Moment-based probability tables for angular anisotropic scattering

This paper presents a strategy for taking into account anisotropy scattering into a Monte Carlo algorithm relying on the subgroup method and developed in the DRAGON lattice code. For the sake of consistency, we limited our Monte Carlo code to the same cross-section libraries available for deterministic methods. However Legendre moments for the transfer cross-sections cannot be directly used during the Monte Carlo random walk, due to the presence of non-positive parts into the distributions. The discrete angle method is proposed to deal with this limitation, following an approach initially introduced in the MORET multigroup Monte Carlo code. We selected a moment approach, originally employed to compute probability tables for resonant cross-sections, to derive consistent sums of Dirac distributions conserving Legendre moments of the angular distributions. A detailed analysis of the applicability of the moment approach is here mandatory. When the moment technique fails due to incoherent Legendre moments, the discrete angle technique is substituted by legacy semi-analytical methods. We illustrate the proposed method using critical benchmarks coming from the ICSBEP handbook by comparison toward SN and other Monte Carlo results. The impact of the anisotropy scattering is also discussed on a PWR MOX assembly case.