3D AND-Type Stacked Array for Neuromorphic Systems.

Area of Science:

• Neuromorphic Engineering
• Solid-State Devices
• Memory Technology

Background:

• Neuromorphic systems require efficient synaptic devices for vector-by-matrix multiplication (VMM).
• Existing NOR/AND flash memory cells are too large for dense neuromorphic integration.
• Memristor devices offer smaller footprints but face challenges in scalability and variability.

Purpose of the Study:

• To propose a novel 3D AND-type stacked array architecture for NOR/AND flash memory.
• To reduce the physical footprint of synaptic devices in neuromorphic computing.
• To enable low-power, high-efficiency VMM operations.

Main Methods:

• Fabrication of a 3D AND-type stacked array using a tilted implantation technique.
• Technology Computer-Aided Design (TCAD) simulations for structural and electrical validation.
• Controlled deposition thickness to mitigate cell-to-cell variations.
• Utilization of Fowler-Nordheim (FN) tunneling for synaptic weight updates.

Main Results:

• Successful formation of conformal sources and drains in each cell via tilted implantation.
• Elimination of cell-to-cell variations by controlling deposition thickness.
• Demonstration of parallel drain and source lines enabling simple program/inhibit schemes.
• Achieved low-power conductance updates for synaptic devices.

Conclusions:

• The proposed 3D AND-type stacked array effectively reduces cell size for NOR/AND flash memory in neuromorphic applications.
• This architecture facilitates efficient and low-power VMM operations through FN tunneling.
• The design offers a promising solution for scalable and reliable neuromorphic hardware.