Silicon nanowires grown on si(1 0 0) substrates via thermal reactions with carbon nanoparticles

Silicon nanowires grown on Si(1 0 0) substrates via
thermal reactions with carbon nanoparticles
S. Botti
a,
*
, R. Ciardi
a
, R. Larciprete
b,c
, A. Goldoni
b
, L. Gregoratti
b
,
B. Kaulich
b
, M. Kiskinova
b
a
ENEA, UTS Tecnologie Fisiche Avanzate, Via Enrico Fermi 45, 00044 Frascati, Italy
b
Sincrotrone Trieste, S.S. 14 Km 163,5 in Area Science Park, 34012 Basovizza (TS), Italy
c
CNR-IMIP, Zona Industriale – 85050 Tito Scalo (PZ), Italy
Received 4 November 2002; in final form 16 December 2002
Abstract
The effect of thermal processing at 1050 °C of a dispersed film of carbon nanoparticles deposited on a Si substrate
with a native SiO
2
layer has been studied by scanning electron microscopy and scanning photoelectron spectromi-
croscopy. It has been found that the thermal processing results in formation of pyramidal-shaped defects of 2–7 lm

with strongly reduced SiO
2
content with silicon wires of diameter ranging between 30 and 50 nm decorating the pyramid
walls. The nucleation of the Si nanowires occurs via reduction of the native oxide layer by the nanosized carbon
particles, without the need of metal catalysts and at temperatures relatively lower than that used in similar techniques.
Ó 2003 Elsevier Science B.V. All rights reserved.
1. Introduction
Silicon nanowires with diameter of several tens
of nanometers and length of tens of micrometers,
exhibit unusual quantum confinement effects and
interesting electrical and optical properties with
promising technological impact in the microelec-
tronic field [1–7]. Several techniques have already
been used for synthesis of silicon nanowires, such
as vapour–liquid–solid growth (VLS) catalysed by
a gold layer on Si [1], laser ablation of metal
containing targets [7] and, most recently, Ni-as-
sisted solid–liquid–solid process [3] and carbo-
thermal reduction of Fe-catalysed SiO
2
particles
[4]. Some of the proposed growth models show
that the formation of Si nanowires does not re-
quire the presence of a metal, because the top SiO
2
layer can play the role of a catalyst in the wire
nucleation [2–5].
We have already shown that Si nanowires can be
synthesised without metal catalyst by simple car-
bothermal reduction of the native oxide layer on a

Si(1 0 0) substrate, assisted by C nanoparticles
(CNPs) deposited by laser pyrolysis [8,9]. Anneal-
ing of the Si substrate covered by a dispersed layer
of CNPs produces square-like pyramidal voids
which act as nucleation sites for growth of Si
nanowires with diameter of 40–50 nm. The Si
nanowire production was favoured at low CNPs
Chemical Physics Letters 371 (2003) 394–400
www.elsevier.com/locate/cplett
*
Corresponding author. Fax: +39-06-9400-5312.
E-mail address: (S. Botti).
0009-2614/03/$ - see front matter Ó 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0009-2614(03)00090-3
coverage, when the carbothermal reduction oc-
curring at the CNPs/substrate interface is termi-
nated after complete consumption of CNPs in
contact with Si. On the contrary, when the CNP
density is high, the excess of nanoparticles reorga-
nise into self-assembled carbon nanotubes [10–12].
Here we used two laterally resolved techniques,
scanning electron microscopy (SEM) and syn-
chrotron radiation scanning photoelectron mi-
croscopy (SPEM), to obtain deeper insight on the
morphology, composition and the lateral distri-
bution of the chemical species on the surface
obtained under conditions that we believe to fa-
vour Si nanowire formation, i.e., annealing of the
low density CNPs/Si(1 0 0) interface at 1050 °C.
2. Experimental

The CNPs were prepared by CO
2
laser-induced
decomposition of acetylene–ethylene mixtures in a
flow reactor [8,9]. The CNPs were amorphous and
largely hydrogenated with a mean size of 50 Æ 30
nm. After synthesis, the CNPs were sprayed on the
Si(1 0 0) substrates and resistively heated up to
1050 °Cat10
À6
atm. The annealed samples were
analysed ex situ by SEM (Jeol 5400, resolution of
3 nm with an accelerating voltage of 30 kV) and
SPEM. The SPEM measurements were carried out
with the experimental station of the ESCA Mi-
croscopy beamline [13,14] at the synchrotron
radiation facility ELETTRA (Trieste, Italy). The
SPEM uses a Fresnel zone plate lens to focus
the photon beam into a submicrometer spot while
the emitted photoelectrons are collected by an
hemispherical analyser mounted at a grazing angle
with respect to the incident beam and the sample
normal. The system can operate in a photoelectron
spectroscopy mode from a microspot and in im-
aging mode. The images are obtained by scanning
the sample with respect to the focused beam col-
lecting simultaneously photoelectrons emitted
from a selected elemental core level. All experi-
ments reported here were carried out with a pho-
ton energy of 500 eV, lateral resolution of 0.12 lm

and overall energy resolution of 0.4 eV. The SPEM
measurements were performed on the Ôas receivedÕ
sample and after annealing in situ at 450 ° C.
3. Results
Figs. 1a and b shows SEM images obtained
after annealing a Si substrate covered with a low
density CNPs at 1050 °C. They reveal the pres-
ence of nanostructured filaments and tubules
(with diameter ranging between 30 and 50 nm)
inside square-like features (ÔsquaresÕ). These
ÔsquaresÕ are the basal plane of void defects, often
observed during carbonisation of Si surfaces
[15–17]. They are hollow inverted pyramids (base
at the interface, vertex in the substrate) formed by
{1 1 1} Si planes and usually have dimension be-
tween 2 and 7 lm. In the SPEM Si 2p maps (see
Figs. 1c and d), obtained by collecting the pho-
toelectrons corresponding to non-oxidised Si, the
ÔsquaresÕ appear as brighter areas (consisting of
four triangles of different grey level), which indi-
cates that locally the native SiO
2
layer has been
partially reduced. Fig. 1c also reveals higher
density of defects on the right side of the Si map
where they even overlap. This distribution reflects
the initial non-uniform distribution of the pristine
CNPs onto the surface. Although the nanostruc-
tures formed inside the defects cannot be resolved
by SPEM, their presence and composition can be

elucidated from the chemical maps and the pho-
toelectron spectra measured inside and outside the
ÔsquaresÕ.
The three left panels in Fig. 2a show raw maps
obtained collecting the C 1s photoelectrons and
the Si 2p photoelectrons emitted from the Si and
SiO
2
states, respectively. In these maps topo-
graphic artefacts obscure the real ÔconcentrationÕ
contrast: because of the photoelectron detection
geometry the emission from defect sides facing
(opposite to) the electron analyser is enhanced
(reduced). In fact, the contrast variations of the
SiO
2
and C maps is very similar to the topographic
maps, obtained by collecting the secondary elec-
trons with kinetic energies negligibly affected by
the Si or C core level emission and used for re-
moval of the topographic artefacts from the im-
ages [18]. The true C, Si and SiO
2
concentration
maps (right panels in Fig. 2), obtained by dividing
the raw maps to the topographic map, clearly
manifest the higher Si content inside the ÔsquaresÕ,
where the relative amount of SiO
2
and C is

S. Botti et al. / Chemical Physics Letters 371 (2003) 394–400 395
reduced. The higher elemental Si signal can be
easily explained assuming formation of Si-based
filaments formed via reduction of the oxide layer,
which decorate the pyramid walls. This result is
confirmed by the XPS spectra shown in Fig. 2b.
The Si 2p spectra from both regions contain a SiO
2
component (the broad band at 103.6 eV, shifted by
$4 eV with respect to Si 2p
3=2
(99.3 eV) and
Si 2p
1=2
(99.9 eV) peaks), which derives from
the native oxide layer terminating the Si substrate.
The major difference is that inside the ÔsquaresÕ the
SiO
2
=Si intensity ratio is $5 times smaller than
that of the surrounding areas, where the SiO
2
component is dominant. The C 1s spectra taken
inside and outside the pyramidal voids exhibit
similar lineshapes peaked at 284.7 eV and a bit
lower intensity inside (see the processed C map),
whereas a much higher C concentration should be
detected if the filaments are not Si-based but
originate from the CNPs as residuals or by-prod-
ucts. In fact, as already reported, much higher

CNP density is required for self-organisation of
these particles into carbon nanotubes [10–12].
The shape of the C 1s spectra change only
slightly upon sample heating to 450 °C which in-
dicates that the detected C cannot be related to
contaminants adsorbed upon exposure to atmo-
sphere, but is a constituent of the near-surface
layer. The BE position of the C 1s peak covers the
energy region corresponding to sp
3
and sp
2
hy-
bridised C emitting at 285 and 284.3 eV [19], and
the peaks measured both inside and outside the
pyramids appear intermediate between sp
3
and sp
2
hybridisation, indicating the presence of poorly
organised carbon species. This C 1s energy posi-
tion excludes the formation of stoichiometric SiC,
which has a component at a much lower BE
(282.6 eV) [20]. This is in accordance with the
Si 2p spectrum, where the SiC component should
appear at 100.4 eV [20].
Fig. 1. Pyramidal-shaped voids formed at the CNPs/Si(1 0 0) interface annealed at 1050 °C. (a) SEM image of pyramidal void and (b)
magnified view of silicon nanowires in the void. (c) (38:5 Â 38:5 lm
2
) SPEM maps taken on the Si 2p peak and (d) a magnification of

the same region (19 Â 19 lm
2
) measured with higher lateral resolution.
396 S. Botti et al. / Chemical Physics Letters 371 (2003) 394–400
The Si 2p image in Fig. 3 reveals the growth of a
relatively large structure, protruding out of the
merged pyramidal voids. This image is taken in a
sample region with a higher density of ÔsquaresÕ,
which coalesce in disordered patterns, resulting in
round voids and occasionally microstructures
which stick out of the ÔsquaresÕ. The spectrum of
this wire-like feature has a rather prominent Si
component at 99.3 eV, whereas the C 1s intensity is
further reduced compared to the spectrum (dashed
line) taken inside the voids shown in Fig. 2. This
confirms that the structures standing out of the
walls are almost C-free and that the detected car-
bon has mostly contamination origin.
Fig. 4 shows the SEM image of a sample ob-
tained by increasing the annealing time, showing a
dense film of silicon nanowires. The wires have
smooth surfaces and exhibit bends and kinks. The
newly formed wires adhere to the ones previously
formed and they mechanically twist or knit to-
gether.
4. Discussion
The SEM and SPEM analysis of the samples
annealed for a short time (Figs. 1 and 2) demon-
strates that the formation of the pyramidal-shaped
defects plays a crucial role in nucleation of Si

nanowires, which grow mostly along the pyramid
walls. These defects are typical features appearing
on the Si side of the SiC/Si interfaces [15–17]
during Si carbonisation, their origin being attrib-
uted to presence of oxygen and oxygen-related
defects in bulk Si wafers [21,22]. However,
microscopic imaging of the pyramidal defects at
Fig. 2. (a) (12 Â 24 lm
2
) SPEM maps taken collecting the Si 2p photoelectrons corresponding to elemental Si (Si) and SiO
2
(SiO2)
components and C 1s photoelectrons (C). The left and right panels show the raw and processed images, respectively. (b) Si 2p and C 1s
core level spectra taken inside (upper) and outside (lower) ÔsquaresÕ.
S. Botti et al. / Chemical Physics Letters 371 (2003) 394–400 397
SiC/Si interfaces revealed mostly smooth walls
without any evidence of nanosized structures
protruding out of the surface [15,21,22]. Often the
pyramidal defects are buried below a rather ho-
mogeneous SiC layer which connects the voids and
are considered as a source of Si atoms outdiffusing
from the hole bottom to the surface.
The results reported here do not show presence
of any homogeneous film covering the pyramidal
voids formed at the CNPs/Si(1 0 0) interface and
the SPEM Si 2p spectra do not contain any
spectroscopic feature related to SiC. Apparently
under these reaction conditions (annealing at 1050
°C in low vacuum) the native SiO
2

layer cannot be
completely removed. On one hand it acts as a
barrier for the direct interaction between hydro-
carbons and Si and formation of SiC, and on the
other hand, being exposed to ÔactiveÕ C containing
species, it is gradually reduced with production of
Si-based nanostructures and volatile CO. This
indicates that despite the similar dimensions (a
few microns) and surface density ($10
6
cm
À2
) the
defects observed in the present study should not
be identical to that formed at Si/SiC interfaces
[15–17].
In brief, the reaction mechanism can be de-
scribed as follows. The first reaction step is partial
reduction of the native silicon oxide layer accord-
ing to the scheme, SiO
2
þ C ! SiO þ CO, where C
stays for the highly active carbon nanoparticles.
Due to the high reactivity of carbon particles, in
this case the temperature required for carbother-
Fig. 4. (a) SEM micrograph showing a high density of Si
nanowires: (b) magnified view of the same sample.
Fig. 3. Top: (19 Â 19 lm
2
) Si 2p image of a region showing

merged pyramidal voids and a spiral wire protruding out from
them. Bottom: XPS spectra taken on the spiral wire and outside
the pyramidal voids (see arrows). The dashed spectrum is taken
inside one of the ÔsquaresÕ in Fig. 2 and is shown for the sake of
comparison.
398 S. Botti et al. / Chemical Physics Letters 371 (2003) 394–400
mal reduction is lower than that usually required
[4], leading to a sizeable SiO formation even at
temperature as low as 1050 °C. The formation of
voids seems to be triggered by the faster con-
sumption of the oxide layer below the adsorbed
nanoparticles. At the sites where severe carbo-
thermal oxide reduction occurs, the developed
surface stress might favour formation of cavities
with lowest energy {1 1 1} symmetry planes. Ex-
trusion of material during such collapse of the
Si(1 0 0) surface lattice seems to be the process
leading to nucleation of Si nanowires along the
{1 1 1} plane walls. Disproportionation of Si
monoxide seems to be the key reaction for the
formation of Si nanowires by thermal evaporation
of pure SiO [23] or mixed Si–SiO
2
[24] powders.
Similarly, in our case a possible reaction route
leading to the nucleation of the Si nanostructures
is the decomposition of silicon monoxide
2SiO ! Si þ SiO
2
at 1050 °C followed by the

precipitation of Si atoms which are expelled from
the mixture as crystalline nanowires. A direct re-
duction of the SiO
2
layer due to the reactive spe-
cies, eventually present in the residual atmosphere
of vacuum chamber, can be excluded as key step in
the formation of Si nanowires since it would act
homogeneously of the sample surface without
originating the observed thinning of oxide layer in
correspondence of the ÔsquaresÕ.
The XPS spectra reported in Figs. 2 and 3 are
a clear evidence that the Si nanowire are less
oxidised than the regions surrounding the
ÔsquaresÕ. After deconvolution of the Si 2p spectra
we determined an oxide thickness of $6.5 and
20

AA for the ÔsquareÕ and outside areas, using the
relationship d
oxide
¼ k
oxide
cosðaÞ ln½1 þðI
4þ
=I
0
Þ=c,
where I
4þ

and I
0
are the integrated intensities of
the Si and Si oxide components, k is the Si 2p
photoelectron mean-free path in SiO
2
(4.7

AA in
our case), a is the acceptance angle of the ana-
lyser with respect to the sample plane, and
c ¼ 0:7 is the intensity ratio of pure SiO
2
and Si
[25]. This result differs from the previously re-
ported 3 nm thick SiO
2
layer surrounding the Si
core of the Si nanowires grown with different
techniques [5], which is assumed to have a crucial
role in limiting the side growth of the nano-
structure.
5. Conclusions
Silicon nanowires can be successfully synthes-
ised without metal catalysts by annealing the Si
substrates at 1050 °C in the presence of active
carbon nanoparticles. The Si nanowire formation
is well described by the oxide-assisted growth
model. The present results have proved that in the
initial reaction stage the produced Si nanowires

decorate the walls of pyramidal voids, formed
during carbothermal reduction of the native SiO
2
layer covering Si substrates.
Acknowledgements
The authors wish to thank D. Lonza for his
technical assistance and C. Cepek for fruitful dis-
cussions.
References
[1] J.D. Holmes, K.P. Johnston, R.C. Doty, B.A. Korgel,
Science 287 (2000) 1871.
[2] Z. Zhang, X.H. Fan, L. Xu, C.S. Lee, S.T. Lee, Chem.
Phys. Lett. 337 (2001) 18.
[3] H.F. Yan, Y.J. Xing, Q.L. Hang, D.P. Yu, Y.P. Wang,
J. Xu, Z.H. Xi, S.Q. Feng, Chem. Phys. Lett. 323 (2000)
224.
[4] X.C. Wu, W.H. Song, K.Y. Wang, T. Hu, B. Zhao, Y.P.
Sun, J.J. Du, Chem. Phys. Lett. 336 (2001) 53.
[5] N. Wang, Y.H. Tang, Y.F. Zhang, C.S. Lee, I. Bello, S.T.
Lee, Chem. Phys. Lett. 299 (1999) 237.
[6] Y.F. Zhang, Y.H. Tang, H.Y. Peng, N. Wang, C.S.
Lee, I. Bello, S.T. Lee, Appl. Phys. Lett. 75 (1999)
1842.
[7] Y.F. Zhang, Y.H. Tang, N. Wang, D.P. Yu, C.S. Lee, I.
Bello, S.T. Lee, Appl. Phys. Lett. 72 (1998) 1835.
[8] S. Botti, R. Coppola, F. Gourbilleau, R. Rizk, J. Appl.
Phys. 88 (2000) 3396.
[9] S. Botti, A. Celeste, Appl. Phys. A 67 (1998) 421.
[10] S. Botti, R. Ciardi, M.L. Terranova, V. Sessa, S. Piccirillo,
M. Rossi, M. Vittori-Antisari, Appl. Phys. Lett. 80 (2002)

1441.
[11] S. Botti, R. Ciardi, M.L. Terranova, V. Sessa, S. Piccirillo,
M. Rossi, Chem. Phys. Lett. 355 (2002) 395.
[12] R. Larciprete, S. Lizzit, S. Botti, C. Cepek, A. Goldoni,
Phys. Rev. B 66 (2002) 121402R.
[13] L. Casalis, W. Jark, M. Kiskinova, D. Lonza, P.
Melpignano, D. Morris, R. Rosei, A. Savoia, A. Abrami,
C. Fava, P. Furlan, R. Pugliese, D. Vivoda, G. Sandrin,
S. Botti et al. / Chemical Physics Letters 371 (2003) 394–400 399
F.Q. Wei, S. Contarini, L. De Angelis, C. Gariazzo,
P. Nataletti, G.R. Morrison, Rev. Sci. Instrum. 66 (1995)
4870.
[14] L. Gregoratti, M. Marsi, M. Kiskinova, Synchrotron Rad.
News 12 (1999) 40.
[15] R. Scholz, U. G
€
oosele, E. Niemann, F. Wischmeyer, Appl.
Phys. A 64 (1997) 115.
[16] L O. Bjorketun, L. Hultman, I.P. Ivanov, Q. Wahab, J E.
Sundgren, J. Cryst. Growth 182 (1997) 379.
[17] J. Hofmann, S. Veprek, J. Heindl, J. Appl. Phys. 85 (1999)
2652.
[18] S. Gunther, A. Kolmakov, J. Kovac, M. Kiskinova,
Ultramicroscopy 75 (1998) 35.
[19] J. D

ııaz, G. Paolicelli, S. Ferrer, F. Comin, Phys. Rev. B 54
(1996) 8064.
[20] P. Reinke, D. Rudmann, P. Oelhafen, Phys. Rev. B 61
(2000) 16967.

[21] K. Chul, C. Park, J. Roh, K.S. Nahm, Y.H. Seo, J. Vac.
Sci. Technol. A 19 (2001) 2636.
[22] A. Leycuras, Appl. Phys. Lett. 70 (1997) 1533.
[23] Z.W. Pan, Z.R. Dai, L. Xu, S.T. Lee, Z.L. Wang, J. Phys.
Chem. B 105 (2001) 2507.
[24] N. Wang, Y.H. Tang, Y.F. Zhang, C.S. Lee, S.T. Lee,
Phys. Rev. B 58 (1988) R16024.
[25] F.J. Himpsel, F.R. McFeely, A. Taleb-Ibrahimi, J.
Jarmoff, G. Hollinger, Phys. Rev. B 38 (1988) 6084.
400 S. Botti et al. / Chemical Physics Letters 371 (2003) 394–400