Benchmark calculations for the validation of the coupled neutron kinetics/thermohydraulic code complex DYN3D-ATHLET are described. Two benchmark problems concerning hypothetical accident scenarios with leaks in the steam system for a VVER-440 type reactor and the TMI-1 PWR have been solved. The first benchmark task has been defined by FZR in the frame of the international association "Atomic Energy Research" (AER), the second exercise has been organised under the auspices of the OECD. While in the first benchmark the break of the main steam collector in the sub-critical hot zero power state of the reactor was considered, the break of one of the two main steam lines at full reactor power was assumed in the OECD benchmark. Therefore, in this exercise the mixing of the coolant from the intact and the defect loops had to be considered, while in the AER benchmark the steam collector break causes a homogeneous overcooling of the primary circuit. In the AER benchmark, each participant had to use its own macroscopic cross section libraries. In the OECD benchmark, the cross sections were given in the benchmark definition. The main task of both benchmark problems was to analyse the re-criticality of the scrammed reactor due to the overcooling. For both benchmark problems, a good agreement of the DYN3D-ATHLET solution with the results of other codes was achieved. Differences in the time of re-criticality and the height of the power peak between various solutions of the AER benchmark can be explained by the use of different cross section data. Significant differences in the thermohydraulic parameters (coolant temperature, pressure) occurred only at the late stage of the transient during the emergency injection of highly borated water. In the OECD benchmark, a broader scattering of the thermohydraulic results can be observed, while a good agreement between the various 3D reactor core calculations with given thermohydraulic boundary conditions was achieved. Reasons for the differences in the thermohydraulics were assumed in the difficult modelling of the vertical once-through steam generator with steam superheating. Sensitivity analyses which considered the influence of the nodalisation and the impact of the coolant mixing model were performed for the DYN3D-ATHLET solution of the OECD benchmark. The solution of the benchmarks essentially contributed to the qualification of the code complex DYN3D-ATHLET as an advanced tool for the accident analysis for both VVER type reactors and Western PWRs.

Benchmark calculations for the validation of the coupled neutron kinetics/thermohydraulic code complex DYN3D-ATHLET are described. Two benchmark problems concerning hypothetical accident scenarios with leaks in the steam system for a VVER-440 type reactor and the TMI-1 PWR have been solved. The first benchmark task has been defined by FZR in the frame of the international association "Atomic Energy Research" (AER), the second exercise has been organised under the auspices of the OECD. While in the first benchmark the break of the main steam collector in the sub-critical hot zero power state of the reactor was considered, the break of one of the two main steam lines at full reactor power was assumed in the OECD benchmark. Therefore, in this exercise the mixing of the coolant from the intact and the defect loops had to be considered, while in the AER benchmark the steam collector break causes a homogeneous overcooling of the primary circuit. In the AER benchmark, each participant had to use its own macroscopic cross section libraries. In the OECD benchmark, the cross sections were given in the benchmark definition. The main task of both benchmark problems was to analyse the re-criticality of the scrammed reactor due to the overcooling. For both benchmark problems, a good agreement of the DYN3D-ATHLET solution with the results of other codes was achieved. Differences in the time of re-criticality and the height of the power peak between various solutions of the AER benchmark can be explained by the use of different cross section data. Significant differences in the thermohydraulic parameters (coolant temperature, pressure) occurred only at the late stage of the transient during the emergency injection of highly borated water. In the OECD benchmark, a broader scattering of the thermohydraulic results can be observed, while a good agreement between the various 3D reactor core calculations with given thermohydraulic boundary conditions was achieved. Reasons for the differences in the thermohydraulics were assumed in the difficult modelling of the vertical once-through steam generator with steam superheating. Sensitivity analyses which considered the influence of the nodalisation and the impact of the coolant mixing model were performed for the DYN3D-ATHLET solution of the OECD benchmark. The solution of the benchmarks essentially contributed to the qualification of the code complex DYN3D-ATHLET as an advanced tool for the accident analysis for both VVER type reactors and Western PWRs.

Durant cette these, j'ai etudie la formation des galaxies grace a des simulations cosmologiques et a un nouveau modele analytique. Dans le cadre du modele standard de formation hierachique des grandes structures, les petites fluctuations primordiales de densite observees sur le fond diffus cosmologique sont amplifiees par la gravite pour donner des halos de matiere noire de plus en plus gros. C'est au sein de ces halos que le gaz s'effondre et se refroidit pour former des disques de gaz froid a support centrifuge. Ces disques sont ensuite convertis en disques stellaires : les galaxies. Le probleme dans ce scenario est celui du sur-refroidissement : une trop grande part du gaz finit sous forme d'etoiles comparativement aux observations. J'ai donc realise une etude de l'evolution des baryons (un gaz d'hydrogene et d'helium) dans l'Univers grace a des simulations numeriques hydrodynamiques haute-resolution. Cependant, ces simulations sont affectees par des effets de resolution finie. J'ai ainsi developpe un modele analytique simple qui possede l'avantage de ne pas etre affecte par de tels effets. Celui-ci predit la quantite de baryons dans chacune des 4 phases suivantes : etoiles, gaz froid dans les disques galactiques, gaz chaud dans les amas et gaz diffus dans le milieu intergalactique. La comparaison des resultats aux observations a montre que la cosmologie controle le taux de formation d'etoiles dans l'Univers. Ce modele a aussi mis en lumiere le role essentiel des vents galactiques qui, ejectant le gaz froid des disques jusque dans les halos de gaz chaud, permettent d'eviter le probleme du sur-refroidissement. Enfin, en une ouverture vers l'astroparticule, j'ai etudie les implications de la physique du gaz sur le fond diffus gamma produit dans l'hypothese de matiere noire legere.