How to Take the Pulse of Aging Warheads?
The thousands of processors that make up one of the world’s two fastest computers fill a 10,000-square-foot building at Los Alamos National Laboratory in the mountains of New Mexico, and they are finally beginning to hum. When it is finished, the giant computer called Blue Mountain will amount to one small step in a mission some say is impossible.
That mission is to verify that the nation’s aging stockpile of nuclear weapons, which can no longer be test-fired or replaced, will work as expected if they are ever needed.
“I can’t find anybody here who will say with certainty that this will work,” said Jas Mercer-Smith, the No. 2 man for nuclear weapons at the sprawling lab that gave birth to the nuclear age. “I can find a large number of people who say we may be able to pull this off. But I can find a large number of people who say we will fail.”
So Mercer-Smith and his colleagues at Los Alamos, as well as a team of scientists developing a similar and equally powerful machine at Lawrence Livermore National Laboratory in California, are faced with a formidable challenge, which comes down to this: The health of the nation’s nuclear arsenal rests with computer technology that no one has tried before, and no one can say will succeed.
The stage for this drama began with the passage of the Comprehensive Nuclear Test Ban Treaty and the decision by U.S. leaders in 1989 to ban further construction of nuclear weapons. That left the country dependent on a huge number of warheads that are, like the rest of us, getting older.
“Along with everything else in the world, nuclear weapons age,” Mercer-Smith said. “They have chips, gaps, cracks and corrosion. These can change the performance of the weapon.”
The primary fuel inside a nuclear bomb is plutonium, a man-made element, although it exists in minute quantities in some ores. And although plutonium has a half-life of 25,000 years (the time required for half its atoms to decay), that does not necessarily mean the element will remain viable indefinitely as a weapons fuel. No one really knows yet just what changes it will go through in the earliest stages of decay.
In the past, the stewards of the stockpile were able to verify the quality of the weapons by occasionally popping one off in an underground test. But now that they can no longer do that, they have turned to computers in hopes of simulating everything that can happen to a weapon as it ages.
But that has left them with two profound problems: How can you be sure the data you feed into the computer is correct if you can no longer compare the results with a test in the real world? The second is nearly as daunting: How do you build a computer that is millions of times faster than the most powerful computers in the world?
With as many as 6,000 parts, a nuclear weapon is a very complex piece of equipment. The housing that holds the fuel, for example, must remain perfectly symmetrical for the fuel to reach criticality, and everything must work exactly right in an incredibly short period of time. The goal is to be able to simulate an explosion from the first signal all the way through detonation, and that requires trillions of instructions to the computer in less than a second.
Scientists at both Los Alamos, which is building Blue Mountain, and Livermore, which is building Blue Pacific, are banking their hopes on linking together thousands of powerful processors to work in parallel at each site as single, giant computers. It is all part of a 10-year, $4-billion program by the Energy Department called the Accelerated Strategic Computing Initiative.
Los Alamos is working with Silicon Graphics Inc. of Mountain View, and Livermore is working with IBM, and the approach at both labs is similar.
Los Alamos received its first 1,000 processors last December, and at first nothing worked. It turned out that one connector out of 10,000 was faulty. Since that time, hundreds of additional processors have been delivered to the lab, and later this year scientists hope to have all 6,144 processors working together.
“A lot of people thought we were crazy,” Mercer-Smith said, because no one was sure such a complex network of powerful computers could even be programmed. “But I’ll tell you right now it’s working, and it’s working far better than I ever thought it would.”
Even though it still isn’t completed, Mercer-Smith said the computer is working on a problem it is expected to finish in two months. It would have taken a Cray supercomputer 100 years to do the same thing, he said.
When completed, Blue Mountain will be a 3.1-teraflop computer. A single teraflop is 1 trillion computer instructions per second. According to Silicon Graphics, Blue Mountain will be able to perform more calculations in one second than a human being with a hand-held calculator could do in 3 million years.
Blue Mountain and the 3-teraflop machine at Livermore are to be followed by a 10-teraflop machine at Livermore, and subsequently a 30-teraflop computer at Los Alamos by 2001 or 2002. The goal is to build a 100-teraflop computer at an undetermined location by 2004. That machine must be able to do a million billion calculations a second.
And then the going will really get tough. Unable to conduct new tests, scientists will rely chiefly on data from about 1,000 underground tests over the last few decades. That will give them a chance to postulate all kinds of things that might go wrong, and then see what the computer tells them. They can then compare the computer results to what happened in the earlier tests.
“Our mistakes [in the past] will probably turn out to be more important than our successes,” Mercer-Smith said. “Like when we didn’t build it quite the way we thought we had built it, like when the engineers weren’t perfect. That becomes a critical part of this validation” because scientists will be able to see if their powerful computer predicts what they now know happened.
That won’t solve every problem, such as what happens inside the bomb with the gradual decay of the plutonium because no one has ever been there before.
Mercer-Smith compares the whole thing to his car.
“I’ve got a really old car,” he said. “It’s a ’72 Ford, and it’s got dents and it’s rusting away. But it’s got a great engine, and I drive it every day.”
It is, one might say, a bomb. But what’s really important, he added, “is knowing the difference between a broken headlight and a cracked engine block.”
“That’s the metaphor we’re talking about here.”
*
Lee Dye can be reached via e-mail at leedye@compuserve.com.
