Les sismologues ont révélé l'existence d'une anisotropie sismique complexe dans la graine terrestre. Cette thèse, en deux parties, présente deux approches différentes pour, d'une part, expliquer l'existence de cette anisotropie avec des modèles de graine numérique et, d'autre part, étudier l'effet des transformations de phase sur cette anisotropie.Dans un premier temps, nous avons construit un outil de simulation de la propagation des rais sismiques à travers une graine numérique, afin d'étudier la possibilité de reproduire numériquement les données sismologiques et de contraindre la dynamique et la minéralogie de la graine. Nous concluons qu'aucune structure du fer cubique ne permet de développer une anisotropie sismique significative. Seule la structure hexagonale compacte permet de le faire. Parmi les modèles géodynamiques quadripolaires étudiés, la croissance équatoriale préférentielle est le processus dynamique qui présente le meilleur accord avec les observations de dépendance à la profondeur et d'anisotropie polaire. L'ajout d'une stratification chimique permet d'amplifier d'environ 40% l'anisotropie globale mais augmente la dispersion des résidus sismiques, ce qui n'est pas conforme aux observations de dépendance à la profondeur. Enfin, une croissance dendritique (texture de solidification) est peu compatible avec une anisotropie principale Nord-Sud. Il apparaît clairement qu'aucun de ces modèles géodynamiques ne permet d'obtenir une amplitude d'anisotropie suffisante en utilisant les propriétés élastiques publiées pour le fer dans la graine. A l'issue de ce travail, de nombreux autres processus dynamiques restent encore à étudier, tels que les forces de Lorentz et la translation-convection thermique, processus bipolaire qui présente un fort intérêt pour la caractérisation de la composante hémisphérique de l'anisotropie sismique.Dans un deuxième temps, nous avons procédé à des expériences de transition de phase α-ε (fer cubique centré vers fer hexagonal compact) dans le fer pur. Nous avons simulé les textures résultantes à l'aide du mécanisme de Burgers. Nous avons confronté nos simulations aux résultats expérimentaux. Nous avons pu confirmer la validité du mécanisme de transformation de Burgers. Nous avons trouvé une forte sélection de variant sous l'effet de la contrainte non-hydrostatique dans le sens direct α→ε, produisant de fortes textures. Dans le sens inverse ε→α, on observe des textures finales très faibles, voire aléatoires. Appliqués avec prudence au cas de la graine terrestre, nos résultats indiquent que la transformation de phase du fer hexagonal à cubique ne peut pas expliquer la forte anisotropie observée. Nous avons aussi noté une mémoire de texture partielle, déjà documentée pour d'autres métaux de transition. Enfin, nous avons procédé à des tests préliminaires en cellule diamant en rotation. Les résultats de ces expériences de cisaillement ont également montré un très bon accord avec les simulations fondées sur le mécanisme de Burgers. En revanche, nous avons pu constater de grands écarts entre nos résultats expérimentaux et les simulations en cisaillement présentées dans la littérature [Levitas et al., 2010], notamment en termes de gradient de contraintes. / Seismologists revealed the existence of a complex seismic anisotropy in the Earth's inner core. In this thesis we took two different approaches in order to characterize the anisotropy: in a first part, we tried to explain this anisotropy using formation models of inner core and, in a second part, we considered the impact of eventual phase transitions on anisotropy.Firstly, we simulate the propagation of seismic rays through a numerically grown inner core, in order to measure the seismic anisotropy to compare with actual observations and to constrain the dynamics and mineralogy of the inner core. We conclude that no cubic structure of iron may produce a significant global anisotropy. Only the hexagonal compact phase of iron may produce a measurable signal. Considering a panel of quadripolar geodynamical models, we observe that simple preferential equatorial growth is the most consistent with seismological observations of a polar anisotropy and depth dependence of seismic residuals. Chemical stratification amplifies the global anisotropy by about 40%, but at the same time increases the scatter of residuals in a way that is poorly compatible with the observed depth dependence. Finally the addition of dendritic growth (solidification texture) prohibits the emergence of a first order North-South anisotropy. Independently of the geodynamical model it appears clearly that none of these geodynamical models the development of an anisotropy consistent with observations when using the published elastic properties of iron at inner core conditions. Following this work, several geodynamical models remain to be studied such as magnetic forcing (Lorentz forces). Models involving translation-thermal convection are of great interest as they may account for the hemispherical component of the seismic anisotropy.Secondly, we performed experiments on the α-ε (cubic centered to hexagonal compact iron) phase transition in pure iron. We compare our experimental texture results with our simulations of the transformation considering Burgers mechanism. We confirm here the Burgers atomic path as the mechanism activated during the α-ε transformation in iron. We find direct evidence of a strong variant selection controlled by non-hydrostatic stresses in the diamond anvil cell during the forward α→ε transformation, producing strong textures. The opposite will occur with the reverse ε→α phase transition where an almost complete randomization is to be expected. Our observations can be applied with some caution to the Earth's inner core and show that the strong seismic anisotropy in the inner core may not be explained by the occurrence of a hexagonal to cubic phase transition in the inner core. A limited texture memory effect was brought into light, already documented for other transition metals. At last, we performed a preliminary study on the effect of shear on the α-ε phase transition in pure iron using a rotational cell. Again the α-ε phase transformation in iron can be modelized by the Burgers mechanism. We find that simulations in the literature [Levitas et al., 2010] fail to reproduce our experimental results, particularly in terms of stress field.

Continental rifting is an important geodynamic process during which the Earth's outer-most rigid shell undergoes continuous stretching resulting in continental break-up and theformation of new oceanic basins. The East African Rift System, which has two continentalsegments comprising largely of the East African Rift (EAR) to the West and the easternmostsegment Madagascar, is the largest narrow rift on Earth. However, the driving mechanismsof continental rifting remain poorly understood due to a lack of numerical infrastructure tosimulate rifting, the lack of knowledge of the underlying mantle dynamics, and poor knowl-edge of mantle rheology. Here, we use state-of-art computational modeling of the upper660 km of the Earth to: 1) provide a better understanding of mantle flow patterns and themantle rheology beneath Madagascar, 2) to elucidate the main driving forces of observedpresent-day∼E-W opening in the EAR, and 3) to investigate the role of multiple plumesor a superplume in driving surface deformation in the EAR. In chapter 1, we simulate EdgeDriven convection (EDC), constrained by a lithospheric thickness model beneath Madagas-car. The mantle flow associated with the EDC is used to calculate induced olivine aggregates'Lattice Preferred Orientation (LPO), known as seismic anisotropy. The predicted LPO isthen used to calculate synthetic seismic anisotropy, which were compared with observationsacross the island. Through a series of comparisons, we found that asthenospheric flow result-ing from undulations in lithospheric thickness variations is the dominant source of the seismicanisotropy, but fossilized structures from an ancient shear zone may play a role in southern Madagascar. Our results suggest that the rheological conditions needed for the formationof seismic anisotropy, dislocation creep, dominates the upper asthenosphere beneath Mada-gascar and likely other continental regions. In chapter 2, we use a 3D numerical model ofthe lithosphere-asthenosphere system to simulate instantaneous lithospheric deformation inthe EAR and surroundings. We test the hypothesis that the∼E-W extension of the EAR isdriven by large scale forces arising from topography and internal density gradients, known aslithospheric buoyancy forces. We calculate surface deformation solely driven by lithosphericbuoyancy forces and compare them with surface velocity observations. The lithosphericbuoyancy forces are implemented by imposing observed topography at the model surfaceand lateral density variations in the crust and mantle down to a compensation depth of 100km. Our results indicate that the large-scale∼E-W extension across East Africa is driven bylithospheric buoyancy forces, but not along-rift surface motions in deforming zones. In chap-ter 3, we test the hypothesis that the anomalous northward rift-parallel deformation observedin the deforming zones of the EAR is driven by viscous coupling between the lithosphereand deep upwelling mantle material, known as a superplume, flowing northward. We testtwo end-member plume models including a multiple plumes model simulated using high res-olution shear wave tomography-derived thermal anomaly and a superplume model (Africansuperplume) simulated by imposing a northward mantle-wind on the multiple plumes model.Our results suggest that the horizontal tractions from northward mantle flow associated withthe African Superplume is needed to explain observations of rift-parallel surface motions indeforming zones from GNSS/GPS data and northward oriented seismic anisotropy beneaththe EAR. Overall, this work yields a better understanding of the geodynamics of Africa. / Doctor of Philosophy / Continental rifting is an important geodynamic process during which the Earth's outer-most rigid shell undergoes continuous stretching resulting in continental break-up and theformation of new oceanic basins. The East African Rift System, which has two continentalsegments comprising largely of the East African Rift (EAR) to the West and the easternmostsegment Madagascar, is the largest narrow rift on Earth. However, the driving mechanismsof continental rifting remain poorly understood due to a lack of numerical infrastructure tosimulate rifting, the lack of knowledge of the underlying mantle dynamics, and poor knowl-edge of mantle rheology. Here, we use state-of-art computational modeling of the upper660 km of the Earth to: 1) provide a better understanding of mantle flow patterns and themantle rheology beneath Madagascar, 2) to elucidate the main driving forces of observedpresent-day∼E-W opening in the EAR, and 3) to investigate the role of multiple plumesor a superplume in driving surface deformation in the EAR. In chapter 1, we simulate EdgeDriven convection (EDC), constrained by a lithospheric thickness model beneath Madagas-car. The mantle flow associated with the EDC is used to calculate induced olivine aggregates'Lattice Preferred Orientation (LPO), known as seismic anisotropy. The predicted LPO isthen used to calculate synthetic seismic anisotropy, which were compared with observationsacross the island. Through a series of comparisons, we found that asthenospheric flow result-ing from undulations in lithospheric thickness variations is the dominant source of the seismicanisotropy, but fossilized structures from an ancient shear zone may play a role in southern Madagascar. Our results suggest that the rheological conditions needed for the formationof seismic anisotropy, dislocation creep, dominates the upper asthenosphere beneath Mada-gascar and likely other continental regions. In chapter 2, we use a 3D numerical model ofthe lithosphere-asthenosphere system to simulate instantaneous lithospheric deformation inthe EAR and surroundings. We test the hypothesis that the∼E-W extension of the EAR isdriven by large scale forces arising from topography and internal density gradients, known aslithospheric buoyancy forces. We calculate surface deformation solely driven by lithosphericbuoyancy forces and compare them with surface velocity observations. The lithosphericbuoyancy forces are implemented by imposing observed topography at the model surfaceand lateral density variations in the crust and mantle down to a compensation depth of 100km. Our results indicate that the large-scale∼E-W extension across East Africa is driven bylithospheric buoyancy forces, but not along-rift surface motions in deforming zones. In chap-ter 3, we test the hypothesis that the anomalous northward rift-parallel deformation observedin the deforming zones of the EAR is driven by viscous coupling between the lithosphereand deep upwelling mantle material, known as a superplume, flowing northward. We testtwo end-member plume models including a multiple plumes model simulated using high res-olution shear wave tomography-derived thermal anomaly and a superplume model (Africansuperplume) simulated by imposing a northward mantle-wind on the multiple plumes model.Our results suggest that the horizontal tractions from northward mantle flow associated withthe African Superplume is needed to explain observations of rift-parallel surface motions indeforming zones from GNSS/GPS data and northward oriented seismic anisotropy beneaththe EAR. Overall, this work yields a better understanding of the geodynamics of Africa.