ABQjournal.com : A service of the Albuquerque Journal
Friday, March 8, 2002
Super-Fast Computer Simulates Nuclear Blast
By John Fleck Journal Staff Writer
U.S. nuclear weapons scientists have for the first time detonated a nuclear weapon without the boom, simulating the blast in a supercomputer.
The Department of Energy on Thursday announced two successful three-dimensional computer simulations of nuclear bomb blasts, one by a team at Los Alamos National Laboratory and the second by a group at Lawrence Livermore National Laboratory in California.
The simulations are part of an effort to replace underground nuclear tests with computer simulations.
U.S. policy has prohibited underground tests since 1992 in keeping with an international effort to restrict the spread of nuclear weapons.
"We've got to move into the future without testing," said Bob Weaver, who headed the Los Alamos effort.
Both calculations were done on the world's fastest computer, a behemoth at Lawrence Livermore built by IBM as part of an effort by the Energy Department to push the state of the art in high-speed computing.
Frank Gilfeather, a supercomputer expert at the University of New Mexico, called the achievement "a major breakthrough in high-performance computing."
It is a demonstration, Gilfeather said, that the DOE supercomputer program is meeting its ambitious objectives.
Three-dimensional physics simulations have long been the Holy Grail of nuclear weaponeers.
In years past, weapons scientists built two-dimensional simulations in their computers to determine how nuclear weapons worked, then built bombs and set them off in underground tests to see if their theories were right.
Three-dimensional simulations require far more computing power, Weaver said.
The simulation tracks the temperature, pressure and other characteristics of the materials in the weapon as it explodes, to help weapons scientists understand what is happening inside the complex devices.
To ensure that their software accurately mimics a real weapon's performance, the recently completed tests simulated an actual nuclear explosive that had been tested in an underground blast, Weaver said.
The results were then compared to data from the real test to ensure the simulation was accurate, Weaver said.
The Los Alamos calculation took more than four months of computer time, spread over eight months last year. It was completed in October, according to lab scientists, but the announcement was delayed while the results were reviewed.
The Los Alamos calculation was the equivalent of a high-end home computer working continuously for 750 years, according to lab scientists.
For their calculation, Livermore scientists used the computer for 39 days straight.
The project was nerve-wracking, Weaver said, because it was not clear until the final weeks of the calculation whether it would be successful.
The next step is to get the simulation software running on an even-faster computer under construction at Los Alamos so large-scale simulations can be run more routinely, said Bob Boland, with Science Applications International Corp., which worked on the project.
The software was loaded onto the Livermore computer and was run from Los Alamos over the Internet.
The researchers used high-end computer graphics systems to analyze the data as the calculations proceeded, Weaver said.
They also tweaked their program as it was running in order to make it faster, he said.
Boland had retired from Los Alamos but came back as a consultant to work on the project, putting off plans to buy a motor home and tour America with his wife.
Boland, Weaver and other researchers gathered in front of a lab computer in late October when it was clear that the calculation had been successfully completed and together typed the instructions to shut the program down.
Copyright 2002 Albuquerque Journal Click for permission to reprint (PRC# 3.4676.620668)
All content copyright © ABQJournal.com and Albuquerque Journal and may not be republished without permission. Requests for permission to republish, or to copy and distribute must be obtained at the the Albuquerque Publishing Co. Library, 505-823-3492, or through Icopyright.com.
Back to story page