Gate-Oxide Trapping Enabled Synaptic Logic Transistor

Abstract

Brain-inspired neuromorphic systems have attracted much attention as a new computing paradigm for energy-efficient computation by enabling massive parallelism in artificial neural networks. The successful realization of a large-scale manufacturable artificial synapse holds the key to a full-fledged neuromorphic hardware application. This work reveals basic synaptic-like responses in the output characteristics of a normal logic CMOS transistor (with EOT < 2 nm), enabled by charge trapping dynamics at oxide defects. In addition, metaplasticity, a higher order synaptic response, is also observed by encoding relative timing. Given the mature transistor technology, a synaptic logic transistor may potentially offer a quicker pathway towards commercial neuromorphic systems.