Nuclear-data uncertainty of keff for 2,843 ICSBEP configurations with ENDF/B-VIII.1 covariances

The keff sensitivity profiles of DICE, the database of the ICSBEP Handbook, that were calculated with SCALE 6.0 were matched to the isotopic TSUNAMI-3D calculations they were made from: 2,843 of 2,849 configurations, each checked against DICE's own element-summed profile to the printed precision. Combined by the sandwich rule with four 56-group covariance libraries, they give the first-order nuclear-data uncertainty of every configuration. With ENDF/B-VIII.1 covariances processed by AMPX the median is 0.59 % Δk/k (5–95 %: 0.53–0.94 %), against 0.96 % with an ENDF/B-VII.1 library and 0.72 % with SCALE's own 56-group library; the ratio to ENDF/B-VII.1 has a median of 0.63. The reduction comes from smaller uncertainties of U-235 ν̄, χ and capture, U-238 inelastic scattering and capture, and from new anticorrelations between H-1 elastic scattering and capture and between Pu-239 fission and capture. Fast plutonium systems are the exception: larger Pu-239 fission and ν̄ uncertainties raise PU-MET-FAST from 0.45 % to 0.62 %. The same covariance data processed by NJOY give the same result within ±0.05 percentage points for 97.7 % of the configurations, and within ±0.1 points for all of them, except 147 configurations that contain one of nine nuclides whose R-matrix resonance-parameter covariances NJOY's ERRORR module mis-processes; for these the NJOY values reach 15.3 %. Mapping DICE's 238 groups onto the 56 groups of the covariances changes no result by more than 0.005 percentage points. Nuclide by nuclide, the ENDF/B-VII.1 and SCALE-library results agree with the results posted for the NEA's NDaST tool, with median ratios of 1.00 and 1.03. PDF of the preprint published at https://hardtohit10.gitlab.io/icsbep-uncertainty-paper.html (version 1, updated 8 October 2026). Generated with AI (Claude, Anthropic). About the authors and the status of this work. Claude is an AI model developed by Anthropic. Ian Hill created the DICE sensitivity profiles used, provided the original SCALE 6.0 calculations they come from, proposed the study and directed the work. Claude wrote the extraction, matching and analysis code, ran the calculations, made the figures and tables, and wrote this article. The article has not been peer reviewed, and the results have not been independently reviewed or verified. Anthropic has not reviewed or endorsed this work. Readers should check the results before relying on them.