The Cryogenic Challenge : A "Real-Life" Parallel Application© Adam Hughes
Jun 18, 1999
The last article described the general principles of revving
up a chemical simulation through the use of high-performance
computing, but no concrete examples were given. This
deficiency will be addressed this time around as we look at a
specific example, namely the simulation of liquid hydrogen.
Liquid hydrogen is an important cryogenic fluid, and the
studies discussed here formed part of my doctoral thesis work
and started me on the career path I'm following now.
In the simulation of a chemical system, the researcher usually
begins by representing each particle as a point in cartesian
space. This generally works fine for large atoms such as carbon
and nitrogen. However, as the atoms get smaller and smaller,
they become more quantum in nature. What this means in layman's
terms is that the atom is more delocalized, or it's harder to
pin down just where it resides in space. The more light atoms
in a system, the more important are its {\em quantum effects }.
Because of this, it's not good enough to represent a mass of
light particles (such as a system of liquid hydrogen) by single
points in space. One way to overcome this problem is with the
use of path integrals. In using path integrals, the scientist
represents each atom as a ring of interconnected points, so as
to account for the delocalized nature of the particle. If you'd
like to learn more about the gory details of path integrals and
their application to condensed matter systems, a good place to
start would be http://voth.chem.utah.edu/indexr.html. The take-home point of the above description is that a system of
path integral particles contains P times as many points as a
system of classical particles, where P is the number of path
integral "beads" per particle. So, if I start with a "box" of
100 hydrogen atoms represented classically, the system grows to
1600 particles ifI want every atom to be represented by 16 path
integral beads. As you might imagine, this increases the
calculation time by a factor of 16. Considering that a good
classical simulation of hydrogen can take a couple of days, and
I needed to perform somewhere around 100 {\em path integral}
simulations to get meaningful results, you might think I was in
grad school for a good decade or so. Although this jobian
demonstration of patience would certainly have reflected well
on my character, I was rescued from such heroics by the
application of parallel programming.
Go To Page:
1
2
The copyright of the article The Cryogenic Challenge : A "Real-Life" Parallel Application in Scientific Computing is owned by . Permission to republish The Cryogenic Challenge : A "Real-Life" Parallel Application in print or online must be granted by the author in writing.
|
