By Paul Rincon
BBC News science reporter, in Washington DC
US researchers plan to use energetic particles that shower Earth from space
to detect smuggled nuclear material held in vehicles and cargo containers.
Muons could easily detect shielded plutonium
These cosmic ray muon particles strike Earth at the rate of 10,000 per square metre every minute.
By tracking the muons, the scientists can see through lead, steel and other heavy shielding that might be used to mask a radioactive source.
The Los Alamos National Laboratory team discussed the plan in Washington DC.
The researchers were attending the annual meeting of the American Association for the Advancement of Science (AAAS).
When cosmic rays hit the upper atmosphere, they produce muons, which are charged particles similar to electrons.
Their electric charges make them very easy to detect and they can penetrate heavy metal and thick rock. Indeed, with an average energy of three billion electron volts, most muons can penetrate about 1.8m of lead.
The researchers say the X-ray and gamma-ray detectors currently used at US borders are inefficient for detecting nuclear materials shielded with lead or steel.
Muons are also harmless, unlike X-rays or gamma-rays.
The prototype detector built by the team records each muon's path before and after it passes through a cargo and then analyses changes in the particles' energy and trajectory. This can be used to build up a three-dimensional map of dense items inside.
"The change in angle tells you what the material on the path of a muon was," explained Chris Morris of Los Alamos' division of physics.
"The scattering of muons is very sensitive to the density and atomic number of a material. It could therefore easily detect uranium, plutonium or the shielding material that would have to surround them to make these materials undetectable by other methods."
Unlike airport baggage screeners, which require people to interpret images and data, the muon detector can be trained with known examples until it can directly decide whether a cargo contains nuclear materials, such as a bomb, or shielding.
"We've shown we can put the data through a machine-learning algorithm and train the system to spot objects of interest with a rate of false positives and false negatives that is less than 3%," said David Chartrand, also of Los Alamos.
"We think we can continue to improve that."
The researchers have applied for funding to the US Department of Homeland Security through an industrial partner.