New infrared spectroscopic ellipsometer for low-k dielectric
characterization
P. Boher9 ML Bucchia, C. Guillotin, C. Defranoux and J. C. Fouere
SOPRA, 26 rue Pierre Joigneaux, 92270 Bois Colombes, France
Abstract. A new infrared spectroscopic ellipsometer devoted to the characterization of silicon microelectronics has
been recently developed at SOPRA. Its main features are the ability to measure on a small spot ( 80x20Qjim) with a high
signal/noise ratio. A original patented optical design suppress back face reflection and insure good quality spectral
measurements in the 600-7000cm~1 range. Excellent signal/noise ratio allows to perform measurements in less than 30s.
All automation and real time analysis are included to offer an operator orientated metrology tool. Hereafter, the
instrument is presented in details, and used to characterize different kinds of low-k dielectrics.
available through the refraction index according to k =
n2 [3]. The experiment shows that the low frequency
value of K generally determined from capacitance
measurements of MIS structures, and representing
both the electronic and nuclear components, can be
deduced. Indeed, for many low-k materials under
consideration, the nuclear component is small relative
to the electronic part of the response. Finally the effect
of the porosity on dielectric constant can be easily
predicted using the Bruggemann effective medium
approximation [4]. So, standard optical models can be
used to deduce the material porosity in an absolute
way [5]. Conventional spectroscopic ellipsometry has
already been used for this purpose [6].
INTRODUCTION
The future scale-down of integrated micro
electronic devices will lead to a need for low dielectric
constant replacement
for silica in device
interconnection array [1]. The replacement material
must attain a relative dielectric constant k lower than
SiO2 (k = 4) while maintaining adequate thermal and
mechanical stability during and after fabrication
process. Generally the types of chemical structures that
impact structural stability are those having strong
individual bonds and a high density of such bonds.
However, the strongest bonds are often the most
polarisable and so polarization increases with the
dielectric constant. In order to reduce the k value
relative to that of SiO2? it is necessary to either
incorporate atoms and bonds that have a lower
polarizability or else to lower the density of atoms and
bonds in the material, or both.
In the following paper, we propose to use infrared
spectroscopic ellipsometry (IRSE) for low-k material
characterization. In contrast with FTIR, IRSE is an
absolute technique which provides independently the
refractive index and absorption coefficient without
need of Kramers Kronig calculations. SOPRA has
already proposed a Fourier Transform Infrared
Spectroscopic Ellipsometer (IRSE) for research and
development in 1993 [7]. For ULSI manufacturing an
automated equipment is required with small measuring
spot, high throughput and operator mode operation. In
this paper, we present a new tool called IRSE300
which fulfill all these requirements. Some
experimental results on low-k dielectric are also
presented.
Optical methods can provide many valuable
information on this kind of material. Individual bonds
have generally a signature in the infrared region and
so, the amplitude of their absorption peaks is a direct
information on the density of bonds. Fourier transform
infrared spectroscopic FTIR has intensively been used
for this type of analysis [2], but its main drawback is
the impossibility to deduce directly the dielectric
constants of the materials. The high frequency value of
K reflecting the electronic component, can be directly
CP683, Characterization and Metrology for ULSI 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
540
EXPERIMENTAL SETUP
Spot size measurement
IRSE300 is an industrial metrology tool especially
devoted to 300mm silicon wafer technology. The
infrared setup can be seen as an option on the UVvisible tool developed in 1997 [8] and successfully
qualified during the gauge study performed at the
International 300mm Initiative (I300I) [9].
The spot size has been measured accurately using
a SiO2(2^im)/Si box on bare silicon. The spot is
scanned along and perpendicular to the incidence
plane by steps of 20 and 5jim and a measurement is
made at each position. The characteristic spectra of
bare silicon and SiO2(2jLim)/Si are very different due to
the occurrence of very well defined interference
fringes for the SiO2 layer. The intermediate region
where the measurement is a mixing of the contribution
of the two regions can be easily evaluated by making
an adjustment on the experimental curves. The error
between experiment and simulation provides a good
evaluation of the intermediate region (cf. figure 2).
The lateral (85±5|nm) and longitudinal (200±20jim)
spot sizes have been measured.
Optical setup
The design of the optical system of IRSE300 has
been guided by three main major points: (i) reduce the
spot size on the sample, (ii) suppress the parasitic
reflection coming from the rear surface of the silicon
wafer and (iii) keep a maximum light flux on the
detector to maximize the signal to noise ratio and
increase measurement speed.
80pm < Lateral size < SOpm
We use a commercial Fourier transform
interferometer (FTIR) with a globar source. The light
beam coming out of the FTIR is focused on a first
aperture slit using parabolic and flat mirror
combination. The image of the slit is then magnified
on the surface of the wafer using an ellipsoidal mirror.
A second ellipsoidal mirror collects the reflected light
from the sample surface and images of the front face
and rear face of the silicon wafer are separated by a
detector slit after reflection on a folding mirror (cf.
figure 1 and SOPRA patent [10]). The angle value and
the angular aperture are selected by a second slit
before the detector. Grid polarisers mounted on
stepping motors are used to fix the polarization before
the sample and to analyze the variation of polarization
after reflection. The light beam is finally focused on a
0.25x0.25 mm nitrogen cooled MCT detector. The
system includes robotic wafer alignment and pattern
recognition.
0.5
-200
200
knife
600pm
Source
400
Figure 1. Removal of back face reflection (Sopra patent)
Figure 2. Spot size measurement along and perpendicular to
the incidence plane. The edge of 2jim thick SiO2 box on
silicon is used. The intermediate region increases the error
function between experiment and simulation.
541
interference fringes.
The intensity of the
interferometric signal versus the sample height is
reported in Figure 4 with and without the detector slit.
The rear face reflection is easily detected when the
detector slit is used and clearly separated from the
front face reflection. So, if the sample is at the right Z
position the rear face reflection is easily removed. We
have verified this point making different spectra at
various Z positions. When the measurement is made
with the Z position at the maximum of signal for the
front face reflection (8.83mm), no interference fringe
is detected. Interference fringes coming from the rear
face reflection appear only if the Z position is increase
above 8.85mm (cf. figure 4).
Measurement speed
We have made different series of measurements on
a 2.53|im SiO2/Si sample to check the repeatability of
the system for various acquisition conditions. The
system is working at fixed polarizer position and the
acquisitions at four analyzer positions are necessary to
get an ellipsometric spectrum. To increase the statistic,
one simple way is to average each acquisition on a
given number of interferometer scans. On figure 3, we
have reported the standard deviation (1 cr) obtained on
30 identical measurements versus the measurement
time. The acquisitions are more or less averaged on 1,
4, 16 and 64 scans of the interferometer. As expected
the repeatability is improved with the average in a
normal way (a factor of \N for N acquisitions). This
ratio is verified for 4 and 16 acquisitions.
Nevertheless, 64 acquisitions do not improve the
repeatability probably because of long term signal
fluctuations. Optimum acquisition conditions are
obtained for about 40s measurement time and the
repeatability is excellent ( 0.34nm for a thickness
value of 2.53jam).
3
-0-with slit
8.8
Z(mm)
1.4
>
Front face reflection
2.5
1.8
1.6-
0)
,—*— without slit
Figure 4. The detector slit removes the rear face
reflection if the sample is at the right Z position.
1
0 0.8
I0'6
0.4
0.2
0
EXPERIMENTAL RESULTS
Figure 3. Standard deviation (1 a) on the thickness of
a SiO2(2.5|^m)/Si sample for various acquisition
conditions
Two kinds of low-k materials are analyzed
hereafter. First carbon incorporate silicon oxides are
explored and we show that carbon content and layer
porosity can be deduced from IRSE measurements.
Then hydrogen silesquioxane system is analyzed and
the thermal behavior of the films is put in evidence.
Removal of back face reflection
Carbon incorporate silicon oxides (SiOC)
To verify that the rear face reflection is really
suppressed when the sample Z positioning is correctly
adjusted, we have measured a 8 inches, TOOjum thick
sample in the following conditions. The wafer is
double side polished to enhance the contribution of the
rear face reflection. A thick SiO2 coating is deposited
on one side of the wafer and the measurement is made
with this face down so that the rear face reflection can
be easily characterized by the occurrence of some
Silicon oxide-based low K dielectric materials
containing alkyl groups have attracted much attention
due to their high thermal and mechanical stability. SiO-C-H films deposited by plasma enhanced chemical
vapor deposition (PECVD) have shown the formation
of nano-sized voids due to Si-CH3 and OH related
bonds included in the film [11]. These features are
controlling the dielectric properties of the films and
must then be measured accurately. As shown in figure
50
100
Measurement Time (s)
542
5, different absorption peaks can be detected in the
infrared region in addition to the O-Si-O bounds
classically detected on silica. A well defined peak
around 1270cm"1 associated to Si-CHs bounds is also
measured and two other peaks at 2350cm"1 and
3 000cm- 1 associated to the water absorbed by the
films. We have used these measurements to extract the
optical indices of different samples has shown in
Figure 6. The relative concentration of carbon
incorporated in the film can be calculated by
normalizing to the area of the absorption peak
corresponding to the O-Si-O stretch using:
amplitude of the O-Si-O absorption peak as shown in
figure 8 for another series of samples.
Sam pie #1
2
Sam pie #2
Sam pie #3
1.5
.*
*
c
1 -I
0.5
where A0 and Ac are the peak areas of the Si-O
stretching vibration mode at 1040cm"1 and the Si-CH3
vibration mode at 1270cm"1 respectively. A summary
of the results for the samples of figures 5 and 6 is
reported in Table 1.
O-Si-O
950
1050
1150
1250
Wavenumber (cm-1)
1350
Figure 6. Measured optical indices of the Si-O-C-H films of
figure 5. The thickness of the films is between 300 and
500nm.
Sample 1
Sam pie 2
Sample 3
0.5
H20
Q
0
3
o
-0.5
-1
600
1600
2600
Wavenumber (cm-1)
Figure 5. SE measurements on three Si-O-H-C low K
dielectric with variable C content. Different absorption
bounds are easily detected.
fl)
Table 1: Carbon content of the Si-O-C-H low K
dielectrics of figs. 3 &4.
Sample 1
Sample 2
Sample 3
Ao
119.4
132.5
150.7
Ac
8.3
10.4
15.6
Carbon (%)
6.5
7.3
9.4
This type of analysis can be made automatically on the
entire surface of a silicon wafer. Such kind of analysis
has been realized on a 200mm wafer as reported in
figure 7. The layer thickness and carbon concentration
are deduced independently with the method explain
above. There are not correlated in this example. The
porosity of this kind of film can be also related to the
number of Si-O bounds in the material and so to the
Figure 7. high resolution mapping of a SiOGH film on
200mm silicon wafer. Scales : 102-115nm for the thickness
and 4-14 (a.u.) for the Si-CH3 peak amplitude.
543
100
600
O-Si-O peak amplitude
1600
2600
3600
4600
5600
Wavenumber (cm-1)
Figure 8. Film porosity versus O-Si-O peak amplitude.
Figure 9. Experimental and simulated ellipsometric spectra
of a HSSQ(0.5um)/Si sample.
Low-k hydrogen silesquioxane (HSSQ)
Commercial hydrogen silsesquioxane (HSSQ)
solutions (FOX®, Dow Corning Corporation [12]) has
shown interesting properties for low dielectric
constants (k = 2.5-3.3) properties [13]. An infrared
ellipsometric spectra for a 0.5jim thick film is shown
in figure 9. In addition to the interference fringe
characteristic of the layer thickness, different
absorption peaks are identified including Si-H
stretching mode (at 2250cm"1), Si-O stretching mode
(around 1100cm"1) and various Si-O-H bending modes
(< 1000cm"1). The optical index of the HSSQ layer is
build using a dispersion law model with different
Lorentz peaks. Adjustment of figure 9 is made on the
different peak parameters and on the layer thickness.
The precursor HSSG oligomers were predominantly
cubic T8 cage-like structures [14]. It is why the Si-O
stretching mode is spitted in one mode at 1140cm"1 for
O in T8 cage structure, and one mode at 1070cm"1 for
Si tetrahedrally coordinated with O in the Si-O like
network as shown in figure 10 where the extracted
optical indices are reported.
600
1100
1600
2100
2600
Wavenumber (cm-1)
Figure 10. Extracted optical indices of the HSSQ layer of
figure 9.
Thermal annealing increases the Si-O network
stretch at the expense of the Si-O cage stretch and the
ratio of the absorption peaks changes as shown in
figure 12. This evolution is especially important
because it controls the mechanical and electrical
properties of the layers.
544
ACKNOWLEDGMENTS
As deposited
Part of this work is supported by the French
ministry of industry through an European MEDEA+
project (DIAMANT, T304)
After annealing
REFERENCES
600
1. P.S. Ho, J. Leu, W.W. Lee, "Low dielectric constant
materials for 1C applications", Springer Verlag, Berlin,
2003.
1100 1600 2100 2600
Wavenumber (cm-1)
2. A.R. Garton, "Infrared spectroscopy of polymer blends,
composites and surfaces", Oxford Univ.Press, New
York, 279, 1992.
3. G.R. Fowles, "Introduction to modern optics", Holt,
Rinehart & Winston, New York, 1968
-As deposited
-After annealing
4. R.M.A Azzam, N.M. Bashara, "Ellipsometry and polarize
light", Elsevier, Amsterdam, 1977
5. C.V. Nguyen, R.B. Byers, C.J. Fawker, J.L. Hendrick,
Polymer preprints, 40, 261, 1995
6. S. Dietrich, A. Haase, Phys. Rep., 260, 1, 1995
600
1100 1600
7. G. Zalczer, O. Thomas, J.P. Piel and J.L. Sterile, Thin
Solid Films, 234 (1993) 356-362
2100 2600
Wavenumber (cm-1)
8. P. Boher, C. Defranoux, S. Bourtault, J.L. Sterile, SPIE
symposium, Santa Clara, 14-19 March, 1999
Figure 11. Optical indices of HSSQ films after deposition
and after thermal annealing.
9. S. Wurm, "SOPRA SE-300 Gauge Study Report",
Technology Transfer 98033491-ENG, I300I, March,
1998
CONCLUSION
10. M. Lutmann, P. Boher, J.L. Stehle, "Ellipsometre a haute
resolution spatiale fonctionnant dans Pinfrarouge",
Patent N° 0009318, (2001)
A new infrared spectroscopic ellipsometer has been
presented in details. In addition to a small spot
(typically 200x85jam) which get ride of any problem
of rear face reflection on silicon substrate, this system
is capable to measure accurately and rapidly which
makes it suitable for industrial control in the
semiconductor industry. Two kinds of low-K
dielectrics have been investigated with this new
techniques and we have shown that chemical and
physical information on such kind of layer can be
extracted in addition to thickness information. Carbon
content and porosity of SiOC films are deduced and
structural information on HSSG versus thermal
annealing have been also demonstrated. We think
strongly that this new instrument will be used in a near
future to control various key fabrication stages of the
1C manufacturing
11. Y.H. Kirn, S.K. Lee, HJ.Kim, J. Vac. Sci. Technol.,
A18, 4, 1216(2000)
12. FOX® Flowable Oxide Technical Brief 2.2, Dow
Corning Corporation, 10, Midland, MI, 1998.
13. Y. Toivola, J. Thurn, R.F. Cook, J. of Electrochem. Soc.,
149,F9,2002
14. M.G. Albrecht, C. Blanchette, J. Electrochem. Soc., 145,
2861, 1998.
545