Catalytic effect of carbon in shaping Si(1 1 1) surfaces

Article abstract of DOI:10.1016/j.susc.2004.05.135

This paper presents an experimental and theoretical study of the effect of carbon in the formation of pits on silicon surfaces. Experimentally we have searched for the best conditions to produce a homogeneous density of single shape pits in Si(111) using a C₂H₄ source. The result is that two steps are required: a rapid "precursor" process at relatively low temperature and high gas pressure and a longer "etching" step at higher temperature and lower gas pressure. Pits obtained with the two step process have a triangular shape that is not only due to the surface symmetry but also to the presence of carbon. In fact, we have found that the holes still remain after complete carbon removal by further annealing, but their shape becomes circular. Motivated by these experimental results we investigated theoretically the beginning of the pit formation mechanism using a tight binding method. We have performed two types of calculations: structural optimization of Si slabs containing carbon at 0 K and high temperature molecular dynamics simulations. We have found that carbon atoms prefer to occupy substitutional sites near the surface rather than adsorbed or interstitial sites. Their presence weakens some Si-Si bonds near the surface and therefore induces surface silicon atoms to move away from their equilibrium position and to diffuse on the surface. This may start the pit formation. We have also found that there is an attractive interaction between substituted carbon atoms and surface vacancy clusters, which must certainly influence the pit growth.

Full text of DOI:10.1016/j.susc.2004.05.135

Surface Science 563 (2004) 48–56
www.elsevier.com/locate/susc
Catalytic effect of carbon in shaping Si(1 1 1) surfaces
a,
a
a
*
Chu-Chun Fu , Jean-Jacques M ꢀe tois , Jean Pierre Astier ,
a
b
Andr ꢀe s Sa uꢀ l , Mariana Weissmann
a
Campus de Luminy, Centre de Recherche en Mati ꢁe re Condens ꢀe e et Nanosciences, CNRS, Case 913,
3288 Marseille Cedex 9, France
1
Comision Nacional de Energ ꢀı a Atomica, Avda. del Libertador 8250, 1429 Buenos Aires, Argentina
b
Received 27 January 2004; accepted for publication 19 May 2004
Available online 17 June 2004
Abstract
This paper presents an experimental and theoretical study of the effect of carbon in the formation of pits on silicon
surfaces. Experimentally we have searched for the best conditions to produce a homogeneous density of single shape
2 4
pits in Si(1 1 1) using a C H source. The result is that two steps are required: a rapid ‘‘precursor’’ process at relatively
low temperature and high gas pressure and a longer ‘‘etching’’ step at higher temperature and lower gas pressure. Pits
obtained with the two step process have a triangular shape that is not only due to the surface symmetry but also to the
presence of carbon. In fact, we have found that the holes still remain after complete carbon removal by further
annealing, but their shape becomes circular.
Motivated by these experimental results we investigated theoretically the beginning of the pit formation mechanism
using a tight binding method. We have performed two types of calculations: structural optimization of Si slabs con-
taining carbon at 0 K and high temperature molecular dynamics simulations. We have found that carbon atoms prefer
to occupy substitutional sites near the surface rather than adsorbed or interstitial sites. Their presence weakens some Si–
Si bonds near the surface and therefore induces surface silicon atoms to move away from their equilibrium position and
to diffuse on the surface. This may start the pit formation. We have also found that there is an attractive interaction
between substituted carbon atoms and surface vacancy clusters, which must certainly influence the pit growth.
ꢀ
2004 Elsevier B.V. All rights reserved.
Keywords: Low index single crystal surfaces; Carbon; Silicon; Catalysis; Etching; Molecular dynamics; Atom–solid reactions
1. Introduction
In recent years there have been many experi-
mental works studying the interaction of carbon
atoms with silicon surfaces. Their aim has been
twofold: either to eliminate carbon contamination
*
Corresponding author. Present address: Service de Recher-
ches de Mtallurgie Physique CEASaclay, Bat 520 CEASaclay,
[
1,2] or to obtain size and shape controlled SiC
structures on silicon [3–9]. In this work we show
that a precisely controlled C gas source may pro-
vide a new method to shape the Si surfaces. The C
9
4
1191 Gif-sur, 91191, GifsurYvette, France. Tel.: +33-1-69-08-
3-49; fax: +33-1-69-08-68-67.
E-mail address: chuchun@cea.fr (C.-C. Fu).
0
039-6028/$ - see front matter ꢀ 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2004.05.135

C.-C. Fu et al. / Surface Science 563 (2004) 48–56
49
atoms act catalytically, as they can be removed
later, making the process a simple and cheap way
to structure Si surfaces.
2. Experimental setup
Using C gas as carbon source, the best
2
H
4
Previous studies have already reported that reg-
ular shape etch pits are obtained on Si surfaces by
reaction with different carbon sources. For exam-
ple, in 1985 Ishikawa et al. [1] obtained them by
annealing a carbon contaminated sample at 900 ꢁC.
In 1990 Yang and Williams [2] observed etch pits in
different silicon surfaces when carbon was depos-
ited by desorption from the vacuum chamber walls.
conditions to produce well defined holes of con-
trolled size and shape on Si(1 1 1), have been sys-
tematically searched for. Experiments have been
performed in an ultra high vacuum chamber, the
3
sample being a small piece (20 · 2 · 0.3 mm ) of a
silicon (1 1 1) wafer. After ex situ cleaning in eth-
anol and hot acetone the sample is flashed in situ
ꢀ
10
(under 10 Torr) by heating at 1250 ꢁC and then
annealed at 1150 ꢁC for 1 h, in this way large
atomic flat areas are obtained. A direct current of
3 8
In 1999 Pacheco et al. [3] used propane (C H ) as
gas source and obtained a thin SiC film. Triangular
and square voids appeared on Si(1 1 1) and Si(0 0 1),
respectively, corresponding to the different surface
symmetries. Some other works have used pyrolitic
graphite as a solid carbon source, which is evapo-
rated by electron beams [10,11].
ꢂ
about 2–3 A and a voltage of around 6–7 V are
used in the heating process. High purity C H is
2 4
then introduced in the ultra high vacuum chamber
by a controlled leak system. The doses can be
controlled varying the gas partial pressure (be-
ꢀ
9
ꢀ6
Most of the works that produce SiC thin films
tween 10 and 10 Torr) or the exposure time.
The sample is later studied ex situ at room tem-
perature using optical microscopy, scanning elec-
tron microscopy (SEM) and atomic force
microscopy (AFM).
2 2 2 4
use either C H or C H gas as carbon source.
These experiments indicate that the molecules dis-
sociate, leaving only atomic carbon on the surface
[
4–9]. De Crescenzi et al. [9] have shown that
crystalline SiC can be pseudomorphically grown on
Si(1 1 1) when the substrate temperature is between
600 and 700 ꢁC, while at higher temperatures a
carbonization process occurs with carbon atoms
3. Method of simulation
also occupying non-substitutional sites.
In order to simulate the atomic scale processes
involved in the experiment described above we
have used a 12 layer Si slab with a two dimensional
unit cell of 16 atoms per layer to represent the Si
surface. C atoms are initially deposited in different
sites of this slab and periodic boundary conditions
in the directions parallel to the surface are applied.
The Si–Si, Si–C and C–C interactions are de-
scribed using a density functional based non-
orthogonal tight binding Hamiltonian. In this
model the Hamiltonian and overlap matrix ele-
ments are obtained from pseudo-atomic orbitals as
a function of distance. Although only a minimal
basis set is used, the method has been proved to be
transferable, giving good results for clusters, sur-
faces and solids [20,21]. It uses only two center
integrals and calculates the total energy as the sum
of a band-structure term and a repulsive term, this
last one being parameterized with experimental
information. When performing the dynamical
simulations at finite temperatures or obtaining the
In spite of the large amount of experimental
work, the atomistic mechanisms involved in each
experiment are not completely understood. In the
case of a molecular gas source this implies several
processes: the surface mediated dissociation of the
molecule, the carbon migration, the hydrogen
desorption, and the carbon incorporation. Many
ab initio calculations have been performed to
study the energetics of C adsorption versus incor-
poration in near surface sites. Most of them were
performed for the (0 0 1) orientation [12–18] and
they all agree that carbon atoms prefer to be
substituted rather than adsorbed, although they
differ in which is the best substitutional site. Very
recently, the interaction with the (1 1 1) surface has
also been studied theoretically [19].
The purpose of this work is to understand the
etch pit formation mechanism on the Si surfaces,
both from the experimental and the theoretical
point of view.

50
C.-C. Fu et al. / Surface Science 563 (2004) 48–56
relaxed equilibrium structures, the atoms are
moved according to the Hellmann–Feynmann
forces and the equations of motion are integrated
using Verlet’s algorithm. Kinetic processes are
studied by constant temperature molecular dynam-
ics using a Berendsen’s thermostat [22], each sim-
ulation run being about 10 ps long.
4. Results
4.1. Experimental results
Table 1 lists some of the different experimental
conditions investigated. We describe them so as to
indicate how we arrived at the final procedure. In
case (I) the surface was exposed to the gas during a
long time and at a high temperature. Pits did not
appear, and the surface ended up covered by a C
film. In case (II) a short ‘‘precursor’’ treatment at
low temperature and high pressure [5] was fol-
lowed by annealing at high temperature and ultra
high vacuum (UHV) but again no pits were
formed. These results suggested that more steps of
gas exposure might be necessary to obtain a
homogeneous density of single shape and similar
size pits, and the results may depend on either the
temperature or the gas pressure at the precursor
stage. To test these hypothesis we considered sit-
uations (III)–(V). We noticed that while in (III)
only irregular shaped pits were formed, in (IV) and
(V) triangular pits were formed. Therefore we
concluded that when a rapid precursor process at
low temperature (either above or below the 7 · 7 to
1 · 1 transition temperature (830 ꢁC)) and rela-
tively high pressure is followed by a longer etching
step at higher temperature and lower pressure,
regular shaped pits can be successfully obtained.
Figs. 1 and 2 show the images of case (V) obtained
by optical microscopy and SEM, respectively. We
The relative stability of substitutional, adsorbed
and interstitial carbon has been investigated by
structural optimizations using successive quenches
and simulated annealing, starting from different
initial configurations. The bond orders or off-
diagonal elements of the Mulliken Overlap Popu-
lation Matrix, which are proportional to the bond
strength between different atoms [23], are calcu-
lated for the optimized structures. Calculations for
systems having different number of carbon atoms
require some approximation concerning the car-
bon chemical potential but calculations for sys-
tems with the same number of carbon atoms are
independent of this parameter. The adsorption or
inclusion energies of C are calculated using as
reference system the 192 atoms slab of Si and as
chemical potentials the cohesive energies of Si and
C in the diamond structure. If EðM; NÞ is the total
energy of a slab with M silicon atoms and N car-
bon atoms the formation energy of the slab with
carbon is
E ¼ EðM; NÞ ꢀ Eð192; 0Þ ꢀ ðM ꢀ 192Þl ꢀ Nl
Si
C
ð1Þ
Table 1
Representative cases in the systematic search of experimental conditions to obtain regular shaped pits on Si(1 1 1)
Case
(I)
(II)
700
(III)
900
(IV)
(V)
700
(VI)
(Fig. 3(b))
T (ꢁC)
P (10 Torr)
Time (sec)
880
1
700
1
700
1
ꢀ6
1
30
0.01
1
30
30
30
30
30
T (ꢁC)
P (10 Torr)
Time (h)
1000
1000
1000
1000
5
1000
1000
ꢀ8
6
5
5
4
6
4
6
3
5
4
21
T (ꢁC)
P (10
Time (h)
1130
1
1130
1250
ꢀ
10
Torr)
1
7
1
2
13
The first three lines are the temperature, C
2 4
H gas pressure and exposure time of the ‘‘precursor’’ step, the fourth to sixth lines
correspond to the ‘‘etching’’ step and the last three lines to the final UHV ‘‘annealing’’.

C.-C. Fu et al. / Surface Science 563 (2004) 48–56
51
Fig. 1. Optical microscopy images of (a) case (V) and (b) case (VI) described in Table 1.
Fig. 2. SEM image of (a) case (V) in Table 1, (b) a single pit (top view) and (c) the same pit tilted 30ꢁ. The horizontal size of the images
are 200 and 14 lm, respectively.
see in Fig. 2 that the pits are rather flat, their
outline is triangular in the three Æ1 1 0æ directions
with {1 1 1} facets at the bottom and limited at the
edge by three facets which have been indexed by
SEM and AFM. One pit is further observed by
SEM under two azimuths: top view and tilted. At a
tilt angle of 30ꢁ, one facet is observed on side view
proving that this facet is {1 1 3} in agreement with
the observations of Ref. [2]. This has been con-
firmed by AFM, where a scan in the Æ1 1 2æ direc-
tion, perpendicular to the side facet, allows to
measure an angle of around 30ꢁ in comparison
with the mean orientation of the sample. The pit
depth is about 0.5 lm.
was performed and the density of pits increased by
an order of magnitude. This is shown in Fig. 1.
In the SEM image of Fig. 2 we observe bunches
of monoatomic steps forming mounds of circular
shape between the pits. One of these is shown in
Fig. 3(a) using AFM. We believe that these
mounds are formed by Si atoms pushed out of
their equilibrium positions by C atoms, as their
shape is close to that of an equilibrium thin flat
pancake of pure silicon at 1000 ꢁC (see Ref. [24]).
We wish to emphasize that the polygonal shape
of the pits is related to carbon incorporation. In
fact, carbon can be completely removed by
annealing during 2 h at 1250 ꢁC and the pits be-
come circular. They are surrounded by a flange,
the outside of which is quite similar to that of the
mounds (see Fig. 3(b)). It is also interesting to note
that the bottom of the pit (the {1 1 1} facet) does
not present any circular step, in contrast to the
positive mound in Fig. 3(a). If no nucleation of
The polyhedral holes observed have sizes be-
5
tween 4 and 6 lm and their density is about 10
pits/cm . We have found that in order to increase
2
the pit density, keeping a homogeneous distribu-
tion, further annealing is required. In fact, in case
(
VI) annealing under UHV at 1130 ꢁC during 7 h

52
C.-C. Fu et al. / Surface Science 563 (2004) 48–56
ꢂ
Fig. 3. (a) AFM image of a Si mound. The diameter is about 5 lm, the height 60 A, the left edge angle 0.6ꢁ and the right one 2ꢁ.
(b) AFM image of a circular pit (after C removal). The diameter is about 10 lm, the depth 0.14 lm and the internal angle between 7ꢁ
and 8ꢁ.
new terraces takes place, the bottom facet remains
perfectly flat, and this has been used to obtain very
flat Si(1 1 1) surfaces on a large scale (100 lm · 50
lm) [25].
the second one we assume the existence of clusters
of surface vacancies, arranged in a way consistent
with the surface geometry, and calculate the
equilibrium energy and the bond orders for carbon
atoms in different positions in the slab. This allows
us to estimate the interaction of C atoms with the
vacancies and to infer the way these holes may
grow.
4
.2. Simulation results
The previously described experimental results
are performed only on Si(1 1 1). However, to be
able to compare our results with previous calcu-
lations we have simulated both Si(0 0 1) and
Si(1 1 1). Buckled dimers are present in the Si(0 0 1)
surfaces but no reconstruction is considered in
Si(1 1 1) as the pits are observed for two ‘‘precur-
sor’’ temperatures, one above and one below the
transition temperature to the 7 · 7 structure (see
Table 1).
4.2.1. First stages of pit formation: dynamics of C
atoms in perfect Si slabs
To study the very first stages of pit formation
we perform molecular dynamics simulations at
high temperature (T ¼ 927 ꢁC and T ¼ 1227 ꢁC)
for one adsorbed carbon atom on Si(1 1 1) and
Si(0 0 1). We find that in all cases the C atom be-
comes substitutional in the first layers practically
without diffusion. It pushes one Si atom to an
adsorbed position, from which it can diffuse on the
surface. This clearly shows that the diffusion bar-
riers for C on Si(1 1 1) and Si(0 0 1) are higher than
those for Si atoms. In all our simulations on
Si(1 1 1) at 927 ꢁC the carbon atom becomes sub-
stitutional in the first layer but at 1227 ꢁC it often
arrives at the subsurface layer. We believe that if
the temperature or the simulation time were in-
creased the probability of the C atoms to penetrate
further in the silicon slab would also increase.
To find the preferential positions for C atoms in
a slab we have calculated the energy of some high
symmetry adsorption and interstitial sites as well
Because the experimental evidence is that the
2 4
C H molecule dissociates at the surface [4–9],
leaving only atomic carbon on it we shall limit our
calculations to this situation. Moreover, some
previous and the present calculations show that
while single C atoms interact strongly with silicon
surfaces if a C2 molecule is formed it desorbs
quickly. Therefore, in the following our discussion
will be concerned only with deposited single C
atoms. Within these hypothesis we proceed along
two lines: in the first one we study the very first
stages of pit formation following the path of a
carbon atom when it is deposited on a Si surface,
using molecular dynamics at high temperature. In

C.-C. Fu et al. / Surface Science 563 (2004) 48–56
53
Table 2
Different configurations studied by 0 K energy relaxation for
the Si(1 1 1) surface
Table 3
Different configurations studied by 0 K energy relaxation for
the Si(0 0 1) surface
Configuration
Si(1 1 1) slab
Configuration
Si(0 0 1) slab
M
N
E (eV)
M
N
E (eV)
Substitutional in layer 1
Substitutional in layer 2
Substitutional in layer 3
Substitutional in layer 4
Adsorbed
191
191
191
191
192
192
1
1
1
1
1
1
0.8
0.5
1.3
1.2
2.5
4.7
Substitutional in layer 1
Substitutional in layer 2
Substitutional in layer 3
Substitutional in layer 4
Adsorbed
191
191
191
191
192
1
1
1
1
1
0.8
0.6
0.7
0.8
1.8 (over
dimer) 2.4
(far)
1.4
Interstitial between layers 2
and 3
Interstitial between layers 1
and 2
192
1
Tri-vacancy
Tri-vacancy + 1C in layer 1
189
188
0
1
1.5
1.4 (near)
2
.2 (far)
1.5 (near)
.6 (far)
Missing dimer
Missing dimer + 1C in layer 1
190
189
0
1
0.1
Tri-vacancy + 1C in layer 4
188
1
1.6 (near)
1.4 (far)
1.4 (near)
0.8 (far)
1.0 (near)
2
Four-vacancy
Four-vacancy + 1C in layer 3
188
187
0
1
1.4
2.1 (near)
Missing dimer + 1C in layer 2
Missing dimer + 1C in layer 3
Missing dimer + 1C in layer 4
189
189
189
1
1
1
2
.9 (far)
0
.7 (far)
0.7 (near)
.0 (far)
E is the energy calculated according to Eq. (1), M and N denote
number of Si and C atoms in the slab. In the case of one sub-
stitutional carbon in presence of surface vacancy clusters the
two energy values correspond to one substitutional carbon
being: (near) as close as possible to and (far) far from the
vacancies.
1
E is the energy calculated according to Eq. (1), M and N denote
number of Si and C atoms in the slab. In the case of one sub-
stitutional carbon in presence of a missing dimer the two energy
values correspond to one substitutional carbon being: (near) as
close as possible to and (far) far from the vacancies.
as the substitutional sites, for both Si(1 1 1) and
Si(0 0 1) slabs. Tables 2 and 3 show the results
obtained.
consistency, but their relative values are quite
reliable. Electronic charge is transferred from Si to
C and from the bulk to the surface atoms. In the
buckled Si(0 0 1) surface dimers it is transferred
from the lower to the higher atom. Bond orders,
being indicative of bond strengths, can show some
possible paths for bond breaking and subsequent
pit formation. For example, in Si(0 0 1) the lowest
energy configuration for two carbon atoms is when
they replace a surface dimer [12]. However, when
two substitutional carbon atoms are first neigh-
bors in the surface of a silicon slab the equilibrium
structure is such that their bond order is large,
The energy values are given with only 0.1 eV
precision, as the different simulated annealings
lead to minima that differ by this amount. It is
clear that systems with one substitutional carbon
atom have lower energy than those with one ad-
sorbed or interstitial carbon, the best site being in
the second layer for Si(1 1 1) and in the second (or
third) for Si(0 0 1) [12–18]. It must be recalled that
Si(1 1 1) presents bilayers, so that the first and
second layers are close to each other, while the
third one is much further away. When more than
one Si atom is replaced by C in the Si(1 1 1) unit
cell, the best configuration is a local SiC structure
in the subsurface layer, that is, carbons not being
nearest neighbors of one another.
2
about 1.2 as if a C molecule was formed, while the
bond orders of each C atom with its other neigh-
bors decreases with respect to that of a single
substituted carbon (from 0.40 to 0.36). This result
predicts what in fact happens in the high temper-
ature simulations, the carbon dimer desorbs easily
and leaves a surface dimer vacancy. This may be
one possible starting process for hole formation.
Tight binding calculations not only give total
energies but also provide information on the
electronic structure, such as charge transfers and
bond orders. Charge transfers are usually overes-
timated if the calculation does not include self-

54
C.-C. Fu et al. / Surface Science 563 (2004) 48–56
The bond order between Si atoms in bulk
almost degenerated with that of a neighbor site in
the first layer and the formation energy of the
triangular cluster of vacancies is appreciably re-
duced when a carbon atom is in one of those
positions.
The lowest energy configuration for a carbon
atom substituted in a Si(0 0 1) slab with a surface
dimer vacancy is in the fourth layer, symmetrically
located below the vacancy, as was also reported in
[26]. The formation energy of a dimer vacancy
is very small and it remains the same under
carbon incorporation, within the error of our
calculations.
Bond order analysis for Si(1 1 1) shows that
with three vacancies one third of the bonds be-
tween surface and subsurface atoms in the unit cell
change their bond order from 0.5 to 0.4, but not in
a symmetric pattern. However, if a carbon atom is
present in the fourth layer more bond orders de-
crease and they do so in a very symmetric way.
Also, the bond orders between the second and
third layer are further reduced, to 0.36.
silicon and in slabs is about 0.5 within this method
of calculation. A carbon atom substituted in the
silicon slab has a smaller bond order with its
neighbor silicon atoms, on the average 0.4. This is
different from bulk SiC, where the interatomic
distances are smaller and therefore the bond
orders are larger. A carbon atom substituted in the
third or fourth layer of Si(0 0 1) weakens the
bonding of a surface dimer close to it (from 0.5 to
0
.4) keeping however the reconstruction c(4 · 2)
and the dimer buckling. A carbon atom substi-
tuted in third or fourth layer of Si(1 1 1) weakens
one or two bonds between surface and subsurface
silicon atoms. This suggests a second possible
mechanism for pit formation: a weak bond breaks
up and leaves a surface vacancy and a mobile
silicon atom on the surface. This one may end up
in a mound.
4
.2.2. Interaction of substituted C atoms with
surface cluster vacancies: pit stability and growth
In this section we show, using 0 K energy cal-
culations, that the interaction between surface
vacancy clusters and nearby substituted carbon
atoms is attractive. This happens both for Si(1 1 1)
and Si(0 0 1) slabs but it is much more pronounced
for Si(1 1 1). It is therefore energetically favorable
to produce surface vacancies when the slab has
carbon already incorporated. We have only stud-
ied the configurations suggested by symmetry, that
is, the surface dimer vacancy in Si(0 0 1), as in [26],
and triangular vacancy clusters on the surface of
Si(1 1 1). These are of two types, either they have a
second layer atom in the center of the triangle, in
which case the cluster vacancy considered was of
four atoms, or it does not have it and the cluster
vacancy considered was of three atoms. Only a
single carbon atom was substituted in different
places of the unit cell.
For Si(0 0 1) the presence of a dimer vacancy
weakens the bonding of neighbor dimers in the
same row and also their interaction with second
layer atoms. Substituted carbon atoms enhance
these effects.
5. Discussion and conclusions
The experimental conditions that produce a
homogeneous density of pits in Si(1 1 1) of the
2 4
same size and shape, using a C H source, have
been found. These consist of a rapid ‘‘precursor’’
process and a second longer ‘‘etching’’ step. The
pit density can be after increased, keeping a
homogeneous distribution, by subsequent anneal-
ing at ultra high vacuum and high temperature.
The pits have a triangular shape that is not only
due to the surface symmetry but also to the pres-
ence of C, as when carbon is removed by further
annealing they become circular.
In this calculation we compare the energy of a
92 atom slab without vacancies with that of the
1
slab with the vacancies, in both cases with and
without one substitutional carbon atom. The re-
sult is obviously independent of the chemical
potentials, but the numerical value depends on the
cell size. For the three vacancy cluster on Si(1 1 1)
the more symmetric position in the fourth layer is
From these experimental results and those in
Ref. [5], the following mechanisms involved in the
reaction of C atoms with Si surfaces can be in-
ferred. If the molecule C H is adsorbed on silicon
2
4
at temperatures between 700 and 900 ꢁC it disso-

C.-C. Fu et al. / Surface Science 563 (2004) 48–56
55
ciates and leaves single carbon atoms deposited on
the surface. At lower temperatures the molecule
does not dissociate and at higher temperatures a C
film is formed. Those single carbon atoms on the
silicon surface initiate the formation of etch pits,
that increase in size if carbon is supplied continu-
ously and if the temperature is high enough for
silicon atoms to diffuse easily on the surface. In
fact, the pits grow when silicon atoms are re-
moved, diffuse on the surface and end up forming
mounds, probably near surface defects. At very
high temperature (1250 ꢁC) carbon disappears
from the vicinity of the surface, probably by dis-
solution in the bulk, and the pits change their
shape, which becomes close to the equilibrium
shape of a Si hole at the same temperature.
vacancy clusters, not comparable in size to the
experimental pits, but the bond order analysis al-
lows us to infer possible growth mechanisms, as
bonds are weakened symmetrically around the
carbon-vacancy cluster structure.
The possibility that carbon incorporation could
induce a new type of surface reconstruction
[19,27], was not considered, but we believe that the
same mechanisms are involved in such recon-
struction and in the pit formation: i.e. the C sub-
stitution, induced Si–Si bond breaking and Si
adsorption on the surface.
The C etching of Si and the elimination of C by
annealing can be a simple and cheap way, in
comparison with usual lithography, to shape the
silicon surface.
Using a tight binding molecular dynamics
method we have performed numerical simulations
in order to confirm the previously suggested
mechanisms. Only the very beginning of the pit
formation process was studied, as the size of the
experimental pits is outside our computational
possibilities. Calculations at finite temperatures
show that an adsorbed C atom has a higher
probability to incorporate in the substrate than to
diffuse on the surface. If another adsorbed carbon
atom comes close to the previously substituted one
they tend to form a molecule and desorbs, thus
starting the formation of a hole. At high enough
temperatures the C atoms move to subsurface
sites, and our calculations show that for Si(1 1 1)
the second layer is energetically preferred. How-
ever, to maximize the configurational entropy, the
atoms will penetrate further into the bulk and in
this case the bond order analysis shows that some
Si–Si bonds on the surface become weaker. These
have a large probability of breaking, thus pro-
ducing more pits and mounds.
Acknowledgements
We wish to acknowledge Drs. Christophe Bic-
hara, and Javier Fuhr for fruitful discussion and
Dr. Hernan Bonadeo for carefully reading the
manuscript.
References
[
[
[
1] Y. Ishikawa, N. Ikeda, M. Kenmochi, T. Ichinokawa, Surf.
Sci. 159 (1985) 256.
2] Y.-N. Yang, H. Williams, J. Vac. Sci. Technol A 8 (1990)
2
481.
3] F.J. Pacheco, A.M. S ꢀa nchez, S.I. Molina, D. Ara uꢀ jo, J.
Devrajan, A.J. Steckl, R. Garci ꢀı a, Thin Solid Films 343–
3
44 (1999) 305.
[
4] L. Simon, L. Kluber, J.L. Bischoff, D. Bolmot, J. Faur ꢀe , A.
Claverie, J.L. Balladore, Phys. Rev. 54 (1996) 10559.
5] I. Kusunoki, T. Takagaki, S. Ishidzuka, Y. Igari, T.
Takaoka, Surf. Sci. 380 (1997) 131.
[
[6] M. Ikeda, T. Muruoka, N. Nagashima, Surf. Sci. 416
1998) 240.
(
Energy calculations at 0 K prove that it is easier
to form a small surface vacancy cluster in Si(1 1 1)
if there is a substituted carbon atom close to it.
Also, because C atoms are attracted by surface
vacancies they will tend to introduce themselves
into the subsurface layers close to the pits and if a
C atom arrives to the third or fourth layer due to
thermal energy the probability of formation of
vacancy clusters nearby will be enhanced. The
calculations have been performed for very small
[
[
[
7] M. De Crescenzi, M. Marucci, R. Gunella, P. Castrucci,
M. Casalboni, G. Dufour, F. Rochet, Surf. Sci. 489 (2001)
1
85.
8] M. De Crescenzi, R. Bernardini, R. Gunella, P. Castrucci,
R. Pizzoferrato, G. Dufour, F. Rochet, Philos. Mag. B 80
(
2000) 669.
9] M. De Crescenzi, R. Bernardini, S. Pollano, R. Gunella, P.
Castrucci, G. Dufour, F. Rochet, Surf. Sci. 489 (2001) 185.
[
10] F. Scharmann, P. Maslarski, W. Attenberger, J.K.N.
Lindner, B. Stritzker, Th. Stauden, J. Pezoldt, Thin Solid
Films 380 (2000) 92.

56
C.-C. Fu et al. / Surface Science 563 (2004) 48–56
[
11] T. W o€ hner, V. Cimalla, Th. Stauden, J.A. Schaefer, J.
Pezoldt, Thin Solid Films 364 (2000) 28.
[19] X. Peng, L. Ye, X. Wang, Surf. Sci. 548 (2004) 51.
[20] R. Gutierrez, T. Frauenheim, T. Kohler, G. Seifert, J.
Mater. Chem. 6 (1996) 1657.
[
[
12] P.C. Kelires, E. Kaxiras, Phys. Rev. Lett. 78 (1997) 3479.
13] O. Leefeld, D. Grytzmacher, B. Muller, K. Kern, E.
Kaxiras, P.C. Kelires, Phys. Rev. Lett. 82 (1999) 972.
14] I.N. Remediakis, E. Kaxiras, P.C. Kelires, Phys. Rev. Lett.
[21] C.-C. Fu, J. Java, R. Weht, M. Weissmann, Phys. Rev. B
66 (2002) 045405.
[
[
[
[22] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren, A.
DiNola, J.R. Haak, J. Chem. Phys. 81 (1984) 3684.
[23] D. Pettifor, Bonding and Structure of Molecules and
Solids, Clarendon Press, Oxford, 1995.
86 (2001) 4556.
15] Ph. Sonnet, L. Stauffer, A. Selloni, A. De Vita, R. Car, L.
Simon, M. Stoffel, L. Kubler, Phys. Rev. B 62 (2000) 6881.
16] L. Simon, M. Stoffel, Ph. Sonnet, L. Kubler, L. Stauffer, A.
Selloni, A. De Vita, R. Car, C. Pirri, G. Garreau, D.
Auber, J.L. Bishoff, Phys. Rev. B 64 (2001) 035306.
17] Ph. Sonnet, A. Selloni, L. Stauffer, A. De Vita, R. Car,
Phys. Rev. B 65 (2002) 085322.
[24] J.M. Bermond, J.J. M ꢀe tois, Surf. Sci. 330 (1995) 48.
[25] P. Finnie, Y. Homma, Phys. Rev. Lett. 82 (1998)
2737.
[
[
[26] W. Kim, H. Kim, G. Lee, J.-Y. Koo, Phys. Rev. Lett. 89
(2002) 106102.
18] C.-L. Liu, L. Borucki, T. Merchand, M. Stoker, A.
Korkin, Phys. Rev. B 62 (2000) 5021.
[27] P. Castrucci, A. Sgarlata, M. Scarselli, M. De Crescenzi,
Surf. Sci. 531 (2003) L329.