Advanced synaptic transistor device towards AI application in hardware perspective

Abstract

For the past decades, the synaptic devices for the inmemory computing have been widely investigated due to the high-efficiency computing potential and the ability to mimic biological neurobehavior. However, the conventional twoterminal synaptic memristors show drawbacks of resistance reduction caused by large-scale paralleling and asynchronous storage-reading process, which limit its development. Recently, researchers have paid attention to the transistor-like artificial synapse. Due to the existence of insulator layer and the separation of input and read terminals, the three-terminal synaptic transistors are believed to have greater potential towards artificial intelligence (AI) application fields. In this work, a summary of recent progresses and the future challenges of synaptic transistors are discussed.