Neutron transport calculations by Monte Carlo methods are finding increased application in nuclear reactor simulations. In particular, a versatile approach entails the use of a 2-step procedure, with Monte Carlo as a few-group cross section data generator at lattice level, followed by deterministic multi-group diffusion calculations at core level.
In this thesis, the Serpent 2 Monte Carlo reactor physics burnup calculation code is used in order to test a set of diffusion coefficient models, as well as neutron leakage methodologies at assembly level. The tests include novel anisotropic diffusion coefficient and heterogeneous leakage models developed and implemented by the author.
The analyses are mainly focused on a sodium-cooled fast reactor system, for which few-group cross section data was generated by stochastic methods with Serpent 2. The quality of the full-core diffusion results is evaluated by contrasting system eigenvalues and power distributions against detailed, full-core reference solutions also supplied by the Serpent 2 code and the same nuclear data library.
Whereas the new anisotropic diffusion coefficient formalism exhibits improved performance in the fast reactor system studied, there are restrictions to its applicability in other reactor designs. The newly proposed leakage model has a similar performance to that one of albedo ite-rations, and provides valuable information about the space-energy coupling of the scalar neutron flux at lattice level. This hitherto unavailable information does not entail a significant computational cost.
In sodium-cooled fast reactor calculations, the quality of diffusion theory results can be improved by either using directional diffusion coefficients and a fine energy mesh, or via leakage-corrected discontinuity factors. These factors can be calculated using net neutron currents supplied by heterogeneous leakage models. Preliminary results from this research also suggest that the studies maybe extended to graphite-moderated, gas-cooled reactors.