Chemistry Reference
In-Depth Information
Application
The transparent interface method has been tested on the
Si
2
O
7
H
6
system and then applied to the study of SiO and SiO
2
molecules
and their interaction with water.
CONCLUSIONS: THE OUTLOOK
The development and use of hybrid, multiscale methods for modeling
solid-state processes is expanding rapidly and has been driven largely by the
burgeoning activity in nanoscience and nanotechnology. Throughout this
chapter, we have concentrated on the various modeling techniques that are
available today, along with their strengths and weaknesses. Looking toward
the future, however, it is equally important to address the question:
What
more is needed
? In other words, are there important questions in computa-
tional solid-state physics that existing methodologies are incapable of answer-
ing? Perhaps not surprisingly, we find that the most sophisticated
computational techniques available today can adequately address only a tiny
fraction of the important problems that require solutions. For example, nearly
all of the applications described in this chapter deal with geometrically simple,
single-component systems composed of just a single element (often Si). While
solving such problems greatly improves our fundamental understanding of
important nanoscale processes, it is also true that the behavior of real-world
materials and devices often critically depends upon impurities and complex
interactions in multicomponent systems. Existing computational methods
are woefully inadequate for modeling such systems.
Many of the important unanswered questions that we cannot adequately
address can be separated into four main modeling categories: (1) multicompo-
nent and multielement models, (2) quantitative models, (3) coupled-mechanism
models, and (4) multiple time-scale models. The primary difficulty in handling
multicomponent and multielement models is that atomistic simulations using
classical potentials generally fare poorly when chemistry effects are important
or when bonds are significantly distorted away from the configurations used
to develop the potential. Multiscale models that incorporate quantum
mechanics methodologies can help, but only when the important physics is con-
centrated within a very small volume. The LOTF method shows great promise
in handling problems in this class since the classical potentials are modified on
the fly using small-scale quantum calculations. It is important to emphasize,
however, that multiscale models that include a QM component generally
require the imposition of boundary conditions on the electronic wave function
at the QM boundary. This problem is particularly severe for metals, which have
highly delocalized bonding, so most CP/QM models (including LOTF) are
currently restricted to modeling insulators and semiconductors.
Quantitative modeling has two primary aspects. First, a quantitative
model must make
predictions
that are ''qualitatively'' correct, meaning that
