Enhancement of long-term memory of IGZO synaptic transistors by the introduction of an Al2O3 charge trapping layer

Abstract

Brain-inspired neuromorphic computing has been widely considered a promising solution to overcome the limitations of traditional von Neumann architecture in the current computer system. As an essential component of the neuromorphic system, the artificial synaptic device exhibits great potential in adaptive learning. Due to their controllable channel conductance and CMOS compatibility, solid electrolyte-gated synaptic transistors (EGSTs) have garnered significant interest as next-generation neuromorphic devices. However, most of the existing EGSTs suffer from rapid self-diffusion of the ions, making it difficult to maintain the stable channel conductance states. In this work, the synaptic transistors were fabricated with indium–gallium–zinc oxide as the channel layer, Al2O3 as the charge trapping layer, and ZrO2 as the solid electrolyte layer. The self-diffusion of the hydrogen ions can be suppressed by the positive charges trapped in the Al2O3 layer, which significantly improves the long-term plasticity (LTP) of the devices. By adjusting the presynaptic spike scheme, the typical synaptic behaviors, including excitatory postsynaptic current, paired-pulse facilitation, and the transition from short-term memory to long-term memory, were simulated. Based on the conductance modulation properties of the channel in the synaptic transistor, an artificial neural network was constructed for pattern recognition, and a high accuracy of 95.4% was obtained. This work demonstrates an effective strategy for the enhancement of the LTP of the synaptic transistor.