Surface Review and Letters, Vol. 5, No. 1 (1998) 181–186
c World Scientific Publishing Company
ATOMIC STRUCTURE OF HEXAGONAL
6H AND 3C SiC SURFACES
J. SCHARDT, J. BERNHARDT, U. STARKE and K. HEINZ
Lehrstuhl für festkörperphysik, Universität Erlangen-Nürnberg,
Staudtstr. 7, D-91058 Erlangen, Germany
The crystallography of hexagonal SiC surfaces prepared ex situ by chemical methods was investigated
by low energy electron diffraction (LEED) structure analysis. The surface morphology was analyzed by
considering mixtures of domains with different surface layer stacking geometries. On the 6H–SiC(0001)
surface ABCACB stacking is the dominating termination, covering about 80% of the surface. On 6H–
SiC(0001̄) all three possible surface stacking sequences are present. The (111) surface of a 3C–SiC film
sample shows linear stacking of layers with only one domain orientation present. In contrast to the
carbon-terminated 6H sample, the silicon-terminated surfaces are covered by an oxygen layer with the
adatoms bound on top of silicon.
our LEED structure analysis of the same surface11
where we found only ABCACB stacking present on
the surface.
1. SiC Growth, Polytypes and
Surface Morphology
The development of semiconductor technology based
on silicon carbide substrate material has drawn
considerable interest in the past decade.1,2 Due to
its unique physical properties SiC can be used for
high frequency and high power devices3,4 and even
at temperatures as high as 900–1000 K.5,6 However,
difficulties in growing SiC material of good enough
purity and crystallinity still hamper the industrial
application. The morphology and surface condition
of the substrate samples seem to play an important
role for the structure of the material obtained, especially when epitaxial films are grown.7,8 However, it
is not clear whether the step morphology determines
the polytype of the film grown as found for off-normal
aligned substrates7 or defects serve as seeds for a
certain polytype.8 In view of this apparent importance of the surface structure for SiC growth, several investigations of its geometric structure and
morphology have been carried out using low-energy
electron diffraction (LEED) experiments9–11 or scanning tunneling microscopy (STM).10,12–17 In our own
earlier studies10,12 of a 6H–SiC(0001) sample with
STM we have found the surface morphology to be
governed by step bunching into triple or multiples
of triple steps. This is in complete accordance with
The basic structural element of SiC crystals is
a hexagonal bilayer containing silicon and carbon
in alternating positions. Different stacking of these
bilayers leads to many different polytypes known
for SiC. In a zinc blende structure all bilayers are
stacked in the same orientation resulting in the 3C,
the only cubic polytype. In a wurtzite structure
consecutive bilayers have opposite orientation. The
respective hexagonal polytype is labeled 2H. The 6H
polytype which is also hexagonal contains sequences
of three linearly stacked bilayers followed by an
orientation change. In Fig. 1 the 3C and 6H stacking sequences are shown in a cross-sectional view.18
The zigzag structure of the Si–C bonds visualizes
the stacking orientation of the planes.19 For the 6H
polytype Fig. 1 also shows that three different surface terminations may be present, distinguished by
the depth of the first orientation change. They are
labeled according to the number of linearly stacked
layers “S1,” “S2” and “S3”. In our LEED analysis11
of 6H–SiC(0001) both single terminations and mixtures of them with equal weight were tested, yielding
a clear best fit for the S3 structure. Obviously, the
other terminations, S1 and S2, are largely suppressed
181
182 J. Schardt et al.
Fig. 1. Stacking of bilayers in a cross-sectional projection parallel to the (112̄0) plane for the 6H and the (12̄1) plane
for the 3C structure. Linear stacking is present in the cubic (3C) polytype (β-SiC). The 6H unit cell contains six bilayers
with sequences of three linearly stacked bilayers followed by an orientation change. Three different stacking sequences
can be found on the surface which are labeled according to the depth of the orientation change (S1 = CACBABC,
S2 = BCACBAB and S3 = ABCACBA). Each termination can be present in two orientations rotated by 60◦ against
each other, thus accounting for the sixfold LEED pattern.
by the step structure. Yet, from the literature there
is evidence that the silicon side (0001) and the carbon
side (0001̄) of as-grown 6H samples show different
step morphologies.20,21 Therefore, we have extended
the parameter variation scheme of the LEED analysis using an automated search algorithm22 in order
to allow different mixing ratios to be tested in the
fit procedure. So, we can account for the possibility
that small amounts of terminations other than S3
may be present on the surface of the 6H–SiC(0001)
due to fluctuations of the step morphology. Besides
this reinvestigation of this structure we have newly
analyzed a 6H–SiC(0001̄) surface and a 3C–SiC(111)
film grown on Si(111).
2. Sample Preparation and LEED
Measurements
The samples used in this investigation were obtained from different sources. The 6H–SiC(0001)
was an epitaxial film grown by chemical vapour
deposition (CVD) on bulk grown substrate material. Its preparation as described previously10 involved a sacrificial oxidation and subsequent oxide
removal in HF. In addition, two different bulk 6H
samples in (0001̄) orientation were available for the
new experiments. They were also prepared by oxidation and HF etching. Due to the large lattice mismatch between the silicon substrate and the SiC film,
the 3C sample23 was very delicate and consequently
just briefly dipped in HF to remove natural surface
oxide.
After the chemical treatment the samples were
introduced into vacuum through a sample transfer
mechanism. With no further treatment, especially
without heating, all samples immediately showed a
(1 × 1) LEED pattern with a periodicity corresponding to the SiC bulk unit cell. Diffraction spot intensities were measured using a video LEED system.24
Due to a substantial radiation sensitivity of the
Atomic Structure of Hexagonal 6H– and 3C–SiC Surfaces 183
surfaces only a few minutes were available for the
measurement before the LEED pattern degraded.
After the inevitable time to adjust normal incidence
we used a video cassette recorder to store the LEED
patterns for a complete energy scan in order to avoid
too large electron doses. Intensity spectra of each individual spot were determined off-line. The similarity of data sets obtained from different samples of
the same orientation demonstrates the reliability
of the preparation procedure. However, LEED intensity spectra from different polarities, i.e. (0001)
and (0001̄) orientations, are clearly distinguishable
(see Fig. 3) and can be used as fingerprints to identify the orientation of a sample. It can be determined
from the symmetry of the LEED pattern whether domains in both possible orientations, i.e. rotated by
60◦ with respect to each other, of a certain stacking sequence are present on the surface. A single
domain in only one orientation would give rise to a
threefold-symmetric LEED pattern in a sense that
the (10) and (01) spots have different intensity spectra. A diffraction pattern rotated by 60 ◦ appears
from the other orientation. Provided the domains
are large than the coherence area of the primary
beam, their diffraction intensities are added together,
which — for a 1:1 orientation mixture — in effect
yields a sixfold-symmetric (LEED) pattern. For 6H
samples this is indeed observed, indicating that both
orientations of S1, S2 and S3 stacking sequences are
equally present on the surface. However, the 3C
sample shows a threefold-symmetric LEED pattern.
From the comparison of two inequivalent first order
diffraction spots, i.e. (10) and (01), we conclude that
no domain twinning is present in the film, i.e. all
layers have the same stacking orientation.
3. LEED Intensity Calculations,
Model Parameters and
Structural Search
Model intensities were generated from full dynamical
calculations25 which were supplemented by tensor
LEED26 in the case of the cubic sample. For
energies of up to 500 eV for the 6H–SiC(0001) and
400 eV for the other two surfaces, atomic scattering was modeled using 11 phase shifts.11 Electron
attenuation was considered by an energy-dependent
imaginary part of the optical potential, V0i ∼ E 1/3 .
As in the earlier study,11 bilayers were treated
as composite layers and their scattering matrices
combined by the layer-doubling scheme.
The geometry optimization included layer spacings between the first and the fourth bilayer (Dij )
and the thickness of the first bilayer (dB ). In addition, we considered the possibility of adatoms covering the first SiC layer in different sites, i.e. T1,
T4 and H3,11 or — in some models — substituting
atoms of the topmost half bilayer. Using the average T -matrix approximation (ATA),27 the concentration of adatoms could be tested including silicon or
carbon segregation in the topmost substrate bilayer.
The Debye temperatures were optimized separately
for each element using the best-fit structure obtained
for the 6H–SiC(0001) and resulted in 860 K for
carbon and 750 K for silicon and oxygen. Other nonstructural parameters, namely the real and imaginary parts of the optical potential, were fitted for
each data set individually. For practical reasons the
large number of parameters necessary to describe
the complex structural model prohibits a conventional scan of the parameter grid in a trial and error
method. Therefore, we employed a new automated
random sampling search algorithm described in
detail elsewhere.22
4. Surface Structural Results
The exact portion of S1, S2 and S3 domains on
the 6H–SiC(0001) surface was determined by mixing
the respective intensities, while adjusting the domain
geometries simultaneously. The best fit develops for
about 80% S3 termination, 15% for S1 and 5% for S2,
as shown by the behavior of the Pendry R-factor28 in
Fig. 2. The dominance of S3 domains is in agreement
with earlier work.11 On these S3 domains oxygen
atoms are bonded to the topmost Si atoms in T1
coordination, i.e. in top sites with a bond length
of 1.66 Å. From our earlier vibrational analysis10
we interpret these oxygen atoms as belonging to
hydroxyl species saturating the dangling bonds. Yet,
the scattering power of hydrogen is too small to
identify their existence in the LEED analysis in this
case. Due to the different bonding environment of
the topmost silicon atoms, the first bilayer is compressed to dB = 0.55 Å as compared to 0.63 Å in the
bulk. A slight expansion to D12 = 1.94 Å is found
for the spacing between first and the second bilayer.
The value obtained for D23 (1.88 Å) is practically
184 J. Schardt et al.
Fig. 2. RP -factor plots for the mixing ratios of domains with different stacking sequence, (a) for 6H–SiC(0001) and
(b) for 6H–SiC(0001̄). For a single point on each curve the ratio of one domain is fixed while all other parameters are
fitted. The best-fit ratios are indicated by vertical dashed lines.
Fig. 3. Comparison of selected (solid lines) and theoretical (dashed lines) best-fit I(E) spectra, (a) for 6H–SiC(0001)
and (b) for 6H–SiC(0001̄).
Atomic Structure of Hexagonal 6H– and 3C–SiC Surfaces 185
equal to the bulk spacing (1.89 Å). Though it is clear
from the variance of the R-factor [Fig. 2(a)] that
S3 domains do not fully account for the measured
intensities, a precise distinction between S1 and S2
is not possible due to the small contribution of those
domains to the LEED intensities. For the same
reason relatively large error limits apply for the
geometry parameters of S1 and S2 domains which
turn out to be bulklike within these error limits. It
is also not possible to determine if oxygen adatoms
are bonded to the S1 and S2 domains. The RP -factor
of 0.13 obtained in this analysis seems very good for
a compound material surface.
On the carbon-terminated sample,
6H–
SiC(0001̄), all three terminations are present with
more or less equal portions. The best-fit values
yielding an R-factor RP = 0.20 are 25% (S1), 30%
(S2) and 45% (S3), although deviations from a 1:1:1
mixture are not statistically significant, as shown in
Fig. 2(b). However, a predominant termination by
one type of stacking sequence as found on the (0001)
surface can be excluded (RP ≥ 0.35). Though we
failed to improve the fit by the inclusion of adatoms,
there are too many possibilities for a safe exclusion.
In particular, in order to avoid too slow convergence
of the search algorithm, we did not test mixtures
of three adatom-covered domain types. However,
we tried the most plausible models including oxygen
adatoms in high symmetry sites (T1, T4, H3) on
single terminations mixed with the other uncovered
terminations. In all cases the R-factors obtained are
above 0.40. Hydrogen atoms are practically of no influence on the intensity curves similar to the (0001)
case. A model of silicon adatoms in T1 position
yields RP > 0.50, which eliminates the possibility
of a half bilayer surface model. Within the limits
of error the best-fit geometry parameters have bulklike values, i.e. between 1.83 Å and 1.95 Å for the
spacings Dij between bilayers for all three termination. The first bilayer is compressed on all three
terminations to values of dB = 0.47 Å (S1) and
0.53 Å (S2 and S3) compared to 0.63 Å holding in
the bulk. In Fig. 3 a selection of I(E) spectra is
shown demonstrating the good agreement between
experiment and model calculations for both (0001)and (0001̄)-oriented 6H–SiC samples. In addition, it
can be seen that spectra from different sample orientations are clearly distinguishable, as mentioned
above.
For the 3C–SiC(111) film a bulklike substrate
geometry is found as best fit. A 50% coverage of
oxygen adatoms in top sites results in RP = 0.18
whereas both full and zero coverage yield R-factors
above 0.32. The topmost bilayer is uncontracted. No
twinning, i.e. presence of opposite stacking direction,
is observed, as already obvious from the LEED
pattern. Due to the lack of space, details of this
analysis will be given elsewhere.29
5. Conclusion
By full dynamical LEED structure analyses using an
automated search algorithm the structures of hexagonal 6H– and 3C–SiC surfaces were determined. On
6H surfaces there are mixtures of domains which
differ by the depth of the first layer orientation
switch. Both the relative weights of these domains
and their structural parameters could be determined.
Also, the polarity of the sample, i.e. whether (0001)
or (0001̄) orientation is present, can be clearly
distinguished. For 6H–SiC(0001), i.e. the siliconterminated surface, a single domain dominates
accompanied by a step bunching to triple or multiple triple bilayer heights, which is also observed
in STM.10,12 Silicon dangling bonds are saturated
by oxygen adatoms on top of silicon. This latter
feature applies also to the 3C–SuC(111) surface,
which otherwise is basically bulk-terminated. On
the 6H–SiC(0001̄) surface all three possible surface
terminations are present. Accordingly, one might
expect a single step morphology on this surface,
which indeed was suggested by growth experiments21
but has yet to be shown by STM.
Acknowledgments
We would like to thank Niels Nordell from IMC,
Kista, Sweden, who provided the 3C–SiC film
sample. This work was supported by the Deutsche
Forschungsgemeinschaft (DFG) through SFB 292.
U. S. gratefully acknowledges additional support by
DFG (Sta 315/4-2).
References
1. W. J. Choyke, in The Physics and Chemistry of
Carbides, Nitrides and Borides, ed. R. Freer (NATO
ASI Series, Manchester, 1989).
186 J. Schardt et al.
2. G. Pensl and R. Helbig, in Festkörperprobleme/
Advances in Solid State Physics, Vol. 30, ed. U.
Rössler (Vieweg, Braunschweig, 1990), p. 133.
3. M. Ruff, H. Mitlehner and R. Helbig, IEEE Trans.
Electron Devices 41, 1048 (1994).
4. R. J. Trew, J. B. Yan and P. M. Mock, Proc. IEEE
79, 598 (1991).
5. J. W. Palmour, H. S. Kong and R. F. Davis, Appl.
Phys. Lett. 51, 2028 (1987).
6. K. Dohnke, R. Rupp, D. P. J. Völkl and D. Stephani,
in Silicon Carbide and Related Materials, eds. M.
G. Spencer et al. (Institute of Physics Publishing,
Bristol, 1994), p. 625.
7. H. Matsunami, T. Ueda and H. Nashino, Mat. Res.
Soc. Proc. 162, 397 (1990).
8. J. A. Powell, D. J. Larkin, L. G. Matus, W. J.
Choyke, J. L. Bradshaw, L. Henderson, M. Yoganathan, J. Yang and P. Pirouz, Appl. Phys. Lett.
56, 1353 (1990).
9. R. Kaplan, Surf. Sci. 215, 111 (1989).
10. U. Starke, C. Bram, P.-R. Steiner, W. Hartner, L.
Hammer, K. Heinz and K. Müller, Appl. Surf. Sci.
89, 175 (1995).
11. J. Schardt, C. Bram, S. Müller, U. Starke, K. Heinz
and K. Müller, Surf. Sci. 337, 232 (1995).
12. U. Starke, J. Schardt, P.-R. Steiner, W. Hartner,
S. Müller, L. Hammer, K. Heinz and K. Müller, in
Proceedings: Symposium on Surface Science 3S’ 95,
eds. P. Varga and F. Aumayr (Inst. f. Allg. Physick,
Tu wien, Wien, 1995), p. 11.
13. F. Owman and P. Martensson, Sirf. Sci. Lett. 330,
L639 (1995).
14. C.-S. Chang, I. S. T. Tsong, Y. C. Wang and R. F.
Davis, Surf. Sci. 256, 354 (1991).
15. L. Li and I. S. Tsong, Surf. Sci. 351, 141 (1996).
16. M. A. Kulakov, P. Heuell, V. F. Tsvetkov and
B. Bullemer, Surf. Sci. 315, 248 (1194).
17. Y. Marumoto, T. Tsukamoto, M. Hirai, M. Kusaka,
M. Iwami, T. Ozawa, T. Nagamura and T. Nakata,
Jpn. J. Appl. Phys. 34, 3351 (1995).
18. R. Verma and P. Krishna, Polymorphism and
Polytypism in Crystals (Wiley, New York, 1966).
19. L. S. Ramsdell, Am. Mineral. 32, 64 (1947).
20. T. Kimoto and H. Matsunami, J. Appl. Phys. 76,
7322 (1994).
21. H. Matsunami, private communication.
22. M. Kottke and K. Heinz, Phys. Rev. B, submitted.
23. The sample was provided by Niels Nordell, IMC,
Sweden.
24. K. Heinz, Prog. Surf. Sci. 27, 239 (1988).
25. M. A. Van Hove and S. Y. Tong, Surface Crystallography by LEED (Springer, Berlin Heidelberg, 1979).
26. P. J. Rous, J. B. Pendry, D. K. Saldin, K. Heinz,
K. Müller and N. Bickel, Phys. Rev. Lett. 57, 2951
(1986).
27. Y. Gauthier, Y. Joly, R. Baudoing and J. Rundgren,
Phys. Rev. B31, 6216 (1985).
28. J. B. Pendry, J. Phys. C13, 937 (1980).
29. U. Starke, J. Bernhardt, J. Schardt and K. Heinz, to
be published.