Just after 10 a.m. during the second of Columbia's 16 full days in orbit, something drifted away from the shuttle at the speed of a brisk walk.
Columbia was flying at 17,500 mph, tail-first and upside down, its open cargo bay pointing toward the planet below.
The object, measuring no more than 1 square foot, began to spin faster and faster. It took 2 1/2 days to fall out of orbit and burn up in the atmosphere.
It twirled and twinkled as it fell to Earth.
Apparently unseen by human eyes, it was recorded by military radar and stored in a computer.
There the reading stayed until after the accident, when Air Force technicians discovered it among thousands of automatic radar sweeps of Columbia in space.
In the surveillance files of the Air Force 21st Space Wing, the electronic traces of the mysterious object amounted to little more than squirts of static.
Had something drifted out of the open cargo bay on Jan. 17?
NASA investigators checked the lost-and-found log at Kennedy Space Center and discovered seven missing tools — an Allen wrench, screwdrivers, a flashlight and pliers.
Had the shuttle been hit by something in orbit?
Investigators checked mission records to see if Columbia had shuddered from the impact of a meteoroid or orbital debris.
They found nothing out of the ordinary.
Only one place had a chance of deciphering the faint radar signal — the Hog Works, a windowless laboratory behind two sets of locked security doors at Wright-Patterson Air Force Base near Dayton, Ohio.
Inside, signs warned against classified discussions in the hallways. A red security light flashed in warning when the janitor came by to mop the floors.
When a visitor entered the building, a researcher made a timeout gesture with his hands like a football referee, indicating that conversations should stop.
Technically, the radar range was called an occult chamber.
The sealed room was the size of a strip mall cinema. Its walls, floor and high ceiling were covered with thick layers of blue sculpted baffles that resembled the sound-scattering acoustic clouds in a concert hall.
In the center was a 40-foot-high pylon. Researchers secured targets at the top and bombarded them with radar waves.
The technical trick of the range was to allow researchers to see how objects appeared on radar at distances of hundreds or even thousands of miles.
Typically, a radar image could contain measurements recorded at 1,600 different frequencies, etching a richly detailed three-dimensional picture. The object leaving Columbia had been recorded in just one wavelength.
In essence, it was a drawing that consisted of a single jagged line.
From the way the unidentified object had scattered the radar beam, the researchers deduced that it was smaller than the wavelength of that frequency, about 27 inches. Analysis of the way it fell to Earth gave them some ideas about its weight and shape.
In April, a team led by Brian Kent, a 46-year-old expert in radar measurements, began screening materials and objects aboard the space shuttle based on how they scattered radar energy along that one wavelength.
One by one, Kent positioned each target at the top of the pylon and bombarded it with radar waves. The team tested 31 items from NASA's inventory — shuttle tiles, seals, spar insulation, cargo bay blankets and other materials. They also tested four scraps of debris recovered from Texas.
The radar signal and the ballistic measurements matched only one item: a fragment of the curved reinforced carbon- carbon panel from the leading edge of the orbiter's wing.
But what had knocked it loose?
It was the first to be suspected, the first to be denied and the most credible possibility remaining after months of sophisticated second-guessing.
During every launch for 22 years, the foam insulation coating the shuttle's 15-story external fuel tank had flaked off like dandruff.
The debris that struck Columbia's wing 81.9 seconds after liftoff Jan. 16 was the largest piece of foam ever to strike a shuttle.
NASA Administrator Sean O'Keefe had publicly dismissed concerns, comparing the debris to "a Styrofoam cooler blowing off a pickup truck ahead of you on the highway."
More than 2 tons of foam insulated the tank. Most of it was applied precisely by computer-controlled robotics. A few hard-to-get-at areas still had to be done by hand. One such area was the source of the foam that hit the wing.
Trying to understand how NASA used the foam, Nobel laureate Douglas Osheroff, a member of the Columbia Accident Investigation Board, watched a video on his laptop computer.
In it, a technician at Lockheed Martin, maker of the shuttle's external fuel tank, sprayed streams of foam in broad, measured strokes across a tank brace.
Despite careful application, the hand-sprayed foam in the demonstration video expanded unevenly, leaving gaps and hollows that could fracture. When investigators cut into the foam on another tank, they discovered three air pockets near the same crucial point on the tank.
"Voids will be voids," quipped a technician to the camera operator.
He caught himself and spoke again into the video lens. "I hope you had the audio turned off."
To Osheroff, it was no laughing matter.
What puzzled the Stanford University scientist most was not any physical property of the foam.
It was a property of mind.
"Foam has been around for 25 years. Why did NASA never do the experiments to understand the process?" Osheroff said. "They hadn't a clue about things that might contribute to foam shedding."
Osheroff won a Nobel Prize in 1996 for his breakthroughs in low-temperature physics. He worked with materials only a few thousandths of a degree above absolute zero, so cold that molecular motion almost ceases and matter reveals its strange, normally hidden character.
To work properly, the foam must cling to the tank in a streamlined blanket. Inside the tank is liquid hydrogen at 423 degrees below zero and liquid oxygen at 297 degrees below zero. During flight, the outside temperature rises to about 350 degrees.
Despite continual tinkering, Osheroff learned, no one actually knew why foam fell off the tank.
NASA engineers had long speculated that nitrogen from the air seeped into the foam and condensed against the chill of the tank through a process called cryopumping. During the heat of ascent, they suspected, this super-chilled nitrogen boiled, expanded and burst the foam into pieces.
Osheroff decided to test the theory in the only way he knew.
He did an experiment.
The kitchen island at his home in Woodside, Calif., became his lab bench.
Instead of high-pressure vacuum pumps, sensors or cryogenic tanks, it held a stainless-steel sink, a KitchenAid dishwasher and a bird cage containing two brown African finches.
His research tools were simple: a bottle of red food coloring, a tire pump, a length of copper tubing, two tubes of epoxy, four squares of brass and an $80 regulator from a hobby shop airbrush.
Osheroff glued pieces of shuttle foam to the brass, chilled it and injected the dye under pressure with the tire pump. He wore goggles in case it exploded.
Almost immediately, Osheroff could see that the foam did not crack in the way NASA had assumed.
When he sawed it apart, he could see that the dye seeped into the foam without building up enough pressure to blow a piece off.
He repeated the test a dozen times just to be sure.
NASA had it wrong all those years. Its theory was "hokum," said Hal Gehman, head of the investigating board.
Like other investigators, Osheroff had witnessed a striking complacency in the federal agency responsible for sending humans into space.
The more he learned, the more upset he became.
"It seared my soul," he said.
He didn't know what caused the foam to break loose.
Neither did anyone at NASA. They still don't.
At an agency that rarely tolerated dissent, Scott Hubbard was caught between loyalty and his conscience.
He was the only NASA executive among the 13 members of the independent board investigating the accident.
The unfailingly affable physicist, a founder of astrobiology, was director of NASA's Ames Research Center in Mountain View, Calif. He was both an insider and an outsider, a space agency veteran stationed far from the nexus of human spaceflight.
As the Columbia inquest progressed, Hubbard, 54, clashed with shuttle operations engineers, demanding what he considered realistic tests of accident theories.
By the end of June, all the lines of investigation — wind tunnel tests, computer simulations, imagery analysis, wreckage studies and radar experiments — converged on the theory that a piece of foam insulation had knocked a hole in a reinforced carbon-carbon panel shielding the left wing.
For all the supporting evidence, the theory had one glaring flaw.
Many shuttle engineers and agency managers refused to believe it.
While Columbia was still in orbit, shuttle managers had silenced internal critics who warned of possible foam damage.
They dismissed the concerns for lack of proof, then ensured that no proof could become available. The managers rejected three requests that intelligence agency observatories and imaging satellites be used to inspect the orbiter.
Over the years, shuttle managers had treated each additional debris strike not as evidence of failure that required immediate correction, but as proof that the shuttle could safely survive impacts that violated its design specifications.
Three months before Columbia's last flight, foam debris had severely damaged the shuttle Atlantis. Program managers met to decide what corrective action they should take.
Rather than suspend all flights while engineers explored ways to fix the problem — as official procedure previously dictated — they decided Columbia should proceed without delay.
They refused to classify the incident as an official "in-flight anomaly." Instead, they adopted a formal "flight rationale" stating that it was safe to fly.
A generation of NASA managers had turned engineering on its head, viewing evidence of failure as signs of success.
For Columbia's last flight, all the paperwork was in order.
In the shade of the San Antonio mesquite, Hubbard took aim at a generation of denial with a high-speed cannon designed to hurl dead birds.
On a hot, hazy day in July, the 33-foot barrel of the nitrogen gas gun at the Southwest Research Institute was loaded with an oblong block of foam about the size of a fanny pack.
The bright blue gun had been built to test aircraft canopies by simulating bird strikes at supersonic speeds. Now, it was cocked and aimed at an $800,000 reinforced carbon-carbon panel from a space shuttle wing, wired with 200 sensors and 15 high-speed cameras.
A countdown echoed over loudspeakers.
Over NASA objections, Hubbard had arranged to demonstrate the actual damage — if any — that a light foam fragment moving at more than 500 mph could do to a real piece of the $1.8-billion space shuttle.
It was a $3.4-million experiment designed to demolish an attitude.
"I felt there was denial," Hubbard said.
The gas gun belched its block of foam at 524 mph.
The hurtling 1.67-pound piece of insulation ripped a hole the size of a space helmet in the underside of wing panel No. 8.
Watching from the shade of two white awnings, a crowd of onlookers gasped as if the violent impact had knocked the breath out of them as well.
Hubbard threw his arms up in an involuntary gesture of vindication and dismay.
Michael Bloomfield, a shuttle commander who had flown three missions in space, watched the test over Hubbard's right shoulder. A second astronaut stood at Hubbard's left.
"It almost appeared like a magic trick," Bloomfield said. "You never saw the foam. It was going so fast, you never saw it hit."
One man poked his head in the hole. There were three large pieces of debris lodged inside.
The size of the shards closely matched the radar signature of the puzzling object that had floated away from Columbia on its second day in orbit.
A technician measured the opening with a yellow tape. It was 16 inches by 17 inches — a death warrant for astronauts.