In this work, two new spintronic memory cell structures are proposed. The first cell uses the diffusion of polarized spins into ferromagnets with perpendicular anisotropy to tilt their magnetization followed by their dipolar coupling to a fixed magnet (Bhowmik et al., 2014). The possibility of setting the magnetization to both stable magnetization states in a controlled manner using a similar concept remains unknown, but the proposed structure poses to be a solution to this difficulty. The second cell proposed takes advantage of the multiple stable magnetic states that exist in ferromagnets with configurational anisotropy and also uses spin torques to manipulate its magnetization. It utilizes a square-shaped ferromagnet whose stable magnetization has preferred directions along the diagonals of the square, giving four stable magnetic states allowing to use the structure as a multi-bit memory cell. Both devices use spin currents generated in heavy metals by the Spin Hall effect present in these materials. Among the advantages of the structures proposed are their inherent non-volatility and the fact that there is no need for applying external magnetic fields during their operation, which drastically improves the energy efficiency of the devices. Computational simulations using the Object Oriented Micromagnetic Framework (OOMMF) software package were performed to study the dynamics of the magnetization process in both structures and predict their behavior. Besides, we fabricated a 4-terminal memory cell with configurational anisotropy similar to the device proposed, and found four stable resistive states on the structure, proving the feasibility of this technology for implementation of high-density, non-volatile memory cells.

<p>A wide range of unstable current driven MHD modes is present in the re- versed τeld pinch (RFP) conτguration. An ideally conducting wall facing the plasma can stabilize the ideal MHD modes. In the presence of a resistive wall characterized by the wall time τw, fast mode rotation with the frequency exceeding the inverse wall time gives stabilization for resistive MHD modes. The ideal MHD modes in the RFP are non-rotating modes and can not be stabilized by the resistive wall. Instead they are converted into resistive wall modes (RWM) growing with a growth rate proportional to the inverse of the wall time τw. EXTRAP T2R is an RFP device equipped with a thin resistive wall having the wall time shorter than the plasma pulse duration τw < τp. This feature allows the study of non-resonant non-rotating resistive wall modes. Resistive wall modes dynamics has been studied in EXTRAP T2R . RWM growth rates has been measured and compared with linear MHD stability calculations. Quantitative agreement is observed. In the case τw < τp the RWM can cause discharge degradation and should be stabilized. Active feedback is the way to stabilize the RWM in the RFP. An intelligent shell scheme is one possible feedback scenario. An active feed- back system including a set of sensors and discrete active coils is installed in EXTRAP T2R. The intelligent shell tries to keep the magnetic flux zero at the positions of the sensor. The analog PID controller for the intelligent shell feedback scheme has been studied. A model of the active control system was developed and comparison with the experimental results showed good agree- ment. Encouraging experimental results on the active feedback stabilization of multiple RWMs in the RFP plasmas were obtained.</p>

A wide range of unstable current driven MHD modes is present in the re- versed τeld pinch (RFP) conτguration. An ideally conducting wall facing the plasma can stabilize the ideal MHD modes. In the presence of a resistive wall characterized by the wall time τw, fast mode rotation with the frequency exceeding the inverse wall time gives stabilization for resistive MHD modes. The ideal MHD modes in the RFP are non-rotating modes and can not be stabilized by the resistive wall. Instead they are converted into resistive wall modes (RWM) growing with a growth rate proportional to the inverse of the wall time τw. EXTRAP T2R is an RFP device equipped with a thin resistive wall having the wall time shorter than the plasma pulse duration τw &lt; τp. This feature allows the study of non-resonant non-rotating resistive wall modes. Resistive wall modes dynamics has been studied in EXTRAP T2R . RWM growth rates has been measured and compared with linear MHD stability calculations. Quantitative agreement is observed. In the case τw &lt; τp the RWM can cause discharge degradation and should be stabilized. Active feedback is the way to stabilize the RWM in the RFP. An intelligent shell scheme is one possible feedback scenario. An active feed- back system including a set of sensors and discrete active coils is installed in EXTRAP T2R. The intelligent shell tries to keep the magnetic flux zero at the positions of the sensor. The analog PID controller for the intelligent shell feedback scheme has been studied. A model of the active control system was developed and comparison with the experimental results showed good agree- ment. Encouraging experimental results on the active feedback stabilization of multiple RWMs in the RFP plasmas were obtained.

We study current-induced domain wall dynamics in a thin ferromagnetic nanowire. We derive the effective equations of domain wall motion, which depend on the wire geometry and material parameters. We describe the procedure to determine these parameters by all-electric measurements of the time-dependent voltage induced by the domain wall motion. We provide an analytical expression for the time variation of this voltage. Furthermore, we show that the measurement of the proposed effects is within reach with current experimental techniques. We also show that a certain resonant time-dependent current moving a domain wall can significantly reduce the Joule heating in the wire, and thus it can lead to a novel proposal for the most energy efficient memory devices. We discuss how Gilbert damping, non-adiabatic spin transfer torque, and the presence of Dzyaloshinskii-Moriya interaction can effect this power optimization. Furthermore, we propose a new nanodot magnetic device. We derive a specific time-dependent current that is needed to switch the magnetization of the nanodot the most efficiently.

Avec les avancées technologiques et la miniaturisation, le réseau d'interconnexions est devenu de plus en plus dense et complexe. Pour les domaines qui utilisent des applications à forts courants, comme l'automobile, les très fortes densités de courant dans les lignes métalliques peuvent conduire à des phénomènes comme l'électromigration, le voltage drop ou encore les surcharges électriques. La conception des circuits doit donc être réalisée en prenant en compte ces contraintes et en adaptant la largeur des lignes aux courants. Ce travail de thèse a eu comme objectif de développer des solutions pour la prise en compte des contraintes en courant lors de la phase de routage de blocs analogiques fort courants. Après une présentation des phénomènes impliqués et de l'état de l'art, une approche algorithmique pour l'aide au routage est introduite. Une méthode de caractérisation du courant est définie, un algorithme exhaustif de routage est présenté, puis utilisé pour effectuer des recherches de critères d'une bonne topologie. Deux algorithmes sont ensuite étudiés et comparés, un algorithme glouton, servant de référence, et un « Divide & Conquer » original. Il présente une amélioration d'environ 10% pour l'aire, et presque 27% en temps CPU par rapport à l'algorithme glouton. La section suivante s'intéresse à la correction du current crowding, avec une méthode basée sur un ensemble de modèles mathématiques. Enfin, un flot basé sur les solutions développées durant la thèse est présenté et validé. / In deep submicron VLSI circuits, excessive current density in interconnects is a major concern for analog high current application. If current over maximum density is not effectively mitigated, this can lead to phenomena like electromigration, voltage drop and electrical overload. It is a hot topic of interest in modern circuits due to the decrease of metal track sizes while high currents are necessary in automotive or mobile applications. This thesis had as goal to develop solutions for the consideration of the constraints in the current phase of routing analog blocks strong currents. After a presentation of the phenomena and the state of the art, an algorithmic approach to current driven net generation is introduced. A method to characterize the current is defined. Then an exhaustive routing algorithm is presented and used to search criteria for a good topology. Next, two algorithms are studied and compared, first a greedy algorithm, used as a reference, and a "Divide & Conquer" original algorithm. It shows results improved on average by about 10% for area and almost 27% for CPU time compared with existing solution. The next section focuses on current crowding correction, with a method based on a set of mathematical models. Finally, a conception flow based on the developed solutions is introduced and validated.