phase, result-

104

"-

0\
0

"'"'
'~ 105

.:5
w

Jo-

\.
\

z
0

i=

C31o"
X

WET 02
DRY 02

SI

Si

SI

1\

0
0

~Si /

5i

'0
\

/ '0
5'
\

pi

HO" 0"
H

51

\\
/51'

51

\0

SI

000

\ /

Si

51

Si

J. Appl. Phys., Vol. 68, No.3, 1 August 1990

/\

5i

FIG. 15. Temperature dependence of thermal oxidation rate fer silicon in

dry &nd wet oxygen between 800 and 1050 C.

/'

1000lT (K-')

~, S~

/' "I \ / "-

10 L-.b--'-~-'--~-'--~-:-':~--:-'-:-'---:-"::--"
0.6
0.7
0.8
O.S
1.0
1. 1
1.2

, 0/ '0
Si

We have demonstrated that the growth of native oxide

on cleaned Si wafers in air and in ultrapure water requires
the coexistence of oxygen and water or moisture. The
growth rate of the native oxide in air and in ultrapure water
can be decreased by lowering moisture concentration in air
and by lowering dissolved oxygen concentration in ultrapure
water. Layer-by-Iayer growth of the native oxide has been

v. CONCLUSION

" Si "

-->

I"Si 0/'Si

ing in layer-by-layer oxide growth. Figure 16 illustrates this

oxidation scheme conceptually.
Figure 17 illustrates the chemical structure model ofthe
native oxide grown in ultrapure water. The Si atoms of native oxide surface are terminated by 0 or OH. The 5i-H
groups exist at the oxide-Si interface.
Three models explaining Si dissolution into ultrapure
water can be proposed: First, S1 compounds are produced at
the oxide-5i interface with the oxidation reaction of8i. These
Si compounds diffuse through the native oxide and dissolve
into ultrapure water. Second, the native oxide grown in ultrapure water is etched by ions and/or dissolved oxygen.
Third, the Si atoms at Si surface dissolutes into ultrapure
water to produce silica and this silica deposits on Si surface.
It is, however, likely that the growth mechanism of native
oxide in ultrapure water is complicated. Further investigation is necessary to clarify the mechanism of the native oxide
growth accompanying the Si dissolution.

1279

Si
/ '

FIG. 16. A model of native oxide growth in air.

\\

0::

H,

-.,.

/ \
SI

51

/ '"

SI

Si

,
I
I

FlG. 17. A model of native oxide growth in ultrapure water.

Morita at al.

1279

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

confirmed to occur on the Si surface in air. The native oxide

growth on an n-Si wafer in ultrapure water accompanies the
Si dissolution into the water and degrades the atomic flatness
at the oxide-Si interface, and produces a rough native oxide
film surface. The growth rate oftlle native oxide on n + -8i in
ultrapure water is high at the initial stage and saturates at the
thickness of 10 Abecause of a field-assisted mechanism. The
native oxide growth on n-Si in a H2 O 2 solution including a
Pt catalyst is controlled by a field assist oxidation because of
a number of oxygen radicals. It is found for the first time that
a field assist oxidation occurs on n -\ -8i in ultrapure water
and on n-Si in (Hz O 2 + Pt) solution at room temperature.
We propose a nitrogen gas tunnel connecting process
equipment as system to prevent native oxide growth, and an
ultrapure water rinsing vessel sealed by nitrogen gas in order
to suppress the native oxide growth in ultrapure water.
ACKNOWLEDGMENTS

The authors wish to thank Y. Morikage and K. Katagiri

at Seiko Instruments Inc. for STM observation, Professor T.
Hattori at Musashi Institute of Technology, and Dr. N. Yabumoto at NTT LSI Laboratories for useful discussions. If it
is noted that preceding work of importance related to this
article is not cited as a reference, it is due to our incomplete
survey of reference. It would be greatly appreciated if such
information or comments are provided to the authors. This
study was carried out at the Super Clean Room of Laboratory for Microelectronics, Research Institute of Electrical
Communication, Tohoku University.

APPENDIX

The thickness of thermal oxide films was measured by

the nulling method using the eUipsometer (Shimadzu AEP1(0) in the range from 70 to 140 A. Figure 18 shows the
thermal oxide thickness measured by ellipsometry as a function of the area ratio of the XPS signal from the oxide (SiOx :
x>O.S) to that from the Si substrate. The accuracy of refractive indices measured in this oxide thickness region is
1.46 0.01 in this ellipsometry measurement. It is seen
from Fig. 18 that the oxide thickness measured by ellipsometry is not scattered as expressed by

d=A t ln(A

~~ + 1),

nsO'sAs

A=---,
n,O'/AI

where d is the thickness of an oxide overlayer, I, and II are

the photoemission intensities for the Si substrate and the
oxide overlayer. respectively, tis and n I are the densities ofSi
atoms, O's and 0'1 are the atomic photoionization cross sections, andAs andA J are the escape depths. Thus the errors in
refractive index and oxide thickness are considered to be
small enough even for these thin films between 70 and 140 A
in our experiments.

R. J. Archer, J. Electrochem. Soc. 104,619 (1957).

F. Lukes, Surf. Sci. 3(}, 91 (1972).
3S. r. Raider, R. Fitsch, and M. J. Palmer, J. Electrochem. Soc. 122,413
(1975).
4 H. R Philip and E. A. Taft, J. Appl. Phys. 53. 5224 ( 1(82).
'E. A. Taft, J, Electrochem. Soc. 135,1022 (1988).
"T. Hattori, K. Takase, H. Yamagishi, R. Sugino, Y. Nara, and T. Ho, Jpn.
J. App!. Phys. 28, L296 (1989).
7 K. Takase, T. Igarashi, N. Miyata, K. Moriki, R. Sugino, Y. Nara, T. Ito.
M. Fujisawa, and T. Hattori, in Extended Abstracts of the 21st Conference on Solid State Devices and Materials, Tokyo, 1989, p. 393.
8 G. Gould and E. A. Irene, J. Electrochem. Soc. 134, 1031 (1987).
9 R. Takizawa, T. Nakanishi, and A. Ohsawa, J. App!. Phys. 62, 4933
(1987).
to A. S. Maeda and M. Ogino, in Extended Abstracts of the 169th Electrochemical Society Meeting, Boston, 1986, p. 372.
11 Y. Matsushita, in Digest, Technical Papers, 1989 VLSI Symposium,
Kyoto, 1989, p. 5,
l2y. Matsushita, in Proceedings ofthe 1st Workshop on ULSI Ultra Clean
Technology, Tokyo, 1989, p. 1.
"T. Shirnono and M. Tsuji, in Proceedings of the 1st Workshop on ULSI
Ultra Clean Technology, Tokyo, 1989, p. 49.
14 T. Ohmi, Oyo Butsuri 58, 193 (1989).
"T. Ohmi, M. Morita, E. Hasegawa, M. Kawakami, and K. Suma, in Ex
tended Abstracts of the 175th Electrochemical Society Meeting, Los An
geles, 1989, p. 227.
"M. Morita, T. Ohmi, E. Hasegawa, M. Kawakami, and K. Suma, in Di
gest, Technical Papers, 1989 VLSI Symposium, Kyoto, 1989, p. 75.
17M. Morita and T. Ohmi, Ultra Clean T echnol. 1, 22 ( 1989).
'"M. Morita, T. Ohmi, E. Hasegawa, M. Kawakami, and K. Suma, App!.
Phys. Lett. 55, 562 (1989).
'9E. A. Taft and L. Cordes, J. Electrochem, Soc. 126,131 (1979).
2D. E. Aspenes and J. B. Theeten, J. Electrochem. Soc. 127,1359 (1980).
21 N. Cabrera and N. F. Mott, Rep. Prog. Phys. 12, 163 (1949).
22G. Lucovsky,l. Phys. (Paris) Colloq. 42, C4-74! (19tH).
2JH. Mishima, T. Yasui, T, Mizuniwa, M. Abe, and T. Ohmi, IEEE Trans.
Semicond. Manuf. 2, 69 (1989).
24 N. Miki, H. Kikuyama, 1. Kawanabe, M. Miyashita, and T. Ohmi, IEEE
Trans. Electron Devices 37, 107 (1990).
1

THERMAL OXIDE FILM

140

o~ 120
'-"

(/)
(f)

ill
Z

ul00

:E

fW

o
X
o 80

60

FIG. 18. Thermal oxide thickness measured by ellipsometry as a function of

the area ratio of the XPS signal from the oxide (SiO,: x>O.5) to that from
Si substrates.
1280

J. Appl. Phys., Vol. 68, No.3, 1 August 1990

Morita et al.

1280

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun, 16 Feb 2014 15:47:09

To Ohmi. K Matsuda, To Shibata, T. Ichikawa, and H. Iwabuchi, Jpn. J.

App\. Phys. 27. 2146 (1988).
lOT. Ohmi, T. Ichikawa, T. Shibata, K. Matsudo, and H. Iwabuchi, App\.
Phys. Lett. 53,45 (1988).
27T. Ohmi, K. Matsudo, T. Shibata, T. Ichikawa. and H. fwabuchi, App/.
Phys. Lett. 53, 364 (1988).
28 T. Ohmi, H. lwabuchi, T. Shibata, and T. Ichikawa, App!. Phys. Lett. 54,
25

253 (1989).

29T. Ohmi, K. Matsudo, T. Shibata, T. Ichikawa, and H. Iwabuchi, in Ex-

1281

tended Abstracts afthe 20th Conference on Solid State Devices and Materials, Tokyo, 1988, p. 49.
,oT. Ohmi, T. Ichikawa, and H. Iwabuchi, App!. Phys. Lett. 54, 523.
(1989).
31 M. Grundner and H. Jacob, Apr!. Phys. A 39, 73 (1986).
"T. Takahagi, L Nagai, A. Ishitani, H. Kuroda, and Y. Nagasawa, J. Appl.
Phys. 64, 3516 (1988).
3' N. Yabumoto, K Minegishi, Y. Komine, and K. Sato, Jpn. J. Appl. Phys.
29, L490 (1990).

Morita et al.

J. Appl. Phys., Vol. 68, No.3, 1 August i 990

1281

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
93.180.53.211 On: Sun,. ,16
. -. Feb 2014 15:47:09
-_ ,-.-.c-." .". --- ., ... no ,
~,

-.-~.-..~.~.~.~.~..,.-:;;..~.~.'.....'Z:.:.:.:.:.:.:.:.:.:::.:.:.-:.~.'.-,;<.'.'.'.';'.'-;O;-'.o;' '.".-;-..-.~.v.

.~_"

~.