Mechanism of low temperature C54 TiSi2 formation bypassing C49 TiSi2: Effect of Si microstructure and Mo impurities on the Ti-Si reaction path

Article abstract of DOI:10.1063/1.122032

X-ray diffraction, high resolution transmission electron microscopy, and resistivity measurements were used to demonstrate a modification of the Ti/Si reaction path consisting of direct nucleation followed by diffusion limited growth of low resistivity C54 TiSi₂ without nucleation of high resistivity C49 TiSi₂, for the reaction of Ti with Mo doped polycrystalline or Mo doped amorphous Si by rapid thermal processing at 650 °C. We also report the mechanism of early C54 nucleation. We demonstrate that MoSi₂ and an unidentified silicide phase lattice matched to C54 TiSi₂, with spacings of 4.15 and 2.26 A, nucleate along the Ti/Si interface at early stages of the reaction and act as templates on which C54 TiSi₂ nucleates and grows epitaxially. In contrast, the conventional phase sequence, nucleation and growth of C49 TiSi₂ preceding nucleation of C54 TiSi₂, was observed for the Ti/Mo doped single crystal (100) Si reaction and for all samples without Mo.

Full text of DOI:10.1063/1.122032

Mechanism of low temperature C54 TiSi 2 formation bypassing C49 TiSi 2 : Effect of Si
microstructure and Mo impurities on the Ti–Si reaction path
J. A. Kittl, M. A. Gribelyuk, and S. B. Samavedam
Citation: Applied Physics Letters 73, 900 (1998); doi: 10.1063/1.122032
View online: http://dx.doi.org/10.1063/1.122032
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/73/7?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Formation of the TiSi 2 C40 as an intermediate phase during the reaction of the Si/Ta/Ti system
Appl. Phys. Lett. 78, 1864 (2001); 10.1063/1.1359142
Influence of molybdenum on the formation of C54TiSi 2 : Template phenomenon versus grain-size effect
Appl. Phys. Lett. 76, 1831 (2000); 10.1063/1.126180
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
Alternative pathway for the formation of C54 TiSi 2
J. Appl. Phys. 86, 704 (1999); 10.1063/1.370789
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:
129.22.89.130 On: Wed, 03 Dec 2014 17:37:52

APPLIED PHYSICS LETTERS
VOLUME 73, NUMBER 7
17 AUGUST 1998
Mechanism of low temperature C54 TiSi2 formation bypassing
C49 TiSi2: Effect of Si microstructure and Mo impurities
on the Ti–Si reaction path
J. A. Kittl,a) M. A. Gribelyuk, and S. B. Samavedam
Silicon Technology Development, Texas Instruments Inc., Dallas, Texas 75243
͑Received 19 February 1998; accepted for publication 15 June 1998͒
X-ray diffraction, high resolution transmission electron microscopy, and resistivity measurements
were used to demonstrate a modification of the Ti/Si reaction path consisting of direct nucleation
followed by diffusion limited growth of low resistivity C54 TiSi2 without nucleation of high
resistivity C49 TiSi2, for the reaction of Ti with Mo doped polycrystalline or Mo doped amorphous
Si by rapid thermal processing at 650 °C. We also report the mechanism of early C54 nucleation.
We demonstrate that MoSi2 and an unidentified silicide phase lattice matched to C54 TiSi2, with
spacings of 4.15 and 2.26 Å, nucleate along the Ti/Si interface at early stages of the reaction and act
as templates on which C54 TiSi2 nucleates and grows epitaxially. In contrast, the conventional
phase sequence, nucleation and growth of C49 TiSi2 preceding nucleation of C54 TiSi2, was
observed for the Ti/Mo doped single crystal ͑100͒ Si reaction and for all samples without Mo.
© 1998 American Institute of Physics. ͓S0003-6951͑98͒02933-7͔
Much work has focused recently on the Ti/Si reaction,
due to the wide spread use of TiSi2 in self-aligned silicide
͑salicide͒ processes in the ultralarge-scale integration ͑ULSI͒
semiconductor industry.1–12 Silicides are used to lower the
contact resistance and sheet resistance of gate and source/
drain areas of complementary metal-oxide semiconductor
͑CMOS͒ devices, allowing faster operation and higher per-
formance. In the salicide process, a deposited metal is re-
acted, typically by rapid thermal processing ͑RTP͒, with Si
surfaces on gate, source, and drain areas to form silicide
films. Residual metal is then removed from insulating sur-
faces by a selective wet etch ͑strip͒. Ti is used due to the low
resistivity of TiSi2 in the thermodynamically stable C54
phase ͑ϳ15 ␮⍀ cm͒. However, in the Ti/Si reaction, nucle-
ation and growth of the metastable, high resistivity ͑60–90
␮⍀ cm͒ C49 TiSi2 phase has always been observed to pre-
cede nucleation of the C54 phase.1–6 A high temperature
anneal step, generally performed after the strip, is then re-
quired to transform the C49 TiSi2 film to C54 TiSi2, and
achieve the required low sheet resistance.1–7
A major problem for scaling Ti salicide to deep-
submicron features is the dependence of the C49 to C54
transformation on lateral feature size.1–3 This was
attributed3,7 to the scarcity or lack of C49 grain boundaries,
in particular triple-grain boundaries, which act as nucleation
sites for C54 TiSi2, in structures where the lateral dimen-
sions are smaller than the C49 grain size. This limits the
application of Ti salicide to ϳ0.25 ␮m ͑ϳC49 grain size͒ for
processes in which Ti is reacted by RTP with crystalline Si.
Preamorphization implants ͑PAI͒ extended this limit to ϳ0.1
␮m due to the smaller ͑ϳ0.07 ␮m͒ C49 grain size achievable
in the RTP reaction of Ti with amorphous Si.7 In addition,
the high temperatures required to transform C49 TiSi2 to C54
TiSi2 on deep-submicron structures, can result in agglomera-
tion and degradation of device contact resistance and drive
current.1,2,9–11
An alternative approach, the addition of Mo impurities
close to the Ti/Si interface, was shown to result in easier
formation of C54 TiSi2 on submicron structures.6,8,10,11 In
situ x-ray diffraction ͑XRD͒ studies showed that the tem-
perature of the C49 to C54 transformation was reduced for
the reaction of Ti with Mo doped single crystal ͑100͒ Si.6
Extension of Ti salicide with low sheet resistance to 0.10 ␮m
features using Mo impurities10 and to 0.06 ␮m using Mo
impurities and PAI11 was demonstrated, for low temperature
single RTP step processes, with excellent contact resistance
and drive currents.10,11 However, the mechanism by which
Mo impurities assist the formation of C54 TiSi2 remained
unexplained. The presence of a ͑Ti, Mo͒Si2 phase in the re-
action of a Ti/Mo bilayer structure with Si, and epitaxial
nucleation of C54 TiSi2 on this phase was postulated,12 but
not demonstrated.
In this work, we report XRD, resistivity and high reso-
lution transmission electron microscopy ͑HRTEM͒ studies of
the effect of Mo impurities and Si crystallinity on the early
nucleation stage and subsequent growth of silicide phases in
the Ti/nϩSi system. We demonstrate nucleation and subse-
quent diffusion limited growth of C54 TiSi2 bypassing the
formation of C49 TiSi2 on Mo doped polycrystalline ͑poly-
Si͒ or Mo doped amorphous Si. For these samples, we report
the existence of silicide phases along the Ti/Si interface,
MoSi2 and an unidentified phase not found in samples with-
out Mo, that nucleate at early stages of the reaction and act
as templates for epitaxial growth of C54 TiSi2, providing the
mechanism for early C54 nucleation. In contrast, we ob-
served the conventional phase sequence ͑C49 TiSi2 nucle-
ation and growth preceding C54 TiSi2 nucleation͒ on
samples without Mo or for Mo doped single crystal ͑100͒ Si.
These results are important for the applicability of Ti sili-
cides to future scaled technologies.
a͒
Electronic mail: kittl@spdc.ti.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:
0003-6951/98/73(7)/900/3/$15.00 900 © 1998 American Institute of Physics
129.22.89.130 On: Wed, 03 Dec 2014 17:37:52

Appl. Phys. Lett., Vol. 73, No. 7, 17 August 1998
Kittl, Gribelyuk, and Samavedam
901
FIG. 2. ␪–2␪ reflection geometry XRD pattern of a Ti/Mo doped poly-Si
sample after 10 s reaction at 650 °C.
doped poly-Si sample after 10 s reaction at 650 °C. As ob-
served using grazing incidence geometry, C54 peaks are
present and C49 peaks are absent.
FIG. 1. Grazing incidence XRD patterns showing the evolution of the
Ti/Mo doped Si reaction at 650 °C for ͑a͒ ͑100͒ Si, ͑b͒ poly-Si, for anneal
times from 10 to 120 s.
Figure 3 shows XRD spectra following the evolution of
the reaction of Ti with ͑a͒ amorphous Si and ͑b͒ Mo doped
amorphous Si. Conventional C49 nucleation and growth is
observed at 650 °C on samples without Mo, while the addi-
tion of Mo results in growth of C54 TiSi2 with no evidence
of C49 TiSi2 even at the short annealing time of 10 s.
The XRD data discussed was obtained after removal of
TiN and unreacted Ti. We verified that the removal of TiN
and Ti films did not affect our results, having no effect on
C49 or C54 peaks, by comparing spectra for samples before
and after Ti/TiN removal.
The difference in phase sequence observed on Mo doped
samples for single crystal ͑100͒ Si ͑C49 grows first͒ com-
pared to either poly-Si or amorphous Si ͑C54 grows bypass-
ing C49͒, suggests that either disordered Si regions such as
grain boundaries or amorphous Si are needed, or special
crystal orientations need to be avoided, for early C54 nucle-
ation.
Samples were prepared on ͑100͒ Si wafers by sputter
depositing Ti films on nϩSi with or without Mo impurities
for ͑a͒ single crystal ͑100͒ Si, ͑b͒ poly-Si, and ͑c͒ amorphous
Si obtained by PAI. Mo was introduced into Si by ion im-
plantation at similar implant conditions as previously
reported,6,10,11 which do not cause amorphization of crystal-
line Si. Samples were reacted by RTP at 650 °C in a N2
ambient, for times varying from 10 to 120 s, and analyzed ͑a͒
as prepared and ͑b͒ after removal of TiN and unreacted Ti by
a high selectivity wet etch that does not attack Ti silicides.
Sheet resistance was measured using a four point probe
method. XRD spectra were taken with a Siemens D500 dif-
fractometer, using ͑a͒ thin film geometry with a grazing in-
cident angle of 2° from the surface of the sample, ͑b͒ ␪–2␪
reflection geometry. Specimens were thinned to electron
transparency by mechanical polishing and ion milling, and
analyzed using a Philips CM 200 ͑FEG͒ transmission elec-
tron microscopy ͑TEM͒. Phases were identified by matching
HRTEM with simulated images.
In addition to C49 and C54 TiSi2 diffractions, we note
the presence of peaks labeled A ͑ϳ2.26 Å͒ and B ͑ϳ4.15 Å͒
in Figs. 1 to 3. Peak A could correspond to the ͑011͒ diffrac-
tion of slightly reacted Ti.
Figure 1 shows XRD spectra of the evolution of the
Ti/Mo doped Si RTP reaction at 650 °C for RTP times rang-
ing from 10 ͑early stage of silicidation͒ to 120 s, for single
crystal ͑100͒ Si ͓Fig. 1͑a͔͒ and poly-Si ͓Fig. 1͑b͔͒. Spectra in
Fig. 1 were taken under a grazing incidence geometry, after
removal of Ti/TiN. For Mo doped ͑100͒ Si C49 TiSi2 is seen
to form at early stages of the reaction, is present after 10 s,
and grows with increasing reaction time. C54 TiSi2 is not
observed at any processing time. In contrast, C54 TiSi2 is
present after only 10 s processing at 650 °C for the Mo
doped poly-Si sample, with the C54 TiSi2 ͑331͒ diffraction
undoubtedly present. Further anneal at 650 °C results in
growth of the C54 phase, with all peaks clearly observed for
longer anneal times ͑80 and 120 s͒. No C49 TiSi2 peaks are
seen in Mo doped poly-Si samples at any processing time.
To rule out film texture or special orientation as causing the
absence of C54 TiSi2 peaks in Fig. 1͑a͒ or C49 peaks in Fig.
1͑b͒, a ␪–2␪ reflection geometry was also used, resulting in
similar diffraction patterns to those in Fig. 1, and the same
FIG. 3. Grazing incidence XRD patterns showing the evolution of the Ti/
amorphous Si reaction at 650 °C for samples ͑a͒ without Mo and ͑b͒ with
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:
conclusions. Figure 2 shows a ␪–2␪ XRD pattern of a Ti/Mo
Mo doping.
129.22.89.130 On: Wed, 03 Dec 2014 17:37:52

902
Appl. Phys. Lett., Vol. 73, No. 7, 17 August 1998
Kittl, Gribelyuk, and Samavedam
FIG. 4. The thickness of the silicide layer, composed predominantly of C54
TiSi2, evolves according to diffusion limited kinetics for the Ti/Mo doped
poly-Si reaction.
TEM measurements of the thickness of the reacted sili-
cide layer ͑on samples without Ti/TiN removal͒ verified that
the samples were only partially reacted, and were in excel-
lent agreement with thickness values obtained from resistiv-
ity measurements ͑after Ti/TiN strip͒, for a resistivity of 22
␮⍀ cm ͑Fig. 4͒. The thickness of the reacted layer evolves
according to diffusion limited kinetics as shown in Fig. 4 for
Mo doped poly-Si samples. The reacted layer is composed
predominantly of C54 TiSi2 as indicated by resistivity, XRD,
and TEM results.
HRTEM analysis confirmed the conclusions of the XRD
studies, and identified the presence of other silicide phases.
After 10 s reaction at 650 °C, a thin reacted layer containing
C54 TiSi2, MoSi2, Mo5Si3, and Ti5Si4 grains was observed
in the Mo doped poly-Si sample. In addition, a silicide phase
with spacings of ϳ2.27 Å and ϳ4.15 to 4.3 Å, lattice
matched to C54 TiSi2, was observed. This phase, and the Mo
silicide phases were observed to nucleate along the Ti/Si
interface with a layer thickness of ϳ50 Å. The unidentified
phase and MoSi2 were found to act as templates for epitaxial
nucleation and growth of C54 TiSi2 ͑Figs. 5 and 6͒, and were
not present in samples without Mo. These template phases
provide the mechanism for early nucleation of C54 TiSi2.
Image simulations suggest that the structure of the unidenti-
fied template phase is based on Mo5Si3. XRD peaks A and B
in Fig. 2 could correspond to this phase. We did not find any
FIG. 6. HRTEM of the Ti/Si interface region after 10 s reaction at 650 °C
for a Ti/Mo doped poly-Si sample. MoSi2 nucleated at the interface and acts
as a template for epitaxial growth of C54 TiSi2.
evidence of C40 ͑Ti, Mo͒Si2.12 HRTEM showed that further
anneal of Ti/Mo doped poly-Si resulted in growth of the
silicide layer, composed predominantly of C54 TiSi2. How-
ever, grains of Ti5Si4, Mo5Si3, and the unidentified template
phase were still present in samples annealed for 80 s. We
also verified by HRTEM that in samples without Mo, C49
TiSi2 is observed after 10 s reaction as well as Ti5Si4 but not
C54 TiSi2, while samples annealed for 80 s resulted in a
layer composed predominantly of C49 TiSi2.
In conclusion, we demonstrated nucleation followed by
diffusion limited growth of C54 TiSi2 without the presence
of C49 TiSi2 in the Ti/nϩSi reaction for Mo doped poly-Si
or Mo doped amorphous Si. The early nucleation of C54
TiSi2 is due to the presence of MoSi2 and an unidentified
phase ͑possibly based on Mo5Si3) lattice matched to C54
TiSi2, that nucleate at early stages of the reaction along the
Ti/Si interface, and act as templates for epitaxial nucleation
and growth of C54 TiSi2. C49 nucleation and growth preced-
ing C54 nucleation was observed for Mo doped ͑100͒ Si and
samples without Mo.
1 K. Maex, Mater. Sci. Eng., R. 11, 53 ͑1993͒.
2 J. B. Lasky, J. S. Nakos, O. J. Cain, and P. J. Geiss, IEEE Trans. Electron
Devices ED-38, 262 ͑1991͒.
3 J. A. Kittl, D. A. Prinslow, P. P. Apte, and M. F. Pas, Appl. Phys. Lett. 67,
2308 ͑1995͒.
4 M. H. Wang and L. J. Chen, J. Appl. Phys. 71, 5918 ͑1992͒.
5 Z. Ma, L. H. Allen, and D. D. Allman, J. Appl. Phys. 77, 4384 ͑1995͒.
6 L. A. Clevenger, C. Cabral, Jr., R. A. Roy, C. Lavoie, R. Viswanathan, K.
L. Saenger, J. Jordan-Sweet, G. Morales, K. L. Ludwig, Jr., and G. B.
Stephenson, Mater. Res. Soc. Symp. Proc. 402, 257 ͑1996͒.
7 J. A. Kittl, Q. Z. Hong, M. Rodder, D. A. Prinslow, and G. Misium, 1996
VLSI Tech. Dig. of Tech. Papers, 14 ͑1996͒.
8 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,
3729 ͑1995͒.
9 J. A. Kittl, A. Chatterjee, I.-C. Chen, G. A. Dixit, P. P. Apte, D. A.
Prinslow, and Q. Z. Hong, 1997 Int. Symp. VLSI Technology, Systems,
and Applications, Proc. Tech. Papers, 23 ͑1997͒.
10 J. A. Kittl, Q. Z. Hong, C. P. Chao, I. C. Chen, N. Yu, S. O’Brien, and M.
Hanratty, 1997 VLSI Tech. Dig. of Tech. Papers, 103 ͑1997͒.
11 J. A. Kittl, Q. Z. Hong, M. Rodder, and T. Breedijk, Tech. Dig. Int.
Electron Devices Meet., 111 ͑1997͒.
FIG. 5. HRTEM of the Ti/Si interface region after 10 s reaction at 650 °C
for a Ti/Mo doped poly-Si sample. A silicide phase with spacings of ϳ2.27
and ϳ4.2 Å nucleated along the interface and grew ϳ50 Å thick, providing
12 A. Mouroux, S. L. Zhang, W. Kaplan, S. Nygren, M. Ostling, and 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:
a template on which C54 TiSi2 grows epitaxially.
Petersson, Appl. Phys. Lett. 69, 975 ͑1996͒.
129.22.89.130 On: Wed, 03 Dec 2014 17:37:52