Outstanding Research on Verifying Plutonium Production in Nuclear Reactors

How can artificial intelli­gence (AI) be used to verify compliance with nuclear arms control agree­ments? Fabian Unruh investi­gated the potential of invertible neural networks (INNs) for re­constructing plutonium pro­duction in nuclear reactors. For this work, he was awar­ded first prize in the J.D. Williams Stu­dent Paper Competition at the 67^th Annual Meeting of the Insti­tute of Nuclear Materials Manage­ment (INMM) in Austin, Texas.

Invertible neural networks (INNs) are a special form of AI: they utilize a mathe­matical structure in which every calculation can be re­versed. This makes it possible to train INNs to solve a specific class of mathe­matical problems. In his paper, Fabian Unruh exa­mines how this approach can be applied to the ana­lysis of highly radio­active waste generated during plutonium production in nuclear reactors. According­ly, INNs could contribute to recon­structing past plutonium production in nuclear reactors – and thus to verifying compliance with nuclear arms control and disarmament agreements.

For this work, the Doctoral Re­searcher in the Cluster for Natural and Technical Science Arms Control Research (CNTR) was awarded the J.D. Williams Best Student Paper Award by the Institute of Nuclear Materials Management (INMM).

With the award, the profes­sional association – which inter­nationally pro­motes the responsible use of nuclear and radio­active materials – annually recog­nizes outstanding student research achieve­ments in the fields of nuclear material manage­ment, non­proliferation, and moni­toring.

The award recog­nizes both the best submitted paper and the contes­tant’s presen­tation at the Annual Meeting. The award is named after former INMM Presi­dent James D. Williams.