Nuclear weapons, while simple in principle, are technically complex devices with a multitude of components. As with any complicated piece of equipment, there may be concern that, over time, a weapon’s reliability could decline. To coordinate efforts to maintain the nation’s existing nuclear weapons, the Department of Energy developed a Stockpile Stewardship Program (SSP). A recently released Federation of American Scientists Occasional Paper, The Stockpile Stewardship Program: Fifteen Years On (PDF), by FAS analysts Anne Fitzpatrick and Ivan Oelrich, reviews the status of the experimental devices that support the SSP, describes how each experiment is supposed to work, and identifies the problems that have been encountered.
All of the expensive SSP experiments were initiated because of the cessation of nuclear testing, with the expectation that they would be essential to maintaining the nuclear stockpile. The major components of SSP are all seriously over budget and seriously behind schedule but, even so, our scientists now have a much better understanding of nuclear weapons and how they age. Now the DOE is proposing moving away from indefinite stockpile stewardship to a “Reliable, Replacement Warhead,” which, if designed for simplicity and with broad performance margins, could avoid the need for the SSP experiments. It is fair then to ask just how essential these megaprojects continue to be.
The SSP supported three major experiments: the National Ignition Facility (NIF) to use laser beams to compress a hydrogen target to densities and pressures where fusion would occur; the Dual-Axis Radiographic Hydrodynamic Test (DARHT) Facility uses x-rays to follow the shape of sections of plutonium when they are compressed as they would be in a nuclear bomb; and the Accelerated Strategic Computing Initiative (ASCI)—renamed Advanced Simulation and Computing (ASC) — to build supercomputers and associated software to use the information from other experiments to model nuclear warheads and predict their behavior. Two other experiments later fell under the SSP: The Joint Actinide Shock Physics Experimental Research (JASPER) facility is a high speed gun used to study shock waves in plutonium and the Z-Machine, or Z-Accelerator, creates the x-ray intensity used to study nuclear explosive conditions.
The National Ignition Facility (NIF) was originally budgeted to cost just a shade over one billion dollars and to be finished four years ago. It is now expected to carry out its first experiments in 2010 and to cost more than another billion dollars to complete, greater than the original estimates of total cost. Based on unclassified sources, it appears that the connection between NIF and the current SSP is at best indirect. We believe that NIF could be ended without reducing the confidence in the existing nuclear stockpile.
The Dual-Axis Radiographic Hydrodynamic Test (DARHT) Facility is designed to have two, hence “dual-axis,” x-ray machines that look at a subcritical plutonium pit as it is compressed by conventional explosives. Only one axis is currently operating and is providing valuable information.
The computer effort, ASC, has also been plagued with problems but is different from the two big physics experiments because it never had a focus on one particular machine. Indeed, it is not at all clear when the ASC program will be “done.” Being able to model a nuclear weapon on a computer is one of the critical substitutes for nuclear testing. The Advanced Simulation and Computing (ASC) program has already made important contributions to understanding the behavior of nuclear weapons and reportedly has resolved some worrying questions.
The ASC initiative has supported the construction of several large supercomputers. To achieve the necessary speeds, thousands of central processing units, CPUs, have been linked together to operate in parallel. These ambitious development programs have all had problems. Construction on some computers was started but never completed while some computers suffered from low reliability because of their complexity. In many cases, Herculean hardware developments were not matched by development of software that could fully exploit the new machines’ capability. Even successes were short lived: the world’s fastest computer today will be overtaken by some rival within months or a year.
To the greatest extent possible, DOE should use new computer capability coming out of industry and the universities and focus its efforts on DOE-specific problems, and get a better balance between hardware and software. DOE must justify why it needs leading edge computers when that edge is inevitably overtaken within a year or two. A decade into the SSP, DOE knows that nuclear weapons are far more stable than initially feared, and our understanding of the aging and stability of nuclear weapons continues to increase. There can and should be less urgency to DOE computer development.
All of the SSP experiments, but NIF in particular, are promoted as a means to attract top new scientific talent to DOE and the SSP. FAS remains deeply skeptical. The universities and industry are now at the cutting edge of scientific and technical advance. Anecdotal evidence strongly suggests that newly minted scientists do not look to the DOE labs as their first choice for doing pioneering research. Yet even if NIF did contribute to this goal to some degree, it is far from being the most efficient means of applying those billions of dollars. The great majority of the resources going to NIF support engineering problems—related to lasers, clean rooms, and power supplies—that have nothing whatsoever to do with either nuclear weapons or basic physics. That money could have gone directly to support university research of interest to DOE or to create smaller but scientifically more interesting experiments within the labs.
The current approach to stockpile stewardship, careful surveillance and monitoring along with judicious replacement of parts, has maintained a nuclear stockpile that is safe and reliable. Of the three major experiments supported by SSP, ASC and DARHT are already making contributions. NIF is more uncertain, both its ultimate success and its contribution to our confidence in the stockpile. But even without NIF, the United States can maintain its existing nuclear weapons without a return to testing.