Comparison of Two Approaches for Nuclear Data Uncertainty Propagation in MCNPX for Selected Fast Spectrum Critical Benchmarks

Nuclear data uncertainty propagation based on stochastic sampling (SS) is becoming more attractive while leveraging modern computer power. Two variants of the SS approach are compared in this paper. The Total Monte Carlo (TMC) method by the Nuclear Research and Consultancy Group (NRG) generates perturbed ENDF-6-formatted nuclear data by varying nuclear reaction model parameters. At Paul Scherrer Institute (PSI) the Nuclear data Uncertainty Stochastic Sampling (NUSS) system generates perturbed ACE-formatted nuclear data files by applying multigroup nuclear data covariances onto pointwise ACE-formatted nuclear data. Uncertainties of 239Pu and 235U from ENDF/B-VII.1, ZZ-SCALE6/COVA-44G and TENDL covariance libraries are considered in NUSS and propagated in MCNPX calculations for well-studied Jezebel and Godiva fast spectrum critical benchmarks. The corresponding uncertainty results obtained by TMC are compared with NUSS results and the deterministic Sensitivity/Uncertainty method of TSUNAMI-3D from SCALE6 package is also applied to serve as a separate verification. The discrepancies in the propagated 239Pu and 235U uncertainties due to method and covariance differences are discussed.