Etching of hexagonal SiC surfaces in chlorine-containing gas media at ambient pressure

Article abstract of DOI:10.1016/j.susc.2006.03.011

The modification of the silicon carbide (4H-SiC) single-crystal surface in a chlorine-containing gas mixture at high temperature (800-1000 °C) and ambient pressure was investigated. The results of silicon carbide chlorination are found to strongly depend on the hexagonal surface orientation. Due to the thermodynamically more favorable reaction of chlorine with silicon rather than carbon, the C-terminated side ( 0 0 0 over(1, -) ) clearly undergoes considerable changes, resulting in coverage by a black-colored carbon film, whereas the Si-side (0 0 0 1) surprisingly remains visually untouched. With using X-ray photoelectron spectroscopy (XPS), angle-resolved XPS and SEM it is shown that this drastic change in behavior is associated with a different structure of oxicarbide/silicate adlayer formed on the C- and Si-terminated sides of silicon carbide surface during experimental pre-treatment and air exposure. The presence of oxygen bridges connecting the silicate adlayer with the bulk SiC in the case of Si-side inhibits the chlorination reaction and makes this surface strongly resistant to chlorine attack. Only some places on the Si-terminated side demonstrate traces of chlorine etching in the form of hexagonal-shaped voids, which are possibly initiated by distortion of the initial crystalline structure by micropipes. In contrast, a thin carbon layer resulted on the C-terminated side as a consequence of the chlorination process. XPS, ARXPS, SEM and Raman spectroscopy study of created film allows us to argue that it consists mainly of sp2-bonded carbon, mostly in the form of nanoscale graphene sheets. The absence of a protective oxygen bridge between the silicate adlayer and the bulk silicon carbide crystal leads to unlimited growth of carbon film on the SiC ( 0 0 0 over(1, -) ) side.

Full text of DOI:10.1016/j.susc.2006.03.011

Surface Science 600 (2006) 2242–2251
www.elsevier.com/locate/susc
Etching of hexagonal SiC surfaces in chlorine-containing
gas media at ambient pressure
*
A.V. Zinovev , J.F. Moore, J. Hryn, M.J. Pellin
Materials Science Division and Energy Systems Division, Argonne National Laboratory, 9700, South Cass Ave, Argonne, Illinois 60439, USA
Received 11 November 2005; accepted for publication 15 March 2006
Available online 4 April 2006
Abstract
The modification of the silicon carbide (4H-SiC) single-crystal surface in a chlorine-containing gas mixture at high temperature
(
800–1000 ꢀC) and ambient pressure was investigated. The results of silicon carbide chlorination are found to strongly depend on the
hexagonal surface orientation. Due to the thermodynamically more favorable reaction of chlorine with silicon rather than carbon,
the C-terminated side ð0001Þ clearly undergoes considerable changes, resulting in coverage by a black-colored carbon film, whereas
the Si-side (0001) surprisingly remains visually untouched. With using X-ray photoelectron spectroscopy (XPS), angle-resolved XPS
and SEM it is shown that this drastic change in behavior is associated with a different structure of oxicarbide/silicate adlayer formed
on the C- and Si-terminated sides of silicon carbide surface during experimental pre-treatment and air exposure. The presence of oxygen
bridges connecting the silicate adlayer with the bulk SiC in the case of Si-side inhibits the chlorination reaction and makes this surface
strongly resistant to chlorine attack. Only some places on the Si-terminated side demonstrate traces of chlorine etching in the form of
hexagonal-shaped voids, which are possibly initiated by distortion of the initial crystalline structure by micropipes. In contrast, a thin
carbon layer resulted on the C-terminated side as a consequence of the chlorination process. XPS, ARXPS, SEM and Raman spectros-
copy study of created film allows us to argue that it consists mainly of sp2-bonded carbon, mostly in the form of nanoscale graphene
sheets. The absence of a protective oxygen bridge between the silicate adlayer and the bulk silicon carbide crystal leads to unlimited
growth of carbon film on the SiCð0001Þ side.
ꢁ
2006 Elsevier B.V. All rights reserved.
Keywords: Silicon carbide; Single-crystal surface; Chlorine; Carbon; X-ray photoelectron spectroscopy; Raman spectroscopy; SEM; LEED
1
. Introduction
cessing of SiC articles. The high hardness, for example, in
which SiC is inferior only to diamond and boron carbide,
makes the mechanical treatment of this ceramic very expen-
sive and time consuming. Silicon carbide is also chemically
extremely stable and shows no dissolution even in boiling
acid (including aqua regia) and base solutions. Therefore
there is considerable importance in finding alternate meth-
ods for etching or modifying the surface of SiC chemically.
Since the etching of SiC surface in different solutions is
very difficult, the development of SiC etching in the gas
phase is considered. The main possible reagents in this case
are hydrogen (molecular or atomic) and halogens (mainly
chlorine). The two principal chemical reactions taking
place for SiC etching in a chlorine-containing gas mixture
include
Growing interest in the investigation of the properties of
silicon carbide and its modification is connected with the
promising perspectives of the use of SiC materials in modern
electronics [1] and also with the broad application of SiC
ceramics in industry [2]. SiC has a set of unique physical
and chemical features including high mechanical, chemical
and thermal stability which makes it irreplaceable and
advantageous in many applications. These outstanding
properties in some cases present new challenges in the pro-
*
Corresponding author. Tel.: +1 630 252 2852; fax: +1 630 252 9555.
E-mail address: zinovev@anl.gov (A.V. Zinovev).
0
039-6028/$ - see front matter ꢁ 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2006.03.011

A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
2243
SiC + 2Cl
2
! SiCl
4
+ C
ð1Þ
after the treatment of the samples was done using different
analytical techniques including X-ray photoelectron
spectroscopy (XPS), micro Raman spectroscopy, scanning
electron microscopy (SEM), ellipsometry, and low-energy
electron diffraction (LEED).
which is thermodynamically allowed, with a Gibbs energy
of DG = ꢀ434.1 kJ/mol, while the other reaction
SiC + 2Cl ! CCl + Si
has a low probability due to the instability of CCl at high
ð2Þ
2
4
The etching of SiC ceramics was conducted in a fused
silica reactor 380 mm tall and 76 mm in diameter vertically
placed in a furnace that could be heated to 1000 ꢀC. A
quadrupole mass-spectrometer (Inficon Quadrex-100) with
a sampling valve and differential pumping arrangement
was attached to the reactor for monitoring changes in the
gas composition during the treatment. The samples were
treated at different temperatures (800–1000 ꢀC) and within
4
temperatures [3].
As the result of these reactions, silicon carbide trans-
forms into carbon that may (depending on experimental
conditions) self-organize into different polymorphic forms
including graphite, amorphous carbon and even nanoscale
diamond [3]. This transformation dramatically changes the
surface properties of silicon carbide and represents a
new approach for manufacturing based on traditional
SiC technology.
However the real situation is more complex. Among the
many interesting features of SiC is very notable its ability
to exist in a variety of different crystallographic forms re-
ferred to as polytypism [1]. Along with the primary cubic
structure denoted as 3C-SiC (or b-SiC), numerous hexago-
nal (H-SiC) and rhombohedral (R-SiC) types exist and they
generally noted as a-SiC. The hexagonal and rhombohe-
dral crystalline structure leads to the anisotropic chemical
and physical properties of SiC along different crystallo-
graphic directions that may constrain the chemical etching
process. In modern electronics, the most widely used are
the hexagonal SiC single crystals of 4H and 6H polytypes
with the z-axis parallel to (0001) direction. In this direc-
tion, the SiC single crystal has a stacking sequence of
double layers of Si and C atoms which differ for 4H and
a wide range of active gases concentrations (Cl —0.2–7%,
2
H —0–3.5%, He—balance).
2
Initially, SiC wafers have only one (Si-terminated) side
polished, whereas the C-terminated side remains rough.
Therefore, before processing, both sides of the SiC samples
were mechanically polished using diamond paste followed
by etching in HF acid (40% solution) for 10 min to remove
a SiO layer from the surface. The cleaned samples were
2
placed into the quartz reactor which was pumped down
ꢀ3
to a vacuum better than 10 mbar and refilled with the
Cl -based gas mixture; the composition was monitored
2
with the quadrupole mass-spectrometer. After adjusting
the gas mixture so that the desired equilibrium concentra-
tions were reached, the reactor was heated with a tempera-
ture ramp rate of 25 ꢀC/min. The constant flow of the gas
mixture was maintained during the treatment process at the
3
level of 500 cm /min. The gas composition was monitored
at intervals of 100 ꢀC (corresponding to approximately
6
H-SiC [1]. The outer surfaces of the crystal are referred
4
min). After achieving a steady state temperature, the tem-
to as (0001) for the Si-terminated side and ð0001Þ for
the C-terminated side. The two sides have different surface
energies. Several studies were devoted to the oxidation of
silicon carbide surfaces and it was mentioned [4,5] that
Si- and C-terminated sides have different reaction rates
and oxide-layer structures, which include both silicon oxide
and different forms of oxicarbides and silicates. It is very
important to note that silicon carbide exhibits the presence
of a silicon oxide film after brief exposure to ambient atmo-
sphere and temperature. Silicon oxide is very resistant to
the chlorine attack and, it might be expected to passivate
the SiC surface and inhibit reactions (1) and (2). Neverthe-
less, numerous experiments devoted to the treatment of sil-
icon carbide in chlorine-containing gas mixture have shown
these reactions to be rather effective [3,7].
perature was maintained with an accuracy of ± 2 ꢀC while
monitoring the gas composition every 4 min.
The monitoring of the gas mixture composition showed
that up to 350 ꢀC the initial chlorine/hydrogen ratio was
maintained. Beginning at 350 ꢀC the hydrogen content in
the mixture strongly dropped (below the detection limit
ꢀ
12
of 10
mbar) along with the simultaneous reduction in
the chlorine content and the creation of HCl. At 900 ꢀC
in the gas mixture, a peak in the mass spectrum corre-
sponding to SiCl appeared. This composition is a fragment
of SiCl and is evidence that reaction (1) has begun. It is
4
important to note that no compounds of carbon with
chlorine were detected during the treatment process. The
amount of residual oxygen in the reactor chamber never
exceeded 10 ppm.
We present in this work a comprehensive study of the
mechanism of gas-phase etching of silicon carbide single
crystals in chlorine gas mixtures.
After the completion of the treatment for a given time,
the reactor was vented with 99.999% electronic grade
helium. The reactor was then cooled down to room temper-
ature in approximately three hours, and the samples were
subsequently extracted for spectroscopic analysis.
2
. Experimental
2
.1. Sample preparation and treatment procedure
2.2. Sample analysis
The samples were single-crystal SiC wafers (4H, N-
XPS spectra were recorded using MgKa (1253.6 eV) X-
doped, from Cree Inc.). The surface analysis before and
rays and a hemispherical electron energy analyzer (HEA).

2244
A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
The spectra were recorded in the fixed absolute resolution
mode of the HEA with a pass energy of 44 eV (for survey
spectra) and 22 eV (for detailed measurements of core level
peaks). The electrons were collected from an area with an
elliptical shape of 4 · 3 mm. Survey spectra with energy
step 1 eV and precision measurements of core level photo-
electron lines C1s, Si2p, O1s, Cl1s and valence band with
the energy step of 0.2 eV were recorded. The spectrometer
calibration was performed using the gold XPS emission
line (Au4f7/2 with binding energy 84.0 eV). Samples were
mounted on adjustable sample holder, which allows
changes to the takeoff angle of the detected photoelectrons
from 0ꢀ to 65ꢀ. In this way, angle-resolved X-ray photo-
electron spectroscopy (ARXPS) measurements were
performed for detailed study of SiC near surface depth-
dependent composition. The residual vacuum in the
ꢀ
10
analyzing chamber was 4 · 10
mbar. For depth profiling
+
of the samples, their surface was sputtered by an Ar ion
beam with the total output current 30 lA cm and beam
energy of 3 kV.
ꢀ
2
Processing of the obtained XPS spectra was performed
with CasaXPS. All measured peaks were corrected for
inelastic scattering by first subtracting the Shirley back-
ground from the raw spectra, followed by fitting of the
peaks by using the asymmetric pseudo-Voigt shape peaks
with different relative content of Gaussian and Lorentzian
components that were optimized for the best fitting.
Micro-Raman spectrometry was performed with a Ren-
ishaw system equipped with a He–Ne laser (k = 632 nm) to
ꢀ
1
obtain spectra of the samples in the range 400–3600 cm
.
The film cross-section and the treated surface were ana-
lyzed using scanning electron microscopy with a Hitachi
S-4700 microscope. By using a variable-angle spectroscopic
ellipsometry system (M2000V, Woollam Inc.) the SiO film
2
thickness on SiC before and after processing were mea-
sured. LEED patterns were obtained using the VG Micro-
tech RVL900 rear view instrument with 104ꢀ viewing angle.
Fig. 1. LEED (1 · 1) pattern for untreated SiC(0001) (a) and SiCð0001Þ
b).
(
2
.3. Obtained results
of ꢁ50% up to 65ꢀ of take-off angle. At the same time Si
content is strongly decreasing function of the angle, which
is more distinctly observed for the Si-side (Fig. 2a). The
oxygen content has the tendency to increase for both sides
but on the Si-terminated side it has a larger absolute value
and a more rapidly increasing rate (Fig. 2b). ARXPS data
can be used for the estimation of the film thickness, using
the relation
Prior to any treatment, the clean surfaces of the SiC
samples were analyzed. Both C- and Si-terminated sides
exhibited a well defined non-reconstructed (1 · 1) LEED
pattern (Fig. 1). However, the elemental surface analysis
conducted with XPS showed that the stoichiometric ratio
of carbon and silicon content is far from the bulk ratio
C:Si (1.00) and differs for the C- and Si-terminated sides.
For the C-terminated side SiCð0001Þ an excess of carbon
is observed and the C:Si ratio was equal to 1.35 whereas
the Si-terminated side was carbon depleted and C:Si was
dðlnðI
O
=ISiÞÞ
h ¼
ð3Þ
ꢀ1
dðcosðaÞÞ
0
.86. A significant amount of oxygen was also detected
where h is the film thickness (expressed in the units of elec-
tron mean free path), IO and ISi are the signals of oxygen
and silicon, respectively, and a is electron take-off angle.
Calculations based on the above expression gave a thick-
ness h = 0.67 and 1.16 for C- and Si-terminated sides
respectively. Assuming that the mean free path of electrons
in silicon oxide at the kinetic energy of 1000 eV is 2.0 nm
on both sides, corresponding to 3.5 and 5.5 at.% for the
C and Si sides respectively. ARXPS also demonstrates a
distinct difference between the Si- and C-terminated sides
of crystalline silicon carbide. The carbon content increases
for the C-terminated side and decreases for the opposite
side in the angle range 0–20ꢀ and stabilize its value at level

A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
2245
carbon, C-side
silicon, C-side
carbon, Si-side
silicon, Si-side
the C and Si-terminated sides. First of all this difference be-
came apparent in the ratio between two wide peaks having
maximums at ꢁ10 and 16 eV which can be connected with
Si3p and C2s levels [8–10] and the ratio between C2s
and Si3p in this case corresponds to C1s/Si2p ratio for
C and Si sides. The flat-topped peak at ꢁ30 eV maximum
is connected with the presence of oxygen and associated
with the O2s level [8]. Prolonged heating of the samples
in vacuo (850 ꢀC, 60 min) did not lead to significant
changes in the spectra. It is in agreement with the results
of [11] where was found that removal of oxygen from
SiC surface requires a temperature above 1000 ꢀC, which
unfortunately was not possible in our apparatus.
5
5
4
4
3
5
0
5
0
5
The position of the O1s maximum (533.1 eV) also corre-
sponds to the chemically bonded oxygen state (532.8–533.5
[
6]). Thus, it may be concluded that despite preparation by
0
10
20
30
40
50
60
etching in HF on both sides of the silicon carbide (C- and
Si-terminated), there remains some oxygen-containing film,
possibly a mixture of different oxicarbides, silicates and sil-
icon oxide. This conclusion is confirmed by ellipsometry.
The measurements of freshly etched SiC surface demon-
(
a)
Angle, deg
1
6
Oxygen
O on C side
O on Si side
1
1
1
1
1
1
5
4
3
2
1
0
9
8
7
6
5
4
3
strate the presence of a thin film. The model SiO /SiC
2
was used and gave the film thickness of 1.4 nm and
2.3 nm for the C- and Si-terminated sides respectively.
These data are within 8% of the previously noted estima-
tions based on ARXPS data. It is noted that high residuals
were obtained from the fit which can be evidence of the
optical properties of the existing film differing from those
of SiO2.
Treatment of a prepared SiC single crystal in a chlorine
atmosphere leads to very distinctive results for Si- and C-
terminated sides. Whereas the C-terminated side clearly
undergoes considerable changes, resulting in coverage
by a black-colored film, the Si side appears visually
untouched.
XPS spectra of the C-terminated side demonstrate that
it consists mainly of carbon (70–80%), silicon (ꢁ5%), oxy-
gen (ꢁ15%), and chlorine (ꢁ5%), which can be sufficiently
reduced after heating of the samples in UHV at 850 ꢀC for
30 min. The amplitude of the Si2p core level decays expo-
nentially with increasing treatment temperature (Fig. 5).
Detailed inspection of the C1s peak has shown that it
has a maximum at 284.5 eV which may be corresponded
to graphite or amorphous carbon structure [5] but has an
additional component at 285.6 eV which can be attributed
to an sp3 bonded carbon [12] or C–H bonding [13]. In con-
trast, the analysis of the C1s peak on the Si-terminated side
demonstrates that the composition there is dominated by
the feature with the binding energy of 282.8 eV (silicon car-
bide) with the addition of less than 5% graphite (284.4 eV)
(see Fig. 6). At the same time, the position of the Si2p peak
strongly corresponds to silicon oxide (103.3–103.5 eV, [5])
and no component, corresponding to SiC or silicon oxicar-
bide compositions were detected.
0
10
20
30
40
50
60
(
b)
Angle, deg
Fig. 2. Absolute content of carbon, silicon (a) and oxygen (b) in
dependence of the electron take-off angle measured by ARXPS spectra
for C1s, Si2p and O1s core level peaks amplitudes on SiC(0001) and
SiCð0001Þ sides.
the film thickness may be estimated as 1.5 nm for the
ð0001Þ side and 2.5 nm for the (0001) side of a silicon car-
bide 4H single crystal.
Detailed analysis of C1s and Si2p core level peaks dem-
onstrated similar shape and positions of these lines for both
sides (Fig. 3). The C1s core level peak can be fitted into two
main components with binding energies of 283.3 eV and
85.2 eV, whereas fitting the Si2p peak shows the presence
of components with the binding energies of 101.3 eV and
02.4 eV (the last is the small component with relative con-
tent less than 10%). These features are discussed in detail
below.
2
1
We should emphasize that neither observed C1s nor
Si2p core levels had a binding energy specific for the pure
SiC (282.5–282.7 eV and 99.5–100.15 eV [5,6]). The appro-
priate components in Si2p spectra can be associated with a
mixture of oxycarbide having the composition Si C O2
Despite of the absence of visible changes, the XPS spec-
tra of the Si-terminated side also demonstrates the drastic
changes in surface composition in comparison with the
4
4ꢀx
and silicon oxide SiO [5]. The study of conduction band
2
XPS spectra also has shown differences (Fig. 4) between

2246
A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
1
1
1
2000
1000
0000
C side, C 1s
6000
4000
2000
0
C side Si 2p
9
000
000
000
000
000
000
000
000
000
0
0
6
8
7
6
5
4
3
2
1
0
0
6
0
45
0
5
4
0
0
0
0
2
95
290
285
280
275
108
106
104
102
100
98
96
Binding energy, eV
Binding energy, eV
9
000
1
1
1000
0000
Si side Si 2p
Si side C 1s
8000
7000
6000
9
000
000
000
000
000
000
000
000
8
7
6
5
4
3
2
5000
0
5
4000
4
3
2
1
000
000
000
0
0
45
0
0
0
0
2
95
290
285
280
275
108
106
104
102
100
98
96
94
Binding energy, eV
Binding energy, eV
Fig. 3. Detailed ARXPS spectra of carbon C1s and silicon Si2p core level lines for SiC(0001) and SiCð0001Þ sides.
4
4
3
3
2
2
000
500
000
500
000
Conduction band
Si Side
2
C side
εF
0
0.0010
0.0012
0.0014
0.0016
0
-1
30
20
10
0
-10
1/T, C
Binding energy, eV
Fig. 5. Temperature dependence of Si2p signal for C-terminated SiC side,
Fig. 4. Conduction band photoelectron spectra for SiC(0001) and
SiCð0001Þ side.
treated in chlorine gas mixture of Cl
represents a linear approximation.
2 2
:H = 4:1 for 30 min. Solid line

A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
2247
2
2
2
1
.4
.2
.0
.8
Si2p, Si-side
2
1
1
000
500
000
282.8
Si-side
0
65
1.6
.4
1.2
1
5
00
0
2
84.4
0
0
5
1
.0
.8
0
0
0
2
2
90
288
286
284
282
280
278
0.6
Binding energy, eV
0.4
0.2
0.0
0
0
284.3
2
2
1
1
500
000
500
000
1
10
108
106
104
102
100
98
96
94
92
C-side
Binding energy, eV
Fig. 7. ARXPS spectra of silicon core level for Si-terminated side of SiC
treated in chlorine gas mixture of Cl :H = 4:1, processing temperature
50 ꢀC for 30 min.
2
2
8
285.6
5
00
0
2
90
288
286
284
282
280
278
Binding energy, eV
Fig. 6. C1s core level spectra of SiC for Si-side (upper graph) and C-side
lower graph) treated in Cl :H 4:1 mixture at 850 ꢀC for 30 min.
(
2
2
untreated sample. It became carbon-depleted (C:Si ratio
.72 after treatment at 800 ꢀC for 30 min) and the oxygen
0
content increased to ꢁ30%. The shape and positions of
C1s peaks remain the same as those of untreated samples,
but the Si2p line changes strongly. Fig. 7 shows ARXPS
spectra of the silicon core level line Si2p recorded at differ-
ent take-off angles. Obviously, it has two components at
Fig. 8. SEM image of carbon film on C-terminated SiC side, scratched by
metal tip. Carbon CDC film was obtained after a 30 min treatment of the
SiC sample in Cl :H mixture at 3:1 gas ratio and temperature 850 ꢀC.
2 2
1
01.3 and at 104.1 eV which represent Si C O2 and
4
4ꢀx
SiO respectively. It should be noted that the 104.1 eV com-
2
ponent rises with increasing take-off angle, which demon-
strates the increase of silicon oxide content near the
surface.
The carbon-rich film grown on the C-terminated side of
SiC upon close inspection looks ‘‘crumbly’’ and has a poor
adhesion with the substrate (in contrast to a film grown on
polycrystalline SiC, which demonstrated excellent hardness
and adhesion [3,14]). The film in this case could not be re-
moved by treatment in an ultrasonic bath, but could be
easily detached by scratching the surface with a metal tip
tinctive features (Fig. 9). Like graphite (which also shown
on this picture for comparison and was obtained from
industrial molded graphite samples) it has two pronounced
bands D and G (1328 cm and 1598 cm accordingly),
but also has two additional bands with maxima near
ꢀ
1
ꢀ1
ꢀ
1
0
ꢀ1
1480 cm (D band) and 1220 cm . The optimal fitting
of the spectra is shown in Fig. 9 by dotted lines. Being com-
pared with CDC Raman spectra, obtained by Gogotsi and
co-authors [19] (see insertion in this figure) it demonstrates
0
strong similarity, even in details. The origin of the D and
ꢀ1
(
Fig. 8).
Being examined by Raman scattering, the film demon-
strates a spectrum close to graphite [18] but has some dis-
1220 cm bands in the created carbon films is unclear, but
they are specific for CDC carbon [7,19]. In accordance with
0
[20] D band may be originated as contribution from the

2248
A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
during etching of silicon carbide in molten KOH [15] and
are likely connected with the presence of micropipes and
dislocations in the SiC single crystal.
D
G
'
D
G
10000
D
3. Discussion
'
D
8
000
000
000
000
0
1000 1200 1400 1600 1800
To understand these essential distinctions in behavior of
Si- and C-terminated sides of silicon carbide in chlorine
environment, it is important to draw attention to the differ-
ence in the initial chemical composition and the structure
of the topmost layers of both sides. Despite the fact that
both sides were polished and etched in HF before treat-
ment, stable oxygen compounds were observed on them.
It is very interesting that (according to XPS) the Si-termi-
nated side has a C-deficit, whereas the C-side has a Si-def-
icit. Since the bulk stoichiometry of SiC corresponds to the
equal ratio of silicon and carbon, the difference must be
caused by a specific stoichiometry of the oxidized surface
layer. Only the C-terminated side at small take-off angles
has a prevalence of silicon, but at the angles higher than
6
4
2
8
00
1000
1200
1400
1600
1800
2000
-1
Raman shift, cm
Fig. 9. Raman spectra of carbon film grown on C-terminated side of SiC
single crystal at 800 ꢀC during 30 min. The dotted lines indicate the fitting
peaks, while the dashed line is the spectra of a molded graphite sample. In
the inset, the data digitally copied with using Origin 7.5 software from [19]
are shown.
20ꢀ, the absolute signal due to silicon is less than carbon
(Fig. 2a). We would like to note that it can not be explained
by any systematic errors because after a few minutes of
sample surface sputtering by 3 kV argon beam, the Si/C
ratio became very close to a bulk 1:1 value, with no oxygen
signal. Thus, the topmost layers of silicon carbide consti-
tute a complicated mixture of different carbon and silicon
compounds which strongly determine the chemical activity
of the surface. More detailed information can be obtained
from the analysis of the core level spectra. Most impor-
tantly, we note that the carbon C1s peak shape (Fig. 3)
clearly reflects the presence of two different carbon compo-
nents which corresponds to oxicarbide species (283.4 eV)
and C–O–H bonds (286.3 eV) [5] and the amount of the lat-
ter increases toward the surface. It may be proposed that
C–O–H composition is the result of wet etching of silicon
carbide in the HF solution and following rinsing in water
when the dangling bonds of surface carbon atoms bind
with hydroxyl groups during processing.
phonon density of states in finite-size crystals of graphite or
contribution of C–H vibrations. It is very important to
note that on the Si-terminated side, the Raman spectra
have exclusively peaks specific for pure silicon carbide.
In contrast to the C-terminated side, the SEM image of
the processed Si-terminated side demonstrates that most of
the crystal surface in accordance with the Raman spectra
data did not have any significant visible changes. However,
there are some areas on the crystal surface which exhibit
distinctive traces of etching process and represent true
hexagon-shaped deep voids or cavities (Fig. 10). Some of
these voids appear to be very deep (the depth was beyond
the ability of the SEM to measure), but some have the
shape of an inverse pyramid with distinctly visible steps
on their walls. Such patterns were previously discovered
Recently it was mentioned that oxidizing of silicon car-
bide surface takes place just at room temperatures [16], so
it should not be surprising to observe silicon and oxygen
compounds on the freshly prepared (or etched) SiC surface
exposed in air even for a short time. The key issue in this
case is the difference in oxidation kinetics for ð0001Þ and
(0001) faces that leads to the difference in composition
and structure of the created oxide layer. As it was men-
tioned earlier [5], the initial oxidation step for silicon car-
bide crystals is the formation of Si C O , which can
4
4ꢀx
2
further oxidize to pure silicon oxide SiO . This reaction se-
2
quence, however, is specific for the oxidation of SiC at high
temperatures (>1000 ꢀC), whereas our samples remain at
room temperatures. Moreover, XPS core level Si2p spectra
(in contrast to carbon) demonstrate the domination of the
single component with binding energy 101.3 eV, which cor-
responds to oxicarbide species rather than silicon oxide [5].
It is important to note the different behavior of Si content
Fig. 10. SEM image of hexagonal-shaped voids on Si-terminated SiC side
after treatment in Cl gas mixture during 30 min at 900 ꢀC.
2

A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
2249
for Si- and C-terminated sides with increasing of the take-
off angles and also the clear excess of carbon in XPS spec-
tra in Fig. 2. This may be connected with the presence of
the carbon hydroxides (C–O–H) on the topmost surface
layer [5], which replace silicon washed out due to the etch-
ing in HF acid. At the same time, the different rate of sili-
con content decay with the take-off angle may be caused by
the different oxicarbide composition and its distribution
along the z-axis for Si- and C-terminated sides. As was
shown in [17] the silicate adlayer on the SiC surface has a
differing structure dependent on the crystallographic direc-
tion. Whereas on the C-terminated surface, the silicate ad-
layer mostly has a planar structure and is connected with
the substrate via Si–Si bonds, in the opposite direction Si
atoms are bridged with the substrate by oxygen bonds
chlorination. On the other hand, this amount of oxygen
does nothing to prevent chlorination of the C-terminated
side. Moreover, the increase of oxygen content in the gas
mixture up to the 10 (0.1%) has a negligible effect on
the results.
We suggest that this strong difference in chlorination
kinetics is directly connected with the difference of the
structure of the topmost oxicarbide layers for the Si- and
C-terminated sides of silicon carbide crystals. By using
quantitative LEED in [17] it was shown that Si atoms on
both sides are differently coordinated and have different
chemical bonding. On SiCð0001Þ, Si atoms together with
oxygen atoms create honeycomb structures, whereas lower
SiC layers remain undisturbed and bonded with oxicarbide
layers via C-Si bonds normal to the surface direction. So,
each Si atom on the surface in the (0001) direction has
three bonds with O atoms and only one with a C atom cre-
ꢀ
3
(
e.g. Si–O–Si). This may explain the excess oxygen ob-
served on the Si-terminated surface in comparison with
the C-terminated one (Fig. 2b). It should be noted that
both surfaces maintain periodic order (1 · 1) which corre-
sponds to bulk SiC. Hence, we can make a conclusion that
the chemical composition of SiC single-crystal surfaces
being mechanically polished and etched in HF consists of
ating a SiCO moiety. This Si–C bond is the ‘‘weak link’’
3
that is successfully attacked by chlorine at high tempera-
tures. The further reactions are followed by (1) and (2) with
the diffusion of chlorine through the created carbon species
and backward diffusion of reactant to the surface. Because
the S–C double layers are parallel to the ð0001Þ surface,
the most favorable organization of the remaining carbon
structures in this case are the graphene sheets with the same
spatial orientation. The sizes of these sheets may be
roughly estimated from their Raman spectra using the ratio
a layer of silicate/oxicarbide with an average thickness of
˚
1
5–20 A and a few monolayers of carbon hydroxides. It
is important that the silicate/oxicarbide layer has a different
structure for C- and Si-terminated sides, with the excess of
oxygen on the Si-terminal side.
˚
As was shown in the previous section, the treatment of
silicon carbide in chlorine-contained gas mixture leads to
strong difference in the resulting surface for the ð0001Þ
and (0001) basal planes. Whereas C-terminated side easily
converts to carbon, at the same time the Si-terminated side
demonstrates surprising resistance to chlorine attack. The
evident reason for this difference is the presence of a signif-
icant silicon oxide layer on the Si-terminated surface.
ARXPS spectra of Si core level spectra for this plane dem-
onstrate that topmost layer composition converts to silicon
oxide from oxicarbide initially present on the surface. This
silicon oxide layer is the protective cover and prevents fur-
ther chlorination of silicon carbide.
of G and D peaks [18] and results in a range of 10–30 A
which is in good agreement with the data obtained by small
angle X-ray diffraction [21]. These graphite layers are
weakly bonded with each other and also with the substrate,
explaining their poor adhesion in comparison with a previ-
ous study of SiC industrial ceramic [14].
In contrast with the ð0001Þ surface the topmost silicon-
containing layer on (0001) is connected with the lower SiC
double layers via an oxygen bridge, and so each surface Si
atom bonds with four O atoms creating a silicate. This sil-
icate layer is the ‘‘shield’’ that protects SiC against chlorine
attack during chlorination. Moreover, during treatment in
chlorine due to the presence of small amount of residual
oxygen in gas mixture, the topmost silicate layer, in agree-
ment with a previous report [5], slowly transforms into a
silicon oxide layer covering the SiC(0001) surface and fi-
nally blocking the reaction with chlorine. Fig. 7 clearly
demonstrates that in the upper layers of the treated
SiC(0001) surface, silicon oxide is the dominant species
and it gradually transforms into an oxicarbide/silicate layer
toward the bulk. This shield covers almost all of the
SiC(0001) surface area except for some defects around
micropipes. The etching of these specific defects leads to
the formation of remarkable hexagon-shaped caverns with
a step-like wall structure (Fig. 10). Some of the caverns
have a round orifice in the center, justifying our assignment
of these etch points to micropipes, which formation have
been observed in SiC single crystals growth process [22].
The borders of the pipes and surface defects break the crys-
tallographic symmetry of (0001) surface and naturally
The ð0001Þ plane as noted is easily converted to a car-
bon film, following an exponential growth rate with the
temperature (Fig. 5). To better understand the mechanisms
that lead to these strong differences in chlorination process
for both sides it is necessary to look into the role of O in
2
this process. During chlorination process the amount of
residual O in the reaction chamber did not exceeded
2
ꢀ
5
1
1
0
partial fraction (or even less). Nevertheless, the rela-
tive amount of oxygen on Si-terminated surface increased
up to 30% and also the film thickness, estimated by ellips-
˚
ometry increased to 190 A, which proves clearly that oxida-
tion proceeds in the reaction chamber during the
1
Because of the limited dynamic range of the quadruple mass-
spectrometer, direct signal ratio measurements were not possible and the
amount of residual oxygen was estimated indirectly by varying the
multiplier gain.

2250
A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
have a different silicate structure which is not chlorine-
resistant. As a result of the SiC etching, these hexagonal-
shaped voids with step-like internal structure are formed.
As was mentioned before, the formation of hexagonal-
shaped voids also was observed upon wet etching of silicon
carbide single crystals in molten KOH [15]. The difference
in the etching behavior between Si- and C-terminated
sides of SiC single crystal in these experiments was also
observed. In [23] the authors proposed that this difference
is related to the difference in surface free energy for the
most layers of the initial material. Both the C- and Si-
terminated sides of SiC, oxidized in ambient atmosphere
just at room temperature form silicate surface layers, but
with different organization. In the case of the silicate ad-
layer on C-side, each Si atom bonds with three O atoms
and the fourth is Si–C bonded, but the silicate on the Si-ter-
minated side connects with the topmost SiC bilayer via
oxygen bridge. The last structure leads to the stronger resis-
tance of the Si-terminated side to chlorine attack and only
the silicon oxide/silicate layer forms on this side during
chlorination. At the same time, on the C-terminated side
the thin layer of nanographite is grown. This layer has a
perfect surface structure with excellent conductivity but
poor adhesion to the substrate, which is likely caused by
a parallel orientation of the graphene sheet and SiC sub-
strate bilayers.
Our results clearly demonstrate the strong dependence
of the silicon carbide chlorination process on crystallo-
graphic orientation of its surface. We have also developed
an understanding of the otherwise wildly unpredictable
behavior that often takes place during processing of SiC
industrial ceramics via chlorination. Because industrial
ceramic consists of silicon carbide microcrystals of different
polytypes and random crystal orientations, the etching rate
becomes very non-uniform along the surface and precludes
obtaining reproducible results.
(
0001) and ð0001Þ SiC basal plates which have estimated
ratio of appropriate surface energies of more than two. The
surface with higher surface energy (in this case it is (0001)
surface) has a lower reaction rate than the opposite one,
which (along with the micropipes and dislocations) may
be attacked by the etchant more easy. Generally speaking
this explanation also may be applicable to our experiments
where the similar behavior was observed. Unfortunately
this approach is essentially qualitative, there being no gen-
eral quantitative relationships between the surface energy
and the surface reaction rate. Therefore the actual surface
mechanisms affected on the surface reaction kinetics should
be the subject of separate study in each particular case. We
strongly believe that in our experiments the direct reason of
the difference in the chlorination process of the Si- and C-
terminated sides of the SiC(4H) single crystal is the differ-
ence in surface oxide/oxicarbide structure. The stated
difference of the surface energies for the (0001) and
Acknowledgements
(
0001) SiC basal plates strongly connected with appropri-
ate surface structure and may be used as the additional
argument at the forecast of the surface behavior in the
chemically aggressive environment. From this point of
view, the observed stability of the Si-terminated side in
comparison of the C-terminated one to the chlorine attack
is expectable, but the specific mechanism of the chlorina-
tion is defined by the specific surface structure of the stud-
ied objects. The hypothesis proposed in this study stated
the major role of the SiC surface silicate structure in the
chlorine etching process and is in a good agreement both
with the other experimental data and general view to the
surface reaction kinetics.
This work is supported by the US Department of En-
ergy, BES-Materials Science, under Contract W-31-109-
ENG-38. The SEM analysis was performed in the Electron
Microscopy Center, Materials Science Division, Argonne
National Laboratory, Argonne, IL.
References
[
[
[
[
1] H. Morkoc, S. Strite, G.B. Bao, M.E. Lin, B. Sverdlov, M. Burns,
J. Appl. Phys. 76 (1994) 1363.
2] S. Somiya, Y. Inomata (Eds.), Silicon carbide ceramic—1, Elsevier
Applied Science, London and New York, 1988.
3] Y. Gogotsi, S. Welz, D.A. Ersoy, M.J. McNallan, Nature 411 (2001)
2
83.
4] Z. Zheng, R.E. Tressler, K.E. Spear, J. Electrochem. Soc. 137 (3)
1990) 854.
4
. Conclusion
(
[
[
[
5] B. Hornetz, H.-J. Michel, J. Halbritter, J. Mater. Res. 9 (1994) 3088.
6] C. Onneby, C.G. Pantano, J. Vac. Sci. Technol. A 15 (3) (1997) 1597.
7] S. Welz, Y.G. Gogotsi, M.J. McNallan, J. Appl. Phys. 93 (2003) 4207.
The treatment of a silicon carbide single crystal in chlo-
rine at high temperatures leads to dramatic differences in
the resulting surface compounds, dependent on the initial
surface orientation. Whereas for the SiC(0001), only a thin
silicon oxide layer was formed and no conversion of silicon
carbide into carbon structure was observed, for the oppo-
site C-terminated SiCð0001Þ side the chlorination caused
the complete chemical transformation of the initial mate-
rial. On this side, as the result of the selective etching of sil-
icon atoms, a dense carbon film consisting of nanographite
[8] M.A. Rooke, P.M.A. Sherwood, Carbon 33 (1995) 375.
[
9] L.I. Johansson, F. Owman, P. Martensson, C. Perrson, U. Lindefelt,
Phys. Rev. B 53 (1996) 13803.
[
[
[
10] M.-H. Lee, F.S. Ohuchi, J. Vac. Sci. Tech. A 19 (4) (2001) 2013.
11] G.N. Yushin, Z. Sitar, Appl. Phys. Lett. 84 (20) (2004) 3993.
12] J. Diaz, G. Paolicelli, S. Ferrer, F. Comin, Phys. Rev. B 54 (1996)
8064.
[
[
[
13] A. Avila, I. Montero, L. Galan, J.M. Ripalda, J. Appl. Phys. 89
(
2001) 212.
14] A.V. Zinovev, J.W. Elam, J.F. Moore, J. Hryn, O. Auciello, J.A.
Carlisle, M.J. Pellin, Thin Solid Films 469–470 (2004) 135.
15] M. Syvajarvi, R. Yakimova, E. Janzen, J. Phys: Condens Matter 11
(1999) 10041.
(
graphene) sheets is formed. This strong distinction of the
chemical activity of the different basal planes of SiC single
crystal is directly connected with the structure of the top-

A.V. Zinovev et al. / Surface Science 600 (2006) 2242–2251
2251
[
[
16] D. Dunham, S. Mehlberg, S. Chamberlin, P. Soukiassian, J.D.
Denlinger, E. Rothenberg, B.P. Tonner, Z.D. Hurich, J. Vac. Sci.
Tech. B 21 (2003) 1876.
[20] J. Schwan, S. Ulrich, V. Batory, H. Ehrhardt, J. Appl. Phys. 80 (1996)
440.
[21] E.A. Smorgonskaya, R.N. Kyutt, A.M. Danishevskii, S.K. Gordeev,
17] J. Bernhardt, J. Schardt, U. Strake, K. Heinz, Appl. Phys. Lett. 74
Carbon 42 (2004) 405.
(1999) 1084.
[22] T.A. Kuhr, E.K. Sanchez, M. Skowronski, J. Appl. Phys. 89 (2001)
[
[
18] F. Tuinstra, J.L. Koenig, J. Chem. Phys. 53 (1970) 1136.
19] D.A. Ersoy, M.J. McNallan, Y.G. Gogotsi, Mater. Res. Innovat. 5
4625.
[23] M. Syvajarvi, R. Yakimova, E. Jenzen, J. Electrochem. Soc. 146
(1999) 1565.
(2001) 55.