Diameter dependent synaptic behaviors of III-V nanowires for neuromorphic image denoising

• This work offers new insight into the modulation of surface states for the coming neuromorphic computing by using the global NWFETs. • With the higher surface-to-volume ratio and much more Sb-O bonds on the surface, the thin NWs possess heavier surface states than thick NWs, resulting in the worse electrical stability and more obvious synaptic behaviors. • Although both thin and thick NWFETs display impressive image denoising ability, the recognition accuracy can reach to 93.9 % by taking use of thin and thick NWFTEs together, which is much higher than that of individual thin (92.1 %) or thick (84.4 %) NWFETs. Diameter is an important geometry parameter for III-V nanowires (NWs) in electronics, optoelectronics and neuromorphic computing. In this work, the electrical stability and synaptic behaviors of thin and thick GaSb NWs are studied in detailed. With the higher surface-to-volume ratio and much more Sb-O bonds on the surface, the thin NWs possess heavier surface states than thick NWs. As a result, the thin NW filed-effect-transistors (NWFETs) display worse electrical stability and more obvious synaptic behaviors. These impressive phenomena result from the surface states related carrier trapping and detrapping processes. By taking use of the thin and thick NWFETs together for neuromorphic image, the recognition accuracy can reach to 93.9 %, which is much higher than that of individual thin (92.1 %) or thick (84.4 %) NWFETs. This work offers new insight into the modulation of surface states for the coming neuromorphic computing by using the global NWFETs.