Towards a highly efficient and unbiased population-control algorithm for kinetic Monte Carlo simulations

¹ Université Paris-Saclay, CEA, Service d’Etudes des Réacteurs et de Mathématiques Appliquées, 91191, Gif-sur-Yvette, France
² École polytechnique fédérale de Lausanne (EPFL), Switzerland
³ Institut de Radioprotection et de Sûreté Nucléaire, 31 avenue de la Division Leclerc, Fontenay-aux-Roses, France 92260

^* Corresponding author: cecilia.montecchio@cea.fr
^** Corresponding author: vincent.lamirand@epfl.ch
^*** Corresponding author: davide.mancusi@cea.fr
^**** Corresponding author: wilfried.monange@irsn.fr
^† Corresponding author: andrea.zoia@cea.fr

Published online: 15 October 2024

Abstract

Population-control methods are key to non-stationary Monte Carlo simulations of multiplying systems: they prevent either the unbounded growth or the disappearance of neutrons, occurring respectively in supercritical and subcritical conditions; furthermore, they contribute to an efficient allocation of computational resources by addressing the unbalance between the neutron and the precursor populations. In this paper, we present two alternative populationcontrol algorithms: the legacy implementation in TRIPOLI-4®, the Monte Carlo code developed at CEA, and an improved version that is currently under investigation, based on the use of a simplified point-kinetics solver. We assess the performance of these methods through the simulation of a $2.2 step reactivity insertion in a fast system (Flattop-Pu), leading to an increase of the neutron population by a factor 200, which is benchmarked against point kinetics. We show that the new implementation not only suppresses the slight bias that was present in the legacy method due to a stochastic normalization factor, but also outperforms the previous algorithm in terms of variance reduction and improvement of the figure of merit.