Development and Assessment of Parent-Specific Multigroup Lumped Fission Product for Fast-Spectrum Transport-Depletion Calculation

Sammanfattning

Continuous-energy Monte Carlo is a high-fidelity tool for reactor-physics analysis, but its cost limits its use in workflows requiring repeated transport-depletion calculations. In this context, lumped fission products provide a model-reduction strategy by representingthe collective neutronic effect of the fission product inventory through a reduced set ofpseudo-nuclides.This thesis develops a methodology for generating and applying burnup-dependentmultigroup lumped fission product cross sections in fast-spectrum Monte Carlo depletion.The method is implemented in OpenMC and is based on parent-specific LFPs. Effective fission yields are obtained through auxiliary depletion calculations in which heavymetal transmutation and decay routes are suppressed, leaving fission as the only outgoingreaction channel of the selected parent. The resulting fission product inventories arethen used to construct yield-weighted multigroup microscopic cross sections for the LFP.A Python framework is also developed to use these libraries in a coupled multigrouptransport-depletion loop, selecting and interpolating the appropriate cross-section data ateach burnup step.The methodology is assessed on a Superphénix-like MOX pin cell and tested for transferability on an ESFR-like pin cell. In both cases the coupled MG LFP calculation reproduces the continuous-energy multiplication factor within about 200 pcm over a burnuprange of 0–100 MWd/kgHM. Integral quantities such as total heavy-metal inventory, total fission product concentration, and aggregate reaction rates are reproduced with smalldeviations, while the fission product capture-rate error remains within about 0.5% of thetotal fuel capture rate at end of life.Comparing with the CE reference, the MG LFP calculation is about 76 times fasterfor the Superphénix case and about 127 times faster for the ESFR case. Although librarygeneration has a non-negligible upfront cost, it can be amortised if libraries are reusedacross multiple calculations in a similar spectrum regime. The results show that parentspecific, burnup-dependent LFPs can provide an effective compromise between neutronicaccuracy and computational efficiency for fast-reactor Monte Carlo depletion analyses.