Similarity of a CANDU-6 lattice cell to criticality benchmarks for k-infinity and coolant void reactivity

A collision-history estimator of continuous-energy sensitivity coefficients of k, weighted by a fission-matrix estimate of the fission-source importance, was added to OpenMC 0.16.0. On Godiva it agrees with a direct perturbation of the U-235 density (0.8050 against 0.7976 ± 0.0067) and with TSUNAMI-3D's fission, ν̄ and capture sensitivities within 2 %, and the importance weighting proves necessary. Applied to a CANDU-6 37-element natural-uranium lattice cell with ENDF/B-VIII.1 data, it gives the sensitivities of k∞ in the cooled (1.12531) and voided (1.14636) states and of the coolant void reactivity, CVR = 16.32 ± 0.13 mk. With ENDF/B-VIII.1 covariances processed by AMPX, the nuclear-data uncertainty is 0.81 % Δk/k for k∞ and 2.5 % of the CVR (0.40 mk), against 1.09 % and 3.2 % with ENDF/B-VII.1. The largest single contributor to the CVR, and the second largest to k∞, is H-2 (n,2n), whose evaluated relative uncertainty exceeds 100 % just above its threshold. None of 2,843 ICSBEP configurations reaches ck = 0.9 with the k∞ of the lattice. The most similar is a heavy-water-moderated low-enriched uranium metal lattice (LMT001-001, 0.853), and 35 configurations exceed 0.7. Of four IRPhE heavy-water lattices, a ZED-2 configuration is the most similar, at 0.89–0.93 depending on the covariance library. No benchmark's k is strongly correlated with the CVR. The correlations lie between −0.33 and +0.05 for the ICSBEP configurations and reach −0.41 for ZED-2, so a single benchmark of k can reduce the nuclear-data uncertainty of the CVR by at most 9 %. PDF of the preprint published at https://hardtohit10.gitlab.io/candu-similarity-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 and the MCNP6 sensitivity calculations of the IRPhE lattices, proposed the study and directed the work. Claude wrote the OpenMC sensitivity code and the analysis scripts, 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 The article has not been peer reviewed, and the results have not been independently reviewed or verified.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.