NASA's top spaceflight official, Bill Readdy, on Friday said air or liquefied nitrogen almost certainly seeped into a crack or void in the foam insulation, or collected around bolts and nuts beneath the foam at the time of the launch.
The trapped air or nitrogen expanded as the shuttle rose, and blew off a chunk of foam the size of a suitcase.
"That is really the root cause that we have been able to discover here," Readdy said.
"In all likelihood, faulty application of the foam created air pockets," he said.
Rather than peeling off, as NASA had assumed from past experience, the foam was pushed off with explosive force.
The space agency also had assumed the foam would fall down along the tank and miss the shuttle, but in reality, the falling foam shot towards Columbia and the left wing rammed into it, resulting in a large fatal gash.
NASA said a new tank design and improved techniques for applying and double-checking the foam should solve the problem.
"In all likelihood, faulty application of the foam created air pockets"
Senior NASA official
Columbia had disintegrated over Texas during re-entry on 1 February last year.
At the time of the tragedy, NASA had run only 400 computer models involving flyaway foam and other launch debris.
Since then, millions of models have been run and the space agency has a much better understanding of what went wrong and how to fix it, Readdy said.
Readdy said future space shuttles like Discovery, expected to fly as early as March 2005, would have rescue ships ready.
The seven Discovery astronauts could also seek refuge at the international space station, where they are headed anyway, for as long as three months.
A rescue mission could be put together in as little as 35 days, Readdy said.