You are on page 1of 3

a

a
a
a
a
a

Journal of Crystal Growth 95 (1989) 461—463 461


North-Holland, Amsterdam

ATOMIC LAYER EPITAXY OF CUBIC SIC BY GAS SOURCE MBE


USING SURFACE SUPERSTRUCTURE

Takashi FUYUKI, Michiaki NAKAYAMA, Tatsuo YOSHINOBU, Hiromu SHIOMI


and Hiroyuki MATSUNAMI
Deparimen: of Electrical Engineering, Kyoto University, Yoshidahonmachi, Sakyo, Kvoto 606, Japan

Cubic SiC was eptiaxially grown by gas source MBE for the first time utilizing the change in surface superstructures when the
source gases of Si
2 ~ and C2 ~2 were introduced alternately. The number of Si atoms forming the surface superstructure determines
the growth rate of the layer, which gives atomic layer control in SiC growth.

1. Introduction In this study, cubic SiC was epitaxially grown


at as low as 1000°Cby gas source MBE for the
Cubic SiC is an attractive material for elec- first time. When the source gases of Si2H6 and
tronic devices operating under severe conditions C2H2 were introduced alternately, the surface
(e.g., high temperature, or radiation field, etc.). superstructure was found to change periodically.
since it is chemically and physically stable, and The number of Si atoms forming the surface
has a wide bandgap of 2.2 eV s.with
2/V~ high electron
Development of superstructure
gives determines
atomic level control intheSiC
growth rate, which
growth.
mobility
device up to 1000
applications hascm
been delayed becuase only
small platelet single crystals were obtained by
former growth techniques such as the sublimation 2. Experiment
method. Recently, successful results [1,2] in
achieving a large single crystalline wafer have A conventional MBE system was used in this
brought about the promotion of such devices as experiment, which was equipped with a reflection
MISFETs [3] and MESFETs [4]. Those pioneer high-energy electron diffraction (RHEED) ap-
works used cubic SiC wafers heteroepitaxially paratus for in situ and real-time diagnosis of the
grown on Si by chemical vapor deposition (CVD) reconstruction of surface superstructures. The
with the carbonization process as proposed by the gases of Si
2H6 and C2H2 were used for Si and C
present authors’ group. Relatively high substrate sources, since these gases decompose at relatively
temperatures around 1350°C are needed. A low temperatures. The source gases were intro-
reduced growth temperature is required to in-
crease the feasibility of device fabrication proc- one cycle I
esses. The molecular beam epitaxy method is c2I’lz — ‘H i--~ ~

expected to get high-quality single crystals at low Close Open Close Open
substrate temperatures utilizing the reactions be- 5j2H6 ‘~ ~-~r--~--
~-~~—--i--i0
tween an active substrate surface and source beams pen ose pen ose
under high-vacuum environment. There have been RHEED Uxi (3x2) (lxi) —

studies on SiC growth by MBE techniques [5,6], pattern (3x2) (lxi) (3~2) (lxi)
and single crystalline cubtc SiC was obtained at
1150°C by Si and C atoms evaporated by electron Fig. 1. Sequence of source supply and change of surface
beam. superstructure.

0022-0248/89/$03.50 © Elsevier Science Publishers B.V.


(North-Holland Physics Publishing Division)
462 T. Fuvuki ci cii. / A/.E of cubic SiC by GSMBE using surface .superslructure

duced alternately following the sequence shown in


fig. 1. Cubic SiC(100) grown ott Si by CVD method
was used as a substrate. Growth temperature was

~
10
~ ~
Torr. Flux was controlled by a variable leak
valve, and monitored by a quadrupole mass spec-
- ~~--~--
troscope and an ion gauge.
0 L=10O0~Cor 980’C
I
i-iD- -
~ T~=iO5O’C
3. Results and discussion

0 5 10 15 20 x10
Heat treatment of the SIC substrate at 1000 o C
under a base pressure of 10 to Torr cleans the (512Hs FLUX)x(SiiHc FLOW DURATION) [Torr secj
,
Fig. 3. Growth rate as a I unction o~total Si l~l~
suppls in one
surface, displaying the (1 X 1) pattern in RHEED cycle

as shown in fig. 2a. When Si2 H5 is introduced, the


RHEED pattern changes from (1 x 1) to (3 >< 2) the SLH~,supply is stopped. When C~H~ begins
(fig. 2b). The transition time of the surface super- to flow, the surface superstructure returns to the
structure decreases with increasing flux of Si2 Hi- initial (I X I ) pattern. The transition time to (I ><
The surface superstructure of(3 x 2) is kept after 1)was 5—100 s. and depended on the flux of CH~

~: Illia
(3x2)~

E.B.s~ [011] [011]


Fig. 2. Change of surface superstructures shown in RHFEE) patterns.

ii
a
a
a
a
a
a
a
a
a
a
a
a
a

T. Fuyuki et al. / ALE ofcubic SiC by GSMBE using surface superstructure 463

and the substrate temperature as well as on the surface superstructures when the source gases of
amount of Si2H6 supplied just before the C2H2 Si,H6 and C2H2 were introduced alternately.
cycle. After an appropriate number of cycles with When Si2H6 is introduced, the surface superstruc-
an optimum growth condition, single crystalline ture of the initial (1 x 1) pattern changes to (3 x 2),
cubic SiC was obtained with a smooth and miT- which is maintained after the Si2 H6 supply is
ror—like surface. stopped. When C2H2 begins to flow, the surface
The growth rate defined as the number of SiC superstructure returns to the initial (1 x 1) pat-
molecular layers per cycle was 2.3—3.2, and ap- tern. After an appropriate number of cycles, single
proached 1.7 with decreasing total amount of crystalline cubic SiC was obtained with a smooth
Si2H6 supply in one cycle at growth temperatures and mirror-like surface. The growth rate was
of both 1000 and 1050 °C(fig. 3). The surface determined by the number of Si atoms forming
superstructure of (3 X 2) contains 10 Si atoms in the surface superstructure, which gives atomic level
the unit mesh of that superstructure [7], while the control in SiC growth.
number of Si atoms on the top surface in the same
unit mesh is 6. Thus, the expected growth rate is
10/6 layers per cycle assuming that all the Si
atoms of the superstructure react with C2H2 to
form SiC. The extrapolated value (1.7 in fig. 3) in
the experimental results agreed well with the References
expected value (10/6). The excess growth rates
added to the expected value may be due to Si
atoms physically adsorbed on the surface super- [1] H. Matsunami, S. Nishino and H. Ono. IEEE Trans. Elec-
structure. The number of physically adsorbed [21iron Devices ED-28 (1981) 1235.
S. Nishino, J.A. Powell and I-l.A. Will. Appl. Phys. Letters
atoms is considered to increase linearly with the 42 (1983) 460
total amount of Si2H6, which yields a linear in- [3] K. Shibahara, T. Saitoh, S. Nishino and H. Matsunami,
crease of the excess growth rates as shown in fig 3. IEEE Electron Device Letters EDL-7 (1986) 692.
[4] S. Yoshida, H. Daimon, M. Yamanaka, E. Sakuma, S.
Misawa and K. Endo. J. App!. Phys. 60 (1986) 2989.
[51S. Kaneda, Y. Sakamoto, C. Nishi, M. Kanaya and S.
4. Conclusion Hannai. Japan. J. AppI. Phys. 25 (1986) 1307.
[6] T. Miyazawa. S. Yoshida, S. Misawa and S. Gonda. AppI.
Cubic SiC was epitaxially grown by gas source Phys. Letters 45 (1984) 380.
MBE for the first time, utilizing the change in [71 M. Dayan, J. Vacuum Set. Technol. A4 (1986) 38.

You might also like