The search for the cause of the Challenger disaster has been narrowed to what for weeks has been the focus of most popular accounts of the investigation -- defects in the seal between two segments of the solid-rocket booster.

The disclosure came at a hearing of the presidential commission investigating the accident. It was the commission's most technical public hearing to date, centering on details of the "failure scenario" that led from a puff of dark smoke from the joint in the first second of the flight through a mysterious healing of the leak to the reopening of the rupture which eventually triggered the final conflagration on Jan. 28.

Although it is still unclear how the leak started in the right-hand booster, the new details confirm that the investigation has evolved, rather quickly observers say, from early bewilderment to a National Aeronautics and Space Administration study, now under way, of new designs for sealing booster joints.

Yesterday's hearing had little of the emotional tension that marked some earlier sessions and even the news corps in attendance had dwindled somewhat.

The only major shuttle component that had not yet been ruled out formally was the huge external tank. If it had leaked, a jet of burning hydrogen gas could have melted through the booster. New experiments, however, now show that even the smallest leak of the supercold gas would could have been seen before liftoff as a cloud resembling a steam jet. Newly released photographs taken just before liftoff showed no such cloud.

As the investigation has closed in on the joints, engineers at NASA and Morton Thiokol Inc., which makes the boosters, have been trying produce a burn-through in laboratory models of booster joints that have been damaged deliberately in various ways.

Wayne Littles, associate director of engineering at the Marshall Space Flight Center in Huntsville, Ala., told the commission that the experiments, though they produce puffs of smoke, leaks that heal and eventual burn-throughs, are inconclusive. The reason is that there is not much correlation between the type of damage and the result.

In one case a gap of one-eighth of an inch was cut out of an O-ring before the joint was closed and the rocket engine fired. There was no visible leakage. In another test an O-ring was shaved slightly to reduce its diameter just three-sixteenths of an inch and the joint burned through in 12 seconds.

One test produced a result remarkably like that of Challenger's booster. The experimental booster produced an initial puff of smoke, then stopped leaking for a few seconds only to burn through later. The joint had merely been scratched on the metal surface where the O-ring touched it.

Some of the tests showed that the putty, which lies between the O-rings and the burning rocket fuel, can hold back the pressure of the hot gases for a few seconds. This finding has led NASA to conclude that its original theory of how the joints work was wrong.

NASA originally thought that the O-rings were pushed into sealing position in the first second after ignition because the putty transferred the pressure of the expanding hot gases to the O-rings. As a result, the rings would already be in sealing position when, a fraction of a second later, the joint is warped slightly as the booster's steel casing bulges from the same pressure. This warping can cause an unseated O-ring to move out of sealing position.

The new finding on the test joints shows that the putty can keep the pressure away from the O-rings until after the joint warps. Thus they are unable to seal and the hot gases can burst easily through the putty and out the unsealed joint.

"That concept of the joint operation which people had believed was correct, is not in fact correct," Littles said.

If the putty in Challenger's booster behaved the way it does in the new tests, the O-rings may have been prevented from going into sealing position before the joint warping made it impossible. This would have allowed burning putty, O-rings and grease to escape as dark smoke in the first second after ignition.

Investigators, however, are still puzzling over how the leak could have sealed itself, as it appears to have done about three seconds later. One possible explanation has emerged from the lab tests. When the experimental joints were opened, technicians found deposits of aluminum oxide, a product of burning rocket fuel, coating the joint walls.

The tests indicate that the tough, brittle deposits can eventually clog a leak, sealing it even though high pressure gases are rushing out. "It seems an unlikely result, but we did get it," Littles said.

If this is what happened to Challenger's booster, the clog apparently broke loose at about 59 seconds into the flight when a plume of white hot gases burst through the booster wall. This happened when the shuttle was experiencing rough vibrations as it passed through a normal period of atmospheric turbulence. Just before this period the craft had also been buffeted by a wind shear. The mechanical forces may have broken the deposits loose.

Once the leak opened again, flames quickly melted a larger hole through the booster wall. According to one calculation reported yesterday, the hole had enlarged to an area of about 45 square inches by about one second before the final explosion.

That is when the booster ripped loose from its lower attachment to the external tank and pivoted, swinging its tail toward the orbiter's right wing.

As it broke loose, it ruptured the liquid hydrogen tank, causing flames to flicker around the tank's base. Almost simultaneously the booster's nose crashed into the liquid oxygen tank, mounted above the hydrogen tank, rupturing it. Spilled fuel and oxidizer ignited, causing the fireball.