Formation and structural properties of the amorphous-crystal interface
in a nanocrystalline system
Cesar R. S. da Silva and A. Fazzio
Instituto de Fı ´sica, Universidade de Sa ˜ o Paulo (USP), Sa ˜ o Paulo, Brazil
~Received 14 February 2000; revised manuscript received 31 January 2001; published 5 July 2001!
Free volume Monte Carlo simulation of the melting and quenching of a Si nanosystem is used to produce an
amorphous layer on top of a nanocrystal, and a completely amorphous system. The interface is kept free of any
temperature gradient during the melting. We find that the crystal-amorphous transition layer is thicker than
previous studies indicated, and keeps memory from the crystal structure. In amorphous bulk, the concentration
of fivefold coordinated atoms is substantial, and is associated with a more open structure than fourfold-
coordinated regions. All calculations were done with an empirical Tersoff potential.
DOI: 10.1103/PhysRevB.64.075301 PACS number~s!: 61.43.2j
I. INTRODUCTION
Silicon, a very important material due to its multiple tech-
nological applications, is regarded as a prototype for
fourfold-coordinated covalent amorphous. Si amorphization
and amorphous Si ( a -Si) have been investigated for over 30
years. Most studies, however, have focused on Si amorphiza-
tion due to the impact of energetic ions, and several models
have been proposed.
1–4
A substantial number of recoils gen-
erate binary collision cascades, producing defects that are
unlikely to be produced by other mechanisms, and even
pressure-induced defect generation and amorphization have
been speculated to occur.
5
The important issue is that out of
thermal equilibrium, point defect accumulation plays an im-
portant role in Si amorphization whether you are using ion
beam or electron irradiation.
6
If the irradiation is not ther-
mally assisted, the above equilibrium defects will not be an-
nealed out. Consequently, this kind of mechanism probably
leaves a higher concentration of overcoordinated and under-
coordinated atoms.
Another way to produce a -Si, which has received little
attention, is fast quench from the melt. In this case the dis-
order is introduced by thermally activated atomic reloca-
tions. In comparison to ion beam amorphization, a smaller
number of non-four-coordinated Si is expected, and the en-
tire structure is likely to be different. This is produced using
laser beam pulses, and during the experiment an amorphous
formation was observed for Si ~Refs. 7–9! and Ge.
10
In any case, a -Si is modeled as a covalent random net-
work of fourfold-coordinated atoms. The experimental aver-
age bond length in a -Si is about 1.9% longer than in c-Si. Its
experimental radial pair distribution function g ( r ) is not sim-
ply crystalline g ( r ) with broadened and shifted peaks.
11
The
peak corresponding to the crystal’s third coordination shell,
which has 12 atoms, is absent in the amorphous g ( r ). Con-
sequently, any model for a -Si needs to show these features
to be acceptable.
The structure and atomic coordination at the amorphous
surface and the crystal-amorphous interface are also of great
interest and are the primary concern of this study. Besides
some early studies,
12–14
there are two very recent molecular-
dynamics ~MD! studies especially concerned with the
interface.
15,16
The work of Weber et al.,
16
and the most recent
work of Bernstein et al. on the subject,
17
focused strictly on
the epitaxial growth of the crystalline phase, while in Ref. 15
the authors concentrated specifically on the interface proper-
ties. This latter work was carried out using a tight-binding
approach for the system relaxation, which ensures more ac-
curate forces. Unfortunately they were limited to a small
system of only 320 atoms without free boundaries. They also
used the Parrinello-Rahman
18
approach to allow for cell re-
laxation. The work of Weber et al., used much larger cells
and a Stillinger-Weber
19
semiempirical potential. It was car-
ried out using a fixed cell with a free surface, allowing for
relaxation in only one direction. In order to produce the in-
terface, both studies kept the atoms in a portion of the cell at
a low temperature and the remaining atoms at a temperature
above melting point, until a homogeneous liquid is formed.
This means keeping a temperature gradient of thousands of
K over interplanar distances. After that, the whole cell was
cooled down to room temperature. Consequently the only
effect of the crystal on the near-interface amorphous layer is
the effect produced by the crystal field. Any memory that the
near-interface amorphous layer could have from the crystal is
lost. This is not the way a crystal melts.
20
Melting initiates at
some extended defect, usually the surface, and propagates to
bulk, without any sharp temperature gradient. Therefore, the
near-interface amorphous layer had been in the melt for only
a short time before the quench, and possibly keeps some
memory from the crystal structure.
An alternative way is to use a free volume approach. This
will allow the crystal to melt from the surface toward the
bulk. In this case, a large system is necessary to ensure that
the inner bulk is not affected by the surface. Unfortunately,
this will prevent the use of first principles or even tight-
binding approaches. Tersoff
21,22
and Stillinger and Weber
19
created two widely used empirical potentials for Si. Despite
the fact that the Stillinger-Weber produces a melting tem-
perature in better agreement with experiment, we have cho-
sen the Tersoff potential because it can be used for com-
pounds. Thus in the future we will be able to compare Si
melting and amorphous transition with Si based compounds
and similar materials
23
on the same basis. The parametriza-
tion used was what Tersoff called Si~C!.
22
The Metropolis method Monte Carlo
24
~MC! is used in-
stead of constant temperature MD. By choosing the MC
method we do not lose great accuracy, since MD will not
PHYSICAL REVIEW B, VOLUME 64, 075301
0163-1829/2001/64~7!/075301~7!/$20.00 ©2001 The American Physical Society 64 075301-1