Development of Scalable Algorithms for Numerical Relativity

Black Hole Collision: Step 1

Objective

The grand objective of the project has been and remains the modeling of the collision of two black holes. One of the supporting objectives has been developing algorithms to enforce certain constraints during the modeling, such as the conservation of mass.

Approach

Our approach has been the team development of high-performance algorithms and software. The team consists of astrophysicists and computer scientists.

Accomplishments

Black Hole Collision: Step 2A family of gravitational codes has been developed, with varying numerical and high-performance programming features. Extensive use of distributed-memory machines has been made. These were the CM-5, CRAY T3D, IBM SP2, and Intel Paragon. Also, distributed-shared-memory machines such as the Convex Exemplar have been used, as well as traditional shared-memory vector supercomputers such as the CRAY C90. A portable relativity code has resulted that achieves nearly 20 gigaFLOPS on a 512-node CM-5 and over 8 gigaFLOPS on a CRAY C916. A formulation for constraint enforcement has been achieved in three spatial dimensions.

Significance

Black Hole Collision: Step 3The collision of two black holes produces gravity waves that radiate away through space. These waves can be detected by a gravitational observatory, the very first of which will become operational in the year 2000: the Laser Interferometric Gravitational Observatory (LIGO). Our era is similar to that of Galileo when he first turned a telescope skyward to observe the universe using light. LIGO will be the first to observe the universe using gravitational waves. An important reason for our project is to find out what a gravitational wave looks like so LIGO can recognize one when it "sees" it.

Aside from the historic significance of the new gravitational observatory, our project has also been significant for the progress in high-performance software for the solution of some of the most complicated mathematical descriptions of nature in all of science.

Black Hole Collision: Step 4

References

  1. "Numerical Method of Constrained Time Evolution Problem," by Ming Hwa Torng. NCSA report under preparation. Available from anonymous FTP at "ftp cs.uiuc.edu" and then "cd /pub/saylor".
  2. "Three dimensional numerical relativity: the evolution of black holes," P. Anninos, K. Camarda, J. Masso, E. Seidel, W.-M. Suen, and J. Towns, Physical Review D, 52, 2059, (1995). An image is in FTP site (loc. cit.) under file name 2BHCol3d_4.tiff.

Point of Contact

Paul Saylor
Department of Computer Science
University of Illinois at Urbana-Champaign
saylor@cs.uiuc.edu
217-333-0256