Reversible, reconfigurable multi-mode synaptic operation of HfAlOx-based memristor for neuromorphic application

Abstract

Neuromorphic computing is a promising technology for data processing due to its extreme energy efficiency. To realize a stable neuromorphic computing system, emulating biological synapses is essential. Metaplasticity is one of a synaptic function that is recently being emulated by memristors, but for practical application, reversibility and reconfigurability is required. Here we demonstrate multi-mode operation artificial synaptic devices consisting of atomic-layer-deposited hafnium-aluminum oxide with oxygen scavenging TiN layer (Pt/HfAlOx/TiN/Pt). Reversible, reconfigurable mode switching between “Large mode” and “Small mode” is observed by applying “mode transition voltage” to the device with both DC voltage sweep operation, and pulse train operation. Additionally, the mechanism for this mode transition is explained by rendering the resistance components with oxygen ion density change and voltage divider, current divider effect. According to the operation mode, the device reacts differently on the identical input signal, thus the metaplasticity is emulated.