In some circumstances (low population or large dominance ratio), positive correlations between neutrons lead to significant underestimation of statistical uncertainties. In the past, the effect of spatial correlations in nuclear reactors has received a lot of attention, and spatial clustering has been shown to occur even in realistic reactor (e.g. during start-up). In this work, we investigate in depth the effect of using branchless collisions instead of regular branching collisions as a way of mitigating correlations in Monte Carlo power iteration. We consider a simple homogeneous system of large dominance ratio, and a highly heterogeneous system of smaller dominance ratio (3D-C5G7), and we focus on the analysis of spatial correlations. Branchless collisions are shown to outperform regular branching collisions in all situations, although the effect is less pronounced for C5G7 due to its smaller dominance ratio.  

Project lead: Theophile Bonnet