Reducing neutron multiplicity counting bias for plutonium warhead authentication

Confidence in future nuclear arms control agreements could be enhanced by direct verification of warheads. It would include warhead authentication. This is the assessment based on measurements whether a declaration that a specific item is a nuclear warhead is true. An information barrier can be used to protect sensitive information during measurements. It could for example show whether attributes such as a fissile mass exceeding a threshold are met without indicating detailed measurement results. Neutron multiplicity measurements would be able to assess a plutonium fissile mass attribute if it were possible to show that their bias is low. Plutonium measurements have been conducted with the He-3 based Passive ScrapMultiplicity Counter. The measurement data has been used as a reference to test the capacity of the Monte Carlo code MCNPX-PoliMi to simulate neutron multiplicity measurements. The simulation results with their uncertainties are in agreement with the experimental results. It is essential to use cross-sections which include neutron scattering with the detector’s polyethylene molecular structure. Further MCNPX-PoliMi simulations have been conducted in order to study bias that occurs when measuring samples with large plutonium masses such as warheads. Simulation results of solid and hollow metal spheres up to 6000 g show that the masses are under predicted by as much as 20%. The main source of this bias has been identified in the false assumption that the neutron multiplication does not depend on the position where a spontaneous fission event occurred. The multiplication refers to the total number of neutrons leaking a sample after a primary spontaneous fission event, taking induced fission into consideration. The correction of the analysis has been derived and implemented in a MATLAB code. It depends on four geometry-dependent correction coefficients. When the sample configuration is fully known, these can be exactly determined and remove this type of bias. As the exact configuration of warheads includes sensitive information which must be protected, it will not be fully declared and the coefficients must be estimated. Assuming that the plutonium component of warheads can be approximated by hollow spheres, it is shown that the coefficients could be approximated if the thickness of the hollow sphere were known. Other parameters such as the radius have a very small influence: Only knowing the thickness introduces a negligible bias of 1% for the fissile mass. If the thickness remained unknown, a less accurate estimate of the correction coefficients could also be estimated from the multiplication which is directly measured by neutron multiplicity counting. Bias from this approach is smaller than using the analysis based on the false assumption. In order to study the influence of further materials in the warhead and the container the warhead would be placed in, correction coefficients were also studied for plutonium spheres surrounded by polyethylene. The result of the polyethylene is that the correction coefficients are somewhat smaller. Applying the correction coefficients that do not take further materials into account to these cases, the fissile masses are overestimated by around 7% in exemplary simulations. Bias would be further reduced if information on further materials could be included in the correction coefficients’ assessment. By having developed a method to reduce bias, the reliability of neutron multiplicity counting measurements has been increased. The obtained level of reliability may suffice in the context of additional verification measures as these also add to the overall confidence