Semiconductor manufacturing is the workhorse for a wide range of industries. It lies at the heart of consumer electronics, telecommunication equipment and medical devices. Most semiconductor electronics are made from Silicon, and are fabricated using CMOS technology. The versatility of semiconductor electronics stems from the ever-reducing cost of integrating more computing and memory functions on chip. The small cost for adding extra functions has been maintained in the past 50 years through transistor scaling. Transistor scaling focuses on shrinking the size of transistors integrated on chip. This reduction in transistor size, while keeping the overall cost of the chip fixed allowed us to reduce the cost per function with scaling, and is what is celebrated as Moore’s law. Scaling has been working gracefully up to the last decade, where the exponential rise in manufacturing cost and diminishing gains of scaling on device performance reduce its economic benefit. To revive the cost reduction trend, different techniques were proposed such as augmenting CMOS manufacturing with new materials (Beyond-CMOS), 3D integration, and integrating more non-transistor elements on-chip (More than Moore). In this work, we focus on the efficient implementation of several circuit functions using an allotropy of carbon known as graphene. Graphene, a single layer of carbon atoms arranged in a hexagonal lattice, has unique electronic properties that has been taken the solid-state electronics community by a storm since its first experimental conception in 2004. Despite its promising electronic properties, namely the very high charge-carrier mobility and reduced scattering by impurities, graphene circuits has been held back by a plethora of nonidealities and technological roadblocks that hamper its use in traditional transistor-based circuits. In this work, we attempt to leverage the unique physical properties of graphene to implement non von-Neumann neuromorphic computing architectures, low-loss diodes and evaluate the behavior of diffusive-transport graphene couplers. We focus on the the design, fabrication and characterization of graphene devices in the presence of the current performance-limiting technological nonidealities in heterogeneous graphene-CMOS systems. We present the design, fabrication and characterization of all-graphene resistive data converters devices and diodes, discussing their performance and application as building elements of all-graphene brain-inspired computing architectures. We evaluate the performance of graphene couplers operating in the diffusive transport regime, which serve as a method to analyze the cross-coupling between adjacent graphene interconnects. We also discuss the current technological limitations hampering the performance of graphene devices, and the roles of different processing non-idealities on the characteristics of graphene devices.

In the era of Internet of Things and Big Data, unconventional techniques are rising to accommodate the large size of data and the resource constraints. New computing structures are advancing based on non-volatile memory technologies and different processing paradigms. Additionally, the intrinsic resiliency of current applications leads to the development of creative techniques in computations. In those applications, approximate computing provides a perfect fit to optimize the energy efficiency while compromising on the accuracy. In this work, we build probabilistic adders based on stochastic memristor. Probabilistic adders are analyzed with respect of the stochastic behavior of the underlying memristors. Multiple adder implementations are investigated and compared. The memristive probabilistic adder provides a different approach from the typical approximate CMOS adders. Furthermore, it allows for a high area saving and design exibility between the performance and power saving. To reach a similar performance level as approximate CMOS adders, the memristive adder achieves 60% of power saving. An image-compression application is investigated using the memristive probabilistic adders with the performance and the energy trade-off.

There are many semiconductors with nominally superior electronic properties compared to silicon. However, silicon became the material of choice for MOSFETs due to its robust native oxide. With Moore's observation as a guiding principle, the semiconductor industry has come a long way in scaling the silicon MOSFETs to smaller dimensions every generation with engineering ingenuity and technological innovation. As per the 2012 International Technology Roadmap for Semiconductors (ITRS), the MOSFET is expected to be scaled to near 6 nm gate length by 2025. However, materials, design and fabrication capabilities aside, basic physical considerations such as source to drain quantum mechanical tunneling, channel to gate tunneling, and thermionic emission over the channel barrier suggest an end to the roadmap for CMOS is on the horizon. The semiconductor industry is already aggressively looking for the next switch which can replace the silicon FET in the long term. My Ph.D. research is part of the quest for the next switch. The promises of process compatibility with existing CMOS technologies, fast carriers with high mobilities, and symmetric conduction and valence bands have led to graphene being considered as a possible alternative to silicon. This work looks at three devices based on graphene using first principles atomistic transport simulations and compact models capturing essential physics: the large-area graphene RF FET, the Bilayer pseudoSpin FET, and the double electron layer resonant tunneling transistor. The characteristics and performance of each device is explored with a combination of SPICE simulations and atomistic quasi static transport simulations. The BiSFET device was found to be a promising alternative to CMOS due to extremely low power dissipation. Finally, I have presented formalism for efficient simulation of time dependent transport in graphene for beyond quasi static performance analysis of the graphene based devices explored in this work. / text

Indiana University-Purdue University Indianapolis (IUPUI) / In this dissertation, a beyond-CMOS approach to Static Random Access Memory (SRAM) design is investigated using exploratory transistors including Tunnel Field Effect Transistor (TFET), Carbon Nanotube Field Effect Transistor (CNFET) and Graphene NanoRibbon Field Effect Transistor (GNRFET). A Figure-of-Merit (FOM) based comparison of 6-transistor (6T) and a modified 8-transistor (8T) single-port SRAMs designed using exploratory devices, and contemporary devices such as a FinFET and a CMOS process, highlighted the performance benefits of GNRFETs and power benefits of TFETs. The results obtained from the this work show that GNRFET-based SRAM have very high performance with a worst-case memory access time of 27.7 ps for a 16x4-bit 4-word array of 256-bitcells. CNFET-based SRAM bitcell consume the lowest average power during read/write simulations at 3.84 uW, while TFET-based SRAM bitcell show the best overall average and static power consumption at 4.79 uW and 57.8 pW respectively. A comparison of these exploratory devices with FinFET and planar CMOS showed that FinFET-based SRAM bitcell consumed the lowest static power at 39.8 pW and CMOS-based SRAM had the best read, write and hold static noise margins at 201 mV, 438 mV and 413 mV respectively. Further, the modification of 8T-SRAMs via dual wordlines for individually controlling read and write operations for uni-directional transistors TFET and CNFET show improvement in read static noise margin (RSNM). In dual wordline CNFET 8T-SRAM, an RSNM improvement of approximately 23.6x from 6 mV to 142 mV was observed by suppressing the read wordline (RWL) from a nominal supply of 0.71 V down to 0.61 V. In dual wordline TFET 8T-SRAM, an RSNM improvement of approximately 16.2x from 5 mV to 81 mV was observed by suppressing the RWL from a nominal supply of 0.6 V down to 0.3 V. Next, the dissertation explores whether the robustness of SRAM arrays can be improved. Specifically, the robustness related to noise margin during the write operation was investigated by implementing a negative bitline (NBL) voltage scheme. NBL improves the write static noise margin (WSNM) of the SRAM bitcells in the row of the array to which the data is written during a write operation. However, this may cause degraded hold static noise margin (HSNM) of un-accessed cells in the array. Applying a negative wordline voltage (NWL) on un-accessed cells during NBL shows that the NWL can counter the degraded HSNM of un-accessed cells due to NBL. The scheme, titled as NBLWL, also allows the supply of a lower NBL, resulting in higher WSNM and write-ability benefits of accessed row. By applying a complementary negative wordline voltage to counter the half-select condition in columns, the WSNM of cells in accessed rows was boosted by 10.9% when compared to a work where no negative bitline was applied. In addition, the HSNM of un-accessed cells remain the same as in the case where no negative bitline was implemented. Essentially, a 10.9% boost in WSNM without any degradation of HSNM in un-accessed cells is observed. The dissertation also focuses on the impact of process-related variations in SRAM arrays to correlate and characterize silicon data to simulation data. This can help designers remove pessimistic margins that are placed on critical signals to account for expected process variation. Removing these pessimistic margins on critical data paths that dictate the memory access time results in performance benefits for the SRAM array. This is achieved via an in-situ silicon monitor titled SRAM process and ageing sensor (SPAS), which can be used for silicon and ageing characterization, and silicon debug. The SPAS scheme is based on a process variation tolerant technique called RAZOR that compares the data arriving on the output of the sense amplifiers during the read operation. This scheme can estimate the impact of process variation and ageing induced slow-down on critical path during read operation of an array with high accuracy. The estimation accuracy in a commercially available 65nm CMOS technology for a 16x16 array at TT, and global SS and FF corners at nominal supply and testing temperature were found to be 99.2%, 94.9% and 96.5% respectively. Finally, redundant columns, an architectural-level scheme for tolerating failing SRAM bitcells in arrays without compromising performance and yield, is studied. Redundant columns are extra columns that are programmed when bitcells in the regular columns of an array are slower or have higher leakage than expected post-silicon. The regular columns are often permanently disabled and remain unused for the chip lifetime once redundant columns are enabled. In the SRRC scheme proposed in this thesis, the regular columns are only temporarily disabled, and re-used at a later time in chip life cycle once the previously awakened redundant columns become slower than the disabled regular columns. Essentially, the scheme can identify and temporarily disable the slowest column in an array until other mitigating factors slow down active columns. This allows the array to operate at a memory access time closer to the target access time regardless of other mitigating factors slowing down bitcells in arrays during chip life cycle. An approximate 76.4% reduction in memory access time was observed from a 16x16 array from simulations in a commercially available 65nm CMOS technology with respect to a work where no redundancy was employed.

Le développement de la technologie CMOS a déclenché une révolution dans la production IC. Chaque nouvelle génération technologique, par la mise à l’échelle des dimensions, a entraîné une accélération de son fonctionnement et une réduction de sa consommation. Cependant, la miniaturisation sera contrainte par les limites physiques fondamentales régissant la commutation des dispositifs CMOS dès lors que la technologie atteint des dimensions inférieures à 10 nm. Les chercheurs veulent trouver d'autres moyens de dépasser ces limites physiques. La spintronique est l’un des concepts les plus prometteurs pour de nouvelles applications de circuits intégrés sans courant de charge. La STT-MRAM est l’une des technologies de mémoires fondée sur la spintronique qui entre avec succès en phase de production de masse. Les opérateurs logiques à base de spin, associés aux métiers, doivent être maintenant étudiés. Notre recherche porte sur le domaine des transistors à effet de champ de spin (spin-FET), l'un des dispositifs logiques fondamentaux à base de spin. Le mécanisme principal pour réaliser un spin-FET consiste à contrôler le spin des électrons, ce qui permet d'atteindre l'objectif de réduction de puissance. De plus, en tant que dispositifs à spin, les spin-FET peuvent facilement être combinés à des éléments de stockage magnétique, tels que la jonction tunnel magnétique (MTJ), pour développer une architecture à «logique non volatile» offrant des performances de hautes vitesses et de faible consommation. La thèse présentée ici consiste à développer un modèle compact de spin-FET et à explorer les possibilités de son application pour la conception logique et la simulation logique non volatile. Tout d'abord, nous avons proposé un modèle à géométrie non locale pour spin-FET afin de décrire les comportements des électrons, tels que l'injection et la détection de spin, le décalage de phase d'angle de spin induit par l'interaction spin-orbite. Nous avons programmé un modèle spin-FET non local à l'aide du langage Verilog-A et l'avons validé en comparant la simulation aux résultats expérimentaux. Afin de développer un modèle électrique pour la conception et la simulation de circuits, nous avons proposé un modèle de géométrie local pour spin-FET basé sur le modèle non-local spin-FET. Le modèle de spin-FET local étudié peut être utilisé pour la conception logique et la simulation transitoire à l'aide d'outil de conception de circuit. Deuxièmement, nous avons proposé un modèle spin-FET à plusieurs grilles en améliorant le modèle susmentionné. Afin d'améliorer les performances du spin-FET, nous avons mis en cascade le canal en utilisant une structure d'injection / détection de spin partagée. En concevant différentes longueurs de canal, le spin-FET à plusieurs grilles peut agir comme différentes portes logiques. Les performances de ces portes logiques sont analysées par rapport à la logique CMOS conventionnelle. En utilisant les portes logiques multi-grille à spin-FET, nous avons conçu et simulé un certain nombre de blocs logiques booléens. La fonctionnalité des blocs logiques est démontrée par le résultat de simulations transitoires à l'aide du modèle spin-FET à plusieurs grilles. Enfin, en combinant le modèle spin-FET et le modèle multi-grille spin-FET avec le modèle d'élément de stockage MTJ, les portes à «logique non volatile» sont proposées. Comme le seul signal de pur spin peut atteindre le côté détection du spin-FET, la MTJ reçoit un courant de pur spin pour le transfert de spin. Dans ce cas, la commutation de la MTJ peut être plus efficace par rapport à la structure conventionnelle MTJ / CMOS. La comparaison des performances entre la structure hybride MTJ / spin-FET et la structure hybride MTJ / CMOS est démontrée par un calcul de retard et de courant critique qui est dérivé de l'équation de Landau-Lifshitz-Gilbert (LLG). La simulation transitoire valide le fonctionnement de la logique non volatile basée sur MTJ / spin-FET. / The development of Complementary Metal Oxide Semiconductor (CMOS) technology drives the revolution of the integrate circuits (IC) production. Each new CMOS technology generation is aimed at the fast and low-power operation which mostly benefits from the scaling with its dimensions. However, the scaling will be influenced by some fundamental physical limits of device switching since the CMOS technology steps into sub-10 nm generation. Researchers want to find other ways for addressing the physical limitation problem. Spintronics is one of the most promising fields for the concept of non-charge-based new IC applications. The spin-transfer torque magnetic random access memory (STT-MRAM) is one of the successful spintronics-based memory devices which is coming into the volume production stage. The related spin-based logic devices still need to be investigated. Our research is on the field of the spin field effect transistors (spin-FET), one of the fundamental spin-based logic devices. The main mechanism for realizing a spin-FET is controlling the spin of the electrons which can achieve the objective of power reduction. Moreover, as spin-based devices, the spin-FET can easily combine with spin-based storage elements such as magnetic tunnel junction (MTJ) to construct the “non-volatile logic” architecture with high-speed and low-power performance. Our focus in this thesis is to develop the compact model for spin-FET and to explore its application on logic design and non-volatile logic simulation. Firstly, we proposed the non-local geometry model for spin-FET to describe the behaviors of the electrons such as spin injection and detection, the spin angle phase shift induced by spin-orbit interaction. We programmed the non-local spin-FET model using Verilog-A language and validated it by comparing the simulation with the experimental result. In order to develop an electrical model for circuit design and simulation, we proposed the local geometry model for spin-FET based on the non-local spin-FET model. The investigated local spin-FET model can be used for logic design and transient simulation on the circuit design tool. Secondly, we proposed the multi-gate spin-FET model by improving the aforementioned model. In order to enhance the performance of the spin-FET, we cascaded the channel using a shared spin injection/detection structure. By designing different channel length, the multi-gate spin-FET can act as different logic gates. The performance of these logic gates is analyzed comparing with the conventional CMOS logic. Using the multi-gate spin-FET-based logic gates, we designed and simulated a number of the Boolean logic block. The logic block is demonstrated by the transient simulation result using the multi-gate spin-FET model. Finally, combing the spin-FET model and multi-gate spin-FET model with the storage element MTJ model, the “non-volatile logic” gates are proposed. Since the only pure spin signal can reach to the detection side of the spin-FET, the MTJ receives pure spin current for the spin transfer. In this case, the switching of the MTJ can be more effective compared with the conventional MTJ/CMOS structure. The performance comparison between hybrid MTJ/spin-FET structure and hybrid MTJ/CMOS structure are demonstrated by delay and critical current calculation which are derived from Landau-Lifshitz-Gilbert (LLG) equation. The transient simulation verifies the function of the MTJ/spin-FET based non-volatile logic.