Stockage d'ènergie local dans les mémoires résistives pour les nouvelles solutions de calcul

par Paola Trotti

Projet de thèse en Physique appliquee

Sous la direction de Gaël Pillonnet et de Gabriel Molas.

Thèses en préparation à Grenoble Alpes , dans le cadre de Physique , en partenariat avec Laboratoire d'Electronique et de Technologie de l'Information (LETI - CEA) (laboratoire) depuis le 22-05-2018 .


  • Résumé

    L'objectif de cette these est d'utiliser des dispositifs mémoire resistive (RRAM) pour stocker de l'énergie. Nous proposons donc de tirer profit des effets electrochimiques des RRAM (en particulier des CBRAM, mémoires résistives ioniques à pont conducteur) pour stocker de l'énergie. Ces dispositifs hybrides peuvent donc stocker de l'information et de l'énergie, en function du mode de programmation utilisé. L'avantage est de bénéficier d'une source très locale d'énergie, en utilisant des matériaux familiers pour la microélectronique, ce qui ouvre un large champ d'applications possibles: applications mémoires autonomes (avec faible source d'énergie extérieure, IoT...), computing, circuits neuromorphiques...

  • Titre traduit

    Local energy storage based in Resistive RAM (RRAM) for new computing solutions


  • Résumé

    The objective of this PhD is to use resistive memories to store not only information but also energy. Energy storage The first objective of the PhD is to investigate the energy storage potential that can be achieve in resistive memories (RRAM). To this aim various RRAM technologies will be characterized: Conductive Bridge Resistive memories (CBRAM) and Oxide Resistive memories (OXRAM). On these structures, energy storage features will be quantified in terms of: capacitance, stored charge and energy, charging current and voltage, electro motion force, reproducibility (cycling), stability (retention), breakdown voltage... In each case, the type of energy storage will be identified, and the involved mechanisms will be studied. Modeling could be used to analyze the electrochemical effects taking place in the devices: memory behavior (formation of a conductive filament between the two electrodes) and battery behavior (ion motion in the electrolyte) will be described and simulated. Based on the obtained results, a benchmark among the various tested technologies will be proposed. Then, the link between device stack (materials, thicknesses...) and energy storage capabilities will be clarified. The objective is to propose new device stacks (new elements, multilayers...) to optimize the device performances. Applications Thanks to these inputs, various potential applications will be envisaged and evaluated. - Memory: Stored energy can be restituted to a memory storage array, reducing the total amount of external energy that has to be provided to the circuit. This could be useful for low power and IoT applications. Energy could be stored in a specific energy storage array, or directly from resistive memory devices. In this latter case, one or several RRAM could be used to provide energy and reprogram a failed RRAM in the memory array (patent pending). - Computing: Local energy storage could be beneficial to reduce constraints on power integration systems, getting energy with a shorter latency time and with reduced parasitic capacitances. Neuromorphic: Local energy storage could be used in neuromorphic circuits to emulate a neuron or transmit spikes along synapses. The objective of this task is thus to prove the potential of various applications. Based on the electrical characteristics of the energy, a model card can be drawn. Then this could serve to design new circuits and to evaluate at the circuit level the validity of these new concepts. Analysis around how single devices could be arranged (matrix in parallel or series), and how energy could be stored and restored, will also be made.