Surface Science 443 (1999) 99104

www.elsevier.nl/locate/susc

Ab initio study of the Ge(211) and Si(211)

bulk terminated surfaces
S. Mankefors *
Department of Physics, Chalmers University of Technology, S-412 96 Goteborg, Sweden
Received 4 June 1999; accepted for publication 20 August 1999

Abstract
Ge(211) and Si(211) surfaces are being investigated by ab initio methods. Extensive relaxations are found to take
. The dangling bonds of the edge atom makes both
place and the atomic geometries are altered as far down as 8 A
surfaces metallic, and several other surfaces states are found. Some smaller definitive differences between the Ge(211)
and the Si(211) in geometry and electronic structure are also investigated. 1999 Elsevier Science B.V. All
rights reserved.
Keywords: Germanium; Molecular beam epitaxy; Silicon

1. Introduction
Silicon and germanium surfaces have been
extensively studied theoretically and experimentally over a number of decades. Most theoretical
work, however, has been devoted to understanding
the low index surfaces properly, while little interest
has been focused on the high index systems. There
has been some theoretical interest in the Si(211)
surface [1] lately though, but this surface still
remains a relatively virgin area for ab initio calculations. This despite the fact that parameterised
calculations on Si were made early on in the 1980s
[2,3]. Germanium especially seems to have
attracted little theoretical interested.
The major reason for this situation is the intense
numerical demands, e.g. this investigation required
some 2000 CPU-hours, about 3 CPU-months. It
* Fax: +46-31-772-3177.
E-mail address: [email protected] (S. Mankefors)

is quite possible to use thinner slabs, lower cut-off

and asymmetrical systems [1], but this seriously
risks the quality of the investigation (see below).
Hence to allow extensive investigations of these
systems, a new level of massive computing power
is required, something which has become available
only recently to the majority of the physical community. This is also the reason why we concentrate
on bulk truncated geometries in this investigation future projects will include reconstructions.
In contrast to the theoretical situation, Ge(211)
and Si(211) have been subject to a rapidly growing
experimental interest recently [410] in connection
with CdTe and HgCdTe grown on these surfaces.
Using molecular beam epitaxy (MBE ) it has,
furthermore, become possible to grow infrared
photo diodes on these (211) substrates [4,9,10].
Thus a proper understanding of the structural and
electronic properties of the outermost atomic
layers of Si(211) and Ge(211) surfaces are of
fundamental importance both from a scientific and

0039-6028/99/$ - see front matter 1999 Elsevier Science B.V. All rights reserved.
PII: S0 0 39 - 6 0 28 ( 99 ) 0 09 1 9 -X

100

S. Mankefors / Surface Science 443 (1999) 99104

Fig. 1. The atomic geometry of the ideal (211) surface. Atomic positions are labeled by primed and unprimed numbers respectively
to denote the two types of atomic sites. The labelling is continued in a systematic fashion downwards into the system, although only
the four first atomic planes are marked in the figure. The largest atomic relaxations are indicated by arrows, and the relaxed Si(211)
surface is shown in a small inset.

a practical point of view. The fact that the (211)

surface can be thought of as a stepped (111)
surface also makes this system highly interesting.
In addition it is the simplest bulk truncated surface
with non-equivalent atomic sites. These two types
of sites are shown in Fig. 1 where they are primed
and unprimed respectively. All this calls for a
proper theoretical treatment of the problem.
In this article we investigate the electronic structure and geometrical properties of the bulk truncated Si(211) and Ge(211) surfaces. We find four
different surface states within the bandgap, which
makes the surfaces metallic. Furthermore we find
a number of surface states and resonances below
the valence band maximum ( VBM ). We especially
note a low lying, almost dispersion free, surface
state around 8 to 10 eV in Ge which can only
be partially traced in Si. Some differences in atomic
relaxation and surface energy are also found,
although the overall similarity of the two systems
stands out. The relaxations are found to go very
deep down into the material.

2. Theory
All wave functions and energy eigenvalues were
calculated ab initio within density functional
theory (DFT ) [11,12] using the local density
approximation (LDA) as implemented by
Ceperley and Alder [13] and Perdew and Zunger
[14]. For the electronion interaction, fully separable, non-local pseudopotentials (PP) were used
[15,16 ] based on self-consistent solutions of the
relativistic Dirac equation for free atoms [1719].
The calculations were performed using the planewave band structure code 94. 1, which is
a heavily modified version of 93 [20] concerning the computational methods. The geometries
were described by the slab supercell method using
)
the theoretical lattice constants for Ge (5.58 A
1 94. is based on 93, purchased from the
Computer Physics Communications Library in 1995 while
the basic computational physics is the same, the changes of
computational nature are extensive.

S. Mankefors / Surface Science 443 (1999) 99104

101

Fig. 2. The top view of the atomic geometry of the ideal (211) surface and the surface Brillouin zone, a denotes the lattice constants.

). A plane-wave cut-off of 12 Ry
and Si (5.39 A
and 10 special MonkhorstPack k-points in the
irreducible Brillouin zone (IBZ ) was used to
sample the wavefunctions in the supercells, corresponding to 27 points in the full zone.
The slab geometries were described by 30 atomic
planes, containing a total number of 60 atoms and
12 atomic planes of vacuum. The surface cell consisted of three atomic planes, each at a different
position along the (211) axis perpendicular to the
surface, resulting in a three level stepped surface
before relaxation, see Figs. 1 and 2 . All atomic
positions were completely relaxed throughout the
entire slab. The equilibrium geometries were considered to be established when all forces were
, corresponding to an estismaller than 0.005 eV/A
.
mated numerical uncertainty of ~0.05 A
Because of the stepped nature of these surfaces,
any disturbances will have very deeply reaching
effects. To explicitly test the quality of the model,
we investigated the local density of states (LDOS ),
the effective and electrostatic potential, and the
electronic charge distribution inside the slab. At
the center of the slabs all these characteristics take
on ideal bulk values well inside any numerical
errors and this bulk region extends for more than
14 atomic layers, hence ensuring the quality of the
systems. Inside this region all atomic displacements
also vanish completely within the numerical uncertainty, thus lending further credibility to the
results. To test the depth of the vacuum region,
we integrated the electronic density of the central
region of vacuum. The charge accumulated in this
volume was found to be negligible.
In order to calculate the surface energy, bulk

systems corresponding to the slabs used above

were calculated using the same numerical conditions, with the vacuum removed. All atoms were
fixed at their ideal bulk positions. By using this
60 atom bulk cell, the numerical error in the surface energy is kept at a minimum. The use of a
smaller cell would have resulted in the bulk related
error being multiplied several times when comparing the systems because of the difference in the
number of atoms.

3. Results and discussion

The bulk truncated (211) Si (Ge) surface contains two three-fold co-ordinated terrace atoms
with one dangling bond each, while the high
situated edge atom (denoted 1 in Fig. 1) only is
two-fold co-ordinated with two dangling bonds.
These three atoms exhibit very large relaxations
relative to their ideal positions, i.e. up to 29% of
the bulk interatomic bond length, see Table 1.
Since the edge atom is relatively loosely bonded,
it is free to move up and to the right along (111)
and enable the other atom in its plane to move
down and follow it to the right. In this way the
entire first atomic plane actually becomes tilted
and lined up with the rightmost atom in the second

plane (2) while the centre moves about 0.4 A

along the (111) direction. The driving force behind
this is most likely the same type of sp hybridisation
as in the simpler IIIV (110) surfaces.
In contrast to the topmost terrace atom, the
lower lying atom (no. 2 in Fig. 1) is not free to
move downwards and can not take part in the

102

S. Mankefors / Surface Science 443 (1999) 99104

Table1
. Distances
The atomic relaxations for Si(211) and Ge(211) in A
are given relative to the ideal bulk truncated positions. Atomic
indexes as in Fig. 1
Position
(see Fig. 1)

Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom
Atom

1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8

Ge(211)

Si(211)

D [111]

D [211]

D [111]

D [211]

0.37
0.49
0.03
0.10
0.09
0.04
0.02
0.00
0.04
0.05
0.00
0.00
0.01
0.01
0.01
0.01

0.28
0.43
0.15
0.30
0.11
0.09
0.06
0.06
0.05
0.06
0.02
0.03
0.04
0.04
0.03
0.01

0.30
0.45
0.05
0.06
0.05
0.01
0.01
0.00
0.04
0.05
0.01
0.01
0.01
0.01
0.01
0.01

0.22
0.49
0.13
0.18
0.08
0.09
0.06
0.06
0.05
0.04
0.02
0.04
0.04
0.05
0.03
0.01

hybridisation. This is so since both atoms 2 and

3 have all their bonds intact and are consequently
relatively hard bonded and will not make room
for the terrace atom. Alternatively one might consider the local environment of the lower terrace
atom to be an altered bulk truncated (111) surface,
and as a such it is well known that the top atoms
do not move down into the surface. Interestingly
enough, however, this atom instead relaxes significantly outwards along the (111) direction indicated
in Fig. 1, see Table 1. The relaxations of atom 2
and to a certain extent atom 1 (the edge atom),
in addition, differ surprisingly much between the
Si and Ge surfaces, while the remaining relaxed
atoms take almost identical positions in the two
systems. The driving force behind these movements
is the dangling bond of the terrace atom. Since
this bond is directed along (111), it will partly
overlap with the charge cloud of the edge atom.
In an attempt to fill the empty orbit, the low lying
terrace atom consequently moves along (111)
towards to edge atom, which in return move along
(111) to meet it (note that the two atoms are not
lined up in the (111) direction though). Because

the orbitals of Ge are larger then those of Si, the

overlap is greater in this case, and the effect larger.
The third atomic plane relaxes slightly outward
in the valley of the surface, while lower lying
planes mainly relax to adjust themselves to the
atoms above, see Table 1. Below the seventh plane,
below the surface, no movei.e. approximately 8 A
ment outside the numerical error bars can be
found. Still, for a homopolar bulk truncated surface, this must be considered a very great depth
for relaxations. In fact it goes far beyond what is
initially assumed in Ref. [1]. Even taking into
account the use of an asymmetric slab (only one

side relaxed ), the thickness of approximately 6 A

used for the slabs definitively imply that the relaxation found there [1] must be incomplete in the
light of the results presented above.
Using the total energy of the bulk slabs
described above in Section 2, the surface energy
now becomes very straightforward to calculate
i.e. one take the differences in total energy and
divide it by the total surface area (both sides of
the slabs). Because of the use of identical numerical
conditions, the error can be kept as low as
<1 meV, not counting possible errors due to the
finite convergence in the relaxation of the atoms.
Including this additional factor, the numerical
uncertainty of the surface energy becomes 2 meV.
Calculating the surface energies of Si(211) and
Ge(211) in this way returns values of
2 and 0.0890.002 eV/ eV/ A
2
0.1200.002 eV/ A
respectively. For the Si(211) ideal surface, this is
2 lower then the result in [1]. It
0.012 eV/ eV/ A
should be noted though that the use of the substantially worse convergence in the atomic forces in
[1] results in a numerical error even larger then
the above difference. In addition, the lack of a full
relaxation in [1] (see the discussion above) most
likely affects the difference in a negative way.
2 for the ideal
Still, the value of 0.120 eV/ eV/ A
Si(211) is too high for it to be likely to be stable.
The calculated surface energies of the reconstructed 21 Si(100) and 77 Si(111) are found
2 [21] and 0.092 eV/ eV/ A
2
to be 0.071 eV/ eV/ A
[22] respectively, which indicate the instability of
the Si(211) ideal surface. The value of
2 of bulk truncated Ge(211) on the
0.089 eV/ eV/ A
other hand seems to suggest that this surface might

S. Mankefors / Surface Science 443 (1999) 99104

be stable, or at least metastable. The reason for

the difference between the Si and Ge systems is to
be found in the orbital overlap of the terrace atom
in the second atomic plane and the topmost edge
atom discussed above. In Ge this effect is significantly larger, which drastically affects the surface
energy since it partly fills the dangling bonds. Thus
the Ge(211) ideal surface consequently becomes
more stable than the Si(211) geometry.
Turning our attention to the electronic structure
of the two (211) surfaces we notice that the
dangling bonds of the two terrace atoms and the
edge atoms two dangling bonds should be
expected to be found in the vicinity of the fundamental bandgap. This also turns out to be the
case, see Fig. 3a and b. Unlike tight binding
schemes, however, the present calculations do not
provide a simple method for identifying the
different bands in terms of atomic orbitals. By
investigating the spatial distributions at different

(a)

(b)

Fig. 3. (a) The electronic structure of Ge(211) together with the

projected bulk bands (grey). Notations as in Fig. 2, the VBM
is taken as zero energy. See the text for the labelling of the
surface states. (b) As (a) but for Si(211).

103

k-points, we have been able to assign the bands

as follows.
One of the edge atoms dangling bonds will give
rise to a bridge-like bond towards neighbouring
edge atoms. This results in the topmost band B ,
1
labelled 1 after the notation in Fig. 1. The remaining dangling bond gives rise to the slowly dispersing band E (edge atom 1) in the central part of
1
the bandgap. Because of its range from the VBM
to the conduction band minimum (CBM ), it makes
the Ge surface, as well as the Si system, metallic.
Below E , we find the two bands associated with
1
the dangling bonds of the terrace atoms, T and
1
T ( Terrace atoms 1 and 2). They have, not very
2
surprisingly, a very similar dispersion and the
lower band T is connected to the terrace atom in
2
the second atomic plane. This should be expected
since the dangling bond of this atom partly overlaps with the charge cloud of the edge atom, and
thus become slightly more filled then in the case
of the upper terrace atom. The T band is also
2
somewhat more lowly situated in the case of
Ge(211) since the overlap is larger (see above). It
should also be mentioned in this context that this
result differs in the details obtained by the authors
of Ref. [3]. This is, however, most likely to be due
to the cruder (unrelaxed) model used there, and
the general features of the electronic structure are
still relatively similar.
Further down in the range of 1 to 2 eV we
find two weak surface resonances associated with
the first atomic plane. They are too weak to follow
throughout the entire surface Brillouin zone (SBZ )
though. At the bottom of the stomach a strong
surface state, R , appears a few tenths of an eV
1
above the projected bulk density of states. Outside
the stomach, this strongly dispersing band turns
into a clear surface resonance. Like the higher
lying resonances, this state should also be associated with the first atomic layer and is believed to
be connected to the bond between the edge and
the terrace atom.
Below R , we find three surface states/
1
resonances, mainly of s-character. Two are too
faint to follow throughout any larger part of the
SBZ and stems from the third atomic layer. The
surface state which crosses the low lying empty
regions, R , originates from the bonds in the
2

104

S. Mankefors / Surface Science 443 (1999) 99104

second atomic layer though. In Si, this state can

only be followed from the midpoint of C-X to the
X-point, while it is easily seen in the entire SBZ in
the case of the Ge(211) surface. We strongly
believe this difference to be due to the slightly
different behaviour of the second terrace atom on
the two surfaces.

4. Conclusions
We find very large surface relaxations in both
Ge(211) and Si(211) geometry. The topmost layer
becomes tilted and lines up with the closest atom
in the second atomic plane in an sp hybridisation.
The terrace atom in the second plane, on the other
hand, moves outwards along the (111) direction
towards the edge atom in the first plane in order
to partly fill its dangling bond. Relaxations of the
. The dangling
geometry are seen as deep as 8 A
bonds of the surface result in four different surface
states within the bandgap. One of the dangling
bonds associated with the edge atom makes the
surfaces metallic. Several lower lying surface
states/resonances are also found, including a surface state stemming from the first atomic layer in
the stomach and another state in the bottom holes
in the projected bulk density. This state we associate with the bond within the second atomic plane.
Differences both in atomic and electronic structure
between Ge(211) and Si(211) are found to be due
to the larger orbit of the dangling bond on the
second Ge terrace atom.

Acknowledgements
This work was supported by the Swedish
Natural Science Research Council and the

Computation Centre of Hogskolan i Skovde,

Sweden.

References
[1]
[2]
[3]
[4]

C.H. Grein, J. Cryst. Growth 180 (1997) 54.

D.J. Chadi, Phys. Rev. B 29 (1984) 785.
A. Mazur, J. Pollman, Phys. Rev. B 30 (1984) 2084.
J.P. Zanatta, N. Luchier, P. Audebert, P. Demars, J.P.
Chamonal, M. Ravetto, M. Wolny, Proceedings of the
SPIE The International Society for Optical Engineering
3379 (1998) 586.
[5] H. Ebe, T. Okamoto, H. Nishino, T. Saito, Y. Nishijima,
M. Uchikosi, M. Nagashima, H. Wada, J. Electronic Mat.
25 (1996) 1358.
[6 ] A. Million, N.K. Dhar, J.H. Dinan, J. Cryst. Growth 159
(1996) 76.
[7] S. Rujirawat, L.A. Almeida, Y.P. Chen, S. Sivananthan,
D.J. Smith, Appl. Phys. Lett. (1997) 1810.
[8] J.P. Zanatta, P. Ferret, P. Duvaut, S. Isselin, G. Theret,
G. Rolland, A. Million, J. Cryst. Growth 184/185 (1998)
1297.
[9] J.P. Fauri, L.A. Almeida, Y.P. Chen, R. Sporken, S. Sivananthan, Proceedings of the SPIE The International
Society for Optical Engineering 2685 (1996) 28.
[10] H.Y. Wei, L.. Slamance-Riba, N.K. Dhar, Infrared Appl.
Semicond. Vol. II Symp. Mater Res. Soc, Warrendale, PA,
1998. pp. xvii and 692.
[11] W. Kohn, L.J. Sham, Phys. Rev. 140 (1965) A1133.
[12] P. Hohenberg, W. Kohn, Phys. Rev. 136 (1964) B864.
[13] D.M. Ceperley, B.J. Alder, Phys. Rev. Lett. 45 (1980) 566.
[14] J.P. Perdew, A. Zunger, Phys. Rev. B 23 (1981) 5048.
[15] B. Hamann, Phys. Rev. B 40 (1989) 2980.
[16 ] X. Gonze, P. Kackell, M. Scheffler, Phys. Rev. B 41
(1990) 12 264.
[17] D.R. Hamann, M. Schluter, C. Chiang, Phys. Rev. Lett.
43 (1979) 1494.
[18] G.B. Bachelet, D.R. Hamann, M. Schluter, Phys. Rev. B
26 (1982) 4199.
[19] L. Kleinmann, D.M. Bylander, Phys. Rev. Lett. 48
(1982) 1425.
[20] R. Stumpf, M. Scheffler, Comput. Phys. Commun. 79
(1994) 447.
[21] N. Roberts, R.J. Needs, Surf. Sci. 236 (1990) 112.
[22] K.D. Brommer, M. Needels, B.E. Larsson, J.D. Joanopoulos, Phys Rev. Lett. 68 (1992) 1355.