Application of the MGAU Code for Measuring 235U Enrichment at the Brazilian Safeguards Laboratory

The MGAU is a software tool for conducting uranium enrichment measurements based on highresolution gamma-ray spectroscopy. The code is capable of analyzing spectra (90 – 120 KeV region) collected from a wide variety of sample geometries and compositions. The main advantage of the code is its ability to perform spectra evaluation without a requirement for calibration with representative standards. However, it does require that the daughter isotopes be in activity equilibrium with the 235U and 238U parent isotopes. In order for the code to be more versatile in overcoming its limitations, a modified version of the traditional “enrichment meter” method has been also added. In order to perform confirmatory uranium enrichment measurements for safeguards purposes at a laboratory environment, the Brazilian Safeguards Laboratory is investigating the performance of a nondestructive technique based on the use of the MGAU code for analyzing of gamma-ray spectra collected from pure uranium samples (primarily natural and low enriched powders and pellets). Several new good practice procedures were implemented in order to optimize the performance of the method at the best achievable level. This includes positioning of both the high-purity germanium detector and the sample inside a lead chamber for reducing background influence, collection of replicate measurements, and application of robust statistical treatment of data to reduce random contributions from counting statistics to the final uncertainty. Also, temperature and humidity inside the laboratory were monitored so that significant influences in results could be observed. Based on the results arising from analysis of certified reference materials, this paper discusses the performance of the MGAU code version 4.0 with focus on the uncertainties related to sample-dependent effects (mass, density, matrix composition and enrichment level). The reliability of the MGAU predicted uncertainty for single measurements and the occurrence of significant biases are evaluated and reported.