Robert Panoff

National Center for Supercomputing Applications
University of Illinois at Urbana Champaign
605 E. Springfield Avenue
Champaign, IL
rpanoff@shodor.org info.

SimSurface: A computational science approach to optimization

SimSurface is a multi-faceted module for use in computational science classes, or in classes which include a computational component. The problem to be solved is the minimization of the electrostatic potential of N electrons constrained to move on a surface bounded by repulsive edges. This is a 2N-dimensional minimization problem, and that in itself can be quite daunting. Our approach is to take a powerful and elegant method--simulated annealing-- and apply it to this problem in such a way that one can learn more about the annealing process, the simulation process, or the computers used to implement the simulations.

While the significant computational load is relieved by running on a remote supercomputer, the visualization of the answer is accomplished by local, public domain software, namely, NCSA DataScope (for macintosh only) or NCSA Collage (for Mac's, PC's or X-window environments) to produce an animation of the annealing process and the relaxation algorithm. In response to suitable prompting questions, the user inputs the number of electrons in the simulation, the charges on the walls, the parameters controlling the annealing schedule, and a control parameter controlling the frequency with which visualized snapshots of the system are generated. Once generated, the animation can be played back, run in reverse or at different speeds, or can have the color palette adjusted to allow more indepth study of the phenomena.

The project materials themselves consist of a set of sample lectures on electrostatics, on calculation of potentials, on minimization as an optimization problem, on annealing, and on visualization. A key material is the source code itself which is provided on request so that students or faculty can adapt the raw materials for their own use.

SimSurface is itself part of a larger effort instituted by Dr. Panoff to develop a variety of public domain codes using public domain visualization software to yield interactive simulations that encourage and enable project-oriented learning in an interdisciplinary,computational science course. These codes and materials have proven highly successful and adaptable for use in undergraduate courses in physics, mathematics, and computer science, along with our course in computational science. It is a striking example for many of the power of scientific visualization to achieve fundamental insight and understanding of complex phenomena, in understanding the role of supercomputers, and in exploring the nature of computational science itself.


Last modified: 17 July, 1997
Thomas L. Marchioro
uces_info@krellinst.org