Of all the stresses the space shuttle undergoes during reentry to Earth's atmosphere, there is one particularly wicked effect that will tilt the ship to one side and cause the undercarriage to heat up prematurely. It has happened on many occasions, but no one at NASA ever expected that it was capable of destroying a spacecraft.

In the last minutes before shuttle Columbia disintegrated over Texas one week ago, NASA instruments recorded excessive "drag" on the shuttle's left wing, and hot spots in several areas of the wing's underside. At first, the onboard computer ordered the spacecraft's wing-mounted elevons to compensate; then it fired the jet thrusters when that wasn't enough.

There is no proof that the loss of Columbia moments later resulted from the phenomenon scientists describe as "asymmetrical boundary layer transition." But the symptoms observed in Columbia are similar to those that occurred in perhaps a dozen shuttle flights over the life of the 20-year program.

In each case, engineers determined that the premature drag occurred because launch damage or faulty maintenance had roughed up the surface of the thermal tiles that cloak the underside of the shuttle. This interrupted the air flow, pulling the wing down, increasing friction and causing the tiles to heat up more rapidly.

NASA had studied the phenomenon for years, but its fear was not that the shuttle would suffer a catastrophic structural failure because of it. "That wasn't even an issue," said aerospace scientist Steven Schneider, of Purdue University, who consulted with NASA on the boundary transition.

Instead, Schneider said, NASA was worried about firing the thrusters so often that there wouldn't be enough fuel to use them to steady the spacecraft on its final approach to landing. The shuttle's "yaw" -- its movement from right to left -- can be fine-tuned by the same small, 40-pound jets that correct the drag at much higher speeds and altitudes.

In studies published by the American Institute of Aeronautics and Astronautics in 1997 and 2002, scientists analyzed more than 80 shuttle flights. They found generally that the boundary transition under ordinary circumstances took place at an altitude of 156,000 feet, with the spacecraft flying at a speed of about 5,000 miles per hour.

But the 1997 study also determined that the transition "tripped" -- occurred prematurely -- in 20 percent of the flights, and in 60 percent -- about a dozen flights -- the transition was "asymmetrical," in that one wing of the spacecraft encountered turbulence before the other, dragging down the afflicted side.

In 1989, Columbia had an asymmetric transition at an altitude of 220,000 feet, while traveling at a speed of 12,500 mph. Columbia was in roughly the same configuration Feb. 1 when Mission Control lost communications.

"Early tripping is not a good idea," said Jerry Grey, director of science and technology policy for the American Institute of Aeronautics and Astronautics. "The consequence of early tripping is an early heat transfer. The spacecraft stays hotter for longer."

The boundary layer transition is a well-known concept in fluid mechanics, easily understood by looking at the water in any flowing stream. Close to the bank the water is virtually still, while it flows most rapidly in the middle. The increase in speed from riverbank to channel is the boundary layer transition.

"That's because of the friction between the edge water and the bank itself," said five-time astronaut Jeffrey A. Hoffman, who teaches astronautics at the Massachusetts Institute of Technology. "The flow can be smooth, or it can become turbulent."

When a returning space shuttle begins to reenter the Earth's atmosphere, the flow of air over the wing is smooth, even though the transition is from zero mph on the surface of the wing to 17,000 mph less than an inch above it.

As reentry continues, however, the combination of increasing friction and atmospheric density will cause the flow to become turbulent, as if Hoffman's babbling brook had become choked by floodwaters.

"A turbulent boundary will cause the wing surface to get hotter," said Grey, but it is not something to worry about, under normal circumstances. "It happens all the time," Grey added, and "you can't do anything about it."

Hoffman explained further that the transition is not something that the shuttle crew watches closely during the 60-minute reentry. Like Mission Control, the astronauts are preoccupied with their upcoming touch-down and whether their speed and altitude correspond to NASA's templates.

"At Mach 8 [about 5,000 miles per hour] you're only about 300 miles from landing," Hoffman said. "You're at 150,000 feet and you have 1.5 Gs [11/2 times the force of gravity] pushing down on you. You're looking at your numbers on the displays and at the numbers you're supposed to have." In five flights on four shuttle orbiters, Hoffman said he never noticed when the transition occurred.

The ideal is for both sides of the shuttle to make the transition to turbulence at the same time. The asymmetry occurs when an anomaly on one wing triggers the effect, the way boulders will create rapids in an otherwise calm stream.

When this occurs, the onboard computer corrects with the elevons, and if that doesn't work, it begins firing the thrusters. "This is what must have happened to Columbia," Grey said. "The asymmetry was severe enough to cause the steering jets to come on." Even so, Grey expressed doubt that an asymmetric transition alone could have caused the Columbia to disintegrate in midair.

It was an opinion shared by NASA. When Schneider, a specialist in transition, was invited to attend a meeting at the Johnson Space Center in 1995, he found that NASA had already concluded that excess heat produced during premature transition was incapable of destroying an orbiter.

Instead, NASA focused on saving the thrusters for landing. Early transition, engineers determined, was caused either by launch debris damaging thermal tiles, or by flaws in the gap fillers, a material similar to caulking that is placed between tiles to seal out heat.

The gap fillers sometimes broke and ended up protruding from the thermal shield like a finger -- an instant source of turbulence, Schneider said. The solution, which NASA implemented, was to develop fillers with bases that extended underneath the tiles themselves, so the tiles would hold them in place.