The understanding and study of electron transport in semiconductor systems has been the instigation behind the growth of semiconductor electronics industry which has enabled technological developments that are part of our everyday lives. However, most materials exhibit diffusive electron transport where electrons scatter off disorder (impurities, phonons, defects, etc.) inevitably present in the system, and lose their momentum. Advances in material science have led to the discovery of materials which are essentially disorder-free and exhibit exceptionally high mobilities, enabling transport physics beyond diffusive transport. In this work, we explore non-diffusive transport regimes, namely, the ballistic and hydrodynamic regimes in a high-mobility two-dimensional electron system in a GaAs quantum well in a GaAs/AlGaAs heterostructure. The hydrodynamic regime exhibits collective fluid-like behavior of electrons which leads to the formation of current vortices, attributable to the dominance of electron-electron interactions in this regime. The ballistic regime occurs at low temperatures, where electron-electron interactions are weak, constraining the electrons to scatter predominantly against the device boundaries. To study these non-diffusive regimes, we fabricate mesoscopic devices with multiple point contacts on the heterostructure, and perform variable-temperature (4.1 K to 40 K) zero-field nonlocal resistance measurements at various locations in the device to map the movement of electrons. The experiments, along with interpretation using kinetic simulations, demarcate hydrodynamic and ballistic regimes and establish the dominant role of electron-electron interactions in the hydrodynamic regime. To further understand the role of electron-electron interactions, we perform nonlocal resistance measurements in the presence of magnetic field in transverse magnetic focusing geometries under variable temperature (0.39 K to 36 K). Using our experimental results and insights from the kinetic simulations, we quantify electron-electron scattering length, while also highlighting the importance of electron-electron interactions even in ballistic transport. At a more fundamental level, we reveal the presence of current vortices in both hydrodynamic and surprisingly, ballistic regimes both in the presence and absence of magnetic field. We demonstrate that even the ballistic regime can manifest negative nonlocal resistances which should not be considered as the hallmark signature of hydrodynamic regime. The work sheds a new light on both hydrodynamic and ballistic transport in high-mobility solid-state systems, highlighting the similarities between these non-diffusive regimes and at the same time providing a way of effectively demarcating them using innovative device design, measurement schemes and one-to-one modeling. The similarities stem from total electron system momentum conservation in both the hydrodynamic and ballistic regimes. The work also presents a sensitive and precise experimental technique for measuring electron-electron scattering length, which is a fundamental quantity in solid-state physics. / Doctor of Philosophy / Electrons are the charged particles that are bound around the nuclei of atoms. But sometimes in a solid material electrons break free away from the nuclei and wander around. They are then the carriers of electric current ubiquitous in our daily lives as in our homes, and in our electronic devices such as smartphones and computers. Often an analogy is made between the flow of electric current in a material and the flow of water in a stream. However, the analogy does not hold well for most materials. In most materials the motion of electrons can be thought of as balls in a pinball machine - their movement hindered and randomized by collisions with the countless defects and impurities present in the material they travel through. However, recently scientists have been able to synthesize ultraclean materials, where electrons can indeed mimic the flow of water under the right conditions. In this aptly-named hydrodynamic regime, electrons predominantly interact with each other and that leads to the formation of current whirlpools or vortices similar to those forming in water. A telling signature of this regime is a negative electrical resistance appearing near the location of the vortex. When the interactions between electrons are weak, such as at very low temperatures, electrons move along straight-line trajectories until they hit and bounce off the device edges, similar to billiard balls. This low-temperature phenomenon is called ballistic transport. In this work we reveal that measurement of negative resistance and formation of current vortices are not unique to the hydrodynamic regime but can occur in the ballistic regime as well. It is indeed counterintuitive that electrons moving like billiards balls can behave similarly to electrons flowing like water. The similarities can be traced back to a fundamental physics conservation law active in both situations, namely momentum conservation. To experimentally realize the tests, we use a very high purity semiconductor material GaAs/AlGaAs and fabricate tiny devices on the material with a cutting-edge design, capable of precisely measuring resistance at various locations along the device to map the movement of electrons. The simulations of the novel physics indeed reveal current vortices of various sizes in the ballistic regime, in agreement with the experimental data showing negative resistance. In another experiment, we apply a magnetic field, making the electrons move in circular paths. If uninterrupted, electrons complete half circles and are collected through an opening in the device, giving resistance peaks in experiments. Due to electron-electron interactions, the electrons on their circular trajectory are interrupted by other electrons which leads to a decay in resistance peaks. This decay is utilized to measure the strength of electron-electron interactions. The work has both fundamental and applied implications. The existence of whirlpools shows that the electron momentum is not lost by collisions, and that in turn means that the conduction of electrical current in these regimes is inherently efficient. This opens up avenues for electronic devices which are faster, more functional and more power efficient than present electronic devices.

Cette thèse aborde le problème du transport diffusif dans les matériaux multicouches lorsque l'épaisseur des couches est comparable voire plus petit que le libre parcours moyen. Nous présentons un formalisme qui à la fois effectue une synthèse et permet d'aller au delà des divers modèles existants, dérive-diffusion, le modèle Valet-Fert, la méthode des flux ou encore le modèle de Fuchs-Sondheimer. Ce formalisme est applicable à deux types de structures: (i) la géométrie dite CPP (Current Perpendicular to Plane) où le courant moyen est perpendiculaire aux interfaces séparant les couches, et (ii) la géométrie dite CIP (Current In Plane) où le courant moyen est parallèle aux interfaces. Ce nouveau modèle de transport est bâti à partir d'une équation de Boltzmann où les collisions dans les couches et aux interfaces sont représentées par des intégrales de collision linéaires pouvant décrire aussi bien des réflexions spéculaires que des collisions aléatoires non nécessairement isotropes. La résolution de cette équation de Boltzmann pour déterminer les quantités macroscopiques locales d'intérêt se fait en trois étapes : pour chacune des couches, (1) la distribution locale des particules est séparée en deux « chiralités » caractérisés par le signe de la projection du vecteur vitesse de chaque particule le long de l'axe perpendiculaire aux interfaces ; (2) la description locale complète de la distribution angulaire des vitesses pour chaque chiralité est obtenue en développant sur une nouvelle base polynômes orthogonaux adaptée à l'existence de deux chiralités ; (3) pour effectuer la moyenne chirale sur la distribution angulaire des vitesses on définit une troncature minimale de ce développement adaptée aux quantités macroscopiques locales d'intérêt.L’étape (1) est nécessaire afin de pouvoir décrire correctement les collisions d'interfaces, l'étape (3) est usuelle mais l'ingrédient clef de ce formalisme est le point (2) qui seul permet de rendre cohérent les étapes (1) et (3) en présence d'interfaces. Pour la géométrie CPP, ce formalisme « Boltzmann chiral » permet d'unir les systèmes balistique et diffusif sous une même approche macroscopique. En présence de polarisation en spin, ce nouveau formalisme permet d'obtenir entre autre les résistances d'interfaces du modèle Valet-Fert en fonction des coefficients de transmission généralisés associés aux collisions d'interface. Pour les structures CIP, ce modèle permet d'obtenir des expressions analytiques pour les conductivités locales par couche (avec ou sans polarisation en spin) et de plus il rend le lien avec le transport CPP plus transparent. Ce formalisme n'étant pas propre au transport électrique, nous montrons sa versatilité sur une application au transport lumineux en revisitant le problème de Milne pour lequel nous retrouvons un résultat exact de façon beaucoup plus simple. Nous présentons pour terminer une méthode variationnelle fournissant une interprétation intéressante du modèle de Fuchs-Sondheimer. / This thesis addresses the problem of diffusive transport in multilayer systems when the layers thickness is of the order of or even smaller than the mean free path. We present a formalism which enables to synthetize and to go beyond various the standard models (drift-diffusion, Valet-Fert model, flux method or Fuchs-Sondheimer model). This formalism applies to two kinds of structures: (i) the so called CPP geometry (Current Perpendicular to Plane) where the mean transport current is perpendicular to the interfaces separating the layers, and (ii) the so called CIP (Current in Plane) geometry in which the mean transport current is parallel to interfaces. This new model of transport is build on the Boltzmann transport equation in which the scattering in the layer or at interfaces is represented by linear collision integrals that can describe specular and random scattering not necessarily isotropic. The resolution of this Boltzmann equation to obtain macroscopic quantities of interest is done in three steps for each layer: (1) the particle distribution is splitted into two “chiralities” characterized by the sign of the projection of the velocity vector of each particle along the axis perpendicular to interfaces; (2) the local description of the complete angular velocity distribution for each chirality is obtained by an expansion over a new orthogonal polynomial basis adapted to the existence of two chiralities; (3) to compute the chiral mean of the angular velocity distribution we define a minimal troncated expansion adapted to the local physical quantities of interest. Step (1) is necessary to describe correctly the interface scattering, step (3) is usual but the key ingredient of our formalism is step (2) which solely allows a coherent description of step (1) and (3) in the presence of interfaces. For spin polarized systems this novel formalism allows, among other things, to express the boundaries resistances of the Valet-Fert model in terms of generalized transmission coefficients associated to scattering at interfaces. For CIP structures, with this new approach we obtain explicit analytical expressions for the local conductivity of each layer (with or without spin polarisation) and we make the link with CPP transport more transparent. This novel formalism is not specific to electrical transport, to show its versatility we present an application to transport of light by revisiting the Milne problem for which we can recover certain exact result in a much simpler way. At last, we present a variational method which gives some interesting interpretation of the Fuchs-Sondheimer model.

Les marches aléatoires en milieux aléatoires constituent un modèle permettant de décrire des phénomènes de diffusion en milieux inhomogènes, possédant des propriétés de régularité à grande échelle. La thèse comportent 6 chapitres. Les trois premiers sont introductifs : le chapitre 1 est une courte introduction générale, le chapitre 2 donne une présentation des modèles considérés par la suite et le chapitre 3 un bref aperçu des résultats obtenus. Les preuves sont renvoyées aux chapitres 4, 5 et 6. Le contenu du chapitre 4 porte sur les théorèmes limites pour une marche aléatoire avec biais sur un arbre de Galton-Watson avec des feuilles dans un régime transient sous-balistique. Le chapitre 5 porte sur le comportement de la vitesse d'une marche aléatoire avec biais sur un amas de percolation quand le paramètre de percolation se rapproche de 1. Un développement asymptotique de la vitesse en fonction du paramètre de percolation est obtenu. On en déduit que la vitesse est croissante en $p=1$. Finalement le chapitre 6 porte sur des estimées de déviations modérées pour une marche aléatoire en milieu aléatoire unidimensionnel. / Random walks in random environments is a suitable model to describe diffusions in inhomogeneous media that have regularity properties on a macroscopic scale. The three first chapters are introductive : chapter 1 is a short general introduction, chapter 2 presents the models considered afterwards and chapter 3 is a brief overview of the results obtained. The proofs are postponed to the chapters4, 5 and 6.The content of chapter 4sheds light on limit theorems for a biased random walk on a Galton-Watson tree with leaves in the transient and sub-ballistic regime. Next, chapter 5 deals with the behaviour of the speed of a biased random walk on a percolation cluster as the percolation parameter goes to 1. An expansion of the speed in function of the percolation parameter is obtained. It can be deduced from this that the speed is increasing in $p=1$. Finally, chapter 6 tackles the problem of moderate deviations for random walks in random environments in dimension $1$.