Αντικείμενο της διπλωματικής εργασίας αποτελεί η μελέτη και ανάλυση των μεθόδων διόρθωσης λαθών με χρήση κώδικα BCH που μπορούν να εφαρμοστούν σε συστήματα αποθήκευσης πληροφορίας τεχνολογίας PCM (Phase-Change Memory). Η τεχνολογία PCM αποτελεί μία νέα τεχνολογία που υπόσχεται υψηλές χωρητικότητες, χαμηλή κατανάλωση ισχύος και μπορεί να εφαρμοστεί είτε σε συσκευές αποθήκευσης σταθερής κατάστασης (Solid State Drives) είτε σε μνήμες τυχαίας προσπέλασης (Random-Access Memories), παρέχοντας μία εναλλακτική πρόταση έναντι μνημών τεχνολογίας flash και DRAM. Ένα από τα μειονεκτήματα της τεχνολογίας PCM είναι η ανθεκτικότητα εγγραφής (write endurance), η οποία μπορεί να βελτιωθεί με τη χρήση μεθόδων διόρθωσης λαθών που θα παρατείνουν τον χρόνο ζωής της συσκευής όταν, λόγω της φυσικής φθοράς του μέσου, αρχίσουν να υπάρχουν σφάλματα στα αποθηκευμένα δεδομένα. Για την εφαρμογή της διόρθωσης λαθών μπορούν να χρησιμοποιηθούν κώδικες BCH, οι οποίοι αποτελούν μια κλάση ισχυρών κυκλικών κωδίκων διόρθωσης τυχαίων λαθών, και κατασκευάζονται με χρήση της άλγεβρας πεπερασμένων πεδίων. Οι κώδικες BCH είναι ιδανικοί για διόρθωση λαθών σε συσκευές αποθήκευσης πληροφορίας όπου η κατανομή των λαθών είναι τυχαία. Αρκετοί αλγόριθμοι έχουν προταθεί για τις λειτουργίες αποδοτικής κωδικοποίησης και αποκωδικοποίησης κωδίκων BCH. Στην παρούσα εργασία μελετήθηκαν λύσεις που μπορούν να υλοποιηθούν με παράλληλες αρχιτεκτονικές, ενώ ειδικότερα για την λειτουργία αποκωδικοποίησης έγινε χρήση ενός παράλληλου αλγορίθμου που δεν χρειάζεται αντιστροφείς πεπερασμένου πεδίου για την επίλυση των εξισώσεων των συνδρόμων, επιτυγχάνοντας υψηλές συχνότητες λειτουργίας. Για την κατανόηση των λειτουργιών κωδικοποίησης και αποκωδικοποίησης απαιτείται η προσεκτική μελέτη της άλγεβρας πεπερασμένων πεδίων και της αριθμητικής της. Οι κώδικες BCH προσφέρουν πλεονεκτήματα όπως χαμηλή πολυπλοκότητα και ύπαρξη αποδοτικών μονάδων υλοποίησης σε υλικό. Στην παρούσα εργασία σχεδιάστηκαν ένας παράλληλος κωδικοποιητής και ένας παράλληλος αποκωδικοποιητής για τον κώδικα BCH(728,688). Τα δύο συστήματα υλοποιήθηκαν ως περιφερειακά σε ενσωματωμένο σύστημα βασισμένο σε επεξεργαστή MicroBlaze, με έμφαση σε μια καλή σχέση μεταξύ της συχνότητας λειτουργίας και των απαιτήσεων σε επιφάνεια υλικού και κατανάλωση ισχύος. Για την υλοποίηση χρησιμοποιήθηκε συσκευή FPGA σειράς Virtex-6. / The objective of this thesis is the study and analysis of BCH error-correction methods that can be applied on PCM (Phase-Change Memory) storage devices. PCM is a new technology that promises high capacities, low power consumption and can be applied either on Solid State Drives or on Random Access Memories, providing an alternative to flash and DRAM memories. However, PCM suffers from limited write endurance, which can be increased using error-correction schemes that will extend the lifetime of the device when, due to medium wear-out, errors start to appear in the written data. Thus, BCH codes (powerful cyclic random multiple error-correcting codes) can be employed. BCH codes are ideal for ECC (Error-Correction Coding) in storage devices, due to their fault model which is random noise. Several algorithms have been proposed for the efficient coding and decoding BCH codes. In the present thesis parallel implementations where studied. For the decoding process in particular, a parallel algorithm was used that does not require finite field inverter units to solve the syndrome equations, achieving high operation frequencies. For the understanding of BCH coding and decoding processes, basic knowledge of the finite field algebra and arithmetic is required. BCH codes offer advantages such as low complexity and efficient hardware implementations. In the present thesis a parallel BCH(728,688) encoder and a parallel BCH(728,688) decoder were designed. The above systems were implemented as peripherals on an MicroBlaze-based embedded system, with emphasis on an optimal tradeoff between area and power consumption. A Virtex-6 FPGA device was used for the final stages of the implementation.

Dans cette thèse, nous proposons une étude détaillée des mécanismes responsables de la perte de données dans les mémoires à changement de phase enrichies en germanium (Ge-rich PRAMs), à savoir la dérive de la résistance au cours du temps et la recristallisation de la phase amorphe. Nous commençons par une présentation du contexte dans lequel s'inscrit cette étude an donnant un aperçu rapide du marché des mémoires à semiconducteur et une comparaison des mémoires non volatiles émergentes. Les principes de fonctionnement de la technologie PRAM sont introduits, avec ses avantages, ses inconvénients, ainsi que la physique régissant le processus de cristallisation dans les matériaux à changement de phase, avant de décrire les problèmes de fiabilité qui nous intéressent.Une caractérisation électrique complète de dispositifs intégrant des alliages de GST enrichi en germanium est ensuite proposée, en commençant par la caractérisation des matériaux utilisés dans nos cellules, introduisant alors les avantages des alliages enrichis en Ge sur le GST standard. Les performances électriques des dispositifs intégrant ces matériaux sont analysées, avec une étude statistique des caractéristiques SET & RESET, de la fenêtre de programmation, de l'endurance et de la vitesse de cristallisation. Nous nous concentrons ensuite sur le thème principal de cette thèse en analysant la dérive en résistance de l'état SET de nos dispositifs Ge-rich, ainsi que les performances de rétention de l'état RESET.Dans la dernière partie, nous étudions les mécanismes physiques impliqués dans ces phénomènes en fournissant une étude détaillée de la structure des cellules, grâce à l'utilisation de la Microscopie Électronique en Transmission (MET). Les conditions et configurations expérimentales sont décrites, avant de présenter les résultats qui nous ont permis d'aller plus loin dans la compréhension de la dérive en résistance et de la recristallisation de la phase amorphe dans les dispositifs Ge-rich. Une discussion est finalement proposée, reliant les résultats des caractérisations électriques avec ceux des analyses TEM, conduisant à de nouvelles perspectives pour l'optimisation des dispositifs PRAMs. / In this thesis we provide a detailed study of the mechanisms responsible for data loss in Ge-rich Ge2Sb2Te5 Phase-Change Memories, namely resistance drift over time and recrystallization of the amorphous phase. The context of this work is first presented with a rapid overview of the semiconductor memory market and a comparison of emerging non-volatile memories. The working principles of PRAM technology are introduced, together with its advantages, its drawbacks, and the physics governing the crystallization process in phase-change materials, before describing the reliability issues in which we are interested.A full electrical characterization of devices integrating germanium-enriched GST alloys is then proposed, starting with the characterization of the materials used in our PCM cells and introducing the benefits of Ge-rich GST alloys over standard GST. The electrical performances of devices integrating those materials are analyzed, with a statistical study of the SET & RESET characteristics, programming window, endurance and crystallization speed. We then focus on the main topic of this thesis by analyzing the resistance drift of the SET state of our Ge-rich devices, as well as the retention performances of the RESET state.In the last part, we investigate on the physical mechanisms involved in these phenomena by providing a detailed study of the cells' structure, thanks to Transmission Electron Microscopy (TEM). The experimental conditions and setups are described before presenting the results which allowed us to go deeper into the comprehension of the resistance drift and the recrystallization of the amorphous phase in Ge-rich devices. A discussion is finally proposed, linking the results of the electrical characterizations with the TEM analyses, leading to new perspectives for the optimization of PRAM devices.

La recherche dans le domaine de l’informatique neuro-inspirée suscite beaucoup d'intérêt depuis quelques années. Avec des applications potentielles dans des domaines tels que le traitement de données à grande échelle, la robotique ou encore les systèmes autonomes intelligents pour ne citer qu'eux, des paradigmes de calcul bio-inspirés sont étudies pour la prochaine génération solutions informatiques (post-Moore, non-Von Neumann) ultra-basse consommation. Dans ce travail, nous discutons les rôles que les différentes technologies de mémoire résistive non-volatiles émergentes (RRAM), notamment (i) Phase Change Memory (PCM), (ii) Conductive-Bridge Memory (CBRAM) et de la mémoire basée sur une structure Metal-Oxide (OXRAM) peuvent jouer dans des dispositifs neuromorphiques dédies. Nous nous concentrons sur l'émulation des effets de plasticité synaptique comme la potentialisation à long terme (Long Term Potentiation, LTP), la dépression à long terme (Long Term Depression, LTD) et la théorie STDP (Spike-Timing Dependent Plasticity) avec des synapses RRAM. Nous avons développé à la fois de nouvelles architectures de faiblement énergivore, des méthodologies de programmation ainsi que des règles d’apprentissages simplifiées inspirées de la théorie STDP spécifiquement optimisées pour certaines technologies RRAM. Nous montrons l’implémentation de systèmes neuromorphiques a grande échelle et efficace énergétiquement selon deux approches différentes: (i) des synapses multi-niveaux déterministes et (ii) des synapses stochastiques binaires. Des prototypes d'applications telles que l’extraction de schéma visuel et auditif complexe sont également montres en utilisant des réseaux de neurones impulsionnels (Feed-forward Spiking Neural Network, SNN). Nous introduisons également une nouvelle méthodologie pour concevoir des neurones stochastiques très compacts qui exploitent les caractéristiques physiques intrinsèques des appareils CBRAM. / Research in the field of neuromorphic- and cognitive- computing has generated a lot of interest in recent years. With potential application in fields such as large-scale data driven computing, robotics, intelligent autonomous systems to name a few, bio-inspired computing paradigms are being investigated as the next generation (post-Moore, non-Von Neumann) ultra-low power computing solutions. In this work we discuss the role that different emerging non-volatile resistive memory technologies (RRAM), specifically (i) Phase Change Memory (PCM), (ii) Conductive-Bridge Memory (CBRAM) and Metal-Oxide based Memory (OXRAM) can play in dedicated neuromorphic hardware. We focus on the emulation of synaptic plasticity effects such as long-term potentiation (LTP), long term depression (LTD) and spike-timing dependent plasticity (STDP) with RRAM synapses. We developed novel low-power architectures, programming methodologies, and simplified STDP-like learning rules, optimized specifically for some RRAM technologies. We show the implementation of large-scale energy efficient neuromorphic systems with two different approaches (i) deterministic multi-level synapses and (ii) stochastic-binary synapses. Prototype applications such as complex visual- and auditory- pattern extraction are also shown using feed-forward spiking neural networks (SNN). We also introduce a novel methodology to design low-area efficient stochastic neurons that exploit intrinsic physical effects of CBRAM devices.

Chalcogenide glass based Phase Change Memories (PCMs) are being considered recently as promising alternatives to conventional non-volatile Random Access Memories (NVRAMs). PCMs offer high performance & low power consumption, in addition to other advantages, such as high scalability, high endurance and compatibility with complementary metal oxide semiconductors (CMOS) technologies. Basically PCM is a resistance variable non-volatile memory in which the memory bit state is defined by the resistance of the material. In this case, the initial ‘OFF’ state (logic zero) corresponds to the high resistance amorphous state and the logic 1 or ‘ON’ state corresponds to low resistance crystalline state. The present thesis work deals with electrical, thermal, mechanical and optical characterization of certain tellurium based chalcogenide glasses in bulk and thin film form for phase change memory applications. A comparative study has been done on the electrical switching behavior of Ge-Te-Se & Ge-Te-Si amorphous thin film samples with their bulk counterparts. Further, electrical switching and thermal studies have been undertaken on bulk Ge-Te-Bi and Ge-Te-Sn series of samples. The composition dependence of switching voltages of bulk and thin film samples studied has been explained on the basis of different factors responsible for electrical switching. The thesis contains ten chapters: Chapter 1 deals with a brief introduction on chalcogenides and their applicability in phase change memories. The glass transition phenomenon, synthesis of chalcogenide alloys, different structural models of amorphous semiconductors and electrical switching behavior are also discussed in detail in this chapter. Further, a brief description of optical and mechanical properties along with the principles of few characterization techniques used is discussed. Also, a brief overview on PCM application of chalcogenides is presented. The second chapter provides the details of various experimental techniques used to measure electrical, thermal, optical and mechanical properties of few tellurium based chalcogenide glassy systems. In the third chapter, the electrical switching behavior of amorphous Al23Te77 thin film devices, deposited in co-planar geometry, has been discussed. It is found that these samples exhibit memory type electrical switching. Scanning Electron Microscopic studies show the formation of a crystalline filament in the electrode region which is responsible for switching of the device from high resistance OFF state to low resistance ON state. The switching behavior of thin film Al-Te samples is found to be similar to that of bulk samples, with the threshold fields of bulk samples being higher. This has been understood on the basis of higher thermal conductance in bulk, which reduces the Joule heating and temperature rise in the electrode region. Electrical switching and thermal behavior of bulk; melt quenched Ge18Te82-xBix glasses (1 ≤ x ≤ 4) are presented in chapter 4. Ge-Te-Bi glasses have been found to exhibit memory type electrical switching behavior, which is in agreement with the lower thermal diffusivity values of these samples. A linear variation in switching voltages (also known as threshold voltages) (Vt) has been found with increase in thickness. The switching voltages have been found to decrease with an increase in temperature which is due to the decrease in the activation energy for crystallization at higher temperatures. Further, Vt of Ge18Te82-xBix glasses have been found to decrease with the increase in Bi content, indicating that in the Ge-Te-Bi system, the resistivity of the additive has a stronger role to play in the composition dependence of Vt, in comparison with the network connectivity and rigidity factors. In addition, the composition dependence of crystallization activation energy has been found to show a decrease with an increase in Bi content. X-ray diffraction studies on thermally crystallized samples reveal the presence of hexagonal Te, GeTe and Bi2Te3 phases. The fifth chapter deals with the electrical switching studies and optical band gap measurements on GexSe35-xTe65 (17 ≤ x ≤ 23) amorphous thin film samples. These thin film samples coated with sandwich geometry are found to switch with very low voltages as compared to bulk samples of the same chalcogenide glasses. The switching voltages and optical band gap are found to increase with the addition of Ge at the expense of Se. High structural cross linking with progressive addition of 4-fold coordinated Ge atoms could be the one of the reasons of increasing switching voltage and stronger Ge-Se bond strength could be the reason of increasing band gap for these chalcogenide glasses. In chapter 6, electrical switching studies on amorphous Ge15Te85-xSix (1 ≤ x ≤ 6) thin film samples have been described and the results are compared with their bulk counterparts. Similar trend has been found for both bulk and film samples when the threshold field is varied with composition. Optical band gap has been measured as a function of composition for these films, which also shows a behavior similar to that of switching voltages. The increasing trend in the variation with composition of electrical switching voltages and optical band gap are due to the increase in network connectivity and rigidity as Si atoms are incorporated into the Ge-Te system. Chapter 7 summarizes the electrical switching and glass forming ability of the Ge-Te-Sn glasses of two different composition tie-lines, namely Ge15Te85-xSnx and Ge17Te83-xSnx. Glasses belonging to both the series have been found to exhibit memory type of electrical switching behavior. The thickness dependence of threshold voltages is also found to support the memory switching behavior of the system. Further, ADSC studies are undertaken to explore the thermal behavior of these glasses which indicates that the crystallization tendency increases as Sn concentration is increased in the Ge-Te network. XRD studies done on two samples from both the series, reveal the fact that Sn atoms do not take part actively to enhance the network connectivity and rigidity. The composition dependence of crystallization temperature, metallicity factor and results of XRD studies are put together to explain the variation with composition of threshold voltages for both the series of samples. In chapter 8, investigations on the electrical switching behavior of Ge15Te85-xSnx (1 ≤ x ≤ 5) and Ge17Te83-xSnx (1 ≤ x ≤ 4) amorphous thin films have been discussed. Both the series of samples have been found to exhibit memory type of electrical switching behavior. The composition dependence of threshold voltage shows a decreasing trend, which has been explained on the basis of the Chemically Ordered Network (CON) model, bond strength and the metallicity factor. The optical band gap variation of both the series also exhibits a similar decreasing trend with composition. The observed behavior has been understood on the basis of higher atomic radius of Sn atom than Ge atom, which makes the energy difference between bonding and anti bonding state less at band edge. Chapter 9 deals with the nano-indentation studies on Ge15Te85-xSix (0 ≤ x ≤ 9) bulk glasses. The composition dependence of young’s modulus and hardness is studied systematically in this glassy system. The density of the samples of different compositions has also been measured, which strongly supports the variation of Young’s Modulus and hardness with composition. The composition dependence of mechanical properties of Ge-Te-Si samples has been understood on the basis of the presence of an intermediate phase and a thermally reversing window in this glassy system. A summary of the significant results obtained in the present thesis work is presented in the last chapter along with the scope for future work.

Chalcogenide glasses are a class of covalent amorphous semiconductors with interesting properties. The presence of short-range order and the pinned Fermi level are the two important properties that make them suitable for many applications. With flash memory technology reaching the scaling limit as per Moore’s law, alternate materials and techniques are being researched at for realizing next generation non-volatile memories. Two such possibilities that are being looked at are Phase Change Memory (PCM) and Programmable Metallization Cell (PMC) both of which make use of chalcogenide materials. This thesis starts with a survey of the work done so far in realizing PCMs in reality. For chalcogenides to be used as a main memory or as a replacement to FLASH technology, the electrical switching parameters like switching voltage, programming current, ON state and OFF state resistances, switching time and optical parameters like band gap are to be considered. A survey on the work done in this regard has revealed that various parameters such as chemical composition of the PC material, nature of additives used to enhance the performance of PCM, topological thresholds (Rigidity Percolation Threshold and Chemical Threshold), device geometry, thickness of the active volume, etc., influence the electrical switching parameters. This has motivated to further investigate the material and experimental parameters that affect switching and also to explore the possibility of multi level switching. In this thesis work, the feasibility of using two chalcogenide systems namely Si15Te85-xGex and Ge15Te85-xAgx in the form of amorphous thin films for PCM application is explored. In the process, electrical switching experiments have been carried out on thin films belonging to these systems and the results obtained are found to exhibit some interesting anomalies. Further experiments and analysis have been carried out to understand these anomalies. Finally, the dynamics of electrical switching has been investigated and presented for amorphous Si15Te85-xGex thin films. From these studies, it is also seen that multi state switching/multiple resistance levels of the material can be achieved by current controlled switching, the mechanisms of which have been further probed using XRD analysis and AFM studies. In addition, investigations have been carried out on the electrical switching behavior of amorphous Ge15Te85-xAgx thin film devices and optical band gap studies on amorphous Ge15Te85-xAgx thin films. Chapter one of the thesis, gives a brief introduction to the limitations in existing memory technology and the alternative memory technologies that are being researched, based on which it can be inferred that PCM is a promising candidate for the next generation non volatile memory. This chapter also discusses the principle of using PCM to store data, realization of PCM using chalcogenides, the material properties to be considered in designing PCM, the trade offs in the process of design and the current trends in PCM technology. Chapter two provides a brief review of the electrical switching phenomenon observed in various bulk chalcogenide glasses, as electrical switching is the underlying principle behind the working of a PCM. In the process of designing a memory, many parameters like read/write operation speed, data retentivity and life, etc., have to be optimized for which a thorough understanding on the dependence of electrical switching mechanism on various material parameters is essential. In this chapter, the dependence of electrical switching on parameters like network topological thresholds and electrical and thermal properties of the material is discussed. Doping is an efficient way of controlling the electrical parameters of chalcogenides. The nature of dopant also influences switching parameters and this also is briefly discussed. Chapter three provides a brief introduction to the different experimental techniques used for the thesis work such as bulk chalcogenide glass preparation, preparation of thin amorphous films, measurement of film thickness, confirmation of amorphous nature of the films using X-Ray Diffraction (XRD), electrical switching experiments using a custom made setup, crystallization study using XRD and Atomic Force Microscopy (AFM) and optical band gap studies using UV-Vis spectrometer. Vt is an important parameter in the design of a PCM. Chapter four discusses the dependence of Switching voltage, Vt, on input energy. It is already established that the Vt is influenced by the composition of the base glass, nature of dopants, thickness of films and by the ambient temperature. Based on the results of electrical switching experiments in Si15Te74Ge11 amorphous thin films a comprehensive analysis has been done to understand the kinetics of electrical switching. Chapter five discusses a current controlled crystallization technique that can be used to realize multi-bit storage with a single layer of chalcogenide material. In case of PCM, data is stored as structural information; the memory cell in the amorphous state is read as data ‘0’ and the memory cell in crystalline state is read as data ‘1’. This is accomplished through the process of electrical switching. In order to increase the memory density or storage density, multi-bit storage is being probed at by having multiple layers of chalcogenide material. However, with this technique, the problems of inter-diffusion between different layers cannot be ruled out. In this thesis work, a current controlled crystallization technique has been used to achieve multiple stable resistance states in Si15Te75Ge10 thin films. Chapter six discusses the mechanism behind multi state switching exhibited by certain compositions of Si15Te85-xGex thin films. Crystallization studies on certain Si15Te85-xGex films have been carried out using XRD and AFM to understand the phenomenon of multiple states. The results of these experiments and analysis are presented in this chapter. Chapter seven discusses the results of electrical switching experiments and optical band gap studies on amorphous Ge15Te85-xAgx thin films. Chapter eight gives the conclusion and scope for future work.

In this review the different concepts of nanoscale resistive switching memory devices are described and classified according to their I–V behaviour and the underlying physical switching mechanisms. By means of the most important representative devices, the current state of electrical performance characteristics is illuminated in-depth. Moreover, the ability of resistive switching devices to be integrated into state-of-the-art CMOS circuits under the additional consideration with a suitable selector device for memory array operation is assessed. From this analysis, and by factoring in the maturity of the different concepts, a ranking methodology for application of the nanoscale resistive switching memory devices in the memory landscape is derived. Finally, the suitability of the different device concepts for beyond pure memory applications, such as brain inspired and neuromorphic computational or logic in memory applications that strive to overcome the vanNeumann bottleneck, is discussed.