Full text of DOI:10.1063/1.115246

Patterning method for silicides based on local oxidation
S. Mantl, M. Dolle, St. Mesters, P. F. P. Fichtner, and H. L. Bay
Citation: Applied Physics Letters 67, 3459 (1995); doi: 10.1063/1.115246
View online: http://dx.doi.org/10.1063/1.115246
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/67/23?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
An investigation of acoustic beam patterns for the sonar localization problem using a beam based method
J. Acoust. Soc. Am. 133, 4044 (2013); 10.1121/1.4802831
An x-ray spectromicroscopic study of the local structure of patterned titanium silicide
Appl. Phys. Lett. 71, 55 (1997); 10.1063/1.119467
Localization of sources by two patternmatch methods
J. Acoust. Soc. Am. 87, S35 (1990); 10.1121/1.2028184
Anodic oxidation of tantalum silicide
Appl. Phys. Lett. 47, 579 (1985); 10.1063/1.96077
Thermal oxidation of silicides
J. Appl. Phys. 56, 2127 (1984); 10.1063/1.334212
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:
129.120.242.61 On: Sun, 30 Nov 2014 17:18:27

Patterning method for silicides based on local oxidation
S. Mantl,a) M. Dolle, St. Mesters, P. F. P. Fichtner,b) and H. L. Bay
¨
¨
¨
Institut fur Schicht- und Ionentechnik, KFA Julich, D-52425 Julich, Germany
͑Received 13 June 1995; accepted for publication 26 September 1995͒
Oxidation of CoSi2 layers on Si͑100͒ using oxidation masks has been investigated. It is shown that
local oxidation can be used to pattern the silicide layer. This method allows the formation of buried
interconnects and metallized silicon mesa structures. Epitaxial CoSi2 silicide layers were grown by
molecular beam epitaxy on Si͑100͒. The SiO2/Si3N4 oxidation mask was patterned
photolithographically with linewidths of typically 1.5 ␮m. During thermal oxidation,
SiO2 forms in the unprotected regions of the silicide layer. The silicide is pushed into the substrate
in these regions. At a critical oxide thickness, the oxidized region of the silicide layer separates from
the unoxidized, in conformance with the structure of the oxidation mask. The oxide capped silicide
maintains its uniform layer structure and its single crystallinity in spite of the large shift into the
substrate. The method should be applicable also to polycrystalline silicides, such as TiSi2 . © 1995
American Institute of Physics.
Transition metal silicides are the materials of choice for
contacts and interconnects in microelectronic devices or as
Schottky contacts in infrared detectors.1 Some silicides, e.g.,
CoSi2 , NiSi2 , can be grown epitaxially and thus integrated
local oxidation of epitaxial CoSi2 layers on Si͑100͒. We will
show for the first time, that local oxidation of a silicide layer
can be used to pattern the silicide layer and form metallized
mesa-structures, simply by the use of an appropriate mask
during thermal oxidation. The development of the method
was guided by the well established technology of local oxi-
dation of silicon ͑LOCOS͒, which is broadly used for lateral
isolation of ULSI devices.8 In addition, we will show that
epitaxial silicide layers preserve their single crystalline qual-
ity during oxidation.
First, we will discuss the oxidation of uniform epitaxial
CoSi2 surface layers on Si͑100͒. The silicide was grown by
molecular beam epitaxy at 525 °C followed by a high tem-
perature anneal as described elsewhere.9 Subsequently, rapid
thermal annealing ͑RTA͒ was performed at 1150 °C for 30 s
in order to form a uniform silicide layer. Such CoSi2 layers
are single crystalline and have a very low specific electrical
resistivity of 14 ␮⍀ cm at room temperature. Figure 1 shows
Rutherford backscattering spectrometry ͑RBS͒ spectra of
in
silicon
based
epitaxial
silicon/metal/silicon
heterostructures.2 Such structures are highly useful for new
electronic and electro-optic devices, in particular for the
realization of new vertical devices.3,4 The stability of sili-
cides against oxidation is essential for device processing and
has been studied in detail.5–7 Most transition metal silicides
form a silicon dioxide layer on top of the silicide layer dur-
ing oxidation. The oxidation process of transition metal sili-
cides consists basically of four major steps:6 ͑1͒ diffusion of
the oxidant through the SiO2 layer; ͑2͒ dissociation of the
silicide at the SiO2–silicide interface; ͑3͒ transport of the
metal or the silicon through the silicide layer; ͑4͒ silicide
formation at the silicide–silicon interface. In the case of
CoSi2 , Co diffuses through the silicide layer and forms again
the disilicide phase at the silicide–substrate interface. The
dissociation of the silicide at the upper interface and its ref-
ormation at the lower interface yield to an apparent inertness
of the silicide layer against oxidation. For device application
it is important that the silicide layer preserves its integrity.
Interestingly, the SiO2 layer on top of the silicide has pro-
perities close to thermally grown SiO2 on Si, i.e., the dielec-
tric constants and the density of the oxide are similar to those
of SiO2 thermally grown on Si.6 However, the dielectric
strength is lower, most probably due to the roughness of the
oxide/silicide interface. The thickness of the SiO2 layer is
fairly constant even on top of a rather rough silicide layer, as
revealed by transmission electron microscopy7 ͑TEM͒. We
expect an improvement of the dielectric strength of SiO2 on
highly uniform epitaxial CoSi2 layers.
In this letter we present investigations of uniform and
FIG. 1. Rutherford backscattering spectra of epitaxial CoSi2 layers on
Si͑100͒ with thicknesses of 90 and 150 nm, before ͑solid line͒ and after
wet oxidation ͑dotted line͒ at 980 °C for 20 min. The spectra were
measured with 1.8 MeV Heϩ ions at a tilt angle of 7° and a scattering angle
of 170°.
a͒
Electronic mail: s.mantl@kfa-juelich.de
Permanent address: Department of Metallurgy, UFRGS, AV. Bento
Goncalves, 9500-Caixa Postal 15051, 91501-970 Porto Alegre, RS,
Brasilien.
b͒
Appl. Phys. Lett. 67 (23), 4 December 1995
0003-6951/95/67(23)/3459/3/$6.00
© 1995 American Institute of Physics
3459
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:
129.120.242.61 On: Sun, 30 Nov 2014 17:18:27

FIG. 2. Random and ͗001͘ aligned RBS spectra of a 150 nm CoSi2 layer on
Si͑100͒ after wet oxidation at 980 °C for 20 min. The channeling spectrum
͑dotted line͒ shows a minimum yield of the silicide of 4%, indicating an
excellent single crystalline quality. The SiO2 layer was removed wet chemi-
cally before the RBS measurements.
FIG. 3. Illustration of the local process of oxidation of silicides ͑LOCOSI͒
allowing lateral patterning of silicide layers. ͑a͒ The silicide layer is covered
with a thin SiO2 layer and a patterned Si3N4 layer. ͑b͒ Oxidation yields to
growth of SiO2 locally. The silicide layer undulates but remains continuous.
͑c͒ At a critical oxide thickness, the silicide layer breaks up and forms a
pattern conform with the oxidation mask. This process produces buried in-
terconnects and/or metallized silicon-mesa structures.
two silicide layers with different thicknesses, 90 nm and 150
nm, before and after wet oxidation. After wet oxidation at
980 °C for 20 min the CoSi2 layers preserved their structural
properties, although they were shifted into the substrate by
about 150 nm. The RBS data show a large shift of the Co
signals but no changes of the Co signal shape. This is an
important point, because it demonstrates the chemical and
structural stability of the layer during the oxidation process.
Interestingly, the O signals of the two samples overlap per-
fectly indicating identical thicknesses of the SiO2 layers of Ϸ
250 nm. Therefore, the oxidation rate is independent of the
silicide thickness, in agreement with previous investigations
showing that the diffusive transport through the silicide is
not the rate limiting step.6 Furthermore, the single crystallin-
ity of the oxidized silicide layer remains unchanged as com-
pared to the virgin sample. The minimum channeling yield
͑Fig. 2͒ of the oxidized silicide layer of 4% is the same as
that of the virigin sample ͑not shown͒, indicating the excel-
lent quality of the single crystalline layer.
The oxidation rates of the single crystalline CoSi2 layers
were measured by Rutherford backscattering during dry oxi-
dation and wet oxidation ͑not shown͒. Generally, we found
that for both dry and wet oxidation the surface layer of
SiO2 grows parabolically with time. In order to reduce the
thermal budget we used wet oxidation at 980 °C for the local
oxidation experiment. At that temperature we measured the
parabolic oxidation rate constant Bϭ50 nm2 sϪ1. The oxida-
tion rates of the silicide is significantly larger than for pure
silicon, in agreement with previous work.6
exactly below the edges of the oxidation mask. This is shown
in
a cross-sectional transmission electron microscopy
͑XTEM͒ micrograph ͑Fig. 4͒ of a sample wet oxidized at
980 °C for 45 min. The oxidized silicide line is perfectly
separated from the silicide surface layer due to its shift into
the substrate by about 150 nm. As visible in Figs. 3 and 4,
metallized silicon mesa structures were obtained, where the
two silicide layers are connected electrically only via the
silicon substrate. The buried silicide line is passivated by the
thermally grown oxide. Before oxidation the thickness of the
silicide layer amounted Ϸ55 nm. In the oxidized region it
decreased in average to Ϸ50 nm whereas in the unoxidized
region it increased up to Ϸ80 nm near the edges ͑Fig. 4͒.
Interestingly, despite the large material redistribution along
the silicide layer, the uniformity of the individual layers is
remarkably good. Furthermore, we observed that longer oxi-
dation or additional annealing reduce the variation in thick-
ness. The dark contrasts under the silicide layer in the XTEM
image show strain contrasts. In spite of the presence of
stress, introduced by the local oxidation process and the lat-
The basic steps of the local oxidation of a silicide
͑LOCOSI͒ process are sketched in Fig. 3. An oxidation mask
consisting of a 20 nm thin, thermally grown SiO2 layer and a
CVD Si3N4 layer with a typical thickness of 200 nm was
deposited. Oxidation leads to growth of SiO2 in the openings
of the oxidation mask, the silicide layer is pushed deeper into
the substrate, and thus deformed at the transition between the
oxidized and protected regions. The silicide layer undulates
but remains continuous. At a critical oxide thickness, de-
pending on the thickness of the silicide, the layer disrupts
FIG. 4. Cross-sectional TEM micrograph of a 55 nm thick CoSi2 layer
patterned by local oxidation. The linewidth of the oxidation mask is 1.5
␮m. The thickness of the upper silicide layer amounts 60–80 nm and that of
the buried Ϸ50 nm after oxidation.
3460
Appl. Phys. Lett., Vol. 67, No. 23, 4 December 1995
Mantl et al.
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:
129.120.242.61 On: Sun, 30 Nov 2014 17:18:27

silicide at the SiO2–silicide interface the Co diffuses no
longer primarly along the shortest path through the silicide
layer, as it is the case in the planar regions, but preferentially
toward the upper and lower silicide layers. Such an effec-
tively anisotropic diffusion flux of Co atoms causes first thin-
ning of the silicide layer and second, break up at the high
stress points. This phenomenon will be investigated further.
In summary, we have demonstrated that local oxidation
can be used to pattern silicide layers. The oxidized silicide
layer is buried under a passivating silicon dioxide layer. The
new method allows the formation of buried, passivated inter-
connects and metallized silicon mesa-structures. In view of
the lack of dry etching processes for CoSi2 , there exist no
appropriate Co-chloride or -fluoride gases, this patterning
method may become highly useful. Technologically, it would
be highly desirable to apply this process on polycrystalline
silicide layers. First local oxidation experiments on polycrys-
talline TiSi2 layers look promising. However, because of the
grain structure and the larger interface roughness, the poly-
crystalline layers are less temperature stable than the single
crystalline layers, and thus the process conditions become
more critical, in order to avoid layer disintegration.
FIG. 5. XTEM micrograph of the transition region of the sample of Fig. 4
showing a smooth SiO2/silicon substrate interface between the two upper
and lower silicide layers. No dislocations are observed.
tice mismatch of the single crystalline CoSi2 layer of
Ϫ1.2%, no dislocations in the silicon were observed. This is
verified by the XTEM image of Fig. 5, showing the transi-
tion region in an enlarged scale. The appearance of disloca-
tions at these regions has been reported for silicon locally
oxidized with non optimized thicknesses of the oxidation
masks.8 The SiO2/silicon interface is strikingly smooth, and
in particular, no traces of silicide precipitates are seen. These
observations are most important for microelectronic or opto-
electronic applications.
The key question of the new patterning method refers to
the driving force for the break up of the silicide layer. At
present, we assume that the disruption of the silicide in con-
formance with the nitride mask is related to the stress field
introduced by the local growth of the SiO2 layer and the
lattice mismatch between the silicide and the silicon sub-
strate. Stress calculations of locally oxidized silicon showed
large stresses, localized at the edges.10 We believe that the
stress field generates a chemical potential gradient in the
transition regions in such a way that after dissociation of the
1 R. W. Fathauer, S. Mantl, L. J. Schowalter, and K. N. Tu, eds., Mater. Res.
Soc. Symp. Proc. 320 ͑1994͒.
2 L. J. Chen and K. N. Tu, Mater. Sci. Rep. 6, 53 ͑1991͒; R. T. Tung, Mater.
¨
Chem. Phys. 32, 107 ͑1992͒; H. v. Kanel, Mater. Sci. Rep. 8, 193 ͑1992͒.
3
¨
¨
A Schuppen, L. Vescan, M. Marso, A. v. d. Hart, H. Luth, and H. Benek-
ing, Electron. Lett. 29, 215 ͑1993͒.
4
¨
J. P. Hermanns, F. Ruders, E. Stein von Kamienski, H. G. Roskos, H.
Kurz, O. Hollricher, C. Buchal, and S. Mantl, Appl. Phys. Lett. 66, 866
͑1995͒.
5 W. J. Strydom, J. C. Lombaard, and R. Pretorius, Thin Solid Films 131,
215 ͑1985͒.
6 H. Jiang, C. S. Petersson, and M.-A. Nicolet, Thin Solid Films 140, 115
͑1986͒.
7 G. J. Huang and L. J. Chen, J. Appl. Phys. 76, 865 ͑1994͒.
8 S. Wolf, Solid State Technol. 35, 53 ͑1992͒.
9 S. Mantl and H. L. Bay, Appl. Phys. Lett. 61, 267 ͑1992͒.
10
¨
I. De Wolf, J. Vanhellemont, A. Romano-Rodriquez, H. Norstrom, and H.
E. Maes, J. Appl. Phys. 71, 898 ͑1992͒.
Appl. Phys. Lett., Vol. 67, No. 23, 4 December 1995
Mantl et al.
3461
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:
129.120.242.61 On: Sun, 30 Nov 2014 17:18:27