Officials at CERN, near Geneva, believe they have found a solution to the latest hitch in the construction of the Large Hadron Collider (LHC), a state-of-the-art particle accelerator that straddles the Franco-Swiss border. That was the upshot of a two-day review held the last week of April at CERN to find a way forward following a major failure during a high-pressure test on critical components of the machine in March. The mishap casts a pall on a situation that otherwise seemed to be all-systems-go.

The LHC is a 27-kilometer-long circular tunnel designed to collide proton beams. Scientists hope that the products of those collisions will lead to better understanding of matter and the universe. The beams are directed through the tunnels by 1700 superconducting magnet assemblies, and it was a small but important subset of these—roughly eight 35-meter-long linear assemblies of focusing magnets called inner triplets—that ran into trouble.

The test was intended to simulate the kinds of forces the magnets will encounter under actual operating conditions and to examine, for instance, the movements of pipes. Pressure was applied to one of the triplet assemblies in steps of 5 bar, says Ranko Ostojic, a CERN engineer who leads the team responsible for the LHC’s special-purpose magnets. On the night of the test, Ostojic was in the tunnel with seven other people, making measurements. At 12 bar, the observers retreated 30 meters from the magnet, in keeping with safety rules. At 20 bar, there was a sudden release of helium gas (used to cool the magnets), something went bang, and sirens sounded, alerting the fire brigade. No one was hurt.

Examination half an hour later showed that the magnets were basically in good order. But the pipes connecting them were another matter. Some had shifted a little, some a lot, and one had fractured, which is what caused the discharge of helium gas. Moreover, the supports for one of the magnets, glass-reinforced plastic structures called spiders, had broken. Post-mortem analysis showed that the supports weren’t strong enough to counteract an asymmetric longitudinal force of about 15 tons.

Fermi National Accelerator Laboratory, in Batavia, Ill., the U.S. national laboratory, along with two other U.S. facilities and the High Energy Accelerator Research Organization, KEK, in Japan, designed the triplet magnets and supplied them to CERN. Fermilab announced the failure and immediately accepted blame for what it calls a design blooper. But the lab is hard put to explain how the problem happened. It isn’t that the forces were unknown, says James Kerby, U.S. LHC accelerator project manager at Fermilab. “We’re aware of them, and we didn’t properly account for them.” Four external design reviews carried out from 1998 to 2002 also failed to detect the flaw.

The magnets have suffered other, minor mishaps. Some of the initial supports arrived in Geneva damaged, after having been transported by train, which was in direct violation of shipping instructions. And in December, another pressure test resulted in the collapse of heat exchangers in the triplets. Both defects have been rectified, says Kerby, and neither had anything to do with the current issue.

Fermilab successfully tested the magnet elements separately before sending them to Switzerland. But the forces that caused the failure could occur only in the collective behavior of several magnets strung together, as they were in the tunnel. In addition to pledging its full support to CERN, Fermilab is conducting its own series of reviews “to fix the process” that led to the oversight, says Kerby, “until we get it right.” The purpose of last week’s meeting, however, was solely to remedy the problem at hand.

A total of nine units (eight commissioned, and one spare) need to be “upgraded.” Although only one unit of three magnets was involved the March incident, says Ostojic, “it’s a systematic problem.” The fix will reinforce the magnets at critical points with “cartridges,” stiff mechanical springs made of invar (a type of steel) and aluminum. Broken spiders will be replaced, but otherwise need no redesign. Both the cartridge alone and cartridge plus magnet will be tested on the surface at appropriate loads. It is hoped that the magnets inside the tunnel can then be modified without having to be brought to the surface.

That will save time, though it looks increasingly unlikely that CERN will meet its deadline for an engineering run at the end of 2007. People are philosophical about the episode, citing the LHC’s enormous complexity. What is annoying is to have it happen at such a late stage. “We will see now how quickly we can recover from this,” says Lyn Evans, LHC project director. “Our real number one priority is to make sure…that we come online as foreseen in spring 2008, and that we get the machine up to top energy as fast as we can. With or without an engineering run before.”