Columbia was brought down by a suitcase-size piece of foam insulation that broke away from an aerodynamically shaped ramp used to keep ice from forming on a strut connecting the shuttle's nose to its external fuel tank. The 1.67-pound piece of foam came off 81.7 seconds after liftoff and struck the underside of Columbia's left wing 0.2 seconds later, smashing into the lower side of a reinforced carbon carbon panel, one of 22 making up the wing's leading edge.

Ground cameras were unable to see the point of impact. One long-range tracker that might have shown the impact site was out of focus. And given the resolution of the cameras in place at the time, it's not clear obvious signs of damage would have been detected.

In any case, without the benefit of high-resolution video of the impact, mission managers were forced to rely on computer modeling and other indirect techniques for determining whether the foam strike could have caused any entry-critical damage. In the end, they wrongly concluded Columbia could safely re-enter as is.

Commander Rick Husband, pilot William McCool, Kalpana Chawla, Laurel Clark, David Brown, Michael Anderson and Israeli astronaut Ilan Ramon were killed Feb. 1, 2003, when hot gas burned its way into the interior of the left wing through a presumed breach on or near the underside of RCC panel No. 9. The wing melted and the shuttle broke up 37 miles above Texas.

In the wake of the disaster, the Columbia Accident Investigation Board ordered sweeping changes, including "an aggressive program to eliminate all external tank thermal protection system debris shedding at the source with particular emphasis on the region where the bipod struts attach to the external tank."

The bipod ramp was intended to keep ice from forming around the struts due to the ultra-low temperatures of the shuttle's liquid oxygen and hydrogen propellants. NASA's solution was to simply eliminate the ramps and to install 300-watt heaters on the strut attachment fittings to prevent ice buildups.

Tank engineers also implemented a variety of other changes to minimize foam shedding and the first upgraded tank was shipped to the Kennedy Space Center in the first week of January.

"Through all our testing, we believe the amount of foam that can come off the tank and not cause serious damage is .03 pounds," said Wayne Hale, deputy director of the shuttle program at the Johnson Space Center in Houston. "If you think about that, that is three one hundredths of a pound. That is something like six tenths of an ounce. So that's a pretty small piece of foam.

"All our investigations of the foam indicate we will not get a piece of foam coming off bigger than .008, eight one thousandths, or almost an order of magnitude smaller than the requirement."

It is not possible to eliminate all foam shedding, but "we're clearly moving toward an era where we expect to see much less damage in the tile and no critical damage that will require a repair," Hale said. "So that's our goal in this and it's beginning to look very positive that we'll be able to accomplish that level of control on the ET foam."

But to make absolutely sure, NASA must be able to inspect the tank and the space shuttle after launch with much greater precision than before. The CAIB recommended that NASA "upgrade the imaging system to be capable of providing a minimum of three useful views of the space shuttle from liftoff to at least solid rocket booster separation."

The panel also told NASA to look into putting cameras aboard ships and/or aircraft to provide additional coverage, to develop a capability to obtain high-resolution images of the tank after separation from the shuttle, to develop techniques for high-resolution imaging of the ship's underside and wing leading edges; and to make arrangements to obtain imagery from spy satellites if needed.

On its own, NASA managers decided to launch the first two post-Columbia shuttle flights in daylight to improve photo coverage and to time the launchings so the external tank, separating from the ship half a world away, also would be lighted by the sun.

To reach the space station, the shuttle must launch within a few minutes of when Earth's rotation carries the launch pad into the plane of the lab's orbit. And as a final complication, the shuttle can only visit the station when the "beta angle" - the angle between the plane of the station's orbit and the sun - ensures the shuttle-station stack will not get too hot.

Throwing all of those requirements together, NASA can only launch a shuttle to the station during relatively short windows.

NASA had hoped to launch Discovery in mid May, but the flight was delayed after a fueling test April 14 because of concern about ice buildups around a flexible liquid oxygen feedline bellows assembly. In addition, two of four hydrogen fuel sensors operated intermittently and a pressure relief valve in the hydrogen section of the tank cycled more often than expected.

At the same time, engineers were struggling to define the threat posed by ice during launch. While engineers had focused on foam impacts in the initial stages of the recovery program, it became clear earlier this year that ice posed a similar threat. Areas of concern centered on the feedline bellows and on so-called stand-off brackets that support the tank's externally mounted liquid oxygen line.

NASA managers ultimately ordered engineers to haul Discovery off the launch pad and back to the Vehicle Assembly Building for attachment to an external tank and boosters originally slated for the second post-Columbia mission. The new tank was equipped with a heater to minimize ice buildups around the feedline bellows.

Before the rollback, however, a second fueling test was conducted May 20, confirming the hydrogen vent valve's unusual behavior. Engineers believe the valve cycling was associated with a jet-like device called a diffuser that injects helium into the tank to help maintain the proper temperature and to pressurize the tank for flight. A new dual-screen diffuser was used in Discovery's original tank and managers decided to switch back to the original single-screen design for the new tank.

As for the hydrogen sensors, which operated normally the second time around, NASA managers decided June 6 to forego a third tanking test, saying they are confident the new tank will behave normally during Discovery's countdown.

And they ultimately concluded the threat posed by whatever ice does fall off the tank is an acceptable risk.

But unlike the threat posed by foam, agency managers refused to provide any numbers defining the smallest allowable ice or the odds of a catastrophic impact. They said the issue was too complex and that such numbers would be meaningless without a thorough understanding of the assumptions that went into the calculations.

"It's a really complex formulation, going from how much ice can you a have on the spacecraft and what velocities, at what vibration levels it shakes loose, how big are the pieces that shake loose and then how they transport through the aerodynamic environment," said John Muratore, the engineer in charge of the debris assessment.

"And then, what's the resistance of the tiles to ice impacts and finally, given an ice crater - which is different from a foam (insulation) crater - how well can we enter (the atmosphere) with regard to that?

"The uncertainties in each of those areas are significant," he said. "There are all sorts of numbers that are floating around. We have nine different estimates for ice on tile from one ice location. ... It is a very complex problem.

"So what we did was, we looked at the relative risks and we're convinced the ... the remaining risks due to ice is around an order of magnitude less than the ones that we fixed."

Muratore said ice represents more of a threat to the shuttle's tiles than it does to its reinforced carbon carbon leading edge panels and that even though engineers used "all the supercomputers at NASA to run it down ... it doesn't lend itself to a single number."

"Impact dynamics are a very complicated engineering discipline," he said.

While he would not discuss the underlying statistics, sources said some estimates of the likelihood of tile damage due to ice impacts - damage that would require repair or some other response - could be as high as 1-in-100 or so or as low as one in tens of thousands, depending on the assumptions that were made.

Engineers do not put much faith in either extreme because they do not have enough reliable test data to more precisely define the actual risk. And that won't change anytime soon. As one briefing chart put it, significant improvement in the understanding of the impact threat from ice and foam would require "another enormous undertaking."

The best way to improve that understanding in the near term, many engineers believe, is to collect actual flight data by launching Discovery as planned. They believe the studies carried out to date, plus NASA's experience in 113 previous shuttle missions, show the risk of a debris strike, what ever it might be, is acceptable.

"We believe it's an acceptable risk at this time," said shuttle program manager William Parsons.

PREVIEW REPORT PART 3 --->