Leaky Integrate and Fire Neuron by Charge-Discharge Dynamics in Floating-Body MOSFET

Area of Science:

• Neuro-inspired computing
• Solid-state device physics

Background:

• Spiking Neural Networks (SNNs) offer efficient learning and recognition.
• Large-scale SNNs require power and area-efficient electronic neurons.
• Previous work demonstrated a 4-terminal gated-INPN device for LIF neurons.

Purpose of the Study:

• To propose and experimentally demonstrate a compact 3-terminal PD-SOI MOSFET as an alternative to the 4-terminal gated-INPN device.
• To utilize impact ionization in SOI-MOSFETs for LIF neuron behavior.
• To enable very-large-scale-integration (VLSI) for biology-scale neural networks.

Main Methods:

• Experimental demonstration of a 3-terminal PD-SOI MOSFET with a 100 nm gate length.
• Utilizing impact ionization (II) induced floating body effect to emulate LIF neuron dynamics.
• Characterizing spiking frequency dependence on input signals.

Main Results:

• Successful experimental demonstration of a compact 3-terminal PD-SOI MOSFET neuron.
• Emulation of Leaky Integrate-and-Fire (LIF) neuron behavior using impact ionization in SOI-MOSFETs.
• Achieved MHz operation, enabling hardware acceleration for neural networks.

Conclusions:

• Conventional PD-SOI-CMOS technology is suitable for VLSI implementation of large-scale SNNs.
• The proposed 3-terminal PD-SOI MOSFET offers a power and area-efficient solution for electronic neurons.
• This approach facilitates the creation of biology-scale neural networks.