The propagation of nuclear data uncertainties in neutron transport simulations is typically performed using the “sandwich” formula, which combines a covariance matrix with the sensitivity of the simulation response to the nuclear data. Since most neutron transport codes yield sensitivities on multigroup cross sections, the covariance matrix must be provided between multigroup cross sections. However, in the resonance region, uncertainties are quantified on resonance parameters, such as widths and energies. Modules of processing codes, like PUFF (AMPX) and ERRORR (NJOY), have introduced features to convert the resonance parameter covariance matrix (RPCM) into a covariance matrix on multigroup cross sections, a conversion that is performed using first-order approximations.

Several studies have demonstrated that the use of multigroup uncertainties is not adapted when the neutron flux has significant contributions in the resonance region and the self-shielding effects become important. In the present study, we adopt a similar approach by sampling resonance parameters based on the RPCM, and we propose a strategy for sampling inherently positive resonance parameters using their reduced form. Additionally, we further focus on ENDF evaluations that use the R-Matrix formalism, which allows for the reconstruction of angular distributions of outgoing neutrons from resonance parameters, thus enabling the propagation of resonance uncertainties into the angular distributions within the resonance region.

Among the few nuclei with an available RPCM, we select 63Cu and employ the HMI-006(1) criticality benchmark from the International Criticality Safety Benchmark Evaluation Project database due to its high sensitivity to 63Cu. The module PUFF is capable of processing a RPCM expressed with R-Matrix formalism, conversely to ERRORR, and thus will be use to compare both methods.