A more accurate way to track the sources of illegally trafficked radioactive materials

A team of nuclear physicists and engineers from Sun Yat-sen University and the China Academy of Engineering Physics, both in China, has developed a more accurate way to track the sources of illegally trafficked radioactive materials. In their paper published in the journal Physical Review Letters, the group describes their new method and its accuracy.

As more countries develop the technology to build nuclear reactors for producing electricity, or as a propulsion source for ships or submarines, the threat of illegally trafficked radioactive materials rises. Such material poses a threat as a component in dirty bombs. As part of the effort to slow or stop such trafficking, scientists have been developing ways to trace the origin of such materials as a means of policing those who have access to such technology and are willing to sell it to terrorist organizations.

Currently, it is difficult to identify spent fuel that came from boiling water reactors (BWRs) versus pressurized water reactors (PWRs), which makes it nearly impossible to trace a sample back to its source. In this new effort, the team in China has developed a process that greatly improves the ability to discern the difference.

To trace the source of a sample of spent fuel, forensic physicists look for clues to help identify the time it was inside of a reactor, the extent of its enrichment and finally, the reactor type. To measure and identify reactor type, the researchers studied the characteristics of spent fuel held in databases containing information collected over the past half-century.

They then developed linear equations that related the quantities to one another. Next, they applied their equations to isotopes found in the databases, using their results to tweak their equations. They found they were able to use some of the measurements of the materials to calculate other characteristics and identify the sources of materials originating from six types of nuclear reactors.

The research team then trained a deep-learning AI network to find the differences between spent fuel from BWRs versus PWRs using their equations and tested it on another dataset. They found it to be 91% accurate in identifying BWR sources and 95% accurate in identifying PWR sources.