Adaptive Monte Carlo for nuclear data evaluation

Abstract

An adaptive Monte Carlo method for nuclear data evaluation is presented. A fast evaluation method based on the linearization of the nuclear model guides the adaptation of the sampling distribution towards the posterior distribution. The method is suited for parallel computation and provides detailed uncertainty information about nuclear model parameters. Especially, the posterior distribution of the model parameters is not restricted to be multivariate normal. The method is demonstrated in an evaluation of the 181Ta total cross section for incident neutrons. Future applications are as an efficient sampling scheme in the Total Monte Carlo method, and the restriction of parameter uncertainties in nuclear models by both differential and integral data.