Polysilicon-Channel Synaptic Transistors for Implementation of Short- and Long-Term Memory Characteristics

The rapid progress of artificial neural networks (ANN) is largely attributed to the development rectified linear unit (ReLU) activation function. However, implementation software-based ANNs, such as convolutional (CNN), within von Neumann architecture faces limitations due its sequential processing mechanism. To overcome this challenge, research on hardware neuromorphic systems based spiking (SNN) has gained significant interest. Artificial synapse, a crucial building block in these systems, predominantly utilized resistive memory-based memristors. two-terminal structure memristors presents difficulties feedback signals from post-synaptic neuron, and without an additional rectifying device it challenging prevent sneak current paths. In paper, we propose four-terminal synaptic transistor with asymmetric dual-gate solution Similar biological synapses, proposed multiplies presynaptic input signal stored weight information transmits result postsynaptic neuron. Weight modulation explored through both hot carrier injection (HCI) Fowler–Nordheim (FN) tunneling. Moreover, investigate incorporation short-term memory properties by adopting polysilicon grain boundaries temporary storage. It anticipated that devised devices, possessing long-term characteristics, will enable various novel ANN algorithms.