A combined Moessbauer and Rutherford backscattering spectroscopy analysis of the influence of nanosized cavities on CoSi2 formation

Article abstract of DOI:10.1063/1.117214

For Co in Si, the competition between CoSi₂ formation and cavity trapping is studied by Rutherford backscattering and Moessbauer spectroscopy. The presence of nanosized voids hampers the formation of a buried epitaxial silicide layer in its initial phase, preventing the small CoSi₂ particles from forming a bulk layer. The Moessbauer spectra show that a pre-existing silicide phase can be partially dissolved in favor of cavity trapping. In addition, channeling measurements provide qualitative information about the voids, showing that the thermal stability of the voids is much higher than for defects resulting from self-implantation.

Full text of DOI:10.1063/1.117214

A combined Mössbauer and Rutherford backscattering spectroscopy analysis of the
influence of nanosized cavities on CoSi2 formation
W. Deweerd, R. Moons, J. Verheyden, K. Milants, G. Langouche, and H. Pattyn
Citation: Applied Physics Letters 69, 3584 (1996); doi: 10.1063/1.117214
View online: http://dx.doi.org/10.1063/1.117214
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/69/23?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Control of Co flux through ternary compound for the formation of epitaxial CoSi2 using Co/Ti/Si system
Appl. Phys. Lett. 69, 3498 (1996); 10.1063/1.117224
Formation of metastable epitaxial CoSi x (x2(111)
J. Vac. Sci. Technol. A 14, 3236 (1996); 10.1116/1.580219
Formation and characterization of Si/SiO2 multilayer structures by oxygen ion implantation into silicon
J. Appl. Phys. 80, 4960 (1996); 10.1063/1.363540
Electrical characterization and physical analysis of epitaxial CoSi2 grown from the Si100/Ti/Co system
J. Appl. Phys. 80, 952 (1996); 10.1063/1.362907
On the formation of inhomogeneities in epitaxial CoSi2 layers grown from the interaction of Co/Ti bilayers with Si
100 substrates
J. Appl. Phys. 80, 762 (1996); 10.1063/1.362884
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:
130.102.42.98 On: Sun, 23 Nov 2014 17:34:31

¨
A combined Mossbauer and Rutherford backscattering spectroscopy
analysis of the influence of nanosized cavities on CoSi formation
2
W. Deweerd,a) R. Moons, J. Verheyden, K. Milants, G. Langouche, and H. Pattyn
Instituut voor Kern- en Stralingsfysika, K.U. Leuven, Celestijnenlaan 200D, B-3001 Leuven, Belgium
͑
Received 10 June 1996; accepted for publication 30 July 1996͒
For Co in Si, the competition between CoSi formation and cavity trapping is studied by Rutherford
2
backscattering and M o¨ ssbauer spectroscopy. The presence of nanosized voids hampers the
formation of a buried epitaxial silicide layer in its initial phase, preventing the small CoSi particles
2
from forming a bulk layer. The M o¨ ssbauer spectra show that a pre-existing silicide phase can be
partially dissolved in favor of cavity trapping. In addition, channeling measurements provide
qualitative information about the voids, showing that the thermal stability of the voids is much
higher than for defects resulting from self-implantation. © 1996 American Institute of Physics.
͓
S0003-6951͑96͒03641-8͔
Metallic impurities, especially 3d metals and their sili-
cides, greatly degrade the performance of semiconductor de-
vices, necessitating controlled and efficient gettering. In this
context, nanosized voids have been demonstrated to be very
dechanneling with direct scattering peaks at depths corre-
sponding to the calculated mean projected ranges. This di-
8
rect scattering comes from displaced Si atoms, formed by the
implantation. This point damage can be annealed out while
the voids themselves are thermally much more stable as will
be seen later. A 300-nm-thick layer was etched off to bring
1
–7
promising.
Introducing voids into c-Si, followed by an-
nealing, leads to the dissolution of the contaminant from a
silicide phase in favor of partial or complete coverage of the
internal surface as shown by Myers et al. In contrast to get-
tering by precipitation, cavity gettering persists up to arbi-
trarily small concentrations since the fractional occupancy of
the surface determines the equilibrium solute concentration
near the voids. We study Co trapping by M o¨ ssbauer spec-
7
the cavity layer to the surface. We then implanted
1ϫ101 Co/cm with an energy of 80 keV at 350 °C, fol-
lowed by thermal treatment. Normally, i.e., when no cavities
are present, the implantation leads to the formation of small
silicide molecules. The subsequent two-step annealing pro-
cedure of 1 h at 600 °C and 0.5 h at 1000 °C then leads to the
formation of a buried silicide layer ͑first step͒ with a sharp
7 59
2
troscopy, giving a more microscopic picture.6 In this letter,
we examine the competition between cavity trapping and
silicide formation under various conditions, showing that Co
can be dissolved from a pre-existing silicide phase.
,7
9
epitaxial interface ͑second step͒. In Fig. 2, the RBS spec-
trum after annealing shows the presence of the Co in a
clearly off-stoichiometric layer centered at a depth of
roughly 60 nm. There is hardly any difference between the
simulated profile after implantation and the measured profile
after annealing. This indicates that silicide formation is ham-
pered in its initial phase by the immediate presence of the
voids, preventing the individual precipitates to form a bulk
layer. There can be several reasons for this. First, there can
be a geometrical effect where the mere presence of a steric
obstacle prevents the formation of a silicide layer by coales-
cence of the precipitates. This could influence the quality of
the epitaxial layer but the small size and the low density of
voids probably makes this effect of minor importance. Sec-
We used high-purity ͑111͒-oriented float zone n-Si with
a thickness of 380 ␮m. Annealing was performed in an UHV
Ϫ9
furnace, the pressure being kept below 5ϫ10 Torr. Exten-
sive studies have been done on the formation of He-induced
cavities in c-Si. We used He concentrations of 6–8 at. % and
desorption was performed at 700 °C during 30 min.
In a high Co dose regime, we performed Rutherford
backscattering spectrometry and channeling measurements
ϩϩ
͑RBS-C͒ with an analyzing beam of 1.97 MeV He
ions,
collimated within 0.05°. M o¨ ssbauer spectroscopy gives
complementary information, differentiating various atomic
sites since the local environment of the probe determines the
interaction with its nearest neighbors, giving a unique isomer
shift ␦, quadrupole splitting ⌬, and recoilless fraction f.
These parameters then serve as a ‘‘fingerprint’’ for a particu-
lar site. The spectra have been recorded in an identical setup
as in previous work: The decay of 57Co is observed in a
source measurement in transition mode geometry, using a
SFC ͓Na Fe͑CN͒ •10H O] absorber with a thickness of 1
4
6
2
2
mg/cm .
In a first sample, we implanted He in four steps with
energies between 16 and 80 keV, producing a void layer with
a thickness of roughly 400 nm. Figure 1 gives the RBS-C
spectrum after desorption, showing a gradual increase of the
FIG. 1. Random and channeled RBS-C spectrum after fourfold He implan-
tation and desorption. The arrows indicate the projected range for each of
the implantation energies calculated by TRIM.
a͒Research assistant for the Belgian National Fund for Scientific Research
͑NFWO͒. Electronic mail: wim.deweerd@fys.kuleuven.ac.be
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:
584 Appl. Phys. Lett. 69 (23), 2 December 1996 0003-6951/96/69(23)/3584/3/$10.00 © 1996 American Institute of Physics
30.102.42.98 On: Sun, 23 Nov 2014 17:34:31
3
1

FIG. 2. RBS spectrum after introducing cavities ͑see Fig. 1͒, Co implanta-
tion, and subsequent annealing. The solid line represents a simulation. The
inset shows the Co depth profile. The squares show the simulated implan-
tation profile, taking sputtering into account. The triangles are the measured
values. The solid line is the concentration for a stoichiometric layer.
ond, there can be a partial cavity trapping, where the Co is
dissolved from the silicide nuclei or where trapping occurs
already during implantation. However, the number of trap-
ping sites in the voids is much lower than the number of
implanted atoms so the local concentration will not pass be-
low its critical value for the formation of a buried layer.1
,9
FIG. 4. Mossbauer spectra ͑a͒ immediately after Co implantation, ͑b͒ after
annealing at 500 °C and ͑c͒ after He implantation and annealing at 900 °C
for 3 h.
¨
Finally, the voids can ‘‘pin’’ the CoSi particles formed dur-
2
ing implantation, preventing the free motion of a large frac-
tion of the silicide nuclei. Taking into account the above
arguments, we think that the last mechanism is the most
probable and important one. Partial silicide dissolution, fol-
lowed by cavity trapping, however, although not the main
reason for the above effect, is hereby not excluded.
M o¨ ssbauer spectroscopy provides an insight into this lat-
ter problem, namely, can a pre-existing CoSi phase be par-
2
tially dissolved in favor of cavity trapping? To answer this
question, we started by observing silicide formation in a sec-
ond sample containing ‘‘regular’’ damage, introduced by
1
4
2
self-implanting 5ϫ10 at./cm at 60 keV, followed by 3
ϫ1014
sample was then isochronically annealed for 3 h starting at
00 °C up to 900 °C in steps of 50 °C. Figure 3͑a͒ shows the
57
Co/cm and 7ϫ10
2
14 59
Co/cm at 50 keV. The
2
5
RBS-C spectrum after the Si implantation and after the final
annealing step. After implantation, damage is present under
the form of self-interstitials. After annealing, most point
damage is gone and only extended defects remain. Also the
M o¨ ssbauer spectra show a clear transition. The spectrum af-
ter implantation ͓Fig. 4͑a͔͒ was fitted with a doublet, giving
7
identical parameters as in previous work. After the first an-
nealing step, the spectrum changes to that of CoSi ͓Fig.
2
4
͑b͔͒, implying the formation of silicide nuclei, even in the
presence of residual lattice damage. The hyperfine para-
meters for the CoSi component ͑Table I͒, containing a
2
single line and a small doublet, are identical to the literature
values. Subsequently, we implanted 1ϫ1017 He/cm at 30
keV, followed by desorption. Again we performed isochro-
nal annealing for 3 h, starting at 700 up to 1000 °C in steps
of 50 °C. Figure 3͑b͒ shows the RBS-C spectra after desorp-
tion and after annealing at 700 °C. After desorption, again a
direct scattering contribution is present, superimposed on a
dechanneling contribution by the voids. After further anneal-
ing, the direct scattering contribution disappears while the
dechanneling remains unaltered. If higher annealing tem-
peratures are employed, the minimum yield only slightly de-
creases to approximately 25%, corresponding to a decreasing
2
FIG. 3. ͑a͒ Random and channeled RBS-C spectrum of the self-implanted
sample after the Si implantation ͑᭡͒ and after the final annealing step at
9
00 °C during 3 h ͑᭢͒ and ͑b͒ after He implantation and desorption ͑᭡͒ and
after annealing at 700 °C for 3 h ͑᭢͒. The arrows indicate the calculated
projected range for the Si, respectively He, implantation.
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. 69, No. 23, 2 December 1996 Deweerd et al. 3585
30.102.42.98 On: Sun, 23 Nov 2014 17:34:31
1

TABLE I. Parameters used for fitting the CoSi , Q , and Q M o¨ ssbauer
components of Fig. 4.
2
1
2
times confirms the picture that this component belongs
to atoms residing at flat surfaces of the faceted voids.
a͒
1,6,7
The final number of trapped Co atoms is approximately
3
CoSi2
ϫ1014 at./cm , somewhat higher than the previously
2
Single line
Doublet
Q1
Q2
observed values but the annealing sequence is now
much longer, using higher temperatures. Nevertheless
the coverage is still below saturation, lying in the order of
␦
⌬
͑mm/s͒
͑mm/s͒
0.02͑1͒
•••
Ϫ0.28͑2͒
0.37͑5͒
0.40͑2͒
Ϫ0.30͑3͒
0.50͑3͒
0.50͑5͒
Ϫ0.30͑5͒
1.5 ͑1͒
0.45͑5͒
W ͑mm/s͒
0.38͑2͒
15
2
5
ϫ10 at./cm . This also confirms the above picture that
a͒References 7 and 9.
silicide dissolution cannot be the main reason for the ham-
pering of the formation of a buried layer as observed for the
first sample since the trapping sites are too small in number
to play a significant role.
total area of the internal surface. This confirms that voids are
thermally much more stable defect structures than intrinsic
defects, changing only their morphology and density upon
longer, respectively higher, annealing. The M o¨ ssbauer spec-
tra did not change until annealing at 900 °C was performed.
The spectra could then no longer be fitted with just the
Although no systematic studies on the thermal stability
of silicide precipitates or epitaxial layers upon high tempera-
ture annealing for several hours have been reported, it is
known that they are very temperature resistant. This does not
mean that the Co is not mobile during annealing. The above
results support the idea that, although Co is most strongly
bound in its silicide phase, cavity trapping still can hamper
Co from forming silicides and partially dissolve a pre-
existing silicide phase. The reason is that the silicides are in
local equilibrium with Co in solution according to an
Arrhenius law, so that the Co concentration near the silicide
phase is never zero. Cavities provide an additional driving
force due to the dependence of the equilibrium concentration
of Co in solution upon the fractional coverage of the internal
CoSi component. Figure 4͑c͒ shows the spectrum, fitted
2
with the CoSi component together with the two quadrupole
2
doublets, Q and Q , previously ascribed to two different
1
2
sites at the internal surface of the voids ͑see Table I for
fitting parameters͒.
This shows that Co can indeed be dissolved from a pre-
existing silicide phase in favor of cavity trapping. After this
observation, additional annealing during 25 h was per-
formed, using the same temperature steps as above. In Fig. 5,
the redistribution among the M o¨ ssbauer components is
given. Q1 populations of less than 10% are difficult to as-
certain since the hyperfine parameters are close to the ones of
the side resonance of the CoSi spectrum. In calculating
these numbers we took into account the differences in bind-
ing strength by using the following formula for the relative
1
surface. Only at high enough temperatures will the Co at-
oms, trapped in precipitates, become mobile enough for this
process, resulting in a redistribution of the solutes between
both trapping sites.
2
In summary, we qualitatively characterized the damage
region produced by high-dose He implantation in c-Si, fol-
lowed by desorption. Cavities are shown to be more ther-
mally stable upon annealing than defects resulting from self-
implantation. RBS results show that the immediate presence
of the voids hampers the formation of an epitaxial buried
layer suggesting that pinning of the diffusing silicide par-
ticles takes place. The M o¨ ssbauer results show that, next to
populations a for each component:
i
͑
I /f ͒
i i
aiϭ
,
⌺j͑I /f ͒
j
j
where I is the area of the corresponding resonance in the
i
M o¨ ssbauer spectrum and f is the recoilless fraction. From
this annealing sequence we see that both time and tempera-
ture influence the relative populations, but annealing at
Tу900 °C clearly increases a(Q ) considerably. The fact
that Q2 only shows up significantly after long annealing
i
this, Co can be actively dissolved from a pre-existing silicide
phase in favor of trapping at the voids after annealing at
moderately high temperatures.
i
1
S. M. Myers, D. M. Follstaedt, G. A. Petersen, C. H. Seager, H. J. Stein,
and W. R. Wampler, Nucl. Instrum. Methods B 106, 379 ͑1995͒.
V. Raineri, P. G. Fallica, G. Percolla, A. Battaglia, M. Barbagallo, and S.
2
U. Campisano, J. Appl. Phys. 78, 3727 ͑1995͒.
B. Mohadjeri, J. S. Williams, and J. Wong-Leung, Appl. Phys. Lett. 66,
3
1
889 ͑1995͒.
4
J. Wong-Leung, J. S. Williams, R. G. Elliman, E. Nygren, D. J. Eagle-
sham, D. C. Jacobson, and J. M. Poate, Nucl. Instrum. Methods B 96, 253
͑
1995͒.
5
6
7
8
9
J. Wong-Leung, C. E. Ascheron, M. Petravic, R. G. Elliman, and J. S.
Williams, Appl. Phys. Lett. 66, 1231 ͑1995͒.
W. Deweerd, R. Moons, J. Verheyden, S. Bukshpan, G. Langouche, and
H. Pattyn, Nucl. Instrum. Methods B 106, 252 ͑1995͒.
W. Deweerd, T. Barancira, G. Langouche, K. Milants, R. Moons, J. Ver-
heyden, and H. Pattyn, Phys. Rev. B 53, 16637 ͑1996͒.
J. F. Ziegler, J. P. Biersack, and U. Littmark, The Stopping and Range of
Ions in Solids ͑Pergamon, New York, 1985͒.
M. F. Wu, A. Vantomme, G. Langouche, K. Maex, J. Vanhellemont, J.
Vanacken, H. Vloeberghs, and Y. Bruynseraede, Nucl. Instrum. Methods
B 45, 658 ͑1990͒.
FIG. 5. Relative population ai of the Q1 and Q2 sites upon isochronal
annealing, first during 3 h and then during 25 h. The f fractions are calcu-
lated using the following values of the M o¨ ssbauer temperature:
␪M(CoSi ͒ϭ450 K; ␪ (Q )ϭ490K; and ␪ (Q )ϭ240 K.
2
M
1
M
2
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:
586 Appl. Phys. Lett., Vol. 69, No. 23, 2 December 1996 Deweerd et al.
30.102.42.98 On: Sun, 23 Nov 2014 17:34:31
3
1