A statistical method for attributing plutonium samples to a reactor type from isotopic data

Attributing plutonium to a specific country or organization is a complex and challenging problem that has great interest in the nuclear forensics and counterproliferation communities. Since plutonium is made in a nuclear reactor, identifying the type of reactor, material age, and other physical or chemical characteristics can help in identifying the material origin. For precise attribution, samples or calculations would be needed that represent all operating conditions, for all reactor types, and all possible fuel variations to create a databased to be queried to identify a reactor of origin from sampled material. A database this complex is currently unachievable so existing material databases and validated models must be leveraged with new approaches for identification. Since the produced plutonium isotopics are a function of multiple reactor operating parameters (fuel type, fuel enrichment, moderator, local fuel and moderator temperatures, reactor power, and irradiation time), a multi-variate approach is necessary to capture the variation. In this work, a novel classification algorithm based on regression models of measured and calculated plutonium isotopic data has been developed. While regression analysis is an established method, this is the first application of this technique to available used fuel plutonium isotopic measurement data combined with calculated data from reactor physics models. The innovative algorithm can quickly identify the most probable reactor type of origin and can ultimately help focus limited resources in the event attribution of interdicted plutonium is necessary. Measured used fuel isotopic data was obtained through the Spent Fuel COMPOsition (SFCOMPO) database and combined with additional plutonium measurements from the Hanford plutonium production reactors to create a catalogue of plutonium isotopics. The augmented dataset includes measurements and uncertainties where available of the plutonium isotopes ²³⁸Pu, ²³⁹Pu, ²⁴⁰Pu, ²⁴¹Pu, and ²⁴²Pu for light water moderated reactors, graphite moderated reactors, and heavy water moderated reactors to develop the classification algorithm for use in discriminating the reactor of origin. The developed algorithm can be used to triage plutonium isotopic information to differentiate materials originating in reactors with different moderators, and the potential to discriminate between reactor types with the same moderator. The new capability provided by the classification algorithm can be applied to realworld scenarios in the nuclear forensic, counterterrorism, and counterproliferation communities. This method can incorporate additional datasets to increase the accuracy of identification as well as expanding the number of different reactor types.