Impact of the Normalization Condition and Model Information on Evaluated Prompt Fission Neutron Spectra and Associated Uncertainties

Low evaluated uncertainties compared to experimental information and a strong model impact were observed in some prompt fission neutron spectrum (PFNS) evaluations that include mean values and covariances stemming from a rigid model. Here, we show by studying the ²³⁹Pu PFNS ENDF/B-VII.1 evaluation via generalized least-squares analyses that strong model correlations in combination with the normalization condition on the estimated PFNS and its covariances result in surprisingly low evaluated uncertainties. Furthermore, the model changes the evaluated results by >1σ of combined experimental uncertainties near the average outgoing neutron energy (~2 MeV). We show both analytically and by means of representative numerical examples that the normalization condition on the spectrum and its covariances naturally leads to uncertainties reduced by a fully positively correlated scaling uncertainty.