GOVERNMENT

By Adam Tanner, Reuters - Leading nuclear scientists with top security clearances will gather next summer at a screening room east of San Francisco and witness the results of the greatest effort ever in supercomputing. Using a computer doing 360 trillion calculations a second, scientists at Lawrence Livermore National Lab will simulate the explosion of an aging nuclear bomb in three dimensions. The short, highly detailed video produced by the world's fastest computer will attempt to illustrate how missiles dating back to the Nixon administration would perform today. "My job ... is to ensure that the nuclear weapons in the stockpile are safe and reliable," said Bruce Goodwin, associate director for defense and nuclear technologies. "Safe means no matter what you do to them they don't go off when they are not supposed to. Reliable means that should the president ever have to use one, it will work exactly as it is supposed to." The United States has about 10,000 nuclear warheads as a deterrent against attack. Washington stopped real nuclear tests in 1992, a year after the collapse of the Soviet Union, and signed the Comprehensive Test Ban Treaty in 1996. That ban means that a huge windowless room at Livermore is becoming a prime testing ground to make sure nuclear weapons dating back decades have not developed fatal flaws. In a room about half the size of a football field, BlueGene/L is a series of interconnected 6-foot-high (183-cm-high) racks holding 16 modules, each packed with massive computing power. The first part of BlueGene, built by IBM, became operational in mid-December at 90 trillion calculations a second; the rest should be ready by April. Even at its ultimate 360 trillion calculations a second speed, the simulation will take two to four months, lab officials say. This same calculation would have taken 60,000 years if done on technology available a decade ago. STILL VITAL AFTER THE COLD WAR? Some analysts say as impressive as BlueGene is, test simulations are not as vital in a post-Cold War world. "Why are we so focused on calculating or knowing the differences in performances of weapons?" asked Christopher Paine, co-director of the nuclear program at the Natural Resources Defense Council, an environmental group. "Just as long as we knew we were above a certain threshold, wouldn't that be enough to provide deterrence?" Paine contrasted present U.S. defense needs with those in the Cold War, when planners needed to assure that nuclear weapons would destroy hardened Soviet targets. "There is less reason to focus on the accuracy of our simulations," he said. "What we need to focus on from a deterrent point of view is just that we have weapons that go off." Nuclear physicist Goodwin responds that negotiated cuts in stockpiles in the years to come without newer replacements make the reliability of any one nuclear bomb more vital. "The question we are asking today is, as the stockpile changes on its own, will it continue to work and be safe?" he asked. Scientists say there is a lot they do not know about the effects of aging on the components of a nuclear bomb -- plutonium, uranium, high explosives, plastic and gases. For example will a bomb's plutonium last 50, 100 or 1,000 years? "They are made out of very corrosive materials," Goodwin said. "Yet the charge from the government is that this warhead -- which is made of these materials which are not happy with each other -- should remain perfectly safe and reliable indefinitely." Another question is how well the software written to simulate the atomic explosion will perform. Even the world's top software engineers routinely release flaws, and critics say nuclear scientists need more oversight. "If this were Microsoft Word, you'd have every hacker in the world trying to find the bugs," Goodwin said. "They're doing quality control for Microsoft in a pejorative sense after the fact. Well, you don't want there to be 1,000 places capable of doing nuclear weapons simulations." He said Livermore's main check came from the U.S. nuclear weapons research lab at Los Alamos in New Mexico. Engineers also do more conventional tests, including taking apart 11 atomic bombs every year, Goodwin said. When the nuclear scientists see the several-minute-long 3-D simulation from the roughly $100 million computer next summer, will it prove the most expensive animation ever? No, lab officials say, pointing to the current Hollywood film "Polar Express" which used computer animation in a production costing $270 million to make and promote.