Defect morphology and strain of CVD grown 3C-SiC layers: Effect of the carbonization process

Article abstract of DOI:10.1002/pssa.200460421

The heterostructure formed by cubic silicon carbide/silicon (3C-SiC/Si) is very promising as substrate for cubic III-N growth and for SiC devices. Optimized Si substrate carbonization process before the epitaxial growth of SiC, leads to a higher quality of the layer. In this paper, transmission electron microscopy is used to analyze the defect morphology and strain of SiC layers grown by chemical vapor deposition on two differently carbonized substrates, with tensile and compressive strain state. Misfit dislocations, stacking faults and antiphase domains are observed in this heteroepitaxial system. Irrespective of the substrate used, the epitaxial relationship between Si and SiC is good. However, the grain misorientation of the mosaic structure and the strain of the overgrowth layer depend drastically on the carbonization conditions of the silicon substrate.

Full text of DOI:10.1002/pssa.200460421

Original
Paper
phys. stat. sol. (a) 202, No. 4, 561–565 (2005) / DOI 10.1002/pssa.200460421
Defect morphology and strain of CVD grown 3C–SiC layers:
effect of the carbonization process
D. Méndez*, 1, A. Aouni1, 2, F. M. Morales1, 3, F. J. Pacheco1, D. Araújo1, E. Bustarret4,
G. Ferro5, and Y. Monteil5
1
Departamento de Ciencia de los Materiales e IM y QI, Universidad de Cadiz, 11510 Puerto Real, Spain
Faculté des Sciences et Techniques, BP 416 Tanger, Maroc
Max-Planck-Institut für Metallforschung, Heisenbergstr 3, 70569 Stuttgart, Germany
LEPES-CNRS, BP166, 25 av. des Martyrs, 38042 Grenoble 9, France
2
3
4
5
LMI, UCB Lyon1, 43 Bd du 11 Nov. 1918, 69622 Villeurbanne, France
Received 24 May 2004, revised 30 June 2004, accepted 15 July 2004
Published online 1 March 2005
PACS 68.35.Gy, 68.37.Lp, 68.55.Ln, 81.05.Hd, 81.15.Gh
The heterostructure formed by cubic silicon carbide/silicon (3C–SiC/Si) is very promising as substrate for
cubic III-N growth and for SiC devices. Optimized Si substrate carbonization process before the epitaxial
growth of SiC, leads to a higher quality of the layer. In this paper, transmission electron microscopy is
used to analyze the defect morphology and strain of SiC layers grown by chemical vapor deposition on
two differently carbonized substrates, with tensile and compressive strain state. Misfit dislocations, stack-
ing faults and antiphase domains are observed in this heteroepitaxial system. Irrespective of the substrate
used, the epitaxial relationship between Si and SiC is good. However, the grain misorientation of the mo-
saic structure and the strain of the overgrowth layer depend drastically on the carbonization conditions of
the silicon substrate.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1 Introduction
Improved 3C–SiC layers on Si substrates are promising templates for, either, high quality III-N growth
or SiC device development. Indeed, assuming 4.357 Å as the lattice constant (a) of 3C–SiC and compar-
ing it with the lattice constants of AlN and GaN reported in the literature [1], the lattice-mismatches are
1.1 and 3.5% respectively. This makes 3C–SiC a very important material to be used as a substrate for
growing III-N, instead of Si(001) or GaAs(001) whose lattice mismatches are around 20% [2, 3] with
respect to AlN and GaN. This high lattice mismatch relaxes via defects formation, degrading the crystal-
line quality of the layer. On the other hand, SiC is also suitable for simple and cheap devices such as
pressure, gas or temperature sensors working in harsh environments, or even microelectromechanical
systems [4].
It has been demonstrated that a proper carbonization step prior to epitaxial growth of 3C–SiC on sili-
con can lead to a higher quality of the layers [5]. In addition, the temperature at which propane is intro-
duced in the reactor during the carbonization plays a crucial role and affects the quality and the strain
state of the subsequent epitaxial layer [6, 7]. In Ref. [8] we have already reported that the microscopic
strain of the carbonized silicon varied from a tensile to a compressive state when the propane introduc-
tion temperature (Tintro) changes from below 1030 °C to 1060 °C.
* Corresponding author: e-mail: david.mendez@uca.es, Phone: +34 956 016 452, Fax: +34 956 016 288
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

562
D. Méndez et al.: Defect morphology and strain of CVD grown 3C–SiC layers
In this work we report the defect morphology and the microscopic strain of SiC layers re-grown by
chemical vapor deposition on two differently carbonized substrates with tensile and compressive strain
state, by means of transmission electron microscopy. Planar view and cross section transmission electron
microscopy (TEM) preparation were studied by Selected Area Electron Diffraction (SAED) and High
Resolution Transmission Electron Microscopy (HRTEM) modes to quantify the residual microscopic
strain evolution vs. the SiC thickness.
2 Experimental techniques
3C–SiC epilayers were grown by atmospheric pressure Chemical Vapor Deposition (CVD) in a vertical
cold wall reactor. Silane and propane were used as reactants and H2 as carrier gas. The standard two-
steps process involved 10 min carbonization of the substrate at 1150 °C under propane followed by an
epitaxial growth at 1350 °C by adding silane (see Ref. [9] for more details). Two sets of samples were
prepared over two different (001) carbonized silicon substrates with Tintro below and above the critical
value of 1060 °C. For both sets of samples, different SiC thicknesses were epitaxially grown at 1350 °C
to investigate the strain change vs. thickness. Samples L1, L2 and L3 had 60 nm, 250 nm and 3 µm thick
CVD grown SiC on a low Tintro carbonized layer (725 °C) while samples H1 and H2 had around 50 and
250 nm thick SiC on a high Tintro (1100 °C) carbonized layer.
Conventional TEM (CTEM) and HRTEM were used to analyze the quality of the interface and the
defects appearing in the epilayers. SAED patterns provided information regarding the crystalline quality,
the strain as well as the defects of the 3C–SiC grown layer. HRTEM analysis was carried out in a Jeol
2000 EX transmission electron microscope operating at 200 kV whereas SAED patterns and CTEM
micrographs were recorded in a Jeol 1200 EX TEM working at 120 kV. Specimens were prepared using
mechanical grinding and polishing until 70 µm thickness. After that, a cone-like hole was made on the
sample until the thickness in the center of the hole was approximately 5 µm. Finally, ion milling was
carried out using Ar+ ions to bombard the surface of the sample at a 10° tilt and a voltage of 5 kV until
electron-transparency.
3 Experimental results and discussion
The Si template used to grow epitaxially the 3C–SiC was, firstly, carbonized with two different Tintro
which, among other things, dramatically affects the crystalline quality of the carbonized layer. Carboni-
zation at low Tintro produces nearly monocrystalline layer consisting of small subgrains with a little
misorientation among them. If a high Tintro is used to the carbonization process, the subgrains have a
higher diameter and a higher misorientation, and the layers are nearly polycrystalline. In order to analyze
the crystalline quality of the re-grown layers, SAED patterns were taken in [001] zone axis. Figure 1a and
1b show diffraction patterns of SiC CVD layers over carbonized silicon with low (L1 sample) and high
Fig. 1 (online colour at: www.pss-a.com) SAED patterns on [001] zone axis of a) sample E17 and
b) E37. The insets show SAED patterns of the carbonized silicon used as substrate in each case.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Original
Paper
phys. stat. sol. (a) 202, No. 4 (2005) / www.pss-a.com
563
Fig. 2 (online colour at: www.pss-a.com) a) Cross section HRTEM micrograph in 〈110〉 zone axis of
sample L3 with low Tintro (725 °C) carbonization showing the SiC/Si interface and b) filtered image of a
part of the interface showing misfit dislocations.
(H1 sample) Tintro respectively. The alignment of SiC spots with respect to the Si spots is perfect for both
samples what shows the epitaxial character of the growth. However, SiC diffracted spots have curved
and elongated shapes, instead of a circular shape, indicating a pseudogranular structure in the SiC with
slight variations on their orientation (mosaic structure). If we compare the elongation of the same reflec-
tion for two different samples (see arrows in Fig. 1) we can get some qualitative idea about the misorien-
tation. The {400} SiC spots are more elongated in sample H1 than in L1 as a result of the higher misori-
entation of the underneath carbonized layer (see insets in Fig. 1a and 1b). However, the subgrain misori-
entation in sample H1 is much lower compared with the carbonized layer over which it has been grown
(inset in Fig. 1b). As no evidence for rings appear in its diffraction pattern, a reconstruction or reorgani-
zation of the carbonized silicon must occur during the epitaxial re-growth process.
In order to relax the stress produced by the high Si/SiC lattice mismatch (~20%), a high density
of misfit dislocations appears along this interface. Figure 2a shows a HRTEM micrograph of the
Si/3C–SiC interface along the 〈110〉 zone axis. In order to determine the dislocation type, the image of
the interface was filtered using a Fast Fourier Transform (FFT), applying a mask to the spots from Si and
SiC and obtaining the filtered image by inverse FFT as shown in Fig. 2b. It shows the linear misfit dislo-
cation density (see arrows) that correspond to one dislocation on each four {111} Si planes (and, then, on
each five {111} SiC planes). These dislocations are mainly Lömer (90° type) ones as observed in the
filtered micrograph, that are much more efficient than 60° type for relaxing the strain produced by the
20% lattice mismatch.
Furthermore, as observed in Fig. 2a, this interface is the origin of a high density of planar defects that
propagate in the {111} planes direction with an angle around 55 and 125° with respect to the {001}
planes. Figure 3a is a SAED pattern along 〈110〉 zone axis of the interface Si/SiC on sample L1. The
presence of some streaks connecting the SiC related spots (see arrows on Fig. 3a) indicates that the tilted
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

564
D. Méndez et al.: Defect morphology and strain of CVD grown 3C–SiC layers
Fig. 3 (online colour at: www.pss-a.com) a) SAED pattern on 〈110〉 zone axis of sample L1 showing
streaks coming from stacking faults and b) SAED pattern on [001] zone axis of sample L2 revealing the
presence of antiphase domains.
contrasts appearing on Fig. 2 can be associated to stacking faults. In addition, the sample with the thick-
est SiC epilayer grown over the low Tintro carbonized silicon, presents antiphase domains. Figure 3b
shows a SAED pattern in planar view orientation of that sample in a thin region where only SiC related
spots appear. Analysing in detail the lowest intense spots in the image, it can be seen that they consist of
a central spot and four more around (see inset). This effect can be associated to the presence of antiphase
domains.
SAED patterns as those in Fig. 1 were used to determine the local strain state of the samples. The
Si spots were used as a reference to calibrate the image and to calculate the 3C–SiC layer lattice param-
eter [10]. Using a3C–SiC = 0.4357 nm as the lattice parameter of the 3C–SiC, the strain is defined as
S3C–SiC = (a3C–SiC measured – 0.4357)/0.4357, being positive if the microscopic in-plane strain is tensile and
negative if it is compressive. The effect of the strain of the carbonized layers used as substrate is ana-
lyzed here. The local strain of those carbonized layers, measured using the same method, is +0.5 for the
low Tintro and –3.5 for the substrate carbonized with Tintro = 1100 °C. The results of the measurements are
summarized in Table 1.
When the carbonized layer presents a tensile strain (due to a low Tintro carbonization process) this is
conserved during the CVD growth with no variations during the first 5 minutes of growth. This tensile
state is the most extensively described in the literature for both the carbonized and epitaxially re-grown
layers [11–13]. However, if 3C–SiC is epitaxially re-grown over high Tintro carbonized silicon, the mi-
croscopic strain state of the layer can change as the re-growth process occurs. After one minute of CVD
growth, the lattice parameter of silicon carbide is the same as the nominal value. After five minutes,
it has changed to a tensile state and the residual strain is near the +0.3% value measured in the low Tintro
Table 1 Summarized results of the strain measured in the different samples.
sample
code
sample
description
temperature of propane
introduction (°C)
SiC thickness strain
(nm)
(%)
M769
L1
L2
carbonized Si
725
725
725
725
–
+0.4
+0.3
+0.3
+0.5
carbonized Si + CVD growth
carbonized Si + CVD growth
carbonized Si + CVD growth
60
250
3000
L3
M764
H1
H2
carbonized Si
carbonized Si + CVD growth 1100
carbonized Si + CVD growth 1100
1100
–
–3.5
–0.0
+0.2
~50
~250
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Original
Paper
phys. stat. sol. (a) 202, No. 4 (2005) / www.pss-a.com
565
samples. It seems then that the tensile in-plane strain is the most stable situation for 3C–SiC epitaxially
re-grown layer over carbonized silicon and those samples grown over compressive buffer layers prepared
under compression, tend to the tensile state as the growth process takes on.
4 Conclusions
In summary, the analysis of 3C–SiC CVD grown over two different carbonized silicon templates has
been reported. The crystalline quality of the SiC epilayers is remarkable in all samples, even in those
with nearly polycrystalline carbonized silicon as an intermediate buffer layer. Diffraction patterns reveal
the presence of both stacking faults and antiphase domains. They show also that the microscopic strain of
the SiC epitaxially re-grown is affected by the local strain state of the buffer layer. Samples grown over
low Tintro carbonized Si vary their strain as the CVD process occurs, changing from compressive to ten-
sile state. Further analyses are in progress in order to understand the mechanism of the strain relaxation
during the crystal growth.
References
[1] C. Stampfl and C. G. van de Walle, Phys. Rev. B 59, 5521 (1999).
[2] T. Lei, T. D. Moustakas, R. J. Graham, Y. He, and S. J. Berkowitz, J. Appl. Phys. 71, 4933 (1992).
[3] H. Okumura, K. Otha, G. Feuillet, K. Balakrishnan, S. Chichibu, H. Hamaguchi, P. Hacke, and S. Yoshida,
J. Cryst. Growth 178, 113 (1997).
[4] G. E. Carter, J. B. Casady, J. Bonds, M. E. Okhuysen, J. D. Scofield, and S. E. Saddow, Mater. Sci. Forum
338–342, 1149 (2000).
[5] S. Nishino, J. A. Powell, and H. A. Will, Appl. Phys. Lett. 42, 460 (1983).
[6] T. Chassagne, G. Ferro, C. Gourbeyre, M. Le Berre, D. Barbier, and Y. Monteil, Mater. Sci. Forum 353–356,
155 (2001)
[7] V. Cimalla, T. Stauden, G. Eichhorn, and J. Pezoldt, Mater. Sci. Eng. B 61–62, 553 (1999).
[8] E. Bustarret, D. Araújo, D. Méndez, F. M. Morales, F. J. Pacheco, S. I. Molina, N. Rochat, G. Ferro, and
Y. Monteil, Mater. Sci. Forum 457–460, 281 (2004).
[9] T. Chassagne, G. Ferro, D. Chaussende, F. Cauwet, Y. Monteil, and J. Bouix, Thin Solid Films 402, 83 (2002).
[10] C. Long, S. A. Ustin, and W. Ho, J. Appl. Phys. 86, 2509 (1999).
[11] S. Rohmfeld, M. Hundhausen, L. Ley, C. A. Zorman, and M. Mehregany, J. Appl. Phys. 91, 1113 (2002).
[12] J. Zhu, S. Liu, and J. Liang, Thin Solid Films 368, 307 (2000).
[13] F. M. Morales, Thesis, University of Cádiz (2003).
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim