Monte Carlo nuclear data adjustment via integral information

Abstract

In this paper, we present three Monte Carlo methods to include integral benchmark information into the nuclear data evaluation procedure: BMC, BFMC and Mocaba. They allow to provide posterior nuclear data and their covariance information in a Bayesian sense. Different examples will be presented, based on 14 integral quantities with fast neutron spectra (keff and spectral indices). Updated nuclear data for 235U, 238U and 239Pu are considered and the posterior nuclear data are tested with MCNP simulations. One of the noticeable outcomes is the reduction of uncertainties for integral quantities, obtained from the reduction of the nuclear data uncertainties and from the rise of correlations between cross sections of different isotopes. Finally, the posterior nuclear data are tested on an independent set of benchmarks, showing the limit of the adjustment methods and the necessity for selecting well representative systems.