Methodology investigations on uncertainties propagation in nuclear data evaluation

In this paper, we discuss distinct ways of calculating nuclear cross-section variance–covariance matrices issued from the spherical optical model such as the deterministic sensitivity method and Monte Carlo method. We also discuss the selection of uncertain parameters using both the local sensitivity and global sensitivity methods (based on spectral methods using a polynomial decomposition) to determine the importance of the input parameters with respect to the final uncertainty. Finally, we attempt to evaluate the differences between the uncertainties generated by methods like least squares method and Bayesian method, both necessitating the use of experimental data. Discussing the results of simulation for the cross-section calculations of the major neutron reactions on 89Y, our interest was particularly focused on the mathematical rigour and quality of the uncertainties that we obtained.