Les calculs Monte Carlo en neutronique-criticité permettent d'estimer le coefficient de multiplication effectif ainsi que des grandeurs locales comme le flux ou les taux de réaction. Certaines configurations présentant de faibles couplages neutroniques (modélisation de cœurs complets, prise en compte de profils d'irradiations, ...) peuvent conduire à de mauvaises estimations du kef f ou des flux locaux. L'objet de cette thèse est de contribuer à rendre plus robuste l'algorithme Monte Carlo utilisé et améliorer la détection de la convergence. L'amélioration du calcul envisagée passe par l'utilisation, lors du calcul Monte Carlo, d'un flux adjoint obtenu par un pré-calcul détermi- niste réalisé en amont. Ce flux adjoint est ensuite utilisé pour déterminer le positionnement de la première génération, modifier la sélection des sites de naissance, et modifier la marche aléatoire par des stratégies de splitting et de roulette russe. Une méthode de détection automatique du transitoire a été développée. Elle repose sur la modélisation des séries de sortie par un processus auto régressif d'ordre 1 et un test statistique dont la variable de décision est la moyenne du pont de Student. Cette méthode a été appli- quée au kef f et à l'entropie de Shannon. Elle est suffisamment générale pour être utilisée sur n'importe quelle série issue d'un calcul Monte Carlo itératif. Les méthodes développées dans cette thèse ont été testées sur plusieurs cas simplifiés présentant des difficultés de convergence neutroniques. / Monte Carlo criticality calculation allows to estimate the effective mu- tiplication factor as well as local quantities such as local reaction rates. Some configurations presenting weak neutronic coupling (high burn up pro- file, complete reactor core, ...) may induce biased estimations for kef f or reaction rates. In order to improve robustness of the iterative Monte Carlo méthods, a coupling with a deterministic code was studied. An adjoint flux is obtained by a deterministic calculation and then used in the Monte Carlo. The initial guess is then automated, the sampling of fission sites is modi- fied and the random walk of neutrons is modified using splitting and russian roulette strategies. An automated convergence detection method has been developped. It locates and suppresses the transient due to the initialization in an output series, applied here to kef f and Shannon entropy. It relies on modeling stationary series by an order 1 auto regressive process and applying statistical tests based on a Student Bridge statistics. This method can easily be extended to every output of an iterative Monte Carlo. Methods developed in this thesis are tested on different test cases.

Dans le cadre de la conception des réacteurs, les schémas de calculs utilisant des codes de cal- culs neutroniques déterministes sont validés par rapport à un calcul stochastique de référence. Les biais résiduels sont dus aux approximations et modélisations (modèle d'autoprotection, développement en polynômes de Legendre des lois de choc) qui sont mises en oeuvre pour représenter les phénomènes physiques (absorption résonnante, anisotropie de diffusion respec- tivement). Ce document se penche sur la question de la pertinence de la modélisation de la loi de choc sur une base polynômiale tronquée. Les polynômes de Legendre sont utilisés pour représenter la section de transfert multigroupe dans les codes déterministes or ces polynômes modélisent mal la forme très piquée de ces sections, surtout dans le cadre des maillages énergétiques fins et pour les noyaux légers. Par ailleurs, cette représentation introduit aussi des valeurs négatives qui n'ont pas de sens physique. Dans ce travail, après une brève description des lois de chocs, les limites des méthodes actuelles sont démontrées. Une modélisation de la loi de choc par une fonction constante par morceaux qui pallie à ces insuffisances, a été retenue. Cette dernière nécessite une autre mod- élisation de la source de transfert, donc une modification de la méthode actuelle des ordonnées discrètes pour résoudre l'équation du transport. La méthode de volumes finis en angle a donc été développée et implantée dans l'environ- nement du solveur Sn Snatch, la plateforme Paris. Il a été vérifié que ses performances étaient similaires à la méthode collocative habituelle pour des sections représentées par des polynômes de Legendre. Par rapport à cette dernière, elle offre l'avantage de traiter les deux représenta- tions des sections de transferts multigroupes : polynômes de Legendre et fonctions constantes par morceaux. Dans le cadre des calculs des réacteurs, cette méthode mixte a été validée sur différents motifs : des cellules en réseau infini, des motifs hétérogènes et un calcul de réflecteur. Les principaux résultats sont : - un développement polynômial à l'ordre P 3 est suffisant par rapport aux biais résiduels dus aux autres modélisations (autoprotection, méthode de résolution spatiale). Cette modéli- sation est convergée au sens de l'anisotropie du choc sur les cas représentatifs des réacteurs à eau légère. - la correction de transport P 0c n'est pas adaptée, notamment sur les calculs d'absorbant B4 C. / In reactor physics, calculation schemes with deterministic codes are validated with respect to a reference Monte Carlo code. The remaining biases are attributed to the approximations and models induced by the multigroup theory (self-shielding models and expansion of the scattering law using Legendre polynomials) to represent physical phenomena (resonant absorption and scattering anisotropy respectively). This work focuses on the relevance of a polynomial expansion to model the scattering law. Since the outset of reactor physics, the latter has been expanded on a truncated Legendre polynomial basis. However, the transfer cross sections are highly anisotropic, with non-zero values for a very small range of the cosine of the scattering angle. Besides, the finer the energy mesh and the lighter the scattering nucleus, the more exacerbated is the peaked shape of this cross section. As such, the Legendre expansion is less suited to represent the scattering law. Furthermore, this model induces negative values which are non-physical. In this work, various scattering laws are briefly described and the limitations of the existing model are pointed out. Hence, piecewise-constant functions have been used to represent the multigroup scattering cross section. This representation requires a different model for the dif- fusion source. The discrete ordinates method which is widely employed to solve the transport equation has been adapted. Thus, the finite volume method for angular discretisation has been developed and imple- mented in Paris environment which hosts the Sn solver, Snatch. The angular finite volume method has been compared to the collocation method with Legendre moments to ensure its proper performance. Moreover, unlike the latter, this method is adapted for both the Legendre moments and the piecewise-constant functions representations of the scattering cross section. This hybrid-source method has been validated for different cases: fuel cell in infinite lattice, heterogeneous clusters and 1D core-reflector calculations. The main results are given below : - a P 3 expansion is sufficient to model the scattering law with respect to the biases due to the other approximations used for calculations (self-shielding, spatial resolution method). This order of expansion is converged for anisotropy representation in the modelling of light water reactors. - the transport correction, P 0c is not suited for calculations, especially for B4 C absorbant.

Perhaps surprisingly, the largest experience in operating nuclear power plants has been in nuclear naval propulsion, particularly submarines. This accumulated experience may become the basis of a proposed new generation of compact nuclear power plant designs. In an effort to de-carbonise commercial freight shipping, there is growing interest in the possibility of using nuclear propulsion systems. Reactor cores for such an application would need to be fundamentally different from land-based power generation systems, which require regular refueling, and from reactors used in military submarines, as the fuel used could not conceivably be as highly enriched. Nuclear-powered propulsion would allow ships to operate with low fuel costs, long refueling intervals, and minimal emissions; however, currently such systems remain largely confined to military vessels. This research project undertakes computational modeling of possible soluble-boron-free (SBF) reactor core designs for this application, with a view to informing design decisions in terms of choices of fuel composition, materials, core geometry and layout. Computational modeling using appropriate reactor physics (e.g. WIMS, MONK, Serpent and PANTHER), thermal-hydraulics etc. codes (e.g. COBRA-EN) is used for this project. With an emphasis on reactor physics, this study investigates possible fuel assembly and core designs for civil marine propulsion applications. In particular, it explores the feasibility of using uranium/thorium-rich fuel in a compact, long-life reactor and seek optimal choices and designs of the fuel composition, reactivity control, assembly geometry, and core loading in order to meet the operational needs of a marine propulsion reactor. In this reactor physics and 3D coupled neutronics/thermal-hydraulics study, we attempt to design a civil marine reactor core that fulfills the objective of providing at least 15 effective full-power-years (EFPY) life at 333 MWth. In order to unleash the benefit of thorium in a long life core, the micro-heterogeneous ThO2-UO2 duplex fuel is well-positioned to be utilized in our proposed civil marine core. Unfortunately, A limited number of studies of duplex fuel are available in the public domain, but its use has never been examined in the context of a SBF environment for long-life small modular rector (SMR) core. Therefore, we assumed micro-heterogeneous ThO2-UO2 duplex fuel for our proposed marine core in order to explore its capability. For the proposed civil marine propulsion core design, this study uses 18% U-235 enriched micro-heterogeneous ThO2-UO2 duplex fuel. To provide a basis for comparison we also evaluate the performance of homogeneously mixed 15% U-235 enriched all-UO2 fuel. This research also attempts to design a high power density core with 14 EFPY while satisfying the neutronic and thermal-hydraulics safety constraints. A core with an average power density of 100 MW/m3 has been successfully designed while obtaining a core life of 14 years. The average core power density for this core is increased by ∼50% compared to the reference core design (63 MW/m3 and is equivalent to Sizewell B PWR (101.6 MW/m3 which means capital costs could be significantly reduced and the economic attractiveness of the marine core commensurately improved. In addition, similar to the standard SMR core, a reference core with a power density of 63 MW/m3 has been successfully designed while obtaining a core life of ∼16 years. One of the most important points that can be drawn from these studies is that a duplex fuel lattice needs less burnable absorber than uranium-only fuel to achieve the same poison performance. The higher initial reactivity suppression and relatively smaller reactivity swing of the duplex can make the task of reactivity control through BP design in a thorium-rich core easier. It is also apparent that control rods have greater worth in a duplex core, reducing the control material requirements and thus potentially the cost of the rods. This research also analyzed the feasibility of using thorium-based duplex fuel in different cases and environments to observe whether this fuel consistently exhibit superior performance compared to the UO2 core in both the assembly and whole-core levels. The duplex fuel/core consistently exhibits superior performance in consideration of all the neutronic and TH constraints specified. It can therefore be concluded from this study that the superior performance of the thorium-based micro-heterogeneous ThO2-UO2 duplex fuel provides enhanced confidence that this fuel can be reliably used in high power density and long-life SBF marine propulsion core systems, offering neutronic advantages compared to the all-UO2 fuel. Last, but not least, considering all these factors, duplex fuel can potentially open the avenue for low-enriched uranium (LEU) SBF cores with different configurations. Motivated by growing environmental concerns and anticipated economic pressures, the overall goal of this study is to examine the technological feasibility of expanding the use of nuclear propulsion to civilian maritime shipping and to identify and propose promising candidate core designs.