Neutronic calculations for light water reactors (LWRs) require determining the power distribution in the core, a computationally expensive task. To address this challenge, deterministic simulations are traditionally performed in two steps. First, isolated 2D calculations are carried out: fuel assembly simulations using transport theory with reflective boundary conditions. Then, full 3D simulations are performed: these use homogenized few-group cross sections generated in 2D to model the entire core via diffusion theory.
However, the deterministic codes used for the first step rely on approximations (e.g., energy dependence of cross sections, treatment of anisotropic scattering), which limit the accuracy of the results. In contrast, Monte Carlo methods offer an approximation-free alternative capable of producing reference-level results.
This thesis aims to develop and qualify Monte Carlo methods for generating diffusion coefficients and multi-group cross sections in the TRIPOLI-5 code (co-developed by the CEA and the ASNR). These data will then be used as input for deterministic codes to perform full-core neutronic simulations.
Few-group diffusion coefficients, which are essential for diffusion solvers, are not available in nuclear data libraries. Current deterministic methods rely on approximations, introducing uncertainties. This thesis will propose solutions to compute these coefficients with high fidelity in a continuous-energy Monte Carlo context.
The thesis will explore various approaches, including methods for calculating diffusion coefficients (e.g., the Cumulative Migration Area method) and leakage models. For each approach, the accuracy of the results will be evaluated with the goal of industrial implementation in TRIPOLI-5.
The implementation of these methods in TRIPOLI-5 will improve the accuracy of diffusion data for downstream solvers and overcome the limitations of current deterministic toolchains. Ultimately, this thesis will contribute to establishing TRIPOLI-5 as a reference platform for generating high-fidelity data for light water reactor core calculations, while paving the way for industrial integration of these methods.