abstract: Moore's law has been the most important driving force for the tremendous progress of semiconductor industry. With time the transistors which form the fundamental building block of any integrated circuit have been shrinking in size leading to smaller and faster electronic devices.As the devices scale down thermal effects and the short channel effects become the important deciding factors in determining transistor architecture.SOI (Silicon on Insulator) devices have been excellent alternative to planar MOSFET for ultimate CMOS scaling since they mitigate short channel effects. Hence as a part of thesis we tried to study the benefits of the SOI technology especially for lower technology nodes when the channel thickness reduces down to sub 10nm regime. This work tries to explore the effects of structural confinement due to reduced channel thickness on the electrostatic behavior of DG SOI MOSFET. DG SOI MOSFET form the Qfinfet which is an alternative to existing Finfet structure. Qfinfet was proposed and patented by the Finscale Inc for sub 10nm technology nodes. As part of MS Thesis we developed electrostatic simulator for DG SOI devices by implementing the self consistent full band Schrodinger Poisson solver. We used the Empirical Pseudopotential method in conjunction with supercell approach to solve the Schrodinger Equation. EPM was chosen because it has few empirical parameters which give us good accuracy for experimental results. Also EPM is computationally less expensive as compared to the atomistic methods like DFT(Density functional theory) and NEGF (Non-equilibrium Green's function). In our workwe considered two crystallographic orientations of Si,namely [100] and [110]. / Dissertation/Thesis / Masters Thesis Electrical Engineering 2016

La miniaturisation des transistors MOSFETs sur silicium massif présente de nombreux enjeux en raison de l'apparition de phénomènes parasites. Notamment, la réduction de la surface des dispositifs entraîne une dégradation de la variabilité de leurs caractéristiques électriques. La technologie planaire totalement désertée, appelée communément FDSOI (pour Fully Depleted Silicon on Insulator), permet d'améliorer le contrôle électrostatique de la grille sur le canal de conduction et par conséquent d'optimiser les performances. De plus, de par la présence d'un canal non dopé, il est possible de réduire efficacement la variabilité de la tension de seuil des transistors. Cela se traduit par un meilleur rendement et par une diminution de la tension minimale d'alimentation des circuits SRAM (pour Static Random Access Memory). Une étude détaillée de la variabilité intrinsèque à cette technologie a été réalisée durant ce travail de recherche, aussi bien sur la tension de seuil (VT) que sur le courant de drain à l'état passant (ISAT). De plus, le lien existant entre la fluctuation des caractéristiques électriques des transistors et des circuits SRAM a été expérimentalement analysé en détail. Une large partie de cette thèse est enfin dédiée à l'investigation de la source de variabilité spécifique à la technologie FDSOI : les fluctuations de l'épaisseur du film de silicium. Un modèle analytique a été développé durant cette thèse afin d'étudier l'influence des fluctuations locales de TSi sur la variabilité de la tension de seuil des transistors pour les nœuds technologiques 28 et 20nm, ainsi que sur un circuit SRAM de 200Mb. Ce modèle a également pour but de fournir des spécifications en termes d'uniformité σTsi et d'épaisseur moyenne µTsi du film de silicium pour les prochains nœuds technologiques. / The scaling of bulk MOSFETs transistors is facing various difficulties at the nanometer era. The variability of the electrical characteristics becomes a major challenge which increases as the device dimensions are scaled down. Fully-Depleted Silicon On Insulator (FDSOI) technology, developed as an alternative to bulk transistors, exhibits a better electrostatic immunity which enables higher performances. Moreover, the reduction of the Random Dopant Fluctuation allows excellent variability immunity for the FDSOI technology due to its undoped channel. It leads to a yield enhancement and a reduction of the minimum supply voltage of SRAM circuits. The variability has been analyzed deeply during this thesis in this technology, both on the threshold voltage (VT) and on the ON-state current (ISAT). The correlation between the electrical characteristics of MOSFETs devices (i.e., the threshold voltage and the standard deviation σVT) and SRAM cells (i.e., the SNM and σSNM) has been investigated thanks to an extensive experimental study and modeling. This purpose of this thesis is also to analyze the specific FDSOI variability source: silicon thickness fluctuations. An analytical model has been developed in order to quantify the impact of local TSi variations on the VT variability for 28 and 20nm technology nodes, as well as on a 200Mb SRAM array. This model also enables to evaluate the silicon thickness mean (µTsi) and standard deviation (σTsi) specifications for next technology nodes.