Laser-induced direct formation of C54 TiSi2 films with fine grains on c-Si substrates

Article abstract of DOI:10.1063/1.124801

In this letter, we report on the direct synthesis of C54 TiSi₂ films with fine grains by pulsed-laser irradiation from Ti deposited on Si substrates, using a Q-switched Nd:YAG laser. The films were characterized using micro-Raman spectroscopy, high-resolution transmission electron microscopy, and atomic force microscopy. In comparison with the C54 TiSi₂ using the conventional rapid thermal annealing (RTA) of 35 nm thick Ti/Si, which has an average grain size of about 110 nm and film thickness of 50 nm, the laser-induced C54 TiSi₂ films vary from 13 to about 42 nm in thickness with different laser scanning speed and the grain size is 85 nm on average. The TiSi₂/substrate Si interface is smooth on the atomic scale. Our results demonstrate the unique advantages of the laser-induced formation technique and its potential in deep submicron semiconductor technology. We propose that the C54 phase is formed by solid-state diffusion, rather than melting.

Full text of DOI:10.1063/1.124801

Laser-induced direct formation of C54 TiSi 2 films with fine grains on c- Si substrates
S. Y. Chen, Z. X. Shen, Z. D. Chen, L. H. Chan, and A. K. See
Citation: Applied Physics Letters 75, 1727 (1999); doi: 10.1063/1.124801
View online: http://dx.doi.org/10.1063/1.124801
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/75/12?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Sequence of Mg segregation, grain growth, and interfacial MgO formation in Cu–Mg alloy films on SiO 2 during
vacuum annealing
J. Appl. Phys. 93, 5966 (2003); 10.1063/1.1566451
Effective enhancement of C54 TiSi 2 phase formation with multi-thermal-shock processing at 600°C
Appl. Phys. Lett. 79, 1139 (2001); 10.1063/1.1382630
Nucleation and growth of C54 grains into C49 TiSi 2 thin films monitored by micro-Raman imaging
J. Appl. Phys. 88, 7013 (2000); 10.1063/1.1326469
Effect of interfacial silicon on the structural stability of C54–TiSi 2 on SiO 2
Appl. Phys. Lett. 76, 3697 (2000); 10.1063/1.126753
Mechanism of enhanced formation of C54–TiSi 2 in high-temperature deposited Ti thin films on preamorphized
(001)Si
Appl. Phys. Lett. 74, 224 (1999); 10.1063/1.123300
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:
152.23.46.217 On: Sat, 06 Dec 2014 04:18:24

APPLIED PHYSICS LETTERS
VOLUME 75, NUMBER 12
20 SEPTEMBER 1999
Laser-induced direct formation of C54 TiSi2 films with fine grains
on c-Si substrates
S. Y. Chena) and Z. X. Shen
Department of Physics, National University of Singapore, Lower Kent Ridge Road, Singapore 119260
Z. D. Chen
Process Technology Division, Gintic Institute of Manufacturing Technology, 71 Nanyang Drive,
Singapore 638075
L. H. Chan and A. K. See
Chartered Semiconductor Manufacturing, Ltd., 60 Woodlands Industrial Park D, Street 2, Singapore 738406
͑Received 21 December 1998; accepted for publication 21 July 1999͒
In this letter, we report on the direct synthesis of C54 TiSi2 films with fine grains by pulsed-laser
irradiation from Ti deposited on Si substrates, using a Q-switched Nd:YAG laser. The films were
characterized using micro-Raman spectroscopy, high-resolution transmission electron microscopy,
and atomic force microscopy. In comparison with the C54 TiSi2 using the conventional rapid
thermal annealing ͑RTA͒ of 35 nm thick Ti/Si, which has an average grain size of about 110 nm and
film thickness of 50 nm, the laser-induced C54 TiSi2 films vary from 13 to about 42 nm in thickness
with different laser scanning speed and the grain size is 85 nm on average. The TiSi2/substrate Si
interface is smooth on the atomic scale. Our results demonstrate the unique advantages of the
laser-induced formation technique and its potential in deep submicron semiconductor technology.
We propose that the C54 phase is formed by solid-state diffusion, rather than melting. © 1999
American Institute of Physics. ͓S0003-6951͑99͒03838-3͔
As the density of microelectronic devices increases, low
resistivity materials become more desirable for intercon-
nects, especially in applications where the critical dimen-
sions of transistors are less than 0.25 ␮m. Titanium disilicide
(TiSi2) is considered to be one of the most promising mate-
rials as contacts and interconnects for source, drain, and gate
electrodes in complementary metal–oxide–semiconductor
devices.1 TiSi2 has already been extensively applied in ultra
large scale integration due to its many excellent characteris-
tics such as low electric resistivity, high adhesion to Si sub-
strate, and good thermal stability.2 TiSi2 can exist in two
forms: the C49 phase and the C54 phase. The C49 TiSi2 is a
high resistivity ͑60–70 ␮⍀ cm͒ metastable phase formed be-
tween 550 and 700 °C, while C54 TiSi2 is a stable phase with
low resistivity ͑15–20 ␮⍀ cm͒ transformed from C49 TiSi2
at 700–850 °C.3 Currently, a two-step rapid thermal anneal-
ing ͑RTA͒ process is used for the silicidation of Ti on Si
substrates. The C49 phase is first formed by low temperature
annealing. After chemical etching of the unreacted Ti and a
second annealing at higher temperature, transformation to
C54 phase takes place. However, such transformation using
RTA is extremely difficult in narrow lines4 and thin layers5
because C49 TiSi2 grains are relatively too large to have
sufficient triple points of grain boundaries which act as the
nucleation sites for C54 TiSi2.6 To make TiSi2 extendable
for sub-0.25 ␮m technology, intensive efforts have been con-
centrated on increasing the density of nucleation sites, i.e.,
reducing the C49 TiSi2 grain size, with moderate success.
The various methods that have been attempted include
preamorphization of Si by Xe,7 and As,8 and implantation
through metal.9 All the above methods succeeded in lower-
ing the transformation temperature by 50–100 °C. In addi-
tion, thin silicide films of around 30 nm are also preferred in
deep subquarter micron technology, where shallow junction
͑Ͻ100 nm͒ is applied. However, it is difficult to control the
silicide layer thickness using conventional RTA. Reducing
the Ti layer thickness is not an adequate approach, since it
makes the transformation to C54 phase more difficult, as
mentioned above.
Lasers have been utilized to induce silicide
formation.10–12 By pulsed-laser melting and intermixing of
the metal and Si in liquid phase, a number of groups were
able to obtain C49 TiSi2, while no C54 phase has ever been
synthesized with pulsed-laser irradiation. Very recently,
Shamma et al.13 have succeeded in the formation of C49
TiSi2 on deep-submicron polysilicon gates, using pulsed ex-
cimer laser radiation. Followed by RTA at 850 °C for 20 s,
C54 phase was formed even on 0.1 ␮m polycrystalline sili-
con ͑poly-Si͒ lines with very smooth film/substrate inter-
faces. In their experiments, a layer of amorphous silicon ͑␣-
Si͒ was formed on the c-Si substrate by 75Asϩ implantation,
prior to the Ti deposition using the physical vapor deposition
͑PVD͒ technique. The presence of the ␣-Si layer is the domi-
nant factor responsible for the reduced interface roughness.
They suggested that the ␣-Si layer melted on laser irradiation
while the c-Si substrate did not, due to the large difference in
the melting point and thermal conductivity between them.
Therefore, the formation of TiSi2 is confined to the ␣-Si
layer. The authors also suggested that the grain size of the
C49 phase was smaller than that achievable using RTA, but
detailed results were not presented.
In this letter, laser annealing is applied to induce C54
TiSi2 formation, which has many advantages over RTA. For
example, with the laser annealing technique, energy is only
a͒
Electronic mail: scip7104@nus.edu.sg
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:
0003-6951/99/75(12)/1727/3/$15.00 1727 © 1999 American Institute of Physics
152.23.46.217 On: Sat, 06 Dec 2014 04:18:24

1728
Appl. Phys. Lett., Vol. 75, No. 12, 20 September 1999
Chen et al.
FIG. 2. High-resolution cross-sectional TEM image of a laser-annealed
sample. The lattice in the middle represents that of a typical C54 phase and
the lattice on the right is that of the Si substrate. Note that the silicide/Si
interface is very sharp and smooth. The stoichiometry of the silicide layer is
uniform and the thickness is measured to be 20 nm.
FIG. 1. Raman spectra of the as-deposited Ti film and laser irradiated
sample. The weak peak at 137 cmϪ1 represents the Ti phase. The three
peaks at 193, 201, and 242 cmϪ1 are due to the C54 TiSi2. For comparison,
Raman spectra of RTA C49 and C54 phases are also presented. Three peaks
at 169, 297, and 333 cmϪ1 are assigned to C49 TiSi2. The Raman spectra
verify the C54 phase formation on Ti/Si samples after laser irradiation.
can be assigned to the C54 TiSi2.5 Our Raman results con-
firm the formation of C54 phase with laser irradiation.
The high-resolution cross-sectional TEM micrograph
shown in Fig. 2 exhibits a typical C54 TiSi2 lattice. It is
obvious that the stoichiometry is very uniform across the
film. The image also shows that the TiSi2 /c-Si interface is
very sharp and smooth, and the transitional region is only a
few atomic layers. The thickness of the films measured by
TEM and AFM ranges between 13 and 42 nm with the scan-
ning speed between 1 and 0.1 mm/s ͑corresponding to expo-
sure times of 0.25 and 2.5 ms, respectively͒. The elemental
compositions of the layers shown in Fig. 2 are characterized
using an energy-dispersive spectrometer ͑EDS͒ attached to
the HRTEM system with the electron beam of 1 nm in di-
ameter and the results are presented in Fig. 3. Calculations
using the EDS data and theoretical k values can clearly iden-
tify the layers shown in Fig. 2 as unreacted Ti, TiSi2.06, and
Si substrates. The EDS data are in good agreement with the
Raman spectroscopy results.
The laser-induced C54 TiSi2 grain size is analyzed by
AFM. The samples were carefully etched using an
NH4OH:H2O2:H2Oϭ1:1:5 solution prior to the measure-
ments to completely remove unreacted Ti. Figure 4 shows
the AFM micrographs of the typical laser annealed and RTA
samples. The average grain size extracted from the AFM
micrographs is 85 nm for laser-induced samples and 110 nm
for RTA samples.
applied to the surface region,11 enabling shallow silicide for-
mation. Moreover, with its very fast ramping rate
(ϳ1010 K/s) and short laser pulse duration, this technique is
expected to be able to generate fine grains.
In our experiments, the Ti/Si samples were fabricated by
the standard PVD process, without preamorphization of the
c-Si substrates. The ͑100͒ oriented p-type c-Si wafers were
dipped into dilute HF to remove native oxide (SiO2) imme-
diately before being loaded into the rf sputtering system,
which was then evacuated to
a pressure below 5
ϫ10Ϫ9 Torr. High purity Ar was subsequently introduced
into the chamber to a pressure of 2 mTorr and Ti films of 35
nm in thickness were then sputter deposited on the Si sub-
strate. To produce the C54 TiSi2, the Ti/Si samples were
irradiated by an NEC Q-switched Nd:YAG laser ͑␭
ϭ1.06 ␮m, ␶FWHMϭ155 ns͒ with a repetition rate of 10 kHz,
the laser spot size of about 160 ␮m in diameter. The laser
has a near Gaussian shape with an average laser power of 1.9
W, which corresponds to a density of 0.93 J/cm2. A motor-
ized stage was utilized to move the Ti/Si samples and the
scanning speed varied from 0.1 to 1.0 mm/s. The experi-
ments were carried out in the air ambient.
The laser-annealed samples were characterized using a
number of techniques. A Renishaw System 2000 micro-
Raman spectrometer, with a 20 mW He–Ne laser (␭
ϭ632.8 nm) as the excitation source, was employed to iden-
tify the phases formed by the laser irradiation. Atomic force
microscopy ͑AFM͒ was utilized to investigate the grain size,
and a Philips CM300 high-resolution transmission electron
microscope ͑HRTEM͒ operated at 300 keV was applied to
study the structure and thickness of the film, as well as the
film/substrate interface.
Figure 1 displays the typical Raman spectrum of the
samples after laser annealing, without further thermal treat-
ment. For comparison, the Raman spectra of the as-deposited
Ti/Si sample, together with the RTA C54 and C49 TiSi2 are
also shown. The as-deposited sample has a weak Raman
peak at 137 cmϪ1, due to the first-order transverse optical
mode of the Ti film,14 which disappears in the Raman spec-
trum of the laser-annealed sample and three new peaks at
FIG. 3. EDS spectra of the layers shown in Fig. 2. The composition of the
layers can be identified as ͑bottom͒ unreacted Ti with a small amount of Si,
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:
193, 201, and 242 cmϪ1 are observed instead, all of which
͑middle͒ TiSi2.06, and ͑top͒ Si substrate.
152.23.46.217 On: Sat, 06 Dec 2014 04:18:24

Appl. Phys. Lett., Vol. 75, No. 12, 20 September 1999
Chen et al.
1729
FIG. 4. AFM image of the C54 TiSi2 phase produced by laser annealing
͑left͒ and RTA ͑right͒, the average grain size is measured to be 85 and 110
nm, respectively.
FIG. 5. Scanning speed ͑mm/s͒ dependence of silicide thickness.
We have performed computer simulation on the tempo-
ral distributions of temperature during laser irradiation. The
maximum temperature at the Ti–Si interface is about
1100 °C during laser treatment, which is well below the
melting points of Ti ͑1660 °C͒ and Si ͑1412 °C͒. The results
indicate that the solid-state process is responsible for the C54
TiSi2 formation in our experiments rather than melting. The
extremely fast ramping rate and the short duration of the
laser pulses, together with the fact that the samples do not
melt, make it possible to form fine grains and the atomic-
scale smooth interface. In this case, the short duration of the
pulses limits the growth of silicide grains.
The reaction kinetics has been examined in our experi-
ments. For the reaction-limited ͑RL͒ film formation process,
the film thickness is proportional to the annealing time, while
it is proportional to the square root of the annealing time for
the diffusion-limited ͑DL͒ process. In our experiments, the
film thickness is expected to be inversely proportional, re-
spectively, to the scanning speed and to its square root for
the two cases. In Fig. 5 the solid line is a least square fit
͑LSF͒ of the experimental data with the imposed condition
that the fitted line goes through the origin. The dashed line is
the result calculated from the diffusion-limited formation us-
ing the preexponential factor U0ϭ29 cm2/s and activation
energy Eaϭ1.8 eV.15 In this calculation, the film thickness
formed by one pulse was obtained by integrating the tem-
perature profile, which is obtained by computer simulation.
The calculated result is very close to the experimental data,
considering the fact that there are no adjustable parameters in
the calculation.
We have also attempted to fit the experimental data us-
ing the reaction-limited model. As there are no data on U0
and Ea available for TiSi2 in the reaction-limited case, we
have simply fitted the experimental data to a linear function
of time, which also has to go through the origin. The fitted
result is also plotted in Fig. 5 as the dotted line. Our results
clearly favor the diffusion-limited mechanism.
It has to be pointed out here that the laser-annealed
samples contain cracks of 1–10 ␮m spacing, which leads to
the unreliability of the electrical measurement data ͑not
shown͒. There are two types of stresses which may be re-
sponsible for the cracks:16 ͑1͒ the intrinsic stress due to the
26% shrink in volume after silicidation, which induces
stresses with the substrate and with the surrounding unsili-
cided area; and ͑2͒ the extrinsic stress caused by the differ-
ence in the thermal expansion coefficients between silicide
͑12.5 ppm/°C͒ and Si ͑3 ppm/°C͒. The elimination of the
cracks induced by the laser annealing will be a giant step
toward the realization of its application in semiconductor
manufacturing.
In summary, direct formation of C54 TiSi2 has been
achieved with a Q-switched Nd:YAG laser, using industrial
standard Ti/Si samples without preamorphization of the c-Si
substrate, via a solid-state diffusion mechanism. The C54
TiSi2 films formed are single phased and thin, with fine
grains and a smooth film/substrate interface on the atomic
scale. Laser-induced formation of titanium silicide and other
materials may find applications in the subquarter micron
technology.
1 R. T. Tung, Appl. Phys. Lett. 68, 1993 ͑1996͒.
2 V. Vraciun, D. Craciun, N. Chitica, I. N. Mihailescu, L. C. Nistor, A. L.
Popa, V. S. Teodorescu, I. Ursu, A. V. Kuzmichev, V. I. Konov, A. M.
Prokhorov, G. Leggieri, A. Luches, and M. Martino, J. Phys. D 22, 1500
͑1992͒.
3 A. Mouroux and S.-L. Zhang, Appl. Phys. Lett. 69, 975 ͑1996͒.
4 J. B. Lasky, J. S. Nakos, O. J. Cain, and P. J. Geiss, IEEE Trans. Electron
Devices ED-38, 262 ͑1991͒.
5 I. De Wolf, D. J. Howard, A. Lauwers, K. Maex, and H. E. Maes, Appl.
Phys. Lett. 70, 2262 ͑1997͒.
6 Z. Ma and L. H. Allen, Phys. Rev. B 49, 13501 ͑1994͒.
7 H. Kuwano, J. R. Philips, and J. W. Mayer, Appl. Phys. Lett. 56, 440
͑1990͒.
8 R. T. Tung, Mater. Res. Soc. Symp. Proc. 402, 101 ͑1996͒.
9 Y. H. Ku, S. K. Lee, D. K. Shih, D. L. Kong, C. O. Lee, and J. R.
Yeargain, Appl. Phys. Lett. 52, 877 ͑1988͒.
10 V. Cracium, D. Cracium, N. Chitica, I. N. Mihailescu, L. C. Nistor, A. L.
Popa, V. S. Teodorescu, I. Ursu, A. V. Kuzmichev, V. I. Konov, A. M.
Prokhorov, G. Leggieri, A. Luches, and M. Martinos, J. Phys. D 25, 1500
͑1992͒.
11 E. D’Anna, G. Leggieri, and A. Luches, Thin Solid Films 218, 95 ͑1992͒.
12 E. D’Anna, G. Leggieri, A. Luches, M. Martino, and A. Perrone, Appl.
Surf. Sci. 54, 353 ͑1992͒.
13 N. Shamma, S. Talwar, G. Verma, K.-J. Kramer, N. Farrar, C. Chi, W.
Greene, and K. Weiner, Mater. Res. Soc. Symp. Proc. 470, 265 ͑1997͒.
14 R. J. Nemanich, R. W. Fiordalice, and H. Jeon, IEEE J. Quantum Elec-
tron. 25, 997 ͑1989͒.
15 V. E. Borisenko and P. J. Hesketh, Rapid Thermal Processing of Semi-
conductors ͑Plenum, New York, 1997͒.
16 S. P. Murarka, Silicides for VLSI Applications ͑Academic, New York,
1983͒.
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:
152.23.46.217 On: Sat, 06 Dec 2014 04:18:24