The Dawn supercomputer (IBM BlueGene P) in the Terascale Simulation Facility at Lawrence Livermore National Laboratory. (Jacqueline McBride/Lawrence Livermore National Laboratory )

— A group of nuclear weapons designers and scientists at the Lawrence Livermore National Laboratory conducted a what-if experiment several years ago, deploying supercomputers to simulate what happens to a nuclear weapon from the moment it leaves storage to the point when it hits a target.

They methodically worked down a checklist of all the possible conditions that could affect the B-83 strategic nuclear bomb, the most powerful and one of the most modern weapons in the U.S. arsenal, officials said. The scientists and designers examined how temperature, altitude, vibration and other factors would affect the bomb in what is called the stockpile-to-target sequence.

Such checks typically have been carried out by taking bombs and warheads apart; scrutinizing them using chemistry, physics, mathematics, materials science and other disciplines; and examining data from earlier nuclear explosive tests. This time, however, the scientists and designers relied entirely on supercomputer modeling, running huge amounts of code.

Then came a surprise. The computer simulations showed that at a certain point from stockpile to target, the weapon would “fail catastrophically,” according to Bruce T. Goodwin, principal associate director at Livermore for weapons programs. Such a failure would mean that the weapon would not produce the explosive yield expected by the military — either none at all, or something quite different than required to properly hit the target.

“So we went in and thoroughly investigated that, and determined that the way the weapon is handled by the military had to be changed, or you would be susceptible to having the weapons fail catastrophically when, God forbid, they should ever be used,” Goodwin said. He added that the fault occurred in the “real dynamics of the vehicle” — a term describing the weapon’s trajectory and behavior — and could not have been revealed by underground explosive testing or by examining the components.

Following the discovery and a multi-year effort, the B-83 bombs and the military’s handling procedures for the weapons have been fixed, officials said.

The episode, details of which remain classified, offers a glimpse into a rarely seen but potentially significant shift in the nuclear weapons era. According to scientists and officials, the United States’ weapons laboratories, armed with some of the fastest computers on the planet, are peering ever deeper into the mystery of how thermonuclear explosions occur, gaining an understanding that in some ways goes beyond what was learned from explosive tests, which ended in 1992.

The Obama administration has said that with computing advances, the United States will never need to resume nuclear explosive testing. Undersecretary of State Ellen Tauscher said in May that “our current efforts go a step beyond explosive testing by enabling the labs to anticipate problems in advance and reduce their potential impact on our arsenal — something that nuclear testing could not do.”

The significant advance in computer modeling is at the center of a debate over the Comprehensive Test Ban Treaty, which was approved by the United Nations in 1996 but rejected by the U.S. Senate in 1999. Signed by 182 countries and ratified by 154, the treaty outlaws nuclear explosive testing and sets up a global monitoring system to detect any tests. The treaty needs several key countries, including the United States, to ratify it before it can enter into force. The Obama administration has urged the Senate to ratify the pact and continues to abide by the test ban.

The simulation of the B-83, a device designed and developed by Livermore in the late years of the Cold War, marked the first time such a major fault in a nuclear weapon was detected largely by computer simulation, Goodwin said. “We have a more fundamental understanding of how these weapons work today than we ever imagined when we were blowing them up,” he added.

But a former nuclear weapons designer, who spoke on the condition of anonymity because he is still in the government, offered a more cautious view. “To say the calculations are better than underground testing is silly,” he said. “If you want to know if something works, you have to test it. The calculations are good, but the issue is one of risk. How good do you think the calculations are?”

Stockpile stewardship

The laboratories, including Livermore in California and Los Alamos National Laboratory and Sandia National Laboratory in New Mexico, are responsible for certifying to the president the safety and reliability of the nation’s nuclear weapons under a Department of Energy program known as stockpile stewardship, run by the National Nuclear Security Administration.

Over the years, various flaws have been detected in the nuclear arsenal, some worse than others. A serious incident occurred in 2003, when traditional checks revealed a problem that, while not catastrophic, was widespread. Details of that problem are also classified. In response to the discovery, Livermore scientists performed a series of computer simulations, followed by high-explosive but nonnuclear experiments at Los Alamos, that showed the weapons did not need a major repair that might have cost billions of dollars, Goodwin said. In an earlier time, he added, the only way to reach that conclusion might have been to resume nuclear testing.

At the time the test ban treaty was defeated, critics said the United States might someday need to return to testing. Six former secretaries of defense in Republican administrations, including Caspar W. Weinberger, Richard B. Cheney and Donald H. Rumsfeld, wrote to the Senate in 1999 that the planned stockpile stewardship program “will not be mature for at least 10 years” and could only mitigate, not eliminate, a loss of confidence in weapons without testing.

Sen. Jon Kyl (R-Ariz.), who has long opposed the treaty, said: “Computer simulation is a part of the stockpile stewardship program, which scientists say has been helpful. One told me it produced good news and bad news. The good news is that it tells us a lot more about these weapons than we ever knew before. The bad news is that it tells us the weapons have bigger problems that we realized. While computers are helpful, they’re not a substitute for testing. That’s why, even though we’re not testing right now, we should not give up the legal right to test.”

The United States and the Soviet Union carried out 1,769 nuclear explosive tests during the Cold War. Many were designed to check the yield and other properties of new weapons. But with the end of the superpower confrontation, weapons designers inherited a new and difficult task: to maintain the arsenal without explosions. To bolster the effort, congressional Republicans pressed President Obama this year for a large injection of money for the nuclear weapons complex, and the president pledged to increase spending by $88 billion over the next decade. Obama requested 10 percent more in next year’s budget for the stockpile stewardship program.

There are approximately 20,500 nuclear warheads remaining in the world. The United States and the Russian Federation together have about 19,500 of them, according to the best estimates.

A new understanding

Jeffrey G. Lewis, a nuclear weapons expert and the director of the East Asia Nonproliferation Program at the Monterey Institute of International Studies, said that years of underground nuclear tests helped show weapons designers that the bombs worked under certain conditions, “but they could never fully explain how or why.”

“The best argument against the test ban was always that we didn’t understand how nuclear weapons really worked and couldn’t simulate them, so underground nuclear explosions were an important reality check,” he added. “But even then, there was never enough testing to establish the kind of confidence that comes from actually understanding the process of a thermonuclear explosion.”

As a result of the computer modeling, he added, “for the first time, nuclear weapons designers understand why and how thermonuclear weapons work.”

In recent years, physicists at Livermore surmounted one of the oldest and most difficult challenges they faced. In many nuclear weapons explosive tests, measurements suggested that the detonating bombs appeared to violate a law of physics, “conservation of energy,” which states that in a closed system, the total amount of energy remains constant, and thus energy cannot be either created or destroyed.

For decades, the nuclear weaponeers puzzled over why the test results appeared to break from this principle. Then, the “energy balance” problem, as it was known, was solved by a Livermore physicist, Omar Hurricane, who won the 2009 E.O. Lawrence Award from the Department of Energy for his work, which remains classified.

The supercomputers at Livermore and the other national laboratories do not do the job alone. Scientists use data from the 1,054 U.S. nuclear tests between 1945 and 1992, of which about 200 are relevant to today’s arsenal. They also cross-check the computer findings with laboratory experiments.

One of the most elaborate and ambitious is the National Ignition Facility at Livermore, housed in a stadium-size building where scientists hope to use 192 laser beams to achieve fusion ignition in a laboratory setting, a process that has never been witnessed. If it works, the lasers will heat a tiny fuel pellet of tritium and deuterium to 100 million degrees, causing some of the nuclei to fuse, generating energy and producing conditions close to those inside the core of stars — and inside a detonating nuclear weapon.

“No one has ever seen hydrogen fusion bare naked in a vacuum. It’s always been buried in the middle of an atomic weapon,” Goodwin said. “You can infer what happened.” If ignition is achieved, he added, scientists will be able to not only see it, but measure it.

The facility has suffered long delays and setbacks, but the target is to achieve ignition by next autumn.

Another, the Dual-Axis Radiographic Hydrodynamic Test Facility at Los Alamos, uses X-rays to follow the shape of sections of plutonium when they are compressed as they would be in a nuclear weapon explosion.

Teraflops and beyond

When the stockpile stewardship program was begun in the 1990s, the goal was to build a generation of supercomputers capable of 100 teraflops. A teraflop is a measure of a computer’s processing speed that is 1 trillion floating point operations per second; an operation could be a single mathematical calculation, such as addition or multiplication. This was accomplished, but new machines are now pushing far beyond.

Next May or June, Livermore plans to put into operation an IBM supercomputer, Sequoia, capable of 20 petaflops. A petaflop is a thousand trillion floating point operations per second. The machine, on 96 refrigerator-size racks, will contain 1.6 million processing cores and will be 10 times faster than what is now the fastest computer in the world. By comparison, all the computing power at Livermore today is about 2.5 petaflops.

With such vast computing capability, scientists can attempt to create a realistic model of what happens inside a nuclear explosion, when tremendous pressures and temperatures squeeze metals, including uranium and plutonium, to set off the nuclear blast. Fred Streitz, director of Livermore’s Institute for Scientific Computing Research, said the ultra-fast machines are “opening doors to new science,” such as models of how atoms behave, or how the crystal or grain structure of a metal changes under pressure.

Streitz said that over time, it became clear that smaller computer simulations were returning incorrect answers; only with finer resolution and more power could scientists grasp what was really happening.

In one example involving molten copper and aluminum, Streitz said, 9 billion atoms were modeled. It took more than 212,000 computer processors more than a week to carry out the simulation, he said, but the result was a near-perfect resolution of how the metals behaved.

“This is millions of times finer than you could ever do in a nuclear test,” Goodwin said. “You could never see this process go on inside a nuclear explosion.”