<div>Electronics have now become a part of our daily life and therefore the reliability of microelectronics cannot be overlooked. As the Moore's law era comes to an end, various new system-level innovations (e.g., 3D packaging, evolution of packaging material to molding compounds) with constant scaling of transistors have resulted in increasingly complicated integrated circuits (ICs) configurations. The reliability modeling of complex ICs is a nontrivial concern for a variety of reasons. For example, ever since 2004, self-heating effect (SHE) has become an important reliability concern for ICs. Currently, many groups have developed thermal predictive models for transistors, circuits, and systems. In order to describe SHE self-consistently, the modeling framework must account for correlated self-heating within the ICs. This multi-scales nature of the self-consistency problem is one of the difficult factors poses an important challenge to self-consistent modeling. In addition, coupling between different physical effects within IC further complicates the problem.</div><div><br></div><div>In this thesis, we discuss three challenges, and their solutions related to an IC's reliability issues. We (i) generalize the classical effective medium theory (EMT) to account for anisotropic, heterogeneous system; (ii) develop computationally efficient a physics-based thermal compact model for a packaged ICs to predict junction temperature in the transistor based on the EMT model, and image charge theory. Our thermal compact model bridges different length scales among the sources and rest of the system. Finally (iii) propose the modeling framework of electrical chip package interaction (CPI) due to charge transport within mold compounds by coupling moisture diffusion, electric distribution, and ions transport. The proposed modeling framework not only addresses the three major modeling challenges discussed earlier, but also provides deep and fundamental insights regarding the performance and reliability of modern ICs. </div>

Dans le contexte des moteurs électriques automobiles, les phénomènes vibratoires d'origine magnétique soulèvent une problématique relativement récente. L'objectif de cette thèse est la mise au point d'un modèle multi-physique pertinent d'une machine électrique afin de réaliser quelques études spécifiques, d'optimiser la machine et de prendre en compte la variabilité du comportement vibratoire. La modélisation numérique s'appuie totalement sur des formulations analytiques afin de bien maîtriser les différentes physiques. Des mesures expérimentales sur la machine permettent une confrontation avec le modèle numérique multi-physique et une validation des choix de modélisation. Dans ce contexte de modélisation multi-physique, un outil de couplage est développé entre les modèles 2D électromagnétique et 3D mécanique afin d'évaluer les comportements vibratoires d'origine électromagnétique de la machine. Une attention particulière a été portée à la prise en compte des forces magnétiques radiales et surtout tangentielles sur le stator de la machine électrique. Une méthode d'optimisation, basée sur le principe d'une surface de réponse dynamique, est appliquée sur le modèle électromagnétique afin d'améliorer des paramètres de conception de la machine. Les incertitudes liées à la conception sont souvent nombreuses, notamment dans le domaine vibratoire. A cet effet, la méthode MSP (Modal Stability Procedure) prenant en compte la variabilité des paramètres matériaux est proposée. La formulation MSP pour l'élément 3D hexaédrique est développée et appliquée au stator électrique afin d'évaluer la variabilité des fréquences propres et des fonctions de réponse en fréquence. / In the context of automotive electric motors, vibratory phenomena of magnetic origin arise relatively recent problems. the aim of this thesis is to develop a relevant multi-physic model of the electrical machine to perform some specific studies, to optimize the design of the machine and to take into account the variability of the vibration behavior. Numerical model is too based on analytical formulations in order to monitor the different physics. Experimental measurements on the machine are used to validate the numerical multi-physics model. In this context of multi-physic modeling, a coupling tool is developed between the 2D electromagnetic and 3D mechanical models, in order to evaluate the vibratory behavior of electomagnetic origin of the machine. A special attention was given in modeling of radial and especially tangential magnetic forces on the electric stator. An optimization method based on a dynamic response surface is applied to the electromagnetic model in order to improve the design of the machine. Uncertainties associated to the design are numerous, especially in the vibratory field. In this context, we proposed the MSP method (Modal Stability Procedure), which taking into account the variability of the material parameters. The MSP formulation for 3D hexahedral finite element is developed an applied to the electric staor, in order to evaluate the variability of the natural frequencies and the frequency response functions.

La réduction de la consommation des véhicules est un enjeu majeur pour les constructeurs. Parmi les stratégies employées, l’utilisation de lois de levée de soupape variables est en voie de démocratisation.Le premier objectif de cette thèse est de présenter la recherche et le pré dimensionnement d’un mécanisme offrant à la soupape trois levées distinctes : une pour la pleine charge, une pour les faibles charges et une permettant de désactiver un ou plusieurs cylindres pour les charges intermédiaires. Le pré dimensionnement comprend notamment les conditions d’usinabilité des cames, le calcul des ressorts de distribution et la contrainte de pression de contact entre les cames et les poussoirs.Le mode de fonctionnement de ce mécanisme est piloté par la position de deux actionneurs électromagnétiques choisis pour leur temps de réponse faible. Le deuxième objectif de la thèse est de proposer une optimisation afin de réduire leurs dimensions et faciliter leur intégration. Cette optimisation nécessite une modélisation multi-physique (magnétique, électrique, mécanique) du comportement de l’actionneur et permet de définir le profil de commande et les ressorts.Compte tenu du nombre de paramètres d’entrée et du temps de calcul d’une itération, l’optimisation présente un coût de calcul important. Nous présentons donc deux algorithmes de création de modèles de substitution par krigeage. Le premier permet d’approximer une fonction fine dans tout le domaine d’étude à un coût limité. Le second est inclus dans une boucle d’optimisation et n’utilise le modèle fin que lorsque l’approximation n’est pas suffisamment fiable. Ils permettent tout deux de réduire le coût d’optimisation. / Abstract : As the constraint on vehicle consumption tightens, variable valve lift strategies are used to increase engines efficiency. This thesis focuses on issues related to the design of a three-stage electromagnetic valve tappet. First of all, the tappet has to fit inside the engine and has to offer three different valve lifts: a full lift for high speed and torque requirements, a low lift for small loads and a lift allowing cylinder deactivation for medium loads. The tappet dimensioning includes, for example, cams feasibility, spring parameters and contact pressure between cams and tappets. Valve lift selection is made by two electromagnetic actuators in order to obtain a short transition time. Their size, however, hinders their integration. Hence, we realize an optimization aiming to reduce its volume and define the command profile and its return springs. This optimization requires a multi-physics model (electric, magnetic and dynamic) to simulate the actuator behavior. Due to the parameter number and the computational time needed per iteration, the optimization is expensive. Hence we suggest two metamodels algorithms to be used in the optimization. The first algorithm, off-line, is able to create iteratively a predictor precise in the entire domain studied with a lesser cost than Latin Hypercube Sample. The second, in-line, refines the predictor inside the optimization loops and uses it when the predicted error is small.

<p>  </p> <p>Electron emission plays a vital role in many modern technologies, from plasma medicine to heavy ion beams for fusion. An accurate theoretical model based upon the physics involved is critical to efficient operation of devices pushing the boundaries of complexity. The interactions between different electron emission mechanisms can severely alter device performance, especially when operating in extreme conditions. This dissertation studies electron emission from the perspectives of increasing geometric and physical mechanism complexities </p> <p>One half of this dissertation derives new relations for space-charge limited emission (SCLE) in non-planar geometries. SCLE is the maximum stable current that may be produced by electron emission before the electric field of the electrons themselves self-limits further emission. In planar devices, this is modeled by the well-established Child-Langmuir (CL) equation. The Langmuir-Blodgett (LB) equations remain the most commonly accepted theory for SCLE for cylindrical and spherical geometries after nearly a century; however, they suffer from being approximations based on a polynomial series expansion fit to a nonlinear differential equation. I derive exact, fully analytic equations for these geometries by using variational calculus to transform the differential equation into a new form that is fully and exactly solvable. This variational approach may be extended to any geometry and offers a full description of the electric field, velocity, and charge density profiles in the diode. </p> <p>SCLE is also an important mechanism for characterizing the operation of devices with an external magnetic field orthogonal to the electric field. This “crossed-field” problem decreases the limiting current as electrons travel longer, curved paths, effectively storing some charge in the gap (moving parallel to the emitter). At a critical magnetic field called the Hull cutoff, electron paths become so tightly curved that the circuit can no longer be completed, a condition called magnetic insulation. Crossed-field SCLE has been accurately modeled in planar devices by Lau and Christenson. Using the variational approach, I replicate their planar results and extend the calculation to cylindrical geometry, a common choice for magnetron devices. Further, I derive additional equations with simplified assumptions that, for the first time, provide an analytic description of experimental results below the Hull cutoff field. Following this I incorporate a series resistor: device resistance (or impedance) changes non-linearly with current and voltage, so I couple Ohm’s Law (OL) to all the models of crossed-field devices. For devices just below the Hull cutoff, I predict analytically and show in simulation novel bi-modal behavior, oscillating between magnetically insulated and non-insulated modes. With crossed-field device assessment, the variational calculus approach to space-charge may be used for numerous applications, including high power microwave sources, relativistic klystron devices, heavy ion beams, Hall thrusters, and plasma processing. </p> <p>The other half of this dissertation derives analytic theories to solve for emission current with three or more electron emission mechanisms simultaneously. In addition to the CL law, SCLE may also occur in neutral, non-vacuum diodes, modeled by the Mott-Gurney (MG) equation. These are the two limiting mechanisms I study; the other major modality of electron emission is direct electron production, the source of current in the device. Electrons are ejected when impelled by high temperature or electric field at the emission surface. These mechanisms are thermionic (or thermal) emission, modeled by the Richardson-Laue-Dushman (RLD) equation, and field emission, modeled by the Fowler-Nordheim (FN) equation, respectively. Additionally, just as I calculated the impedance of devices operating in a crossed-field configuration, all these models can be similarly coupled to OL. I derive models unifying FN, MG, and CL (with an extension linking OL, mentoring an undergraduate) and RLD, FN, and CL. These models are relevant for modern device design, especially as micro- and nano-scale devices seek to eliminate vacuum requirements and as space and military applications require higher temperature tolerances.</p> <p>While multi-physics models, like the ones described above, are important, the single-physics models (FN, RLD, MG, CL, OL) are still valid (and much easier to use) in their respective asymptotic limits. For example, a circuit behaves purely according to OL for very high resistances, according to MG for very high pressures, and so forth. Importantly, when devices operate in transition regions between these asymptotic limits, <em>none </em>of the asymptotic equations match the predictions of multi-physics models. Yet, intersections between the asymptotic equations are easily found, say for a certain set of voltage, gap distance, and pressure, CL=MG. Since both asymptotic equations give the same prediction, we may conclude that both must be inaccurate for those physical parameters! This gives rise to what I term “nexus theory:” solving two or more asymptotic equations simultaneously to rapidly and accurately predict sets of physical parameters at which multi-physics models (specifically, the physics leading to the “nexus point” parameters, points or curves at which nexus conditions are satisfied) are required for accurate device predictions. In fact, I show that multi-physics models are necessary within roughly one to two orders of magnitude from a nexus. In effect, nexus theory provides a simple, powerful tool to determine how complex a model is necessary for a particular device. Both nexus theory and multi-physics results in this dissertation have been successfully used to design devices to operate in specific transition regimes and identify the resulting device behavior.  </p>

Ce mémoire contribue à l’amélioration des performances d’une source de stockage hybride embarquée alimentant un véhicule électrique. La solution investiguée est composée de l’association de batteries Li-ion et de super condensateurs, dans le but d’obtenir, par rapport aux solutions classiques, un gain en masse et en durée de vie pour une certaine plage d’autonomie du véhicule. Notre objectif est de mettre à profit l’utilisation de nouvelles méthodes de gestion de la source hybride et de quantifier le gain obtenu. Un modèle multi-physique incluant les aspects électrique, thermique et vieillissement a été développé et intégré dans l’algorithme de gestion d’énergie afin d’évaluer la dégradation progressive des performances des éléments de stockage au cours des cycles de conduite selon la stratégie de gestion implantée. De nouvelles stratégies de gestion ayant pour objectif d’agir sur la durée de vie ont été évaluées. Leur impact sur les performances de la source en termes de masse, coût et durée de vie a pu être quantifié et montre bien que par une meilleure gestion des puissances, il est possible de mieux utiliser le stockeur hybride, ouvrant ainsi la voie à de nouvelles approches de gestion d’énergie pour ces systèmes. / This thesis contributes to the improvement of hybrid embedded source performances supplies an electric vehicle. The studied solution is composed of Li-ion batteries and supercapacitors hybridization, with an aim to achieve improved performances in terms of weight and lifetime over traditional solutions. Our main goal is to take the best advantage of new energy management strategies of the hybrid embedded source and quantify obtained improvements. A multi-physic model including electric, thermal and aging behaviors is developed and integrated into the algorithm of energy management in order to evaluate the gradual degradation of storage components performances during driving cycles and implemented control strategy. New energy management strategies intended to act on the lifetime of hybrid embedded source have been evaluated. Their impact on the performances of the source in terms of weight, cost and lifetime has been quantified and clearly shows that it is possible to make better use of hybrid embedded source thanks to a good power sharing, thus opening the way to new approaches of energy management for these systems.