Growth of TiSi2 from codeposited TiSix layers and interfacial layers

Article abstract of DOI:10.1063/1.118880

The silicide formation characteristics from composition profiles created on Si by codepositions was investigated. It was shown that a thin interfacial amorphous TiSi_x layer, with x～0.5-1, deposited between Si and the Ti film led to a significant reduction in the observed C49→C54 TiSi₂ transformation temperature. The presence of the amorphous interfacial TiSi_x layer slowed down the initial silicidation rate, but promoted the nucleation of the final C54 TiSi₂ phase. Predeposition and preannealing were also found to facilitate the growth of C54 TiSi₂, as was growth from codeposited full TiSi_x layers with Ti-rich compositions. The efficacy of the (interfacial) TiSi_x layer was demonstrated for different temperature ramp rates and for a variety of substrates including undoped α-Si, preamorphized n⁺-Si, and preamorphized p⁺-Si. But this effect was found to be absent on single crystal Si. Possible mechanisms of the observed effects were discussed.

Full text of DOI:10.1063/1.118880

Growth of TiSi 2 from codeposited TiSi x layers and interfacial layers
R. T. Tung, K. Fujii, K. Kikuta, S. Chikaki, and T. Kikkawa
Citation: Applied Physics Letters 70, 2386 (1997); doi: 10.1063/1.118880
View online: http://dx.doi.org/10.1063/1.118880
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/70/18?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Thermal stability and mechanical properties of amorphous coatings in the Ti-B-Si-Al-N system grown by cathodic
arc evaporation from TiB2, Ti33Al67, and Ti85Si15 cathodes
J. Vac. Sci. Technol. A 32, 061508 (2014); 10.1116/1.4897170
Epitaxial growth of VO2 by periodic annealing
Appl. Phys. Lett. 104, 063104 (2014); 10.1063/1.4864404
Growth kinetics of amorphous interlayers in Ti thin films on epitaxial Si–Ge layers on silicon and germanium
J. Appl. Phys. 89, 6110 (2001); 10.1063/1.1365439
Differences between interfacial and surface molybdenum in the formation of Ti Si 2
J. Appl. Phys. 89, 1641 (2001); 10.1063/1.1333736
Effect of Mo doping on accelerated growth of C-54 TiSi 2 : Evidence for template mechanism
J. Appl. Phys. 86, 2571 (1999); 10.1063/1.371094
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:
69.166.47.134 On: Fri, 28 Nov 2014 05:10:21

Growth of TiSi from codeposited TiSi layers and interfacial layers
2
x
R. T. Tunga)
Bell Laboratories, Lucent Technologies, Murray Hill, New Jersey 07974
K. Fujii, K. Kikuta, S. Chikaki, and T. Kikkawa
ULSI Device Development Laboratories, NEC Corporation, Sagamihara, Kanagawa 229, Japan
͑
Received 11 December 1996; accepted for publication 10 March 1997͒
The silicide formation characteristics from composition profiles created on Si by codepositions was
investigated. It was shown that a thin interfacial amorphous TiSi layer, with xϳ0.5–1, deposited
x
between Si and the Ti film led to a significant reduction in the observed C49→C54 TiSi
2
transformation temperature. The presence of the amorphous interfacial TiSi layer slowed down the
x
initial silicidation rate, but promoted the nucleation of the final C54 TiSi phase. Predeposition and
2
preannealing were also found to facilitate the growth of C54 TiSi , as was growth from codeposited
2
full TiSi layers with Ti-rich compositions. The efficacy of the ͑interfacial͒ TiSi layer was
x
x
demonstrated for different temperature ramp rates and for a variety of substrates including undoped
ϩ
ϩ
␣
-Si, preamorphized n -Si, and preamorphized p -Si. But this effect was found to be absent on
single crystal Si. Possible mechanisms of the observed effects were discussed. © 1997 American
Institute of Physics. ͓S0003-6951͑97͒04318-0͔
Ti silicide is presently the most widely used self-aligned
silicide ͑salicide͒ for high-performance ultralarge scale inte-
gration ͑ULSI͒ devices. However, it is difficult to process
devices with a gate length shorter than 0.25 ␮m using
the Ti salicide technology, because of the well-known
on amorphous and crystalline Si was studied. A novel depen-
dence of the formation of the C54-TiSi phase on the com-
2
position of the TiSi layer was discovered.
x
Si͑100͒ wafers, implanted with either 70 keV
2
2
3E15/cm As or 30 keV 3E15/cm BF , annealed at 900 °C
2
problem with the polymorphic transformation, C49-TiSi2
for 10 min, preamorphized ͑PAM͒ with a 70 keV
3E14/cm As implantation were used in the majority of these
1
2
→
C54-TiSi , in narrow silicide lines. The reaction between
2
a Ti thin film and a Si substrate starts with the formation of
a precursor interfacial amorphous silicide layer with a graded
experiments. Oxidized Si wafers with a 200-nm-thick
plasma-enhanced chemical vapor deposition ͑PECVD͒-
grown amorphous-Si layer and other types of wafers were
also used. Substrates were cleaned chemically and given a
dilute HF etch before loading in an ultrahigh vacuum ͑UHV͒
system. Ti and Si were deposited using individually con-
2
,3
composition. The C49-TiSi phase is then nucleated at the
2
interface between the precursor amorphous silicide layer and
3
,4
Si.
Subsequently, the nucleation of the desired, low-
resistivity, C54-TiSi phase is believed to take place at mul-
2
3,5
tiple grain boundaries of the C49 TiSi film. It is the low
trolled e-beam sources. Throughout this article,
a
2
density of these favorable sites in narrow C49-TiSi lines
‘‘1 nm* TiSi ’’ is defined as a codeposited TiSi layer which
2
x
x
which leads to the observed linewidth dependence. By add-
ing impurities, it has been shown that the C49-C54 transfor-
mation temperature can be lowered.6
Recently, the combination of high temperature sputter-
ing ͑HTS͒ of Ti (ϳ450 °C) and the use of preamorphized
contains the equivalent of 1 nm Ti. Anneals were done in the
UHV chamber, without a break of vacuum, by resistively
heating. Unless otherwise specified, a temperature ramp rate
of ϳ10 ° C/s was used for these anneals.
A dependence of the the sheet resistance on the deposi-
tion temperature was found for silicide layers grown from
the deposition of 15 nm Ti. After a 680 °C 10 s anneal, sheet
resistances of 17 and 7.8⍀/ᮀ were measured for Ti layers
originally deposited on As-doped PAM Si at room tempera-
ture and at 460 °C, respectively. These results were in good
agreement with the dependence on the deposition tempera-
ture observed previously, using inert gas sputtering, thicker
Ti layers, and a nitrogen annealing ambient.10 Amorphous
–8
͑
PAM͒ Si was shown to significantly reduce the C54-TiSi
2
9,10
transformation temperature on narrow lines.
The easier
nucleation of the C54-TiSi phase was speculated to arise
2
from the formation of a thick ͑ϳ6 nm͒ ␣-TiSi layer, during
x
the HTS process, at the interface between PAM Si and the
sputtered Ti. In separate experiments,11 it was shown that
deposition of Ti at temperatures (у500 °C) above the
amorphous-to-crystalline transition temperature12 for Si-rich
TiSi led to the direct nucleation of the C54-TiSi phase on
x
2
TiSi interfacial layers were artificially recreated on PAM Si
x
␣
-Si. The easy nucleation of the C54 phase, in the latter
by codeposition at room temperature. Sheet resistances of
case, was attributed to an avoidance of the amorphous
TiSi films grown with a 4 nm* interfacial layer displayed a
2
TiSi phase. These apparently different effects of the amor-
x
dependence on the composition of the interfacial layer, as
phous TiSi interfacial layer, though not necessarily mutually
x
shown in Fig. 1. With a Ti-rich TiSi , xϭ0.5–1, a signifi-
x
conflicting, pointed to the complexity of the Ti silicide prob-
lem and the importance of understanding the roles played by
cant fraction of the silicide film was converted to the
C54-TiSi phase after a 30 s ͑RTA͒ at ϳ650 °C. Layers
2
the amorphous TiSi interfacial layer. In this work, the sili-
x
with no interfacial layer (xϭ0) or with a Si-rich TiSi inter-
x
cide reaction of amorphous TiSi layers and interfacial layers
x
facial layer remained largely in the C49-TiSi phase, after
2
the same anneal. Codeposited amorphous TiSi layers were
found not to have a significant effect on the sheet resistance
x
a͒Electronic mail: rtt@clockwise.bell-labs.com
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:
386 Appl. Phys. Lett. 70 (18), 5 May 1997 0003-6951/97/70(18)/2386/3/$10.00 © 1997 American Institute of Physics
9.166.47.134 On: Fri, 28 Nov 2014 05:10:21
2
6

FIG. 1. Sheet resistances of Ti silicide layers grown from the depositions of
nm* TiSi and 11 nm Ti at room temperature and a 30 s RTA.
FIG. 3. Sheet resistance of Ti silicide layers grown from preannealing of
predeposited Ti films.
4
x
of TiSi layers grown on single crystal Si substrate, under
similar annealing conditions.
To determine the dependence of the silicide reaction on
the thickness of the amorphous interfacial layer, codeposited
was known to lead to the growth of amorphous silicide at a
Ti/Si interface without the nucleation of crystalline silicide
phases. Epitaxial regrowth of the PAM layer was also known
to be insignificant at this temperature. A Ti layer with a
thickness of t (ϭ15 nmϪt ) was then deposited at room
2
TiSi with xϭ1.0 were used. Figure 2 shows planview trans-
x
B
A
mission electron microscopy ͑TEM͒ images of TiSi layers
temperature, and a final in situ RTA at 685 °C then com-
2
grown from interfacial layers with thicknesses of 0, 2, 4, and
pleted the growth of the Ti silicide. As shown in Fig. 3,
6
nm*. In these micrographs, the majority of grains with
TiSi layers formed by preannealing were also effective in
x
small sizes (р50 nm) were C49-TiSi and those with large
lowering the sheet resistance of the TiSi layer. Preannealing
2
2
grain sizes (у200 nm) were mostly C54-TiSi in nature. A
the entire 15-nm-thick Ti layer, however, led to a sheet re-
sistance which was higher than that from a layer grown with-
2
clear increase in the C54-TiSi volume fraction was ob-
2
served in these 650 °C-annealed films, when thin interfacial
TiSi1.0 layers were used. However, the dependence of the
film microstructure on the thickness of the interfacial layer
was not strong for TiSi1.0 interfacial layers over 2 nm* thick.
out preannealing (t ϭ0). When the predeposition and pre-
A
annealing procedures were applied to silicide films grown on
single-crystalline Si substrates, no reduction in the silicide
sheet resistance was observed.
TEM-observed formation characteristics of the C54-TiSi2
The formation of silicide layers from TiSi thin films, 15
x
phase were in good qualitative agreement with the sheet re-
sistances measured in these films ͑see Fig. 2 caption͒.
*
nm in thickness, codeposited at room temperature was stud-
ied. Results on PAM and singel crystal As-implanted Si,
obtained after a 650 °C 30 s anneal were shown in Fig. 4. On
Interfacial amorphous TiSi layers with different thick-
x
nesses and/or compositions were also generated by predepo-
PAM Si, the sheet resistance of a TiSi film showed a strong
2
sition and preannealing. A small amount, t , of Ti was pre-
A
dependence on the composition of the starting TiSi layer.
x
deposited at room temperature on PAM Si and was
preannealed at 460 °C for 5 min. This annealing condition
There was a rapid drop in the sheet resistance when a small
fraction of Si was added to the Ti deposition. Ti-rich codepo-
sitions led to a nearly complete transformation to the
C54-TiSi phase, as evidenced by the low sheet resistance
2
observed for TiSi0.25. The higher sheet resistances for pure
Ti (xϭ0) and Si-rich codeposited layer (xϭ2) was due to
an incomplete transformation to the low-resistivity
C54-TiSi phase. On single crystal Si, the measured sheet
2
resistances suggested that little C54-TiSi reaction had taken
2
place for all the layers. The decrease of sheet resistance with
x, observed on single crystal Si, was likely influenced by a
decrease in dopant incorporation in the ͑majority
C49-TiSi ͒ silicide film.
2
Ti silicide layers grown on BF -implanted PAM Si by
2
codeposition also displayed a dependence of their sheet re-
sistances on the codeposition composition, as shown in Figs.
5
and 6. The formation of the C54-TiSi phase was found to
2
be easier with a starting composition of xϳ0.5–1 than with
pure Ti layers. When the temperature ramp rate was changed
from 10 ͑RTA in Fig. 5͒ to 0.2 °C/s ͑slow ramp in Fig. 5͒,
there was an increase in the sheet resistance. But a clear
FIG. 2. Planview TEM images of TiSi layers grown on As-doped PAM Si
2
with TiSi1.0 interfacial layer thicknesses of ͑a͒ 0, ͑b͒ 2, ͑c͒ 4, and ͑d͒ 6 nm*,
respectively. The measured sheet resistances were ͑a͒ 12.2, ͑b͒ 7.5, ͑c͒ 6.2,
and ͑d͒ 6.7⍀/ᮀ, respectively.
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:
Appl. Phys. Lett., Vol. 70, No. 18, 5 May 1997 Tung et al. 2387
69.166.47.134 On: Fri, 28 Nov 2014 05:10:21

FIG. 6. Planview TEM micrographs of TiSi2 layers grown from
5-nm*-thick TiSi layers codeposited on BF -implanted PAM Si. Codepo-
1
x
2
sition compositions were ͑a͒ xϭ0, ͑b͒ xϭ1.0, and ͑c͒ xϭ2.0
when an interfacial TiSi layer was present.10 Even though
x
the present results showed a dependence on the doping and
annealing conditions, TiSi layers with slightly metal-rich
compositions, xϳ.2–1, generally led to the easiest formation
x
FIG. 4. Sheet resistance of Ti silicide layers grown from the codeposition of
5 nm* TiSi at room temperature and annealing at 650 °C for 30 s. Results
are plotted for layers grown on heavily As-doped single crystal Si and PAM
Si.
1
x
of the C54-TiSi phase. The absence of an effect on crystal-
2
line Si was suggestive that the amorphous Si-TiSi interface
x
was important for the enhanced nucleation of C54-TiSi2.
The precise role played by the TiSi layer is not clear at the
x
dependence of the sheet resistance on the codeposition com-
position was still present using the very slow ramp rate. The
microstructures of silicide layers grown on PAM Si, from
present. The nucleation of C54-TiSi is expected to be en-
2
hanced from C49-TiSi films with a smaller average grain
2
size, but direct evidence for a dependence of the grain size
codeposited TiSi were revealed by planview TEM in Fig. 6.
x
on the presence of ͑interfacial͒ TiSi layer is presently lack-
x
Silicide films consisting of predominantly small C49-TiSi2
grains were found from the annealing of a pure Ti layer, as
shown in Fig. 6͑a͒. The annealing of a nearly stoichiometric
TiSi layer (xϭ2) led to predominantly C49-TiSi grains,
ing. It was suggested that the C49-TiSi crystals grown from
2
such an interfacial layer were more defective, making the
10
ensuing transformation into the C54-TiSi phase easier.
2
x
2
Silicide thin films grown from ␣-TiSi could possibly have a
x
but with a larger average grain size, Fig. 6͑c͒. C54-TiSi2
grains were observed to occupy a high areal fraction only in
the film grown from a codeposited TiSi1.0 layer, Fig. 6͑b͒.
higher compressive stress, thus favoring the nucleation of
C54-TiSi . The transient formation of TiSi or metal-rich sil-
2
icide phases could also have an effect on the disilicide reac-
tion. In any event, the present results suggest the addition of
a small amount of Si in the deposited Ti for easier salicide or
polycide applications.
The authors thank F. Schrey for his assistance with the
experiments.
A thick (Ͼ3 nm*) TiSi layer or interfacial layer was
x
presently shown to lead to an easier C49→C54-TiSi phase
2
transformation on amorphous Si. This observation strongly
supported the attribution of the observed reduction in the
C54-TiSi nucleation temperature in the HTS process to the
2
10
formation of a thick interfacial TiSi layer. It was also
x
known that the initial silicidation rate was slowed down
1
J. B. Lasky, J. S. Nakos, O. J. Cain, and P. J. Geiss, IEEE Trans. Electron
Devices ED-38, 262 ͑1991͒.
K. Holloway and R. Sinclair, J. Appl. Phys. 61, 1359 ͑1987͒.
Z. Ma, Y. Xu, L. H. Allen, and S. Lee, J. Appl. Phys. 74, 2954 ͑1993͒.
M. H. Wang and L. J. Chen, Appl. Phys. Lett. 59, 2460 ͑1991͒.
2
3
4
5
J. A. Kittl, D. A. Prinslow, P. P. Apte, and M. Pas, Appl. Phys. Lett. 67,
2
308 ͑1995͒.
6
7
X.-H. Li, J. R. A. Carlsson, S. F. Gong, and H. T. G. Hentzell, J. Appl.
Phys. 72, 514 ͑1992͒.
R. W. Mann, G. L. Miles, T. A. Knotts, D. W. Rakowski, L. A. Clevenger,
J. M. E. Harper, F. M. d’Heurle, and C. Cabral, Jr., Appl. Phys. Lett. 67,
3
729 ͑1995͒.
8
9
0
A. Mouroux, S.-L. Zhang, W. Kaplan, S. Nygren, M. Ostling, and C. S.
Petersson, Mater. Res. Soc. Symp. Proc. 427, 511 ͑1996͒.
K. Fujii, K. Kikuta, and T. Kikkawa, Proceedings of VLSI Technology
Digest, 1995, p. 57.
K. Fujii, R. T. Tung, D. J. Eaglesham, K. Kikuta, and T. Kikkawa, Mater.
Res. Soc. Symp. Proc. 402, 83 ͑1996͒.
R. T. Tung, Appl. Phys. Lett. 68, 1933 ͑1996͒.
S. F. Gong, A. Robertsson, H. T. G. Hentzell, and X.-H. Li, J. Appl. Phys.
68, 4535 ͑1990͒.
1
1
1
1
2
FIG. 5. Sheet resistance of Ti silicide layers grown from 15 nm* TiSix
codeposited at room temperature and annealed in situ for 30 s. Temperature
was ramped to 650 °C using ramp rates of 0.2 and 10 °C/s.
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:
388 Appl. Phys. Lett., Vol. 70, No. 18, 5 May 1997 Tung et al.
9.166.47.134 On: Fri, 28 Nov 2014 05:10:21
2
6