surface science
ELSEVIER
Surface Science 352-354 (1996)71-76
Surface quality and atomic structure of MBE-grown GaAs(100)
prepared by the desorption of a protective arsenic layer
U. Resch-Esser a,* N. Esser a, D.T. Wang b M. Kuball b, j. Zegenhagen
B.O. Fimland e, W. Richter a
b
a lnstitutfftr Festkfrperphysik, Technisehe Universit~t Berlin, Hardenbergstrasse 36, D-10623 Berlin, Germany
b Max-Planck-lnst#utffir Festki~rperforschung, Heisenbergstrasse 1, D-70569 Stuttgart, Germany
c Department of Physical Electronics, Norwegian Institute of Technology, N-7034 Trondheim, Norway
Received 5 September 1995; accepted for publication 31 October 1995
Abstract
Scanning tunneling microscopy (STM) was used to study clean GaAs(100) surfaces prepared by thermal desorption of a
protective Arsenic layer. The GaAs samples were grown by MBE using an As 2 cracker cell. After transfer through
atmosphere and insertion into the UHV chamber, clean, well ordered surfaces with different reconstructions were prepared
by themaal annealing. The atomic structure and the morphology of the surfaces were found to depend sensitively on the
annealing procedure. For several differently reconstructed surfaces STM images with atomic resolution were obtained
similar to those recently published for in-situ-MBE investigated surfaces. These results demonstrate that the As-decapping
technique is, in fact, a versatile tool for preparing well-defined GaAs(100) surfaces.
1. I n t r o d u c t i o n
In recent years much interest has been dedicated
to investigations o f the GaAs(100) surface. Due to
dimerization this polar surface reveals a number of
different reconstructions depending on surface stoichiometry and arrangement of the As- or Ga-dimers
[1-4]. Many details o f the structures are still unknown or under discussion. The surface preparation
is not straightforward; good quality surfaces can be
prepared by M B E growth o f G a A s in (100) orientation. However, in most cases there is (besides
* Corresponding author. Fax: +49 30 31421769; e-mail:
[email protected].
RHEED) no surface analytic equipment available in
the M B E chamber. On the other hand, in many U H V
systems, capable for detailed surface analysis, a G a A s
M B E apparatus can not be incorporated. This drawback may be overcome by the As-passivation technique where the M B E layer is covered with an
Arsenic cap for surface passivation against contamination during ambient transfer [5-7]. The As layer
can be desorbed after insertion into an U H V analysis
chamber in order to obtain a clean GaAs(100) surface. It has been proven by several authors, that in
such way it is possible to achieve surfaces producing
good-quality L E E D patterns of different reconstructions [8-10]. However, during the last time there has
been a lot o f discussion whether the quality and the
microscopic structure of these surfaces are compara-
0039-6028/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved
SSDI 0039-6028(95)01093-9
72
U. Resch-Esser et a l . / Surface Science 352-354 (1996) 71-76
ble to those of surfaces investigated in-situ after
MBE growth [11-14]. Especially for the c ( 4 X 4)
reconstruction it was stated, that Well ordered surfaces cannot be achieved from As passivated samples [15].
To clarify the structure of As-depassivated
GaAs(100) surfaces we have carried out LEED and
STM investigations on surfaces of different reconstructions. The STM results provide information
about the roughness as well as the microscopic structure of the decapped surfaces. These results are
discussed in comparison to literature data taken insitu from MBE grown samples.
2. Experiment
Homoepitaxial GaAs layers (1 /xm thick) of a
doping concentration of n = 1 X 1018 c m - 3 (Si) were
grown by MBE on on-axis GaAs(100) substrates and
capped with a 60-70 nm thick As-layer deposited by
an As2-cracker cell. After storage in air, the samples
were introduced into the UHV analysis chamber
(base pressure 5 x 10 -11 mbar) equipped with an
UHV-STM and LEED-optics. STM images were
taken in constant current mode. For the c(4 X 4)
reconstructed surface it was possible to receive stable images with positive and negative sample biases,
whereas for the (2 X 4), (6 X 6) and (4 X 2) reconstructions stable images were only obtained using
negative sample biases (imaging the filled electronic
states of the surface) as previously reported for clean
GaAs surfaces by other authors [16].
observed in accordance with Ref. [10]; besides some
variation in intensity all integer and fractional order
spots are clearly visible. In contrast, desorption of As
from surfaces covered with a n A s 4 passivation layer
leads only to weak additional spots in the LEED
pattern, as already described in Ref. [15]. Thus the
use of an As-cracker cell in order to provide As 2
instead of As 4 seems to be of crucial importance to
prepare c(4 X 4) reconstructed surfaces from As-passivated samples.
Fig. 1 shows an STM image of such a surface,
which is of a comparable quality to those prepared
by in-situ MBE [12] or MEE [17]. Large areas with a
"brickwork-like" pattern are observed. Over a range
of 1000 A usually 2 - 3 terrace levels appear, the
topmost in form of large, isotropic islands thus forming a relatively smooth surface. The number of
height levels (i.e. the maximum height difference
between two points at the surface) is similar to that
seen on (2 x 4)-surfaces. In contrast the shape and
number of the islands, i.e. the number of steps over a
certain distance yielding the surface-roughness, is
smaller for the c(4 x 4) than for the (2 X 4) reconstructions. According to Ref. [3] the bright features
in the "brickwork-like" pattern can be interpreted as
blocks containing three As-dimers parallel to the
3. Results and discussion
3.1. c(4 X 4) reconstructed surfaces
Surfaces with a c(4 X 4) reconstruction were prepared by slowly ramping ( 5 - 2 0 ° C / m i n ) the temperature of the As2-capped GaAs(100). The As-cap-desorption was indicated by a strong increase in the
pressure of the UHV chamber (up to around 1 x 10 -7
mbar). To achieve the c(4 X 4) reconstruction the
annealing was either stopped immediately after passing this pressure maximum, or was kept at this state
for about 5 min. After cooling down the samples to
room temperature a LEED pattern of good quality is
Fig. 1. Filled states (-3.0 V, 0.05 nA) STM image (400X 400 A)
of the GaAs(100)-c(4X4) surface prepared by desorption of a
protective As-layer, which was grown by an As-crackercell.
73
U. Resch-Esser et aL/ Surface Science 352-354 (1996) 71-76
to the missing dimers of the (2 X 4) structure. The
corrugation between the bright and dark rows, i.e.
the dimer/missing-dimer rows, is 2.1 ,~ (in good
Fig. 2. High resolution image (50 × 50 A) of the (2 X 4) reconstructed GaAs(100) surface ( - 2.7 V, 0.2 hA). The two As-dimers
of the topmost layer are clearly resolved.
[011] direction which are shifted by 2 a 0 along the
[0~1] direction against each other. In higher resolution [18] the dimer in the middle appears less bright
in comparison to the outer ones, just as found for
MEE grown c(4 X 4) structures [17].
The STM images prove that, by thermal desorption of an As-layer, it is possible to prepare c(4 X 4)
reconstructed GaAs(100) surfaces, whose microscopic and macroscopic structure is in agreement
with that of the same surfaces investigated in-situ
after MBE or MEE growth.
M
3.2. (2 X 4 ) / c ( 2 X 8) reconstructed surfaces
Either annealing of the c(4 X 4) surface or annealing of As-capped samples leads to the formation of a
(2 x 4) LEED pattern. The temperature range to
obtain this reconstruction is rather wide, from about
380 to 450°C. On STM images taken from these
surfaces bright rows along the [0~1] direction are
visible consisting of As-dimer blocks. Similar as
reported for in-situ investigated MBE grown samples
the blocks contain mostly 2 As-dimers (see Fig. 2).
The rows are separated by dark lines corresponding
Fig. 3. STM images (400 × 400 ~.) of two (2 × 4) reconstructed
surfaces prepared by annealing to (a) 450°C, (b) 470°C. An
increase in disorder is clearly visible.
74
U. Resch-Esser et aL / Surface Science 352-354 (1996) 71-76
agreement with Ref. [17]), close to the double layer
spacing of 2.8 ,~. This finding corroborates the recent structure models proposed by Hashizume et al.
according to which the surface unit mesh should
contain 2 As-dimers in the outermost layer orientated
along the [011] direction [19]. The presence of a third
dimer located one layer below [19] can neither be
excluded nor be proven with certainty from our
experiments. It should be mentioned, that in none of
our STM images regions containing Ga-atoms or
Ga-dimers at the step edges were found. This was
assumed in Ref. [13] to occur on As-decapped surfaces due to As-deficient conditions. The (2 X 4)
surfaces exhibit two (sometimes three) height levels
corresponding to one double layer each (step height
2.8 A) over a range of 1000 X 1000 A. In addition
small islands are located on top, leading to a higher
roughness as observed at the c(4 X 4) reconstructed
surface. Most of these small islands show an
anisotropic shape, with a width in the [011] direction
much larger than that in the [011] direction. This
anisotropic shape of the islands is attributed to
anisotropic diffusion on the surfaces [20].
Depending on annealing temperature a variation
of the degree of order in the As-dimer rows is
observed, as previously suggested for the (2 X 4) a-,
/3- and y-phases [12,13]. The disorder is generated
by an increase of kinks in the rows and an increasing
number of dimer-blocks containing only one As-dimer (Fig. 3).
3.3. (6 X 6)- and (4 × 2) / c(8 X 2)-reconstructed
surfaces
(6 X 6) reconstructed surfaces were prepared by
annealing (2 × 4) reconstructed ones at 470°C. The
corresponding STM images and the surface structure
have recently been discussed in detail [21].
A distinct type of (6 X 6) surface structure is
found to coexist with the (4 X 2)/c(8 x 2)-structure
after annealing to higher temperatures (around
540°C). For slow cooling rates the (6 X 6)- and for
more rapid cooling the (4 x 2)-reconsmaction dominates (see also Ref. [22]). By LEED a superposition
of a (4 X 2) pattern and additional sixth order spots
in the [011] direction is observed.
The STM images show large terraces of (4 X 2)symmetry and regions with sixfold symmetry close
Fig. 4. Filled state STM image of a ( 4 × 2) reconstructed surface
(800 × 800 A, - 3 . 3 V, 0.05 nA), prepared by annealing of an
As-capped sample at 540°C for 10 min. Besides the regions with
( 4 × 2) symmetry also areas with a 6-fold symmetry along the
[011] direction are observed.
to the step edges (Fig. 4), as previously reported by
Skala et al. [23] on (4 X 2)-surfaces prepared by
As-decapping. The sixfold symmetry is built up by
bright rows in a distance of 6a 0, which most probably consist of As-dimers. These As-dimers are partly
shifted against each other along the [011] direction,
leading to a certain distortion of the sixfold symmetry. Between the rows, in the dark grooves, a threefold periodicity is observed along the [0~ 1] direction.
Following Biegelsens [3] explanation of the (2 X 6)
reconstruction the features in the dark grooves should
be correlated to the formation of Ga-dimers. H o w ever, as the corrugation exhibits a threefold symmetry there must be a difference in the arrangement of
the Ga-dimers as compared to the model given by
Biegelsen et al.
On the (4 X 2)-reconstructed regions, the STM
results show the same features as known from in-situ
investigations of MBE grown samples [3,23,24]. Accordingly, the microscopic structure of the surface
unit mesh refers to these surfaces. In contrast to the
U. Resch-Esser et a L / Surface Science 352-354 (1996) 71-76
As-terminated (2 × 4) surface, kinks and anisotropic
islands are absent and the surface is comparably
smooth showing only one or two large terraces separated by a double layer step over a lateral range of
1000 X 1000 ,~. These differences in surface roughness in dependence of the surface reconstruction
were discussed in Ref. [25].
As reported in Ref. [22] by varying the cooling
rate after annealing it is possible to switch reversibly
between (4 x 2) and (6 X 6) dominated LEED patterns. This was reproduced in our experiments, however, by STM an increasing roughness of the surface
is observed. It is worth mentioning that this roughness is not necessarily related to the formation of
Ga-droplets, instead STM reveals the formation of
three-dimensional GaAs-islands on the (6 × 6) reconstructed regions. This finding corroborates the
appearance of spots in the LEED pattern due to facet
formation after several annealing steps (at temperatures above 540°C).
4. Conclusions
In summary we have investigated the formation of
differently reconstructed GaAs(100) surfaces prepared by As-decapping in order to reveal similarities
and differences in the microscopic and macroscopic
surface structure in comparison to in-situ MBE investigated surfaces. For the first time it was shown
by STM that qualitatively good c(4 × 4) surfaces
with large ordered regions can be prepared by As-decapping. It was further shown, that the microscopic
and macroscopic surface structure of the As-rich
reconstructions (i.e. the c(4 × 4) and the (2 × 4))
prepared in this way are not noticeable different
from the surfaces investigated in-situ. For the (2 × 4)
surface there can arise slight differences in surface
roughness due to the higher atomic surface mobility
during MBE growth that takes place at higher sample temperatures than the decapping procedure.
However, in contradiction to what is proposed in
Ref. [13], the microscopic structure of the As-decapped (2 X 4) surfaces is the same as observed on
MBE or MEE grown samples. For the (4 X 2) reconstructed regions again the same microscopic structure and a comparable surface quality is observed as
for MBE and MEE prepared samples. Depending on
75
preparation conditions, a metastable (6 x 6)reconstruction at the step edges coexists with the
(4 X 2).
It should be mentioned that growth of three-dimensional GaAs islands on t h e (6X 6) regions is
induced by (extended) annealing of the (4 × 2 ) / ( 6 x
6) surfaces. Therefore, the preparation by As-desorption in a single annealing step should lead to superior
results for the (4 × 2)-reconstructed surface. Desorption of the As-cap in an As-atmosphere might further
allow to modify the surface morphology in order to
avoid the coexistence of different structural phases,
for the less As-rich reconstructions, and to enhance
surface diffusion leading to smoother surfaces for the
As-rich (2 X 4) reconstruction.
Acknowledgements
This work was supported by the DFG under the
project Ri 2 0 8 / 1 5 - 2 and by the EC under contract
ERB CHR XCT 930563 (PESSI). Technical assistance of W. Stiepany is gratefully acknowledged.
References
[1] P. Drathen, W. Ranke and K. Jacobi, Surf. Sci. 77 (1978)
L162.
[2] M.D. Pashley, K.W. Haberern, W. Friday, J.M. Woodall and
P.D. Kirchner, Phys. Rev. Lett. 60 (1988) 2176.
[3] D.K. Biegelsen, R.D. Bringans, J.E. Northtrup and L.-E.
Swartz, Phys. Rev. B 41 (1990) 5701.
[4] L. Diiweritz and R. Hey, Surf. Sci. 236 (1990) 15.
[5] S.P. Kowalczyk, D.L. Miller, J.R. Waldrop, P.G. Newman
and R.W. Grant, J. Vac. Sci. Technol. 19 (1981) 255.
[6] D.L. Miller, R.T. Chen, K. Elliott and S.P. Kowalczyk, J.
Appl. Phys. 57 (1985) 1922.
[7] J. Kawai, T. Nakagawa, T. Kojima, K. Otha and M.
Kawashima, Electron. Lett. 20 (1984) 47.
[8] U. Resch, N. Esser, Y.S. Raptis, W. Richter, J. Wasserfall,
A. F~rster and D.I. Westwood, Surf. Sci. 269-270 (1992)
797.
[9] W. Chen, M. Dumas, D. Mao and A. Kahn, J. Vac. Sci.
Technol. B 10 (1992) 1886.
[10] R.W. Bernstein, A. Borg, H. Husby, B.O. Fimland and J.K.
Grepstad, Appl. Surf. Sci. 56-58 (1991) 74.
[11] M.C. Gallagher, R.H. Prince and R.F. Willis, Surf. Sci. 275
(1992) 31.
[12] A.R. Avery, D.M. Holmes, J. Sudijono, T.S. Jones and B.A.
Joyce, Surf. Sci. 323 (1995) 91.
[13] A.R. Avery, D.M. Holmes, T.S. Jones, B.A. Joyce and G.A.
Briggs, Phys. Rev. B 50 (1994) 8098.
76
U. Resch-Esser et a L / Surface Science 352-354 (1996) 71-76
[14] M.D. Pashley, K.W. Haberem and J.M. Gaines, J. Vac. Sci.
Technol. B 9 (1991) 938.
[15] R. Duszak, C.J. Palmstrom, L.T. Florez, Y.-N. Yang and J.H.
Weaver, J. Vac. Sci. Technol. B 10 (1992) 1891.
[16] V. Bressler-Hill, M. Wassermeier, K. Pond, R. Maboudian,
G.A.D. Briggs, P.M. Petroff and W.H. Weinberg, J. Vac.
Sci. Tectmol. B 10 (1992) 1881.
[17] T. Hashizume, Q.-K. Xue, A. lchimiya and T. Sakurai, Phys.
Rev. B 51 (1995) 4200.
[18] U. Resch-Esser, N. Esser, D.T. Wang, B.O. Fimland and J.
Zegenhagen, to be published.
[19] T. Hashizume, Q.K. Xue, J. Zhou, A. Ishimiya and T.
Sakurai, Phys. Rev. Lett. 73 (1994) 2208.
[20] K. Ohta, T. Kojima and T. Nakagawa, J. Cryst. Growth 95
(1989) 71.
[21] M. Kuball, D.T. Wang, N. Esser, M. Cardona, J. Zegenhagen
and B.O. Fimland, Phys. Rev. B 51 (1995) 13880.
[22] L.K. Verheij, M.K. Freitag and F. Wiegershaus, Surf. Sci.
334 (1995) 105.
[23] S.L. Skala, J.S. Hubacek, J.R. Tucker, J.W. Lyding, S.T.
Chou and K.Y. Cheng, Phys. Rev. B 48 (1993) 9138.
[24] Q. Xue, T. Hashizume, J.M. Zhou, T. Sakata and T. Sakurai,
Phys. Rev. Lett. 74 (1995) 3177.
[25] U. Resch-Esser, N. Esser, C. Springer, J. Zegenhagen, W.
Richter, M. Cardona and B.O. Fimland, J. Vac. Sci. Technol.
B 13 (1995) 1666.