Synaptic Segmented Transistor with Improved Linearity by Schottky Junctions and Accelerated Speed by Double-Layered...

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• Neuromorphic computing aims to overcome von Neumann architecture limitations for artificial intelligence.
• Synaptic devices are crucial for hardware integration in neuromorphic systems, requiring high compactness and silicon-based memory for mass production.
• Existing silicon-based synaptic devices face challenges in linearity and operational speed.

Purpose of the Study:

• To propose a novel segmented transistor for artificial synapses to enhance the linearity of potentiation and depression (P/D).
• To improve the operational speed of synaptic devices through innovative material engineering.
• To advance the development of CMOS-compatible synaptic devices for potential mass production.

Main Methods:

• A complementary metal oxide semiconductor (CMOS)-compatible segmented transistor architecture was designed and fabricated.
• Non-ohmic Schottky junctions at the source and drain were utilized to improve weight linearity.
• A double-layered nitride (SiN) charge trap layer with varied stoichiometries was employed to boost operation speed.
• Drain current-gate voltage transfer characteristics were analyzed to identify distinct operational segments.

Main Results:

• The proposed segmented transistor exhibits three distinct segments in its drain current-gate voltage transfer characteristics, attributed to the Schottky junctions.
• The use of double-layered nitride with different stoichiometries significantly enhanced the operation speed of the artificial synapse.
• The device demonstrated improved linearity in potentiation and depression, crucial for accurate synaptic weight updates.
• The CMOS-compatible nature of the device facilitates integration with existing semiconductor fabrication processes.

Conclusions:

• The novel segmented transistor design successfully addresses key limitations in artificial synaptic devices, particularly linearity and speed.
• This silicon-based device represents a significant step towards the realization of mass-produced, hardware-based neuromorphic systems.
• The integration of Schottky junctions and tailored nitride layers offers a promising pathway for future neuromorphic hardware development.