Vacuum/volume 41/numbers 1-3/pages 602 to 604/1990 0042-207X/9053.00 + .00
Printed in Great Britain © 1990 Pergamon Press plc
Structure and energetics of Ge(111 )c(2 x 8)
reconstruction
W S Verwoerd and P Badziag, Department of Physics, University of South Africa, P.O Box 392, Pretoria
0001, Republic of South Africa
We present self-consistent MNDO calculations of atomic clusters modelling the structure of adatom generated
2 x 2, ~/3 x x/3 and 2 x 8 reconstructions of the Ge(l 11) surface. Comparisons are made of the T4 and H3
adsorption sites and of closed shell singlet and open shell (spin triplet) RHF wavefunctions. We find that
~/3 x ~/3 prefers the T4 site energetically, but 2 x 2 prefers H3. Subsurface relaxation up to three layers deep
plays a crucial role in the energetics and we also find significant charge transfers between rest atoms and
adatoms as well as between layers. A consistent interpretation can be given of the similarities and differences
between the configurations in terms of bond rehybridization and polarization effects as well as the different
surface topologies.
Introduction
Considerable experimental '-6 and some theoretical 7 effort has
been expended to understand the c(2 x 8) reconstruction of the
Ge(111) surface, establishing a simple adatom model in a three-
fold coordinated site, but the details remain uncertain. In particu-
lar, it is not yet clear whether the adatom site is on top of the
second layer (T4), or rather on the hollow above the fourth layer
(H3). Furthermore, neither the role of charge redistribution close
to the surface, nor the relation between the c(2 x 8) adatom
terminated and the surface terminated by foreign atoms (usually
in a x/3 x ,,/3 pattern) have been investigated in detail so far.
This contribution addresses these unresolved issues from the
point of view of semiempirical (self-consistent electronic struc-
ture) total energy atomic cluster calculations. First we present the
method of calculation and describe the investigated surface
structures and model clusters, then the numerical results are
listed and discussed, and finally the conclusions are summarized.
Calculation and results
We model the surface by atomic clusters representing a surface
unit cell (SUC) and containing two to four atomic layers. The
atomic geometry which minimizes the energy is calculated,
subject to symmetry constraints appropriate to the correspond-
ing surface periodicity. The energy calculation is done fully self
consistently using the MOPAC 8 implementation of the MNDO 9
quantum chemical molecular Hamiltonian.
Compared to previous surface cluster calculations, we use two
innovations. Firstly, using the polymer options of the MOPAC
program, the cluster is periodically repeated along one surface
mesh unit vector, so that it represents an infinite ribbon from the
surface. Secondly, the bonds embedding the ribbon in the crystal
lattice are saturated by artificial 'capped bond' atoms (CB's).
These are parametrized in MOPAC to form perfect covalent
bonds, eliminating the problem of a polarized boundary which
would occur with traditional bond saturating hydrogen atoms.
By careful choice of the CB bond distance we have succeeded in
602
reducing all Ge atomic net charges in fully saturated bulk-like
clusters to well below 0.1 e, which sets an accuracy standard for
charge assignments calculated in the surface clusters.
We have used two families of molecular clusters. The 2 x 2 and
~/3 x x/3 cases are modelled by including four 'substrate'
atomic layers plus the adatoms, totalling about 22 Ge atoms. For
the larger 2 x 4 and 2 x 8 unit cells we can only include two
substrate layers, and have confirmed by repeating the 2 x 2 case
with these shallow clusters that consistency with the deep clusters
is maintained.
In Figure 1 we show examples of the relation between the
ribbon-like clusters used and the various reconstruction patterns
modelled. In all clusters, the deepest layer is kept fixed, while
atoms in all other layers are allowed to move vertically and first
layer atoms are in addition free to move radially towards the
adatom.
The calculations reported here are based on restricted Hartree-
Fock (closed shell singlet) wavefunctions. We also repeated the
calculations with wavefunctions representing singly occupied
dangling bond (open shell) states. No detailed conclusions could
be made from the latter about the relative energies of the two
kinds of states, because of certain inconsistencies in the open shell
energy values, to be discussed in more detail elsewhere. Neverthe-
less the atomic displacements, charges and hybridization in the
open and closed shell calculations turn out to be quite similar
and the results discussed below apply at least quantitatively to
the open shell states as well.
A comparison of our results for closed and open shell states, T4
and H3 sites, and x/3 × x/3 and 2 x 2 reconstruction patterns,
may be summarized as follows:
(a) The relative ordering of the four sites investigated, is the same
for closed shell and open shell states.
(b) The 2 x 2 reconstruction has a lower surface energy than the
,,/3 x ,,/3, in agreement with observation.
(c) For x/3 x x/3, the T4 (on top) site is most favourable, as
observed for foreign adatoms and as calculated by Northrup ~°
also for Si(111)x/3 x x/3.