Forensic Measurements for Nuclear ArchaeologyTrawsfynydd revisited

For the success of any lasting efforts towards nuclear disarmament, reliable and widely accepted verification measures are indispensable. Different aspects of the disarmament process pose their own specific challenges and require their own appropriate verification methods. One such aspect is the accountancy and control of fissile material stockpiles. Suitable verification methods for this context can be provided by the toolbox of ‘nuclear archaeology’- a set of methods focused on reconstructing the production and removal history of fissile material. A core method is the reconstruction of a nuclear reactor’s past plutonium production based on isotopic measurements of its structural components. Neutron radiation during reactor operation leaves behind traces through interactions with these components. Measurements of suitable isotopic ratios thus allow the deduction of neutron fluence and the derivation of produced plutonium. In the 1990s a joint U.S.U.K. effort has carried out a practical trial of the methodology using samples taken from the core of the U.K. Trawsfynydd reactor. Their analysis was able to successfully deduct total plutonium production based on titanium isotope ratios measured in the reactor’s moderator graphite. In this contribution we will revisit the example case of the Trawsfynydd reactor, but apply a new methodological approach to the analysis. This approach aims to strengthen the potential of nuclear archaeology by analyzing several isotopic ratios at once, using them to break down past reactor operations in more detail. As a first step, a computer model of the reactor is created, allowing the simulation of the reactor’s neutronic behavior as well as of nuclear interactions leading to changes in the reactor materials’ isotopic compositions. We then perform such simulations based on varied hypothetical operational histories, obtaining a set of isotopic composition data for our analysis. In a second step, we performasensitivity analysis on our data set to identify isotopic ratios containing information on relevant operational reactor parameters, e.g. operating period or power level. In addition to sensitivity, we also consider practical limitations, such as minimal measurable quantities or potential contamination. While modern reactor simulation codes provide reliable and accurate results, the simulations remain mathematically complex, and as such impossible to simply inverse- obtaining operational parameters from isotopic ratios. We therefore employ the method of Bayesian inference to numerically solve the inverse problem. Since the numerical solution requires many forward predictions, and reactor simulations are computationally expensive, as an intermediate step we construct surrogate models for the relevant isotopic ratio predictions using Gaussian Process Regression. This contribution will build on our previous contribution to INMM in 2022, further fleshing out the presented methodology and illustrating its potential by comparison with the practical trial.