/ gears / sp / software / opendx / pastprojects / barb

Molecular Dynamics

Molecular dynamics simulations of atomic motion is increasingly being used to mimic events in a large number of processes. The computer power has grown to the point where one can model very complex events with systems of 10E4 to 10E7 atoms. The challenge, then, is to analyze the results of the motions. Some properties, such as temperature, energy distributions, angular distributions, velocity autocorrelation functions, are relatively easy to turn into x-y or contour plots. Other properties are virtually impossible to analyze without first visualizing the motions of all the atoms.

There are two traditional graphical approaches of plotting atomic and molecular structures. First, wire frames are often used by the biological scientists where a line connects 2 atoms. Second, a "ball" is drawn per atom, an approach preferable to the group of researchers who examine processes in solids in which specific molecular identity and bonds do not exist. The group of Barbara Garrison in the Chemistry Department belongs to the subset of people who want to draw balls to represent atomic motions generated from integrating the classical equations of motions of ~10E4 particles or balls. That is, they want to make animations of the results of their calculations. A critical aspect here is that they use the animations to diagnose, understand, and debug their simulations. Thus making the animation must be easy and must be displayed on the screen at speeds that the brain can assimilate what is happening. One must also be able to easily rotate the configuration. In addition, for presentations they may want to make "nice balls" for a handful of frames.

Their approach since 1988 has been to use AVS on a Stardent GS2020 computer which had a hardware sphere primitive thus drawing spherical balls even with lighting was fast. Several thousand atoms could be easily animated and rotated. Stardent folded several years ago but another company continued to maintain the computer until December 1995. Since then they have tried to hobble along with AVS on an IBM 43P RISC computer but with no sphere primitive, the atoms must be drawn as "coneheads" to get sufficient speed for diagnostic animations. The directionality of the object visually distracts from the directionality of the motion. In another application, a group member wants to draw even one frame with 20,000 atoms but this is too large for AVS. The group has collaborators at several institutions and the AVS graphics are not portable at reasonable cost.

What is really needed is streamlined, machine independent graphics catered to the specific needs of molecular dynamics simulations of atoms and molecules. AVS is very general and powerful but it is more than required. The main issue is to design a "ball" that can be drawn faster than a sphere of many polygons (i.e., lots of computer time). The Visualization Group at the CAC has developed an approach to draw the "balls" in pixel format rather than multiple polygons. Moreover, the "balls" can be rings, concentric rings (bulls eyes) or nice spheres which are user chosen depending on the number of atoms and the computer time required. The code is being authored in ansi C utilizing OpenGL.

Once developed this program should of interest to several groups at Penn State. Groups in the Chemistry Department that use the Garrison AVS/IBM system include Allara, Maroncelli, Steele, Weiss and Winograd. Other researchers doing molecular dynamics simulations of motions in solids include Fichthorn in Chemical Engineering, Banavar in Physics, Chen and Kumar in Materials Science, etc.

Anecdotally, at a recent meeting there was a presentation by Peter Lomdahl of LANL. He had a movie of crack generation with 10E7 atoms which took 12 days to graphically render. This is unacceptable if one is using the graphics for diagnostics and understanding.

Please send questions or suggestions about this web page to: sp@rcc.its.psu.edu

ASET | ITS | Penn State