Convergence acceleration of neutronic Monte Carlo calculations

Abstract

Often neutrons are produced in nuclear reactors with high energies, but they are needed at low energies for uses like activation analysis and neutron capture therapy. The evaluation of the slowed down neutron amount by using the Monte Carlo method is very expensive in computation time and the variance is large for natural simulation. In order to reduce the variance and the computation time, we used two biasing techniques to accelerate the calculation convergence. We have used the adjoint flux in the considered system as an importance function in the neutron slowing down equation. In this study, we have considered a homogeneous medium that contains a mixture of U238 (absorber) and hydrogen (scatterer). By handling the adjoint slowing down equation, we have used an analytical approximation of the fine structure of the adjoint flux, as a neutron importance function in the Monte Carlo simulation, for selecting the nuclide with which neutrons interact during their slowing down without absorption. For the second method, we modified the neutron slowing down equation by multiplying it by the adjoint flux. This allowed us to select neutron energies after collision and to avoid the energies corresponding to the absorption resonance. In fact, this was accomplished by assigning an appropriate statistical weight to the neutron, since its birth. For the two methods, a correction in the statistical weight was made after each neutron collision and a Fortran program was used to perform these calculations.