My interests span the development of materials, structures, electronic devices, and circuits for memory applications and new computing paradigms, like brain-inspired and neuromorphic computing in general.
In the last years, my research was targeted toward the exploitation of material and device properties to develop computational functionalities for the electronic emulation of biological synapses, neurons and overall brain activity and demonstrate their potential in computational platforms by system-level simulation.
One basic function of brain operation is the reconfiguration of the connections between neurons through plastic and tunable synapses. In this respect, we demonstrated that the conductance of memristor devices (or resistive memory devices) can be tuned in analogue manner thanks to a specific engineering of the interfaces of the metal/interface/metal structure of the devices themselves. Such engineered devices was used to modulate the communication between two artificial CMOS spiking neurons realized in 350 nm technology. The experimental device features have been validated in spiking neuronal network with various learning rules through system level simulations. (ref. project RAMP and NEURAM3)
On the otherside, the great potential of the brain comes form the employment of dynamical processes, usually modeled in diverse and heterogeneous functionalities like synaptic and neuron integration, firing adaptation, dendrite communication,... Volatile memristors are dynamical devices that can switch from a high to a low resistance state upon voltage application and relax back to the pristine state after voltage release. We demonstrate useful integrative functionalities of the devices useful to emulate biological processes. (ref. project MeM-Scales)
A completely different perspective towards neuromorphic computing is the idea of emulate some network-level feature of the brain rather than building a large network of artificial neurons and synapses. In particular, it is known that the brain based its efficiency in a so-call "edge-of-chaos" dynamical conditions. At the edge between ordered and chaotic behaviour the processing ability of a dynamical system are maximized. In this perspective, we developed and build a nonlinear dynamical circuits that uses the nonlinearity of a memristor device to manipulate the chaotic dynamics of voltage outputs and foresee to use it for computing purposes.(ref. project COSMO)
Physical Implementation of a Tunable Memristor-based Chua's Circuit
HfO2-based resistive switching memory devices for neuromorphic computing
S Brivio; S Spiga and D Ielmini
Improving HfO2-Based Resistive Switching Devices by Inserting a TaOx Thin Film via Engineered In Situ Oxidation
Tao Wang; Stefano Brivio; Elena Cianci; Claudia Wiemer; Michele Perego; Sabina Spiga; Mario Lanza
Analogue Computing with Dynamic Switching Memristor Oscillators: Theory, Devices and Applications - COSMO
Reference person for CNR-IMM: Stefano Brivio
Memory technologies with multi-scale time constants for neuromorphic architectures - MeM-Scales
Funding: H2020 - ICT
Reference person for CNR-IMM: Sabina Spiga