Descrizione del progetto
Sinapsi artificiali per il calcolo neuromorfico grazie a transistor ferroelettrici a barriera Schottky
Secondo le stime, nel cervello umano adulto sarebbero presenti all’incirca 100 miliardi di neuroni, ciascuno dei quali è collegato attraverso 7 000 sinapsi con altri neuroni, formando una rete di elaborazione dati ad alta efficienza energetica. Il calcolo neuromorfico si prefigge di sfruttare questa efficienza utilizzando neuroni, sinapsi e circuiti ispirati al cervello per ridurre l'enorme fabbisogno energetico dovuto al grande aumento odierno nelle capacità di elaborazione dei dati. I transistor a effetto di campo (FET, ferroelectric transistor) ferroelettrici sono tra le opzioni più promettenti per le sinapsi allo stato solido. Con il sostegno del programma di azioni Marie Skłodowska-Curie, il progetto EASIFeT si concentrerà su un nuovo FET ferroelettrico, nello specifico un FET ferroelettrico a barriera Schottky. L’équipe si avvarrà di un noto strumento di progettazione e di simulatori per esplorare i comportamenti e implementarli in un circuito modello.
Obiettivo
The exponential growth of demand for data processing requires increasingly large computational resources and, consequently, prohibitively high energy consumption. To sustain this evolution, a paradigm switch from conventional computing architectures to data-centric platforms is needed. Neuromorphic computing aims at reaching this goal by realizing brain-inspired circuits based on artificial neurons and synapses, which are extremely energy efficient. The objective of this project is to explore a novel type of artificial synapse to be employed in neuromorphic chips. Among the technological options for solid state synapses, memories based on ferroelectric field-effect transistors (FeFET) are considered very promising due to their energy efficiency and to their compatibility with a Back-End-Of-Line implementation and thus a 3D integration. A FeFET is a field-effect transistors that employs a ferroelectric (FE) material as gate oxide. FE materials have a spontaneous electric polarization that can be reversed by the application of an electric field, and in conventional FeFETs this is used to modulate the threshold voltage and thus resistance in the channel region. This project will address an alternative physical mechanism to obtain a synaptic behavior in FeFET, namely a polarization-induced tuning of the resistance at source/drain Schottky contacts. In the Ferroelectric Schottky barrier FETs (Fe-SBFETs), the FE material overlaps the Schottky contact region, hence in this region the FE material is placed between two metals resulting in an effective and low voltage ferroelectric switching. For this reason, Fe-SBFETs are expected to operate as low energy synaptic devices. In this project, Fe-SBFETs will be extensively studied and modeled, by means of TCAD simulations. A design-space for optimal synaptic operation will be derived. Finally, a compact model for SPICE simulations of neuromorphic circuits based on Fe-SBFETs will be developed.
Campo scientifico
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringanalogue electronics
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringcomputer hardware
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectrical engineering
- natural sciencescomputer and information sciencesdata sciencedata processing
Parole chiave
Programma(i)
- HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) Main Programme
Meccanismo di finanziamento
HORIZON-TMA-MSCA-PF-EF - HORIZON TMA MSCA Postdoctoral Fellowships - European FellowshipsCoordinatore
33100 Udine
Italia