UC Berkeley

Since the height of the Cold War, arms reduction treaties have reduced the number of nuclear weapons from an estimated 70,000 to around 14,000. Future treaties may continue to lower those numbers, address non-deployed or inactive warheads, or address weapon dismantlement procedures. This raises the unique challenge of confirming that items themselves are warheads rather than using a proxy for warheads, such as an entire delivery system. There is a fundamental tradeoff between the confidence a warhead confirmation measurement provides and how intrusive it is. An inspector must have confidence in the accuracy of a declaration of treaty items. If a host’s need to limit sensitive information creates measurement requirements that reduce an inspector’s confidence in the equipment, then their overall confidence may be worse. We are looking for a solution that enables an “instrusive” measurement while meeting a host’s requirements, thus increasing the maximum achievable confidence.

Recent work applied Zero-Knowledge Proof (ZKP) concepts from cryptography to warhead verification as a way to confirm declarations without revealing sensistive information. We developed a verification system using those principles and advanced concepts in radiation detection and imaging, most importantly an advanced rotating mask for Time-Encoded Imaging (TEI) and the availability of stilbene detectors for Pulse Shape Discrimination (PSD). A proof-of-concept fast neutron Time-Encoded Imaging (TEI) system with a cylindrical anti-symmetric mask was built. It demonstrated the ability to distinguish between matching and non-matching items by measuring Cf-252 sources and PuO2 hemispheres in several configurations. Simulations were then used to examine the design space of the system shape and size. New metrics and methodology were developed to better evaluate the performance based on the specific tasks of confirming matching objects and distinguishing between different objects. These showed that a thick spherical mask with a large detector performed the best, but changing the shape of the detector or the shape of the mask elements does not improve performance. A second system was built with a mask design based on the outcome of the simulations. As a spherical mask would have been extremely challenging to manufacture by conventional methods, the mask was 3D-printed using polycarbonate, which was deemed the most suitable 3D-printing material. Measurements with Cf-252 sources were used to evaluate its performance and compare with simulations and the first system. It performed well at confirming matching sources and was able to distinguish between sources with more than a 5 cm difference in diameter. The combination of the initial proof of concept measurements, parametric simulation optimizations, and confirming measurements using the second generation system prove the feasibility of this concept as a verification method and its usefulness to future development and deployment.