Beruflich Dokumente
Kultur Dokumente
892897
Fabrication and Process Simulation of SOI MOSFETs with
a 30-nm Gate Length
Won-ju Cho,
E-mail: chowj@etri.re.kr
cess simulation and a device characteristics simulation
for the various dimensions of the SOI MOSFETs. A
simple solid-phase diusion (SPD) technique to form an
ultra-shallow and low-resistance source/drain junction
was developed. Based on the optimized device structure
and fabrication processes, we successfully fabricated a
30-nm-gate-length SOI nMOSFET.
II. EXPERIMENT
The SOI nMOSFETs used in this study were fabri-
cated on p-type (100) UNIBOND SOI wafers with a 100-
nm top silicon layer and a 200-nm buried oxide. The pro-
cess conditions and two-dimensional (2-D) device char-
acteristics were simulated using ATHENA (Silvaco
TM
)
and ATLAS (Silvaco
TM
), respectively, before the fabri-
cation of 30-nm-gate-length SOI MOSFETs. Boron-ion
implantation at 10 keV was carried out to introduce an
impurity into the channel region. Then, the top sili-
con layer was thinned to 30 nm by thermal oxidation.
Mesa-isolation, which is the most widely used isolation
technique for thick lms as well as ultra-thin-lm SOI
MOSFETs due to the advantages of high integration den-
sities and process simplicity, was adopted in this study
to electrically isolate SOI devices. The active region was
dened in the top silicon layer by using a combination of
-892-
Fabrication and Process Simulation of SOI MOSFETs Won-ju Cho et al. -893-
photolithography and reactive ion etching (RIE).
The etching process started with CF
4
-based RIE (iso-
topic etching) for smoothing the sharp corners of the
silicon active edge in order to prevent the subthresh-
old hump phenomena and ended with Cl
2
-based RIE
(anisotropic etching) to obtain a vertical sidewall pro-
le. A phosphorus-doped 150-nm polysilicon lm was
deposited on the 4.5-nm gate oxide as a gate electrode by
using low-pressure chemical-vapor deposition (LPCVD),
and a 40-nm silicon-dioxide hard mask was formed on
the polysilicon gate by using thermal oxidation. SAL
601, a negative, chemically amplied resist, was used
as an electron-beam (e-beam) resist to form the 30-nm
gate line. The etching of the gate electrode was done
by using Cl
2
-based RIE to obtain a vertical prole. The
phosphorus-doped oxide lms, as solid-phase diusion
(SPD) sources, were usually prepared by using chemical-
vapor deposition (CVD), but a liquid-state dopant source
containing silicon, oxygen, and phosphorus atoms was
used in this study. Some wafers were coated with a
dopant source by spin coating after the RCA standard
surface-cleaning processes. A liquid-type dopant source
was changed to a solid-state phosphorus-doped silicon-
oxide layer on the top of the wafer by using a baking
process.
A rapid thermal annealing (RTA) system was used
to diuse the phosphorus atoms into the silicon [11].
Other wafers were doped by phosphorus-plasma-doping
method (PLAD) [12], and then activated by using the
RTA system. Secondary-ion mass spectroscopy (SIMS)
was utilized to analyze the depth proles of the phospho-
rus atoms for various RTA process temperatures. After
the formation of interlayer dielectric (ILD) lms by using
undoped silica-glass (USG) lms and metal interconnec-
tions, n
+
-p junction diodes and 30-nm-gate-length SOI
nMOSFETs were fabricated. Finally, the electrical char-
acteristics were measured using the fabricated devices.
III. RESULTS AND DISCUSSION
1. Process and Device Simulation
In order to investigate the nano-scale SOI MOSFET
device physics and characteristics, we performed an ex-
tensive two-dimensional process simulation and device
simulation by using ATHENA and ATLAS, respectively.
Figure 1 show the ATLAS simulation results for the sub-
threshold Id-Vg curves, the subthreshold swing, and the
threshold voltage of SOI MOSFETs as functions of the
gate lengths for various channel thicknesses. The chan-
nel doping concentration was 1 10
15
cm
3
in these
simulations. As the gate length was decreased to 20
nm, the short-channel eects became serious because
the roll-o of threshold voltage and the subthreshold
swing increased. On the other hand, thinner SOI channel
was more eective for suppress the short-channel eects.
Fig. 1. Simulation results for the (a) subthreshold I
ds
-
V
g
curves, the (b) subthreshold voltage swing, and the (c)
threshold voltage of SOI MOSFETs as functions of the gate
lengths for various channel thicknesses.
However, because control of top silicon layer and of the
threshold voltage of SOI MOSFETs is very sensitive to
silicon-layer thickness variations, channel doping is es-
sential to suppress the short-channel eects in nano-scale
SOI MOSFETs.
Figure 2 shows the ATLAS simulation results for the
subthreshold I
ds
-V
g
curves for the 30-nm-gate-length
SOI nMOSFET as a function of the channel doping con-
centration. The channel thickness in this simulation was
xed at 30 nm. A channel doping concentrations of
3 10
18
cm
3
was found to be required to suppress
-894- Journal of the Korean Physical Society, Vol. 43, No. 5, November 2003
Fig. 2. Simulation results for the subthreshold I
ds
-V
g
curves for the 30-nm-gate-length SOI nMOSFET for various
channel doping concentrations.
Fig. 3. Schematic cross-section of a 30-nm-gate SOI MOS-
FETs and impurity proles along the channel as simulated
by the ATHENA process simulator. The source/drain doping
was performed by using the phosphorus solid-phase diusion
technique.
the short-channel eects on the characteristics of 30-nm-
gate-length SOI MOSFETs.
Figure 3 shows the process simulation results obtained
by using the ATHENA process simulator. A schematic
cross-section view of the 30-nm-gate-length SOI MOS-
FETs is shown in Fig. 3(a). The source/drain junction
was formed by using SPD and phosphorus-doped oxide
lm because the SPD technique is known to be an eec-
tive method for fabricating ultra-shallow and defect-free
junctions. The gate oxide thickness, the channel thick-
ness, and the channel doping concentration were xed at
5 nm, 30 nm, and 3 10
18
cm
3
, respectively. Figure
3(b) shows impurity proles along the channel surfaces
of SOI MOSFETs. A sucient phosphorus concentra-
tion at the source/drain extension region and an eec-
tive channel length of 20 nm were obtained from the
optimized process simulation.
Figure 4 shows the ATLAS device simulation results
for the 30-nm-gate-length SOI nMOSFETs shown in Fig.
Fig. 4. I
ds
-V
g
and I
ds
-V
d
characteristics of the simulated
30-nm-gate-length SOI MOSFETs using a phosphorus SPD
source/drain region.
3. Good I-V characteristics were obtained as shown in
Fig. 4(a) and Fig. 4(b). The curves of the drain current
versus the gate voltage at drain biases of 0.1 V and 1 V
are shown in Fig. 4(a). The threshold voltage is 0.28 V,
and the subthreshold slope is about 120 mV/dec. Fig-
ure 4(b) shows the simulated output characteristics of
30-nm-gate-length SOI nMOSFETs for several gate bi-
ases. The output current of the device is 750 A/m at
V
gs
-V
th
= 1.2 V and V
ds
= 1.2 V. Using these results
from the process and the device simulations, we opti-
mized the device structure and the process conditions
for fabricating 30-nm-gate-length MOSFETs on ultra-
thin SOI substrates and estimated the characteristics of
device operation.
2. Device Fabrication
According to the process and device characteristics
simulations, we performed a boron-ion implantation with
a dose of 2 10
13
cm
2
, at an acceleration energy of 10
keV, and a tilted angle of 7