Gated Schottky Diode-Type Synaptic Device with a Field-Plate Structure to Reduce the Forward Current

Area of Science:

• Semiconductor Devices
• Materials Science
• Artificial Intelligence Hardware

Background:

• Gated Schottky diodes are explored as synaptic devices for hardware neural networks.
• Unwanted forward current in these devices can lead to malfunctions and increased power consumption.
• Existing designs struggle to mitigate this leakage current effectively.

Purpose of the Study:

• To propose and investigate a new low-power synaptic device using a gated Schottky diode with a field-plate structure.
• To suppress the forward current in gated Schottky diodes for improved neural network performance.
• To reduce power consumption in hardware-based neural network systems.

Main Methods:

• A field-plate structure was integrated into a gated Schottky diode design.
• The device was fabricated using a poly-Si active layer.
• Electrical characteristics, specifically forward current suppression, were measured and analyzed.

Main Results:

• The field-plate structure successfully formed a virtual pn junction within the poly-Si active layer.
• This virtual pn junction effectively suppressed the forward current of the gated Schottky diode.
• The unwanted forward current was reduced to an exceptionally low level, less than 1 pA/µm.

Conclusions:

• The proposed gated Schottky diode with a field-plate structure is a viable low-power synaptic device.
• This design effectively addresses the critical issue of forward current leakage in hardware neural networks.
• The achieved sub-pA/µm forward current demonstrates significant potential for energy-efficient neuromorphic computing.