Monte Carlo source convergence acceleration with multi-group inactive cycles

This paper proposes a novel source convergence acceleration method that speeds up the inactive cycles of Monte Carlo eigenvalue calculations. It consists of simulating the bulk of the inactive cycles with multi-group cross sections and switching to continuous energy cross sections during the active cycles to maintain full fidelity. This method was implemented in the Monte Carlo code SCONE, and tested on two 3D full length problems with a large dominance ratio. In both models, a PWR pin and a PWR assembly, the combination between multi-group and continuous energy Monte Carlo successfully converged to the correct fission source, monitored using Shannon entropy. The inactive cycles of the two calculations were sped up respectively by a factor of 5 and 3. These factors do not account for the time spent generating cross sections. A precise estimation of the preparation time was not reported in this paper, since optimising the cross section generation process will be the subject of future work. Finally, the axial flux produced by the two calculation routes was compared in order to ensure that a bias was not introduced