A comparative study of leakage and diffusion coefficient models for few-group cross section generation with the Monte Carlo method

The quality of few-group cross section data generated for core diffusion solvers depends on the treatment of the neutron leakage at lattice level. In this work, we propose a study about the relative performance of three assembly leakage models in the Monte Carlo code Serpent. Additionally, the extension of a novel method for the generation of directional diffusion coefficients to different reactor types is studied. Using Monte Carlo full core results as reference values, leakage and diffusion coefficient models are contrasted in terms of system eigenvalue and power distributions in a simplified heavy-water-moderated reactor system. Heterogeneous leakage models yielded the best results. Standard diffusion coefficients showed the best performance, due to difficulties arising from the volume-averaging of directional diffusion coefficients in cluster geometry. The results of this work suggest that the generation of directional diffusion coefficients could be extended to graphite-moderated gas-cooled reactors.