Atomic switch : synaptic functionalities and integration strategies
Résumé
Nowadays, neuro-inspired computing has generated a lot of interest and seems to be a promising candidate to complement and to provide enhanced performances and new functionalities to the existing CMOS/Von Neumann processor. Engineering of memristive nano devices with specific functionalities for the implementation of synaptic operation and their integration into high density parallel network are still two big challenges for the development of neuromorphic hardware. We focus on a particular class of CB-RAM, the atomic switch that is composed by an ionic conductor material, the Ag2S, sandwiched between two metal electrodes (Ag and Pt). We characterize the tunable device volatility due to the formation/disruption of silver filaments into the ionic-conductor material under applied bias leading to short-term memory (STM) and long-term memory (LTM). Based on this simple two terminal synaptic device, we then investigate alternative routes toward their integration into crossbar network.