Nanograin network memory with reconfigurable percolation paths for synaptic interactions

Area of Science:

• Materials Science
• Nanotechnology
• Neuroscience

Background:

• Efficient computation requires memory devices that can process and store data simultaneously.
• Artificial synaptic devices are crucial for neuromorphic computation and hybrid networks with biological neurons.
• Existing electrical synaptic devices suffer from irreversible aging and performance degradation.

Purpose of the Study:

• To develop a novel nanograin network memory device with simultaneous data processing and storage capabilities.
• To overcome the limitations of current artificial synaptic devices, particularly their aging and performance degradation.
• To demonstrate electrical and photonic control for analog and reversible adjustment of persistent current levels.

Main Methods:

• Fabrication of a single silicon nanowire with reconfigurable percolation paths, featuring solid core/porous shell and pure solid core segments.
• Utilizing electrical and photonic control to manipulate current percolation paths.
• Demonstrating synaptic behaviors including memory, erasure, potentiation, habituation, and elimination.

Main Results:

• The nanograin network memory exhibited analog and reversible adjustment of persistent current levels through electrical and photonic control.
• Photonic habituation was achieved via laser illumination on the porous nanowire shell, resulting in a linear decrease in postsynaptic current.
• Synaptic elimination was successfully emulated using two interconnected devices on a single nanowire.

Conclusions:

• The developed silicon nanowire device demonstrates effective memory behavior and current suppression.
• Electrical and photonic reconfiguration of conductive paths in Si nanograin networks offers a promising approach for next-generation nanodevice technologies.
• This technology paves the way for advanced neuromorphic computing applications.