Note: this WU is being continually updated as more evidence comes in.

Update, August 27, 2003The final CAIB report is now out. I am slowly digesting it, and will post relevant information here soon. You can download the whole thing (10 MB) in PDF format here. The text below deals mainly with the contents of chapter 3 of that report (download the 1.6 MB PDF file here): the direct, technical causes of the accident. The WU does not deal with any management issues at NASA, nor how the accident might have been avoided or the crew saved.

A picture is emerging (February 28, 2003) as to the likely sequence of events which may have led to the destruction of the Columbia. At present, most of this is a working theory, rather than confirmed proof, but nevertheless, it is probably close. Two months later (April 17, 2003), this is increasingly seen as the best description of what happened to Columbia.

Update, April 17, 2003 In broad terms, the immediate cause of the disaster appears to have been a piece of foam which burst loose from the external fuel tank at 81 seconds afer launch. This hit the leading edge of the left wing. A long, thin, U-shaped seal, or one of the bolts holding it in place, was damaged in the impact. During the mission the shuttle performed a series of manouvres. During one of these, on day 2, radar picked up a piece of material floating away from the shuttle. Although the radar images were not examined until after the disaster, this was almost certainly a carbon fibre seal from the leading edge of the left hand wing. It left a slit-like gap in the leading edge. The seal was almost certainly less strong than when originally fitted due to pitting and ageing.

During the descent phase on Feb 1, sensors within the wing started recording unusual temperatures almost as soon as the Columbia hit the earth's atmosphere. The area behind the loose seal jumped by 350 degrees F (200C) within two seconds. It is almost certain that a plume of air, heated by re-entry friction to thousands of degrees, entered the wing through the damaged leading edge and worked its way through the structure, melting everything in its path. The plume eventually emerged near the landing bay doors. As time progressed, the plume melted the aluminium structure of the craft, leading to the loss of control, and eventually the loss of the shuttle and all crew.

What led up to this disaster?

Right at the start is the manufacture of the external fuel tank, months ago. Apparently, when the silicone-based ablative foam (see below) was being applied in the bipod area—where the tank structure is attached to the top of the shuttle—the epoxy adhesive used has a 90 minute pot life. This means that if the pot is open for more than 90 minutes, then the adhesive may not work as well as intended. Operatives at the Lockheed Martin plant in Michoud, New Orleans report that the time of opening is normally written on a sticky label, which is applied to the open pot. It seems, however, that the timing may have got confused, resulting in the operatives applying the adhesive after its pot life has expired. Also, there are reports that in some cases, those applying the adhesive covered too large an area, and some of the ablative foam was not stuck until after the 90 minute lifetime of the glue.

If there was poor adhesion, then small bubbles of air would remain between the surface of the tank and the foam insulation. When the external tank was filled with cryogenic fuel, these small pockets of air would have liquefied, creating a local low pressure. More air would flow into the region to balance the pressure, and that, too would liquefy.

Next is the history of the tank. it was, apparently, fitted to a shuttle, then removed, and then fitted once more. There is concern that these procedures could have resulted in damage to the foam around the bipod area.

Then we come to the launch on January 16. After 60 seconds of flight, the shuttle encountered a particularly violent wind shear, which may have led to excessive stress on the airframe or the wing tiles. Approximately 20 seconds after that, one or more large pieces of white material broke loose from the bipod area of the tank and collided with the shuttle's left wing. It is likely that any liquid air trapped under the foam surface would remain liquid until the cryogenic fuel drained from the internal tanks, when it would warm up, and expand rapidly, blowing off the poorly-attached foam. This process is known as 'popcorning' or 'cryopumping'.

Update, March 26, 2003 New film footage, subject to computer analysis and enhancement indicates that a lump of foam did hit the leading edge of the left hand wing about 81 seconds after launch. Three of 22 reinforced carbon carbon panels were hit, as were two connecting carrier panels and adjacent underside tiles.

Film from on-board cameras and ground based cameras definitely show two large lumps of material, and possibly three, hitting the wing, with impacts located on the leading edge and somewhere on the underside. The material is seen to shatter on impact. Investigators are looking for evidence that the shattered material may have contained tile debris, as well as pieces of foam or ice. It appears likely that at least one piece of the ablative foam, carrying some polyurethane insulation and possibly some ice, hit the left-hand landing gear door at 81 seconds after launch, at a closing speed of around 500 mph (230 m/s). The debris is described as being the size of a briefcase. At one stage it was thought the debris was just polyurethane foam, and the weight was estimated at 1.2 kg (2.7 lb). However, the ablative foam is up to 10 times denser than the polyurethane insulation, so the weight may have been 5 kg or more. This almost certainly led to significant damage to the door and the heat-resistant tiles surrounding it.

During the mission, a series of events occur, but none appears to have much impact on the ultimate fate of the mission: a piece of material is seen detaching itself from the shuttle and subsequently burning up over the Pacific Ocean. By April 1, 2003investigators were saying this was "almost certainly" a so-called carrier panel from the underside of the wing's leading edge. They added that if one of these panels was missing during re-entry, then— since the tiles are stuck to it—the hot gases would certainly have swept into the wing, eventually destroying it. Meanwhile on the ground, engineers exhange a series of e-mails about possible damage to the tiles and landing gear doors and the tyres. None of these e-mails reaches management levels.

Moving forward to the descent phase on February 1, the laminar aerodynamic flow which is established early in re-entry gives way to a turbulent aerodynamic flow pattern much earlier than normal, leading to unusually rapid temperature build-up on the exterior of the shuttle. Normally the ceramic tiles covering the outside of the shuttle craft protect the fragile inner structure from this heat.

At some point, a quantity of super-heated air, or possibly plasma, enters the leading edge of the left hand wing, snakes through the wing structure, and exits through the landing gear bay. The investigators are speculating that some of the tiles on the leading edge had become porous with age and repeated temperature cycling, allowing a small breach to occur. It is not clear if the air is a small jet which then leads to further damage, such as bursting one or both of the tyres on the landing gear, or if there is a already a large breach in the wing. Either way, the breach becomes a hole around 30cm or more in size. As more superheated air and/or plasma enters this hole, first the sensors go off-line and eventually the structure of the craft is melted, leading to the loss of the shuttle.

March 27, 2003 Investigators have recovered a flight data recorder. Although the recorder is not as robust as those found in commercial airliners, it appears to have survived the incident almost intact. The recorder has valid data up to 18 seconds past 0900. All previous data stopped at 4 seconds past 0900, when communication between the shuttle and the ground stations was broken. The shuttle is recorded as breaking up at 21 seconds past 0900.

March 30, 2003 The data on the recorder shows that the left wing was experiencing unusual temperatures almost as soon as the Columbia entered the atmosphere. Just 16 seconds after the craft experienced maximum re-entry temperatures, sensors near the left wing's leading edge recorded high temperatures. This is over a minute before anyone on the craft or the ground was aware of a problem.

Investigators hope that the data recorded on magnetic tape will reveal the size of the post-launch impact on the wing, which will enable them to get a better idea of the forces involved. Also, that it contains more data from temperature sensors in the wing and landing gear areas.

Separately, NASA has also revealed that the tiles on the leading edge deteriorate with time, and that all but two of the 44 tiles (22 on each side) on the leading edge were the original tiles, fitted when Columbia was built in the late 1970s. Thus, they were almost certainly less capable of withstanding an impact than new tiles. One theory is that a zinc-based primer paint on the launch pad superstructure damaged the tiles. As rain fell on the structure, runs the theory, some of the zinc would have leached into the water, and some of this would have found its way onto the shuttle's tiles. This zinc compound would have reacted with the carbon tiles, leading to microscopic holes. over time, such damage wouod have weakened the tiles significantly. Columbia spent a total of 2.5 years on the launchpad.


On the day of the launch, NASA cameras show that a piece of white material measuring approximately 500mm x 400mm x 150mm fell away from the 50m-long tank about a minute into the flight. It is not clear if the material was foam insulation or ice, or a combination of the two. If it was foam, then this represents a weight of around 1.5 kg (3lb). However, if the material was ice, it would have weighed over 10 times as much. The film shows the material hitting the underside of the shuttle's left wing 80 seconds after the launch of Columbia and shattering after the impact. At this stage in the flight, the shuttle was travelling at between Mach 2 and Mach 4. NASA indicated that the material probably hit the shuttle's wing at a speed of around 500 mph (230 m/s).

Two weeks, later, on Feb 1, Astronomers in California report seeing fragments breaking off the shuttle as it entered the earth's atmosphere, while military film appears to show irregularities in the airflow over the left wing, with a trail of some kind visible aft of the irregularity.

Meanwhile, flight data around the same time indicate a rapid rise in temperature in the left wing a few minutes prior to the loss of the shuttle, followed by all 11 sensors in that area going offline, and then that the autopilot was struggling to cope with excess drag on the left side of the craft.

Investigators have recovered a number of pieces of the left wing from around the landing bay door area. These objects were recovered from an area further west (and hence earlier in the flightpath) than most of the other debris. Some of the tiles are seen to have unusual damage, and others have strange orange markings the tiles from the wing and also pieces of the landing bay. The insulation foam on the external fuel tank is coloured orange.

Five of the six shuttle tyres—made by Michelin at around $100 000 each—have also been found. This is the first time any item made from rubber has been found after a spacecraft broke up on re-entry. The two from the left hand, rear landing gear are much more severely damaged than their equivalents on the right side, and show signs that they may have exploded at a different time and in a different manner from either the forward tyres or those on the right hand side. It is not kown whether this damage, or other evidence of burning seen on the tyres, were incurred before the shuttle broke up, or during the tyre's descent to earth.

Ground-based pressure sensors detected a bang, which is consistent with a rapid decompression of air, at about the time the shuttle broke apart. The numbers appear to show that the sudden decompression is compatible with the pressurised crew compartment rapidly venting to atmosphere.

After analysing the last two seconds of data before the ground crew lost contact with the shuttle, it appears that the crew had realised there was a major problem, and may have tried to over-ride the auto-pilot. Other evidence indicates that the shuttle was out of control, and that the crew were almost certainly aware of the fact. The craft's yaw sensors were off the scale, indicating a yaw rate of at least 20 degrees per second. Down on the ground, meanwhile STS-107 Flight director Leroy Cain apparently ended all attempts at landing the craft around 7 minutes before radio contact was lost. His focus was switched to saving as much flight data as possible in a bid to help the investigators of any subsequent crash.

Glossary and definitions

Ablative foam

This is both badly named and appears to have little function. An ablative layer is one designed to burn away when subjected to heat. In this case, although the insulation is called an ablative layer, it is never exposed to that kind of environment. it is a thin layer of soft, squishy foam which is glued to the main fuel tank wall.

External fuel tank

The external fuel tank is the large, brown-coloured pod slung beneath the shuttle at take-off. It contains liquid oxygen and liquid hydrogen at cryogenic temperatures which are fed to the shuttle's main engines during the shuttle's ascent. During manufacture, the outer surface of the tank is sprayed with a 1-inch (25mm) thick layer of semi-rigid polyurethane foam. The foam is applied in different layers and the precise composition varies according to the position on the tank. NASA has said up to six companies supply foam to Lockheed-Martin, which manufactures the 50m long by 8m diameter tanks.

Previous incidents show that insulating foam can have a significant impact on the heat-resistant tiles. The tiles, made from a ceramic, are designed to dissipate the heat generated by air friction as the shuttle slows from mach 25 to subsonic speeds high in the atmosphere. They have astonishing heat performance, but are mechanically weak. They will break even under the pressure of one hand.

An investigation revealed that the foam came loose as a result of 'popcorning' when the bubbles in the foam expand and then explode as the surrounding pressure drops. The proposed solution was to drill holes into the foam in order to allow air to escape quickly through the foam. It is thought that the tank on Columbia was using the newer foaming agent. It is not known whether the insulation was drilled prior to launch.

Columbia Accident Investigation Board (


  • Admiral Hal Gehman , US Navy

Board Members:

  • Rear Admiral Stephen Turcotte , Commander, Naval Safety Center
  • Maj. General John Barry , Director, Plans and Programs, Headquarters Air Force Materiel Command
  • Maj. General Kenneth W. Hess , Commander, Air Force Safety Center
  • Dr. James N. Hallock , Chief, Aviation Safety Division, Department of Transportation, Volpe Center
  • Mr. Steven B. Wallace , Director of Accident Investigation, Federal Aviation Administration
  • Brig. General Duane Deal , Commander, 21st Space Wing, USAF
  • Mr. Scott Hubbard , Director, NASA Ames Research Center Mr. Roger E. Tetrault , Retired Chairman, McDermott International, Inc.
  • Dr. Sheila Widnall , Professor of Aeronautics and Astronautics and Engineering Systems, MIT
  • Dr. Douglas D. Osheroff , Professor of Physics and Applied Physics, Stanford University
  • Dr. Sally Ride , Professor of Space Science, University of California at San Diego
  • Dr. John Logsdon , Director of the Space Policy Institute, George Washington University