Control of plasma-surface reactions
for next generation semiconductor devices
Tetsuya Tatsumi1,2
1
Semiconductor Technology Development Division, SBG, PDSG, Sony Corporation,
2
Department of Electronics and Electrical Engineering, Keio University
Abstract: Dry etching technologies for various devices have been continuously
improved for over 30 years as the most advanced application of the low
temperature plasmas. In this paper, a brief history of the development of dry
etching systems, current issues, and the future technologies for the quantitative
control of atomic layer reactions will be discussed.
Keywords: Dry etching, Semiconductor, CMOS
1. Introduction
According to “Moore’s law”, transistor size will be
shrunk to below several tenth nanometers. In
fabricating high performance ultra-large-scale
integrated (ULSI) devices, it is necessary to suppress
both the variation of the critical dimension (CD) of
the gate electrode and the degradation of the Si
substrate (dislocation of Si atoms in the Si – Si
network) to within several atomic layers.[1-3] The
length of gate electrodes (Lg) depends on the mask
profile, including line width roughness (LWR), etch
rate uniformity within a wafer and/or within a lot,
and the pattern (space) width. The damage to the Si
substrate occurs both during gate etching and the
etching of sidewall dielectric materials (see Fig. 1).
LWR
Poly-Si
SiO2 (Tox=1.4 nm)
Poly-Si
SiO2
Lg
Si-sub.
Sidewall
damage
Si-sub
Td
Si recess
CD variation
Fig. 1 Si recess
The “Si-recess” is a problem caused during etching
to fabricate gate electrodes. Although the gate oxide
remains when a highly selective etching process is
used, some damage occurs under the gate oxide and
the damaged Si is recessed during the post-wet
treatment, [4,5] as shown in Fig. 1. The change in
profile of source/drain region (near the gate
electrode) can be origin of fluctuations in transistor
properties.[6] Furthermore, when we apply the same
plasma process for gate etching, the depth of the “Sirecess” is not changed although the transistor size
will be smaller in the next generation devices and
beyond. Consequently, this damage will have more
serious impact on transistor properties in the near
future. In this report, we clarify the mechanism of
formation of the “Si-recess”.
2. Experiment
A dual frequency (60/13MHz) capacitively
coupled plasma (CCP) system was used to study Sirecess. The gas pressure was 60mTorr, the ion
energy was controlled using substrate bias power,
and Vpp was varied from 50 to 400 V. We estimated
the ion energy distribution function (IEDF) using a
Monte Carlo simulation. Penetration depths of ions
were estimated using a molecular dynamics
simulation. A gate SiO2 layer (1.4 – 300 nm) was
formed on the Si substrate and exposed to HBr/O2
and O2 plasma. The damage (degradation of Si–Si
network) occurred below the gate SiO2. This damage
was analyzed by spectroscopic ellipsometry (SE),
high-resolution
Rutherford
backscattering
spectroscopy (HRBS), and transmission electron
microscopy (TEM).
4. Discussions
3.1 Degradation of Si substrate
4.1 Model for damage formation
In a sample with a thin gate oxide (Tox =1.4 nm), we
observed thick damage after HBr/O2 plasma
exposure. Figure 2 shows the damage thickness (Td)
under the gate oxide as a function of HBr/O2 plasma
exposure time. As clear in the figure, Td gradually
increased to about 10 nm at 600 s. However, with O2
plasma, Td did not depend on time and was less than
3 nm. The use of HBr plasma results in thicker and
more significant damage.
The results indicate that the Si substrate (under the
gate oxide film) was affected by both: (i)
degradation of the Si-Si network by H ions (from
HBr), and (ii) diffusion of O radical supplied to the
degraded layer. A model of this is illustrated in Fig.
4. The Td is assumed to depend on the O diffusion
time and to be limited by the H ion penetration depth.
Damage thickness, Td (nm)
3. Results
10
Exposure time
60 mT, Vpp =400 V
HBr/O2,
Tox=1.4 nm
5
O2 , Tox =1.4 nm
HBr/O2, T ox=3.0 nm
0
Fig. 4 Model for ion assisted diffusion of O radicals
HBr/O2, T ox=300 nm
400
200
0
600
Time (s)
Fig. 2 Damage thickness as a function of etch time
3.2 Surface analysis
Depth from the surface (nm)
Figure 2 shows the TEM images and depth profiles
of HRBS spectra for the Si substrates exposed to
HBr/O2 plasma (Vpp=400eV, 600s). Oxygen is
diffused deeper than 10 nm along with the dislocated
Si generated by the hydrogen penetration.
0
0
atomic%
4.2 Calculation of H penetration depth
To quantitatively predict the ion penetration through
the gate oxide and the Si degradation, we developed
a series of new potential functions for molecular
dynamics (MD) simulation for Si, O, C, F, and H
systems. Based on the Stillinger-Weber potential
[7,8], we prepared three-dimensional potential
functions and adjusted them to the first-principal
results using Gaussian98. We fabricated a large (2 x
2 x 20 nm) sample structure of a Si substrate topped
by a SiO2 layer (1.4 nm).
Af ter dose of 1000atoms
100
Ion energy (eV)
Br
4
O
Oxidized layer
O
Dislocated
Si
Oxidized
layer
5 nm
16
50
100
300
Rp: 0.5 nm
2 nm
5 nm
10 nm
5
8
12
10
0
HRBS
TEM
10
15 nm
Fig. 3 HRBS and TEM results
Fig. 5 MD calculation of H penetration
Hydrogen atoms with energies of 10-300 eV were
dosed onto the substrate (Fig. 5). As expected, H
penetrated deeply into the Si layer (more than 10
nm), and part of the Si-Si bond degraded.
4.2 Analysis using multi-beam injection system
Reaction
Chamber
Beam line
Mass Selector
TMP Deflector
CP
TMP
QMS
Analyzing
Magnet
FC1
Einzel Lens
Decelerator
Fig. 7 H ion injection
ii) Ion assisted diffusion of O radical
Next we simultaneously injected O radical with Ar+
and H+ ion beam as shown in Fig.8. It has been
found that oxygen (O) diffusion is enhanced in the
alteration layer due to amorphization of Si by H+
beam.
100
80
60
40
20
Composition [%]
To confirm the damage formation model, the author
and his collaborators used a multi-beam injection
system [9]. The multi-beam system consists of three
parts, i.e., a mass analyzed ion beam injector, a set
of two independently controllable neutral
radical/molecular beam injectors, and a reaction
chamber in which a sample substrate can be placed
(Fig.6). In this system, a monochromatic and monoenergetic ion beam as well as independently
controlled radical/molecular beams can be
simultaneously injected into a given substrate
surface. The ion and radical sources are
differentially pumped and therefore the chamber can
be maintained at ultra-high vacuum. Using this
system, Si surfaces were irradiated by H+, or Ar+ ion
beam at 500eV each as well as atomic oxygen (O)
radical beams. [10, 11]
0
100
80
60
40
20
0
100
80
TMP
25 KV
60
40
Ar, CH4
20
Ion Source
0
0
Fig. 6 Multi-beam system
2
4 6 8
Depth [nm]
10
Fig. 8 Multi-beam injection
i) Si-dislocation formed by H ion beam
Figure 7 shows the Si dislocation formed by H+
injunctions under energy of 500eV. Dislocation was
increased by increasing dosage of H+ ion and the
depth of dislocation depended on ion energy.
Our multi-beam injection experiments corroborates
the hypothesis that the Si recess during HBr/O2
plasma etching processes is caused by H+ ion
injections from HBr plasmas and O radical diffusion.
4.3 Suppression of “Si recess”
An ion energy distribution function (IEDF) in the
CCP system was calculated using Monte Carlo
simulation (Fig.9a). Through MD simulation, we
found that it was necessary to maintain the ion
energy at a level below 50 eV in order to suppress
the O penetration through the 1.4-nm gate oxide. By
controlling the high energy peak of IEDF below this
threshold energy level, we were able to successfully
suppress the formation of the Si recess (Fig. 9b). [7]
a)
Counts (arb. unit)
25000
IEDF (O 2+)
Vpp=50 Ｖ
20000
400 V
5000
0
50
100
150
200
250
300
Energy (eV)
b)
15
HBr/O2, 60 mT
Tox=1.4 nm
Td (nm)
The author would like to thank Prof. Satoshi
Hamaguchi, Prof. Kazuhiro Krahashi, graduate
student Ms. Tomoko Ito of Osaka University for the
corroboration on the surface analysis using the MD
simulation and the beam experiment system. The
author also acknowledges Mr. Masanaga Fukasawa,
Dr. Shoji Kobayashi and Mr. Tomokazu Ohchi for
their support on the experiments.
References
270 V
10000
Vpp= 400 V
10
270 V
160 V
5
50 V
0
6. Acknowledgements
160 V
15000
0
resulting in thick oxidation of the layer.
By
lowering the high energy peak of IEDF the Si recess
was successfully minimized.
0
400
800
Time (s)
Fig. 9 Relationship between IEDF (a) and damage thickness (b)
Thus, it is necessary to clarify the penetration depth
and to control the ion energy distribution
quantitatively to ensure stable device fabrication in
mass production.
5. Conclusions
The mechanism of the Si recess that forms during
gate poly Si etching was studied. Hydrogen in HBr
plasma penetrates through the thin gate oxide film
and induces deep degradation (>10 nm @ 400 eV) in
the Si substrate. Simultaneously injected oxygen is
diffused to the damaged layer caused by H+ ions,
[1] A. Ando et al., Thin Solid Films 515 (2007) 4928.
[2] Y. Nakamura et al., J. Vac. Sci. Technol. A25(4),
(2007) 1062.
[3] M. Matsui et al., J. Appl. Phys. Vol. 38 (1999) 2124.
[4] S. A. Vitale et al., J. Vac. Sci. Technol. B21(5),
(2003) 2205.
[5] T. Ohchi, et al., Jpn. J. Appl. Phys. 47 (2008)
5324.
[6] Y. Nakakubo et al., Jpn. J. Appl. Phys. 49 (2010)
08JD02.
[7] H. Ohta et al., J. Chem. Phys. 115 (2001) 6679.
[8] F. H. Stillinger et al., Phys. Rev. B31 (1985) 526.
[9] K. Karahashi et al., J. Vac. Sci. Technol. A22
(2004) 1166.
[10] T. Ito et al., Proc. of 32nd Dry Process Symp.,
Tokyo, J-2 (2010) 187.
[11] T. Ito et al., Jpn. J. Appl. Phys (submitted).