Expand EDL transistors for neuromorphic applications

Abstract

Recent research and developments have suggested that there are numerous benefits in mimicking biological synapses. Implementing such technology enables the fabrication of electronic devices with high advanced memory and self-learning capabilities, which are key components of bioinspired neuromorphic systems. Several artificial synaptic thin film transistors (TFTs) have been developed using mostly inorganic materials and conventional semiconductor device fabrication methods. This carves the way for further developments in the technology, with its ultimate goal being to create a highly effective system that consumes little energy but provides high performances. This project aims to investigate the neuromorphic performance of electric double layer (EDL) TFT using solution processed amorphous indium zinc oxide (IZO) active layer and a high ĸ ionic liquid as the gate dielectric which is located at the plane of the semiconducting active layer. The high polarizing capability induced by the ionic liquid (IL) is vital in encouraging synaptic behaviours in accordance to the information traveling between the source and drain electrodes. To showcase such synaptic performances, the active layer of the thin film transistor is deposited using solution processing technique to promote a controllable amorphous crystal structure that ensures lower defect density. The devices are then measured using a low pre synaptic voltage of 0.01V to 10V. Such voltages are used to accumulate carriers at the semiconductor/dielectric interface. This creates a channel for electrons to travel between the metal electrodes. Ultimately, the synaptic strengths are tested with various pulse width and intervals to mimic biological synapse environments and behaviours. The results from this project have proven that ionic liquid is the superior candidate for gate dielectric as compared to other dielectric materials, particularly SiO2. Therefore with these results, future studies can be conducted to showcase the possibility that an artificial neural network with two presynaptic terminals can be manufactured. This opens the door to more complicated synaptic learning techniques which has a closer and more realistic replication of the biological synaptic performance.