Cameras tracked it from liftoff until it disappeared into space. For 16 days, delicate sensors sent back streams of details about its mechanical systems and human crew. Teams of the world's best engineers monitored its heartbeat, second by second.
So why is it so hard to determine the cause of death for the space shuttle Columbia?
Because the $2 billion spacecraft, with its 2.5 million parts and 3,500 critical systems, is one of the most complex and fragile flying machines ever built. And so far, it has left few unambiguous clues.
In the past two days, NASA officials have all but ruled out their initial theory on the cause of the catastrophe and disclosed that a camera that might have offered some answers malfunctioned on the launch pad.
"We're scratching our heads," shuttle manager Ron Dittemore said yesterday. That candid assessment makes this case different from the 1986 Challenger explosion, when investigators had a good hunch about the cause within hours.
Solving the mystery of Columbia's breakup might be more like investigating an airline crash, which often takes weeks, months or years to resolve, experts say. That's because accidents rarely result from a single failure.
"These kinds of catastrophes are hardly ever the cause of a single stupid mistake or oversight, but are the multiplication of several factors," said James Oberg, a former shuttle flight controller who has written about the space program.
"The key is going to be to have the discipline to identify all the possibilities, not jump to conclusions, and go through each one," said Tommy Holloway, retired manager of the space shuttle program.
NASA investigators are painstakingly assembling "fault trees," engineer-speak for a chain of possible causes that have to be examined one by one.
Nothing is ruled out
"Let [me] emphasize again that we have not ruled out any possible cause," Dittemore said at a news briefing yesterday. "You have to look at each branch, each block, each possible cause, to determine that you have done a thorough job."
One branch on the tree will involve the chunk of foam insulation that broke away from the shuttle's external fuel tank on liftoff and struck the critical tiles that protected the underside of the shuttle's left wing.
An analysis by NASA engineers while the shuttle was in orbit concluded that the incident posed no serious risk to the spacecraft. In the days after the breakup, NASA engineers reconsidered and said the flying insulation had to be a leading suspect, but they ultimately rejected it again after reviewing their calculations.
Even so, Dittemore said yesterday, "it still exists in the fault tree until we can close it."
That task will be more difficult because investigators don't have a sharp video of Columbia's underside during takeoff, which might have shown damage to the craft's heat shield tiles. Dittemore said the camera aimed at the shuttle's belly was out of focus, forcing engineers to rely on video shots from other angles that are inconclusive.
Investigators will also concentrate on Columbia's left wing and wheel well, which began to show signs of heating while the craft was passing high over California on its way to a scheduled landing at Cape Canaveral.
They know the craft began to yaw sharply to the left just before it broke up - a sign of unusual drag on the left wing.
They're looking at the possibility that some other kind of debris - possibly even a meteoroid - might have struck the wing during liftoff, in orbit or during the landing approach, fatally damaging the wing or heat shield. The agency is also trying to determine the validity of an amateur photograph that shows a streak of what could be lightning striking the craft over California, Dittemore said.
Investigators are starting to examine more than 1,000 pieces of debris that have been assembled so far at Barksdale Air Force Base in Texas. That's a fraction of the 12,000 pieces that residents of Louisiana and Texas have reported.
Debris might be the ultimate clue to where the shuttle started to break up, which would help investigators recreate Saturday's events. Dittemore said Columbia appeared to be flying well until it was over Texas. No debris has been found west of Fort Worth, but amateur video and witness accounts suggest the shuttle might have started to break up over California.
NASA officials are beginning to review those accounts carefully. "We need more time to pull it together, sort it out, get it to the right experts so we can put the pieces of the puzzle together," Dittemore said.
There are so many pieces because so many systems have to work perfectly to protect the shuttle during liftoff, support life in orbit and bring the crew safely home through ferociously hostile conditions.
At liftoff, the shuttle is a flying bomb. Its solid-fuel boosters put out 6.6 million pounds of thrust and burn 5 tons of fuel per second. Once ignited, they can't be turned off.
Its three main engines add 1.2 million pounds. They're powered by 704 tons of volatile liquid hydrogen and oxygen fuel. The consequences of an explosion were demonstrated by Challenger.
In orbit, the astronauts fly with still more liquid hydrogen and oxygen powering the fuel cells used to generate electricity and water. There is dangerous hydrazine thruster fuel on board and a variety of explosives needed to deploy the landing gear or jettison other parts in emergencies.
Outside are the near-vacuum and deep cold of space. Any failure of critical seals, any puncture by a speeding bit of space junk or a meteoroid would depressurize the shuttle and kill the crew. There are no lifeboats.
On re-entry, the shuttle encounters Earth's atmosphere at 17,000 mph in temperatures up to 3,000 degrees. Its heat shield is composed of silicate tiles so fragile that they can be dented by raindrops. A single flaw could allow the craft's aluminum airframe to melt.
The shuttle's orientation as it encounters the atmosphere and its path through it at many times the speed of sound are so critical that its flight surfaces must be controlled by computers. A mistake could incinerate the ship or send it skipping off into space.
And because the ship glides to Earth without power, its guidance computer and crew get only one chance to hit the runway.
Dittemore said engineers eventually will reassemble as much of the shuttle as they can find at the Kennedy Space Center in Florida.
This is a common procedure during the investigation of airline crashes by the National Transportation Safety Board, which is also participating in the Columbia probe. Although a special independent panel will oversee the Columbia investigation, NTSB officials noted yesterday that crashes can take months or years to resolve.
Since 1967, when the NTSB was created, there have been three fatal crashes for which causes have never been determined, a spokesman said.
Bernard Loeb, the agency's former director of accident investigations, noted that it took four years to officially determine that a rudder control malfunction was responsible for an accident that killed 132 people on a Boeing 737 jetliner. Boeing contested the original finding.
"These investigations are extremely complicated and sometimes there are a lot of players who have their own views of what may have happened," Loeb said.
Sun staff writer Alec MacGillis and the Associated Press contributed to this article.