Covariance matrices for cross sections issued from optical model calculationsP. Dossantos-Uzarralde1, P. Mounier2, M. Brassart2 and M. Couvert3
1 CEA/DIF BP. 12, 91680 Bruyères-le-Châtel, France
2 CEA/Saclay DM2S/SERMA, 91191 Gif-sur-Yvette, France
3 CEA/Cadarache DTAP/SET, Bât. 220, 13108 Saint-Paul-lez-Durance, France
Published online: 21 May 2008
Our interest is particularly focused at the study of uncertainties propagation within nuclear models. In this paper, a comparison between the most convenient approaches for generating covariance matrices for evaluations based on nuclear modeling calculations is presented. The first method involves the conventional matrix error propagation approach based on sensitivity matrix calculation. In the second one, uncertainties associated to the model parameters are propagated with a Monte Carlo simulation. To perform a quantitative uncertainty analysis, probability distributions must be assigned to each of the varied parameters, Monte Carlo procedures are used to simulate alternative distributions of parameter values.
© CEA 2008