A Biomimetic Tunnel FET-Based Spiking Neuron for Energy-Efficient Neuromorphic Computing With Reduced Hardware Cost
Jin Luo, Cheng Chen, Qianqian Huang, Ru Huang
Abstract
In this work, utilizing the unique features of conventional Si-based tunnel FET (TFET), a TFET-based leaky integrate-and-fire (LIF) neuron with higher energy efficiency and reduced hardware cost is proposed. Compared with traditional CMOS-based LIF neuron, the proposed TFET-based LIF neuron can produce an additional bio-plausible after-hyperpolarization (AHP) behavior and relative refractory period without extra hardware cost by exploiting the features of large Miller effect and forward p-i-n current in TFET. Moreover, the typical ambipolar effect and superlinear onset behaviors in conventional Si-based TFET enable the lower hardware cost and lower energy consumption ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\sim 10\times $ </tex-math></inline-formula> reduction) for TFET-based neuron. Furthermore, the proposed TFET neuron-based spiking neural network (SNN) is demonstrated for pattern recognition tasks, showing its advantage of significant energy efficiency. This work provides a promising highly integrated and energy-efficient solution for the hardware implementation of spiking neuron for neuromorphic computing.