Sun in a Bottle - Charles Seife [76]
In the late 1970s, the LASNEX simulations told Nuckolls and the Livermore team that not that many divisions—not all that much laser power—needed to be brought to bear on the pellet to reach breakeven. The Shiva laser system, with its twenty beams, could pour about 10,000 joules into a pellet within a billionth of a second.60 According to LASNEX, this energy should be enough to ignite the pellet, and the reaction would produce as much energy as it consumed.
LASNEX was wrong. Very wrong. After a year’s worth of experimentation, the New York Times reported that “energy release at the giant $25 million Shiva fusion machine at the Lawrence Livermore National Laboratory in California fell short of the more optimistic estimates by a factor of 10,000.” LASNEX was a quite a few divisions short of a victorious army.
What went wrong? Because LASNEX is classified, it is hard to tell for certain. But what is certain is that a computer simulation is only as good as what its programmers know. When they write the code, they try to include everything they can think of, but there might be, and usually are, unanticipated phenomena lurking around the corner. It’s probable that as researchers experimented with ever more powerful lasers, they discovered difficulties that they didn’t expect and that weren’t programmed into the code. LASNEX wasn’t a perfect reflection of reality, because its programmers didn’t have perfect understanding.
Furthermore, even if every possible phenomenon were somehow programmed into LASNEX, the code wouldn’t necessarily predict the actually progression of a given experiment. There is a built-in limitation in the LASNEX code: it is only two-dimensional.
Running a LASNEX simulation takes an enormous amount of computing power; even the simplest simulated experiments require super-computers to chug away for a long time, moving imaginary electrons and nuclei and photons around in vast memory banks. The calculations are so complex, in fact, that a full-scale, three-dimensional simulation was way too much for the computers to handle. Instead, the programmers decided to simulate only a flat, two-dimensional slice of an imploding pellet. It was a necessary simplification; it brought the calculations into the realm of possibility. But at the same time, it meant that the LASNEX code couldn’t do a true simulation. If the real three-dimensional implosion behaved even slightly differently from a two-dimensional mockup, then LASNEX could not predict its behavior very well.
Nevertheless, LASNEX’s programmers were justly proud of their accomplishment. Under many circumstances, the code did predict precisely the outcome of a given experiment. However, once Shiva became fully operational, it became dreadfully apparent that LASNEX didn’t give quite the right answer all the time—and its promises of success didn’t come true. Its failed predictions were a blow to the Livermore scientists, but they refused to be derailed. A much bigger project, Nova, was already under way. The LASNEX code—presumably tweaked to take into account the results from the Shiva experiments—apparently predicted a resounding success with Nova, which was ten times more powerful than its predecessor.
Then Nova, too, failed to achieve breakeven. The laser was certainly generating fusion reactions. By the mid-1980s it was achieving about ten trillion fusion neutrons with each shot. But again, the laser consumed one thousand to ten thousand times as much energy as the fusion reactions produced. Once more, LASNEX had failed, and the scientists’ optimistic expectations were crushed. This time, though, their failure had cost almost $200 million.
Even though LASNEX has failed and