Response to adverse weather conditions by cotton and other major crops are likely to be heterogeneous across varieties, but it is unclear whether this translates into yield risk heterogeneity across varieties. Crop insurance is the dominant agricultural policy instrument and will play an important role as farmers adapt to climate change. However, climate change impact on the performance of the crop insurance programs is not well established and currently the Risk Management Agency (RMA) does not offer alternative premium rates across varieties; nor has there been any public acknowledgement that it plans to adjust rates in anticipation of climate change. In this study, we identify whether there are heterogeneous premium rates across varieties; we also measure the impact of warming on these premium rates. Our findings show heterogeneities of premium rates across varieties and different warming scenarios, as well as heterogeneities in expected yield and yield risk.

Reservoir heterogeneities exhibit a wide range of length scales, and their interaction with various transport mechanisms control the overall performance of subsurface flow and transport processes. Modeling these processes at large-scales requires proper scale-up of both heterogeneity and the underlying transport mechanisms. This research demonstrates a new reservoir modeling procedure to systematically quantify the scaling characteristics of transport processes by accounting for sub-scale heterogeneities and their interaction with various transport mechanisms based on the volume averaging approach. Although treatments of transport problems with the volume averaging technique have been published in the past, application to real geological systems exhibiting complex heterogeneity is lacking. A novel procedure, where results from a fine-scale numerical flow simulation reflecting the full physics of the transport process albeit over a small sub-volume of the reservoir, can be integrated with the volume averaging technique to provide effective description of transport at the coarse scale. In a volume averaging procedure, scaled up equations describing solute transport in single-phase flow are developed. Scaling characteristics of effective transport coefficient corresponding to different reservoir heterogeneity correlation lengths as well as different transport mechanisms including convection, dispersion, and diffusion are studied. The method is subsequently extended to describe transport in multiphase systems to study scaling characteristics of processes involving adsorption and inter-phase transport. Key conclusions drawn from this dissertation show that 1) variance of reservoir properties and flow responses generally decrease with scale; 2) scaling of recovery processes can be described by the scaling of effective mass transfer coefficient (Keff); in particular, mean and variance of Keff decrease with length scale, similar in the fashion of recovery statistics (e.g., variances in tracer breakthrough time and recovery); 3) the scaling of Keff depends on the underlying heterogeneity and is influenced by the dominant transport mechanisms. To show the versatility of the approach for studying scale-up of other transport mechanisms, it is also applied to derive scaled up formulations of non-Newtonian polymer flow to investigate the scaling characteristics of the apparent viscosity and effective shear rate in porous media. / text

Reconhece-se, por estudos prévios, as complexas heterogeneidades de fácies nos depósitos sedimentares siliciclásticos fluviais, presentes na exploração de hidrocarbonetos. Devido à amostragem, custosa e insuficiente, na indústria, os modelos geológicos construídos possuem alta incerteza associada. Visando auxiliar tanto a exploração quanto a recuperação, foram executados estudos de análogos de reservatório na Formação São Sebastião, na Bacia do Tucano Central. Para isso, com base na dinâmica de processos fluviais, integraram-se análises de micro (petrografia e diagênese) e mesoheterogeneidades (fácies e elementos arquiteturais) tridimensionalmente, junto à permoporosidade, culminando em modelos geológicos de elementos arquiteturais, fácies e permeabilidade. As características permoporosas da Formação São Sebastião estão sujeitas tanto a controles deposicionais (elementos arquiteturais) como a controles diagenéticos (petrofácies), os quais se correlacionam. Os elementos arquiteturais, portadores de características mais efetivas como reservatório são as barras unitárias, seguidas pelas compostas. Em mesoescala, ambas são representadas majoritariamente por cruzadas tabulares e acanaladas de diferentes dimensões. A mediana para permeabilidade das barras unitárias e compostas é de 6000 mD e 4500 mD, enquanto para porosidade é de 21,0 % e 16,9 %, respectivamente. Em contrapartida, as barreiras de permeabilidademais importantes são representadas pelos topos de barra contínuos que majoritariamente apresentam estruturas de marca ondulada e estruturas plano paralelas, em mesoescala, e petrofácies com alta concentração de micas. A mediana de permeabilidade dos topos de barra é 200 mD, já a de porosidade é 13,5%. Como barreiras de permeabilidades subordinadas, encontram-se os canais abandondados (pouco freqüentes nos afloramentos), em mesoescala, representados por estruturas de marcas onduladas e convolutas; e, em microescala, relacionadas a altas concentrações de micas e pseudomatriz. A mediana da permeabilidade nos canais abandonados é 50 mD e a da porosidade é 16,1 %. Os modelos geológicos gerados com os dados de permeabilidade e fácies apontaram para anisotropias de baixa permeabilidade nos topos de barra e canais abandonados. Os topos de barra possuem continuidade lateral maior que 200 m em alguns locais, resultando em grande impacto no fluxo dentro do reservatório. As anisotropias de alta permeabilidade podem ser observadas principalmente na base dos depósitos de barras. Essas anomalias estão relacionadas com arenitos médios a grossos, portadores de cruzadas tabulares. Deste modo, nos depósitos analisados, a permeabilidade é controlada principalmente pela granulometria e concentração de pseudomatriz e micas, com influência subordinada das estruturas sedimentares. Já a porosidade, também controlada pela concentração de pseudomatriz e micas, possui forte influência das estruturas sedimentares e seleção granulométrica, mas não tem correlação com a granulometria. / The complex facies heterogeneities in fluvial deposits are present in the oil and gas industry. Due to expensive and insufficient sampling, in this industry, geological models which are created to assist the hydrocarbon exploration contain many uncertainties. Aiming to assist in this situation, studies on outcrops reservoirs analogues were performed in Sao Sebastião Formation, in Central Tucano Basin. Based on fluvial sedimentology, micro (petrography and diagenesis) and mesoheterogeneities (facies and architectural elements) analyses were conducted. Porosity and permeability were likewise analyzed. The data obtained were used to create architectural, facies and permeability models. Porosity and permeability are mainly controlled by two correlated variables: depositional controls (architectural elements) and diagenetic controls (petrofacies). The architectural elements with the most effective characteristics as reservoir are the unit bars, followed by the compound bars. Both architectural elements are represented by small to large (0.2 m up to 1.5 m) planar and through cross strata. The median permeability of unit bars is 6000mD, and the median of compound bars is 5000mD. The median porosity of unit bars is 21.0%, and the median of compound bar is 16.9%. On the other hand, the most important permeability barriers are the continuous bar tops, they are represented in mesoscale by climbing ripples structures and horizontal bedding. Bartops have shown a high mica assembly in microscale analyses. The median permeability of this element is 200mD, and the median porosity is 13.5%. The abandoned channels are subordinated permeability barriers in São Sebastiao Formation. In mesoscale they are mainly represented by climbing ripples structures and convolute bedding. In microscale they are related to high mica and pseudomatrix content. The permeability median of abandoned channels is 50 mD, and the median porosity is 16.1%. The geological models created also pointed out to bar tops and abandoned channels as permeability barriers. The bar tops can reach 200 m or more of lateral continuity, representing the highest flux barrier inside the reservoir. The positives anisotropies of permeability are linked to medium to coarse sandstones with planar cross strata. Therefore, in Sao Sebastiao Formation the permeability is controlled primarily by grain size, pseudomatrix and mica content. Sedimentary structures also impact the permeability subordinately. The porosity is handed likewise by the pseudomatrix and mica content, but also by the sedimentary structures and grain sorting. Nevertheless, porosity is not correlated with grain size.

Made available in DSpace on 2014-06-11T19:22:20Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-02-07Bitstream added on 2014-06-13T19:48:35Z : No. of bitstreams: 1 fusca_cj_me_sjrp.pdf: 504460 bytes, checksum: ec0f2136aa847064c00b95badf979b51 (MD5) / Este trabalho propõe investigar a abreviação em salas de bate-papo abertas (chats) da internet, frequentadas por escreventes que afirmam ter entre 15 e 20 anos. O conjunto do material é formado por duas “conversas” virtuais, com duração de 60 (sessenta) minutos cada uma. Reconhecida como uma das características do chamado “internetês”, busca-se refletir que a abreviação não consiste em mero “corte de palavras”, mas em recurso que apresenta regularidade e sistematicidade linguísticas. Esse processo é tomado como indício de (novo) gênero de discurso em emergência (BAKHTIN, 1997; MARCUSCHI, 2005). Com a finalidade de verificar os processos formadores de abreviaturas em bate-papos, ou seja, as regularidades linguísticas desse recurso, utiliza-se, como ferramenta de análise, a teoria de sílaba fonológica (SELKIRK, 1982). Acredita-se que, por meio do estudo da estrutura da sílaba, é possível avaliar que a escolha dos grafemas das abreviaturas é fundada na heterogeneidade da escrita (CORRÊA, 2004), visto que o escrevente pode se basear tanto em práticas orais/faladas quanto em práticas letradas/escritas para sua composição estrutural. Considera-se também que o funcionamento e a emergência de um modo de enunciação digital apenas são possíveis em meio a uma sociedade líquida (BAUMAN, 2004), caracterizada pela não permanência e fragilidade das relações entre os sujeitos. A consideração desses fatores parece ser indispensável para o entendimento de prática social ainda pouco (ou não) reconhecida / This research’s propose is to investigate the word’s abbreviation which appears on the internet chats, used by writers who affirm to be around 15 or 20 years old. The material’s amount is composed by two virtual “conversations” that spends around 60 minutes which one. Recognized as one characteristic of the well-known “internetês”, we consider that this abbreviation is not a simple attempt to “cut the words”, it is a resource that presents linguistics’ regularity. This process is seeing as a clue of a (new) speech genres that is borning (BAKHTIN, 1997; MARCUSCHI, 2005). In order to verify the process that forms possible abbreviations used on chats, in other words, the linguistic regularities of that resource, it is used, as a tool of analyze, the theory of phonological syllable (SELKIRK, 1982). It is believed that, based on the syllable structure’s study, it is possible to assess that the choice of the abbreviation’s graphemes is based on the writing heterogeneity (CORRÊA, 2004), since the writer can be based on oral/spoken practices or on literate/written practices. It is also considered that the operation and the emergency of a digital way to enunciate is only possible in a liquid society (BAUMAN, 2004), that is not characterized because of the fragile and permanent relationship among subjects. Considering these factors it seems to be indispensable to comprehend the social practices which are rarely (or never) recognized

La présente étude porte sur les interactions entre deux types de matériaux susceptibles d'être mis en jeu au sein d'un stockage géologique de déchets radioactifs : les matériaux métalliques, constituant notamment le surconteneur de déchets de haute activité, et les matériaux argileux, telle que la roche hôte argileuse. Les interactions entre ces deux matériaux en conditions de stockage pourraient en effet modifier leurs propriétés de confinement initiales. Les travaux présentés ont visé à déterminer l'influence d'hétérogénéités (vides technologiques et fractures) et d'activités bactériennes sur ces interactions, notamment en termes d'évolution de propriétés chimiques et hydrauliques de l'argile. Dans cet objectif, deux expériences intégrées en cellules de percolation ont été mises en œuvre, à 60 °C, pendant 13 mois : la première en présence de deux souches bactériennes (BSR, BFR), la seconde constituant un témoin abiotique. Ces expérimentations ont permis la circulation d'une eau synthétique dont la composition est représentative de celle de l'eau porale de Tournemire au travers d'un compact de fer pulvérulent puis d'une carotte d'argilite du Toarcien de Tournemire, artificiellement fissurée. L'une des deux demi-carottes d'argile contenait également un cylindre de fer massif. Les caractérisations post-mortem (MEB, MEB/EDS, Raman, DRX, tomographie aux rayons X) ont permis d'étudier deux interfaces : l'interface fer pulvérulent/argilite et l'interface fer massif/argilite.Concernant le déroulement des expérimentations, il convient de noter en premier lieu que la fissure a probablement été circulante durant la totalité de l'essai, ce que tendent à confirmer les modélisations couplées chimie/transport avec le code HYTEC. Toutefois, aucune phase néoformée n'a été mise en évidence à son niveau. Par ailleurs, la survie bactérienne dans la cellule biotique a été confirmée au cours de l'expérimentation par le suivi régulier de la population et par une analyse de la diversité génétique à la fin des essais. Une nette diminution de la concentration en sulfates en sortie de cellule confirme l'activité des BSR.Les caractérisations solides en fin d'essais ont mis en évidence dans la zone de fer pulvérulent, avec et sans bactéries, de la magnétite et de la chukanovite, cette dernière étant localisée majoritairement à l'interface avec l'argilite. Une zone enrichie en fer (10 µm) a été identifié à l'interface dans l'argilite. La vitesse moyenne de corrosion du fer pulvérulent a été estimée à 0,2 µm/an (valeur basse). A l'interface fer massif/argilite, deux faciès ont été mis en évidence. Le premier faciès, identifié dans les deux dispositifs, est composé de chukanovite en couche interne et de sidérite en couche externe. L'extension de la perturbation de l'argilite est de l'ordre de 30 µm. Le second faciès, mis en évidence seulement en présence de bactéries, montre la présence de sulfure de fer (mackinawite) et d'avancées locales de corrosion. Enfin, les simulations HYTEC réalisées ont permis de mieux comprendre les mécanismes biogéochimiques observés expérimentalement, notamment l'effet du pH, et de mieux quantifier certains paramètres cinétiques clefs. / This study focuses on the interactions between two materials that may be introduced in a geological disposal of radioactive waste: metallic materials such as the high-level waste overpack, and clay materials such as the clay host rock. Indeed, the interactions between these two materials in such conditions could induce a change of their initial confinement properties. This work aimed at determining the influence of heterogeneities (technological gaps and fractures) and bacterial activities on these interactions, in terms of evolution of chemical and hydraulic properties of clayey materials. To this end, two percolation cells have been conducted during 13 months: the first one with two bacteria (SRB, IRB), the second one without bacteria. These experiments, carried out at 60°C, involved circulating synthetic water representative of the Tournemire pore water through iron powder and through Toarcian artificially cracked argillite from Tournemire. An iron rod was also placed into the argillite. Thus, solid characterizations (SEM, SEM/EDS, Raman, XRD, X-ray tomography) allowed the study of both interfaces: the iron powder/argillite interface and the iron rod/argillite interface.The water probably circulated into the crack during the entire test, which was confirmed by reactive transport modeling with the HYTEC reactive transport code. However, no secondary phase was identified in the crack. In addition, bacteria survival in the biotic cell was confirmed during the experiment by monitoring their population and by analyzing their genetic diversity at the end of the experiment. A strong decrease in sulfate concentration was measured in the output, which confirms the SRB activity.Solid characterization conducted at the end of the experiments have highlighted, with and without bacteria, the occurrence of magnetite and chukanovite in the iron powder, the latter being mainly located close to the argillite interface. In the argillite, a Fe-enriched zone (10 µm) was identified. The mean corrosion rate was estimated at 0.2 µm/y (lower bound). At the iron rod/argillite interface, two corrosion facies were observed. The first, identified in both cells, is mainly constituted of chukanovite in the inner layer and siderite in the outer layer. Extent of the argillite perturbation reaches about 30 µm. The second, only observed with bacteria, highlights the presence of iron sulfide precipitation (mackinawite) and localized corrosion patterns. Finally, HYTEC simulations have enabled us to better understand the observed biogeochemical processes, such as the pH effect, and to better quantify some key kinetic parameters.

This thesis provides a study of different aspects of the mechanical and vibrational properties of disordered and amorphous solids. Resorting to the theoretical framework of non-affine lattice dynamics the attention is focused on the analysis of disordered networks and lattices which serve as tractable model systems for real materials. Firstly, we discuss the static elastic response and the vibrational spectra of defective fcc crystals. The connection to different types of microstructural disorder in the form of bond-depletion and vacancies is described within the context of the inversion symmetry breaking of the local particle configurations. We identify the fluctuations of the local inversion symmetry breaking, which is directly linked to the non-affinity of the disordered solid, as the source of different scalings behaviours of the position of the boson peak. Furthermore, we describe the elastic heterogeneities occurring in a bond-depleted two- dimensional lattice with long-range interactions. The dependence of the concomitant correlations of the local elastic moduli are studied in detail in terms of the interaction range and the degree of disorder. An analytical scaling relation is derived for the radial part of the elastic correlations in the affine limit. Subsequently, we provide an argument for the change of the angular symmetry of the elastic correlation function which was observed in simulations and experiments on glasses and colloids, respectively. Moving to the dynamical behaviour of disordered solids, a framework is developed based on the kernel polynomial method for the approximate computation of the non- affine correlator of displacement fields which is the key requirement to describe the linear viscoelastic response of the system within the quasi-static non-affine formalism. This approach is then extended to the case of multicomponent polymer melts and validated against molecular dynamics simulations at low non-zero temperatures. We also consider the dynamical behaviour of metallic glasses in terms of its shear elasticity and viscosity. A theoretical scheme is suggested which links the repulsive strength of the interatomic potential to the viscoelasticity and fragility in metallic glasses in the quasi-affine limit.

Il est bien établi que les propriétés mécaniques et rhéologiques d'une large classe de matériaux vitreux amorphes met en jeu - contrairement aux dislocations dans les cristaux - des rearrangements structuraux localisés formant par un processus de cascade des bandes de cisaillements. Cette localisation de la déformation est observée dans divers systèmes vitreux ainsi que dans des simulations numériques. Cette réponse mécanique complexe reste mal comprise à une échelle microscopique et il n'est pas clair si l'écoulement plastique peut être associé à une origine structurale locale ou à des processus purement dynamiques.Dans cette thèse nous envisageons ces problématiques à l'aide de simulations atomiques athermales sur un système Lennard-Jones modèle. Nous calculons le tenseur élastique moyenné localement sur une échelle nanométrique. A cette échelle, le verre est assimilable à un matériau composite comprenant un échafaudage rigide et des zones fragiles. L'étude détaillée de la déformation plastique à différents taux de cisaillement met en évidence divers régimes d'écoulement. En dessous d'un taux de cisaillement critique dépendant de la taille du système, la réponse mécanique atteind une limite quasistatique (effets de taille fini, cascades d'événements plastiques, contrainte seuil) alors que pour des taux de cisaillement plus importants les propriétés rhéologiques sont fixées par le taux de cisaillement imposé. Dans ce régime nous mettons en évidence la croissance d'une longueur de coopérativité dynamique et discutons de sa dépendance avec le taux de cisaillements.

Η ακτινοθεραπεία αποτελεί μια απο τις βασικές θεραπείες κατά του καρκίνου. Για τη σωστή θεραπεία η ακριβής δόση ακτινοβολίας θα πρέπει να έχει υπολογιστεί. Ο υπολογισμός αυτός της δόσης ακτινοβολίας δυσκολεύει λόγω της διαφορετικής πυκνότητας ιστών του ανθρώπινου σώματος (πνεύμονες, αέρινες κοιλότητες, κόκαλα, κτλ). Η μελέτη της επίδρασης αυτών των ετερογενειών αλλά και μέθοδοι διόρθωσης της δόσης είναι μέρος αυτής της διδακτορικής διατριβής. Επίσης η θεραπεία όγκων στην περιοχή της κεφαλής με στεροτακτική ακτινοθεραπεία εξετάζεται στη διατριβή αυτη. Οι υπολογισμοί της δόσης στην στεροτακτική ακτινοθεραπεία απαιτούν μεγάλη ακρίβεια και στη διατριβή αυτή μελετάται η μέθοδος Μόντε Κάρλο για το υπολογισμό της δόσης / Radiotherapy is one of the principal modalities for used in treatment of malignant disease. For a correct therapy, the calculation of the delivered dose should be performed with great accuracy. This calcilation becomes difficult when heterogeneities of the human body are involved (lungs, air cavities, bones, etc). In this thesis, we studied the effect of such heterogeneities. We have also proposed methods of accuratley calculating the dose when heterogeneities are present inthe radiation field. Another aspect of radiotherapy is stereotactic radiotherapy for head and neck tumors. The demand of accurate dose calculation is even greater in this cases. W have studied methods of calculting the doses for stereotactic radiotherapy fields. The Monte Carlo methos proved to be the mostaccurate in this case.

Resumo: Este trabalho propõe investigar a abreviação em salas de bate-papo abertas (chats) da internet, frequentadas por escreventes que afirmam ter entre 15 e 20 anos. O conjunto do material é formado por duas "conversas" virtuais, com duração de 60 (sessenta) minutos cada uma. Reconhecida como uma das características do chamado "internetês", busca-se refletir que a abreviação não consiste em mero "corte de palavras", mas em recurso que apresenta regularidade e sistematicidade linguísticas. Esse processo é tomado como indício de (novo) gênero de discurso em emergência (BAKHTIN, 1997; MARCUSCHI, 2005). Com a finalidade de verificar os processos formadores de abreviaturas em bate-papos, ou seja, as regularidades linguísticas desse recurso, utiliza-se, como ferramenta de análise, a teoria de sílaba fonológica (SELKIRK, 1982). Acredita-se que, por meio do estudo da estrutura da sílaba, é possível avaliar que a escolha dos grafemas das abreviaturas é fundada na heterogeneidade da escrita (CORRÊA, 2004), visto que o escrevente pode se basear tanto em práticas orais/faladas quanto em práticas letradas/escritas para sua composição estrutural. Considera-se também que o funcionamento e a emergência de um modo de enunciação digital apenas são possíveis em meio a uma sociedade líquida (BAUMAN, 2004), caracterizada pela não permanência e fragilidade das relações entre os sujeitos. A consideração desses fatores parece ser indispensável para o entendimento de prática social ainda pouco (ou não) reconhecida / Abstract: This research's propose is to investigate the word's abbreviation which appears on the internet chats, used by writers who affirm to be around 15 or 20 years old. The material's amount is composed by two virtual "conversations" that spends around 60 minutes which one. Recognized as one characteristic of the well-known "internetês", we consider that this abbreviation is not a simple attempt to "cut the words", it is a resource that presents linguistics' regularity. This process is seeing as a clue of a (new) speech genres that is borning (BAKHTIN, 1997; MARCUSCHI, 2005). In order to verify the process that forms possible abbreviations used on chats, in other words, the linguistic regularities of that resource, it is used, as a tool of analyze, the theory of phonological syllable (SELKIRK, 1982). It is believed that, based on the syllable structure's study, it is possible to assess that the choice of the abbreviation's graphemes is based on the writing heterogeneity (CORRÊA, 2004), since the writer can be based on oral/spoken practices or on literate/written practices. It is also considered that the operation and the emergency of a digital way to enunciate is only possible in a liquid society (BAUMAN, 2004), that is not characterized because of the fragile and permanent relationship among subjects. Considering these factors it seems to be indispensable to comprehend the social practices which are rarely (or never) recognized / Orientador: Fabiana Cristina Komesu / Coorientador: Luciani Ester Tenani / Banca: Lourenço Chacon Jurado Filho / Banca: Fernanda Correa Silveira Galli / Mestre

Le calcul de la distribution de dose en Radiothérapie externe se fait en routine clinique à l'aide de Systèmes de Planification de Traitement (TPS) commerciaux. Les algorithmes de calcul de ces TPS ont énormément progressé ces dernières années. Cependant ils sont basés sur des approximations qui restent acceptables pour la plupart des conditions cliniques mais qui montrent leurs limites dans certains cas notamment avec des petites tailles de champ d'irradiation et/ou des faibles densités massiques dans un milieu. Or ces deux conditions sont pourtant réunies dans le cadre de la radiothérapie stéréotaxique des tumeurs bronchiques. Si quelques études ont été réalisées pour des densités massiques classiques de poumon, aucune n'a été réalisée pour des densités pulmonaires très faibles comme par exemple lorsque le patient est traité en inspiration profonde (" Deep Inspiration BreathHold ", i.e. DIBH). Mes travaux de recherche de thèse proposent une étude du calcul de dose pour différentes densités massiques et différentes tailles de champ en se basant sur un modèle Monte-Carlo (MC). La première étape modélise un accélérateur de type TrueBeam(r) (Varian, Palo Alto, CA) en utilisant les données du constructeur. Le modèle est construit à l'aide de la plateforme GATE basée sur la librairie Geant4. Les éléments principaux de la tête de l'appareil sont modélisés. Les espaces de phases (fichiers de particules) fournis par le constructeur au format " .IAEAphsp " sont situés en amont des mâchoires. Pour valider ce modèle, une série de champs simples (3x3 à 20x20 cm2) dans un fantôme d'eau sont implémentés pour des faisceaux de photons de 6X FF (" Flattening Filter "), 6X FFF, 10X FF et 10X FFF (" Flattening Filter Free "). Les résultats (profils, rendements de dose) sont comparés à des mesures de référence obtenues dans une cuve d'eau : respectivement 99% et 97% des points de dose des rendements et des profils respectent les critères de gamma index de 2%-2mm. Une fois le modèle validé, nous avons réalisé une série de simulations pour des champs de petites tailles (3x3 à 8x8 cm2) avec des fantômes hétérogènes de formes simples, pour lesquels la mesure reste accessible. Pour cette dernière, ont été insérés des films radio-chromiques dans des fantômes composés de plaques de PMMA et de deux types de liège de densité 0,12 et 0,24 correspondant respectivement aux poumons en DIBH et en respiration libre. Les résultats du modèle MC pour les quatre énergies ont été confrontés aux mesures expérimentales et aux algorithmes AAA et Acuros (Varian). De façon générale l'algorithme AAA surestime la dose au sein de l'hétérogénéité pulmonaire pour les petites tailles de champ et les faibles densités massiques. Par exemple, pour un champ de 3x3 cm2 et une densité de 0,12 au sein de l'hétérogénéité, une surestimation de la dose absorbée dans le poumon de 16% est mise en évidence pour l'algorithme AAA. Enfin, le modèle est utilisé pour trois cas non mesurables : un objet-test numérique cylindrique hétérogène, des données tomodensitométriques d'un patient en DIBH pour un champ fixe et en arc-thérapie en condition de stéréotaxie pulmonaire. Les résultats ont démontré respectivement pour les études sur TDM une surestimation de la dose dans la tumeur de 7% et 5,4% et dans le poumon de 14% et 9,6% par AAA. D'un point de vue clinique, cela se traduit par un sous-dosage du patient et donc un risque de récidive. / For clinical routine in external Radiotherapy, dose computation is achieved using commercial Treatment Planning Systems (TPS). Since ten years, TPS algorithms have been improved. However they include approximations that are acceptable in most of the clinical cases but they show their limits in some particular conditions for example in presence of small fields and/or low mass y media. And these two conditions are found in the context of stereotactic body radiation therapy (SBRT) of lung tumor. Some studies were published for standard lung densities but none for very low y like in lung during Deep Inspiration Breath Hold (DIBH). This work is a study of dose computation based on a Monte Carlo (MC) model, for different field sizes and mass densities. The first step was to model a TrueBeam(r) linac (Varian, Palo Alto, CA) using data furnished by the manufacturer. This model is built using the Geant4-based GATE platform. The main compounds of the linac head are modeled. Space phase files (i.e. particles files) are furnished by Varian in "IAEAphsp" format and are integrated to the model above the main jaws. To validate this model, a set of simple fields (from 3x3 to 20x20 cm2) in a water phantom is implemented for different photon energies: 6FF, 6FFF, 10FF and 10FFF (FFF = "Flattening Filter Free"). Percentage depth dose (PDD) and lateral profiles are compared to reference measurement in a water tank: respectively 99% and 97% of all the points of these curves passed the Gamma Index test (2% 2mm). Once this validation was completed, a set of simulation was achieved with small field sizes (3x3 to 8x8 cm2) for simple heterogeneous phantoms for which the measurement was achievable. For this purpose, radiochromic films were inserted in phantoms made of PMMA slabs and two types of cork. Cork densities were 0.12 and 0.24 that correspond respectively to lungs during DIBH and free breathing. Results of the MC model for four energies are compared to experimental measurements and to AAA and Acuros Varian's algorithms. AAA algorithm overestimates the dose inside the lung heterogeneity for small field sizes and low density. As an example in the case a 3x3 cm2 field, inside the heterogeneity of density 0.12 an over estimation of 16% in the lung is observed for AAA. The model is finally used for three non-measurable cases: a cylindrical digital reference object and computerized tomography data of a patient during DIBH with a static and stereotactic arc field. Results showed respectively for CT studies an overestimation of dose in the tumor of 7% and 5.4% and in the lungs of 14% and 9.6% by AAA. From a clinical point of view, this means under-dosing the patient and thus a risk of recurrence.