The investigation into the Columbia disaster is focusing on a huge reverse-engineering analysis. NASA staff are working to pinpoint where and when each of the 12,000 burnt and twisted pieces of debris landed. The trajectories of these pieces, which are scattered over some 70,000 square kilometres of Texas and Louisiana, will then be plotted to create a virtual video of the craft breaking up.

Known as ballistic trajectory analysis, the technique is among the panoply of forensics being used in the crash investigation by NASA and an independent inquiry board, chaired by retired US Navy admiral Harold Gehman. The investigators will map the coordinates of each piece of wreckage using the Global Positioning System, and estimate its trajectory with the help of data on wind speed and the weight and shape of each part. This should shed light on the sequence in which key parts of Columbia broke off — those that fell first are most likely to be related to the cause of the crash.

Such techniques are used to analyse airline crashes. But the high altitude and speed — 60 kilometres and more than 20,000 kilometres per hour — at which Columbia broke up complicate matters. The debris is strewn over a much wider area than the few square miles typical in an aircraft crash, says William Waldock, a flight-crash analyst at Embry-Riddle Aeronautical University in Prescott, Arizona. Little accurate data on wind speed are available for altitudes above 15 kilometres, he adds.

Useful information may also be extracted from individual pieces of debris by forensic scientists at Barksdale Air Force Base in Louisiana, where all of the debris is being assembled. Electron-microscope images of the grain structure of metal debris will reveal tell-tale signs of the heating that each part experienced, says Arvid Pasto, director of the High Temperature Materials Laboratory at Oak Ridge National Laboratory in Tennessee. Knowing the shape and conductivity of the shuttle's components will also allow the spreading of heat throughout the craft to be modelled. And if pieces are missing, they probably either blew apart or were vaporized, again giving clues as to where fires broke out.

NASA is likely to send debris to Oak Ridge for such tests, says Pasto. Parts that experienced the most heating were probably situated close to where the malfunction originated. But the ultimate goal, he says, is to simulate the heat to which all of the individual pieces were exposed and then reconstruct the entire break-up with the aid of a supercomputer.

NASA staff will check their break-up models with radar and satellite images of the disintegrating shuttle and the wealth of photos and videos taken by the public and astronomers. They will also make use of data from the shuttle's autopilot, which took corrective action to counter a drag on the left wing minutes before the craft broke up. Models of the shuttle's aerodynamics could shed light on what caused the autopilot to take the action that it did.

Little other concrete information about the moments before the accident has emerged so far. Just before contact with the craft was lost, mission control detected temperatures rising by around 2 °C per minute in the well that contains the wheel on the left wing. Temperatures across the wing and fuselage then rose, and sensors of pressure, temperature and other parameters failed.

The investigation is still in its evidence-gathering phase, says Waldock, and firm conclusions are weeks, if not months, away.