A Reproducible Methodology for Measuring the Impact of Common Assumption Violations on SAMMY Evaluated Cross-Section Covariance Estimates

View at publisher Summarise with VIRAT
Authors: Jacob Forbes, Noah Walton, Oleksii Zivenko, Amanda Lewis, William Fritsch, Vladimir Sobes
Cited by: 0
References: 6

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Accurate uncertainty estimates for neutron reaction cross sections are essential to many applications across engineering and the fundamental sciences. SAMMY, an open-source tool used by many evaluators for estimating evaluated cross-section covariance matrices (ECSCMs), has known limitations and inaccuracies. This paper provides the nuclear data community with a methodology for measuring the impact of common generalized least squares (GLS) assumption violations on the accuracy of SAMMY covariance estimates. The methodology benefits from access to high-fidelity synthetic data, making it reproducible and flexible and is used to validate SAMMY ECSCM estimation when all GLS assumptions are met. Additionally, three case studies are performed to highlight the kinds of research questions that can be answered with the proposed approach. The case studies measure the sensitivity of SAMMY ECSCM accuracy to the size of missing resonances in underspecified models, the use of the Single-Level Breit-Wigner approximation to fit data when interference effects are present, and the degree to which systematic uncertainties from the data reduction process are underreported or overreported. Finally, we propose and demonstrate a method for mapping a single-value Frobenius norm metric to familiar metrics with interpretable units.

Fields of Study & Subjects

Nuclear reactor physics and engineering Nuclear physics research studies Nuclear Physics and Applications

Publication details

Journal / source
Nuclear Science and Engineering
Pages
1–13
Publication date
25 March 2026
Language
EN