Uncertainties Caused by Surface Adsorbates in Estimates of
the Thickness of SiO2 Ultrathin Films
Yasushi Azuma, Ruiqin Tan, Toshiyuki Fujimoto, Isao Kojima,
Akihito Shinozaki* and Mizuho Morita*
Materials Characterization Division, National Metrology Institute of Japan, AIST
Tsukuba Central 5, Higashi 1-1, Tsukuba-shi, Ibaraki 305-8565, Japan
*Precision Science and Technology Division, Graduate School of Engineering, Osaka University
Yamada-oka 2-1, Suita-shi, Osaka 565-0871, Japan
Abstract. Grazing incidence X-ray reflectivity (GIXR), ellipsometry and X-ray photoelectron spectroscopy (XPS) have been
used for measuring the thickness of ultrathin SiO2 films on Si(lOO). SiC>2 films were fabricated at a constant temperature to
obtain a fixed interface structure for films with different thicknesses. The thicknesses obtained by GIXR and ellipsometry were
in good agreement with each other, however, ellipsometry showed slightly larger values. The thickness of the adsorbed
overlayer was also compared using GIXR and XPS. Uncertainties included in the XPS measurements of the carbonaceous layer
thickness were estimated. The thicknesses of the carbonaceous layer obtained by XPS were slightly smaller, by about 0.16 nm,
than those of the adsorbed overlayer obtained by GIXR. About 0.3-0.4 monolayer of adsorbed water molecules is believed to
account for the differences in overlayer thicknesses between the GIXR and XPS measurments.
parameters. Especially, XPS requires the effective
attenuation length (EAL) or inelastic mean free path
(IMFP) as an essential parameter in order to estimate
film thickness, where such parameters require
extensive study to attain its reliability.
So far, many reports have been published on how to
obtain accurate SiC>2 thickness measurements [1-4].
Up to now, methods such as XPS, sectional
transmission electron microscopy, ellipsometry and
RBS have been used extensively and have been found
to be in rather good agreement with each other.
However, the accuracy of these measurements has not
yet reached the 4% (3o) required from the
international technology roadmap for semiconductors
(ITRS) [5]. Therefore, further research is required in
the area of comparing the differences between
different methods.
In this study, ultra thin SiO2 films with thicknesses
between 3-10 nm have been investigated by using
GIXR, ellipsometry and XPS techniques. The
influence of the adsorbed layer in evaluating the SiO2
thickness was also investigated by comparing the
thickness of the overlayer obtained from GIXR with
INTRODUCTION
Characterization of ultra thin films is very important
for the development of advanced technologies such as
in semiconductor devices where their size is
continuously shrinking. Much attention has been paid
to the accurate measurement of SiO2 thickness as well
as reference materials because of their use as gate
oxides in silicon devices. The methods used to
measure ultrathin film thicknesses are roughly
classified into two groups. Methods such as GIXR and
ellipsometry give results that directly related to length.
On the other hand, methods related to surface analysis
such as XPS and Rutherford backscattering
spectroscopy (RBS) give results mainly related to the
amount of material in the films. Each has its merits
and demerits. For example, the thickness derived by
the former methods usually includes undesirable
quantities arising from an adsorbed layer. The nonuniformity of the film in the depth direction makes it
difficult to analyze the layered structure accurately.
The signals from the latter methods have a less direct
relation to SI units and depend more on experimental
CP683, Characterization and Metrology for VLSI Technology: 2003 International Conference,
edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula
© 2003 American Institute of Physics 0-7354-0152-7/03/$20.00
337
that of carbonaceous overlayer estimated by XPS.
Measurements by GIXR and ellipsometry were
effected somewhat by the existence of the adsorbed
layer and the interfacial layer. In the usual thermal
oxidization of Si surfaces, since oxides are formed
with considerable thickness at increasing temperatures,
ultrathin film growth is done at relatively low
temperatures. In such cases, there is a danger that there
can be substantial differences in interfacial structure
and film property between films of different
thicknesses. In this study, we employed a rapid
oxidation procedure equipped with an infrared lamp
heater in order to avoid the confusion caused by the
differences in the interfacial structure and film
property.
beams were collimated with 0.05 mm slits and the
reflection intensity was measured by a scintillation
counter. The 29 angular resolution of the instrument is
0.0001°. The specular reflectivity curves were
recorded with a 9-29 scan. Spectroscopic ellipsometry
measurements were performed by using GES-5M and
GESP-5 instruments (SOPRA Co.). XPS spectra
were recorded by an ESCALAB 220i-XL (VG Co.).
Measurements were carried out collecting normaliyemitted photoelectrons from the sample surface using
Al Kamonochromated X-rays.
A "large area"
measurement mode was used since it has a smaller
acceptance angle than other modes.
t min.
9oo°a
2.EXPERIMENTAL
+->I
cd
2.1 Preparation of SiO2 Films
5°C/sec
50°C/sec
o>
H
SiO2 films were fabricated with an ultra-thin oxide
film preparation system set up in an ultra-clean room
(Class 1). The system has a cold-wall reaction
chamber where a silicon wafer is heated from both
sides by infrared lamps through a quartz window.
Since only the wafer is heated, oxidation is carried out
in very clean conditions. Purified oxygen and nitrogen
gases were supplied for oxidizing the silicon wafers
and for filling the reaction chamber as an inert gas,
Prior to oxidation, silicon wafers were cleaned using
the following sequence: a rinse in ultra pure water
containing ozone gas, an ultrasonic cleaning in
HF+H2O2+H2CH-surfactant solution, etching in dilute
HF solution, and an ultrasonic cleaning in ultra pure
water [6]. Oxidation was carried out at atmospheric
pressure in oxygen. The temperature during oxidation
was controlled as shown in Fig. 1. Oxidation occurs
very little during the rise in temperature, and the
thickness is regulated by time after the temperature
reaches 900 °C. Figure 2 is a plot of the oxide
thickness versus time for a constant temperature of 900
O2 consentration : 100%
flow rate : IL/min.
atmosper pressure
Time
FIGURE 1. Temperature control for the oxidation.
10
Q)
I
0
5
10
15
20
25
Time (min.)
2.2 Measurements by GIXR, Ellipsometry
And XPS
FIGURE 2. The relationship between SiO2 thickness and
time for a constant oxidation temperature of 900 °C.
3. RESULTS
GIXR measurements were performed using a highresolution x-ray diffractometer (Rigaku Co., ATX-G2).
A rotating anode Cu Koc source (18kW) was used. A
parabolic multilayer mirror collected X-rays to form a
parallel beam. Then, the beam was compressed and
monochromated with a non-symmetric grating-type
Ge(lll) monochromator. The incident and reflected
AND
DISCUSSION
3.1 Changes in Thickness for an
Exposed Surface in Air
When thickness measurements are carried out in air,
the thickness is expected to change because water and
338
Figure 5 compares the SiO2 thicknesses from GIXR
and spectroscopic ellipsometry for samples kept in air
until the ellipsometric thickness became stable. The
dotted line shows a straight line with a slope of 1. The
thicknesses obtained by the two methods are found to
be in good agreement with each other
organic molecules adsorb continuously on the surface.
In order to investigate this process, ellipsometry
measurements were taken on a continuous basis
immediately after oxidation was finished. Figure 3
shows the thickness results when calculated employing
an ideal SiC>2 film model, namely, assuming that the
film does not contain water or organic molecules on its
surface. The thickness increases with time. The
thickness for the first data point (after 21 min and 5
second) was 4.67 nm, and the thickness for the 180 th
data point (24 hour and 37 min) reached 5.03 nm. The
increase is about 0.36 nm.
5.3
5.2
5.1
r 5
;4.9
i 4.8
1
4.7
28 (degree)
4.6
FIGURE 4. GIXR profile for SiO2 films. These nominal
thicknesses are (a) 10, (b) 8, (c) 5, (d) 3 nm, respectively.
4.5
4.4
0
20000
40000 60000
Time(s)
80000
100000
S 10-
FIGURE 3. SiO2 thickness measured by ellipsometry as a
function of time in the atmospher after fablication.
3.2 Characterization of Film Structure by
I
GIXR
Figure 4 shows the data from the GIXR
measurements. The data were analyzed using a three
layer-model, consisting of an adsorbate overlayer, a
SiO2 layer and an interfacial layer. The solid lines in
the figure were obtained using a non-linear leastsquares calculation. Table 1 summarizes the results
for the thicknesses. In the table, the thicknesses for the
SiO2 layer are given as the sum of SiO2 and interfacial
layers. When data analysis using a 2-layer model that
neglects the adsorbate layer was carried out, the
thickness of SiO2 layer was slightly larger than that
obtained by the 3-layer model. This is due to the
adsorbate layer since it is included in the SiO2 layer.
The increase in thickness depends on the ratio of
densities of the adsorbate and SiO2. Since the
calculated density of the adsorbate layer by GIXR is
about 0.8 g/cm3 (which is about 1/3 that of SiO2), the
thickness of SiO2 obtained by the 2 layer model was
greater by about 1/3 that of the real thickness of the
adsorbate layer.
0
5
1.0
Thickness estimated by XRR (nm)
FIGURE 5. Comparison of thickness results for SiO2
films from GIXR and Ellipsometry. The dotted line shows a
straight line with a slope of 1.
3.3 Thickness Measurements Using XPS
For a layered system of SiO2 on Si, the
photoemission intensities and the thickness of the film
are related by the following equation [7, 8]:
— —— I expO
F
I, S
339
TABLE 1. Thickness of contamination layer and SiO2 layer by GIXR using a calculation model with three
layers ( a contamination layer included) and with two layers (contamination layer neglected).
c
d
a
b
3 layer
2 layer
2 layer
3 layer
2 layer
3 layer
3 layer
2 layer
model
model
model
model
model
model
model
model
Contamination
0.54
0.65
0.71
0.59
layer [nm]
3.24
5.44
3.41
SiO2 layer [nm]
8.15
5.73
10.36
10.23
8.18
than 3 nm, the XPS thicknesses are slightly smaller, by
0.2 - 0.3 nm, than the GIXR thicknesses. Although
the reason for this difference in thickness is not known
exactly, there are several possibilities such as: a nonuniformity in thickness over the film where this nonuniformity will influence XPS results to a greater
extent, or there is a larger GIXR uncertainty in the
thinner films, etc.
where 7S and 70 are the intensities of the Si2p peaks for
the Si substrate and the SiO2 layer, respectively, and /I
o
/
is the effective attenuation length for SiO2. fi = °/<? ,
/ ^s
where 50, Ss are the intensities of Si2p peaks for bulk
Si and SiO2, respectively. 9 is the angle of
photoelectron emission with respect to the surface
normal. Equation (1) can be written as follows.
—^-
(2)
£ 10
X
Equation (1) was used to calculate
and f t .
20 "*
0(exp(d/Acos0)-1)
10
0
5
10
Thickness estimate by XRR (nm)
FIGURE 7. Comparison of thickness results for SiO2
films between GIXR and XPS. The dotted line shows a
straight line with a slope of 1
o10
3.4 Thickness of Carbonaceous Adsorbate
Thickness (nm)
FIGURE 6. Plot of L %
vs. GIXR thickness. Solid
Layer Using XPS
/ *s
line is a least-squares fit using equation (1)
Figure 6 shows a plot of
The photoelectron intensities of Si2p peaks from a
clean SiO2 film on a Si substrate are given as
follows. [7, 8 ]:
Q
/. vs. GIXR thickness.
/ *s
The solid line is a least-squares fit using equation (1)
where A and /? are variables, and the thickness is
obtained using GIXR. The intensities of 7S and /0 were
calculated from spectra with backgrounds subtracted
using Shirley's method. The variables A and /3 were
calculated to be 3.4 and 0.70, respectively. Thus A
and /? were used to calculate the XPS thickness using
equation (2). Figure 7 shows the plot of the XPS
thickness vs. the GIXR thickness. The two thicknesses
are in close agreement, which indicates that the
calibration of thin film thickness based on the GIXR
measurement is useful. However, for films thinner
l02
1(% cos#) }
si02
(3)
(4)
where X*s and X*0 are the attenuation lengths of Si2p
photoelectron for the Si substrate and SiO2,
respectively. Assuming the relation, A° = X*0 , the
photoelectron
intensity,
340
is independent of film thickness is given by the
following equation using Is(dSi02) and I0(dSi02).
thickness and its uncertainty calculated as above as
well as the adsorbate layer thickness obtained using
GIXR. The thicknesses of the carbonaceous overlayer
obtained using XPS are slightly smaller than those of
the adsorbate overlayer obtained using GIXR. This is
mainly because the thickness in GIXR includes an
additional adsorbate. Hirashita et al. reported, that by
using a thermal desorption technique, the SiO2 surface
adsorbs water molecules up to approximately 3.9><10"14
H2O/cm2 [10]. This amount corresponds to 0.3 - 0.4
monolayers, or about 0.1 nm in thickness. Therefore,
we conclude that amount of adsorbed water molecules
accounts for this difference in the overlayer
thicknesses between GIXR and XPS.
(5)
TOO
/
Here, the value for/?= °/^ was obtained in the
/ *s
previous section. When the surface has a carbonaceous
overlayer, the intensity, Is+0 , is reduced to Ix. The
ratio Ic /Ix is expressed as follows where Ic is the
intensity of Cls from the contaminant layer
(6)
TABLE
2.
Value
of
parameters for estimation of
contamination thickness by
XPS.
3.8 nm
where dc is the thickness of the carbon overlayer, and
AC and A# are the effective attenuation lengths of Cls
and Si2p photoelectron in the carbonaceous overlayer,
respectively. When hccl hcsi ~1, a equation (6) can be
transformed into equation (7).
c=
4 cos*In
- {'c **
4
%
%
3.4 nm
3.5 nm
y
(7)
0.76
/nc
Ts+0crs+0/
/TC°C
(8)
1 f-i Yl f
TABLE 3. Thickness of contamination layer calculated by
XPS with standard uncertainties (2a) and the thickness
obtained by GIXR.
GIXR
XPS
carbonaceous overlayer
adsorbate overlayer
[nm]
[nm]
a
0.59
0.44±0.31
0.54
b
0.32±0.24
c
0.65
0.56±0.40
d
0.71
0.54±0.38
where T0 and Tc are photoelectron transmission
functions, and n0 and nc are the atomic
concentrations of Si in SiO2 and C in the carbonaceous
layer, respectively. /L°0 is the effective attenuation
length of Si2p photoelectrons in SiO2, respectively. a0
and <rc are the photoionization cross sections of Si2p
and Cls photoelectron, respectively. Table 2 lists the
parameter values used to calculate dc .
n
°/
was
calculated by assuming the density obtained by GIXR
and the molecular structure of polystyrene for the
carbonaceous layer, /if. and /i£ were calculated by
Cumpson's formula for IMFP proposed for polymers
[9]. /L°0 is 3.4nm as obtained before.
4. CONCLUSION
SiO2 films were fabricated at a constant temperature
to attain a fixed interfacial structure among films with
different thicknesses. These films were then used in
comparative thickness measurements by GIXR,
ellipsometry and XPS. The thicknesses obtained by
GIXR and ellipsometry were in good agreement with
each other, however, ellipsometry showed slightly
larger values. The thickness of the adsorbed overlayer
was also compared using GIXR and XPS. The
thicknesses of the carbonaceous layer obtained by XPS
are slightly smaller, by about 0.16 nm, than those of
T
s+ocrs-
was obtained from data in the XPS software.
The uncertainty of dc is calculated from the
propagation rule. Uncertainties for each parameter
were estimated as follows, uncertainties for 3?Si, A°0
and A£ are 20% of their values, and uncertainties for
and
* s+o^s+o/
0.74
are
10%. Table
3
summarizes the results for the contamination-layer
341
the adsorbed overlayer obtained by GIXR. The
amount of adsorbed water molecules seems to account
for the difference. Further measurements using RBS
are now in progress. Further comparative studies will
lead to a more accurate determination of the thickness
ofultrathin films.
REFERENCES
1.
Lu, Z. H. , McCaffrey, J. P, Brar, B, Wilk, G. D.,
Wallace, R. M., Feldman, L. C., and Tay, S. P., Appl.
Phys. Lett. 71, 2764-2766 (1997).
2.
Toshiharu, K., Yamamoto, H., Ikeno, M., Mashiko, Y.,
Kawazu, S., and Umeno, M., Jpn. J. Appl. Phys. 38,
4172-4179(1999).
3.
Cole, D. A., Shallenberger, J. R., Novak, S. W., Moore,
R. L., Edgell, M. J., Smith, S. P., Hitzman, C. J.,
Kirchhoff, J. F., Principe, E., Nieveen, W., Huang, F.
K., Biswas, S., Bleiler, R. J., and Jones, K., J. Vac. Sci.
Technol. B 18, 440-444 (2000).
4.
Gross, Th., Lippitz, A., Unger, W., and Guttler, B.,
Surf. Interface Anal. 29, 891-894 (2000).
5.
Semiconductor Industry Association. International
Technology Roadmap for semiconductors, 2001, 2001
edition. Austion, TX: SEMATECH, 2001.
6.
Ohmi, T., J. Electron. Soc. 143, 2957-2964 (1996).
7.
Hill, J. M., Roye, D. G, Fadley, C. S., Wagner, L. F.,
and Grunthaner, F. J., Chem. Phys. Lett. 44, 225-231
(1976).
8.
Hochella, Jr, M. F., and Carim, A. H., Surf. Sci. 197,
L260-268(1988).
9.
Cumpson, P. J., Surf. Interface Analy. 31, 23-34
(2001).
10.
Hirashita, N. and Uchiyama, T., Bunseki Kagaku (in
Japanese), 43, 757-764 (1994).
342