IMPACT OF NUCLEAR DATA LIBRARIES ON PREDICTED FAST REACTOR PERFORMANCE

¹ Argonne National Laboratory 9700 South Cass Ave, Lemont, IL 60439
² Georgia Institute of Technology 770 State St. NW, Atlanta, GA 30313

agnelson@anl.gov
kmramey@gatech.edu
fheidet@anl.gov

Published online: 22 February 2021

Abstract

The nuclear data evaluation process inherently yields a nuclear data set designed to produce accurate results for the neutron energy spectra corresponding to a specific benchmark suite of experiments. When studying reactors with spectral conditions outside of, or not well represented by, the experimental database used to evaluate the nuclear data, care should be given to the relevance of the nuclear data used. In such cases, larger biases or uncertainties may be present than in a reactor with well-represented spectra. The motivation of this work is to understand the magnitude of differences between recent nuclear data libraries to provide estimates for expected variability in criticality and power distribution results for sodiumcooled, steel-reflected, metal-fueled fast reactor designs.

This work was specifically performed by creating a 3D OpenMC model of a sodium-cooled, steel-reflected, metal-fueled fast reactor similar to the FASTER design but without a thermal test region. This OpenMC model was used to compare the differences in eigenvalues, reactivity coefficients, and the spatial and energetic effects on flux and power distributions between the ENDF/B-VII.0, ENDF/B-VII.1, ENDF/B-VIII.0, JEFF-3.2, and JEFF-3.3 nuclear data libraries. These investigations have revealed that reactivity differences between the above libraries can vary by nearly 900 pcm and the fine-group fluxes can vary by up to 18% in individual groups. Results also show a strong variation in the flux and power distributions near the fuel/reflector interface due to the high variability in the ⁵⁶Fe cross sections in the libraries examined. This indicates that core design efforts of a sodium-cooled, steel-reflected, metalfueled reactor will require the application of relatively large nuclear data uncertainties and/or the development of a representative benchmark-quality experiment.