We develop object-oriented codes using C++ to perform molecular dynamics
simulations with atomistic potentials
(classical/tight-binding/first-principle). For a given system,
the potential and efficient algorithms
will be chosen based on
Different problems should be investigated with appropriate data structures
and algorithms, while maximizing reuse of the code. This can be
accomplished efficiently by developing code in object-oriented frameworks.
The MD codes are built on POOMA (Parallel Object Oriented Methods and Applications) developed by Los Alamos
National Laboratory which provide underlying objects responsible for
parallelism and node-level
simulations for efficiency, portability and re-usability. Our plan includes
implementations of recent algorithms in solving large scale electronic
structure calculations and in performing accelerated molecular dynamics
simulations to handle a wide range of length and time scales of simulations.
In parallel, we have been working on implementation of
linear scaling method on a tight-binding MD code in collaborations with
F. Kirchhoff. This code will be mainly used for a test tool for tight-binding
Hamiltonians for transition metals and compound systems.
On-going projects using Atomistic Simulator based on POOMA
Rod-like {311} defects are commonly observed in ion-implanted
silicon and are believed to play important roles in boron transient
enhanced diffusion (TED) by providing interstitials during annealing
processes. However, even under the implantation
conditions to suppress the formation of extended defects, TED has been
observed, suggesting that microscopic interstitial defects
exist. While the structural properties of rod-like
{311} defects are well characterized by experiments, e.g.,via
high-resolution transmission electron microscopy
(HRTEM), the existence and properties of
small clusters are not well understood. Only recently, the presence
of interstitial clusters has been evidenced by deep level transition
spectroscopy.
We found trends in the formation energies of n-interstitial
defects. Using first-principle calculations within the local density
approximation (LDA), we consider relevant interstitial defects - small
clusters (n=2-5), chains, and planar {311} defects -- and we
relate them to growth of interstitial defects in interstitial
supersaturated silicon as achieved in ion-implanted samples. In
general, the stability of interstitial clusters increases as the size
n increases. The formation energy per interstitial ranges from
Ef(n=2) = 2.46 eV for di-interstitials
to
Ef(infinity) = 0.7 eV for rod-like {311} defects.
Preferred elongation of rod-like defects in the [011] direction is
predicted in accordance with experiments. We find that the stable
configuration for a given size can be (a) compact for small
clusters, (b) elongated for medium clusters and (c)
planar for large clusters.
For more details, check out
- Thermally activated reodientation of di-interstitial defects in silicon [reprint,pdf]
- Stability of Si-inetrstitial defects: from point ot extended defects [reprint,pdf]
- Animation of a di-interstitial diffusion at 1200 K: View-A[gif-17M] View-B[gif-11M]
We study thermal properties of the {311} extended defects by molecular
dynamics simulations. In particular, we investigate the role of the
extended defects in the interstitial diffusions to clarify the transient
enhanced diffusion in ion-implanted silicon.
Recently developed hyper-MD method [A. Voter, Phys. Rev.
Lett.