Wafer-Scale All-Silicon Self-Rectifying Memristor for Synaptic Response and Reservoir Computing

Area of Science:

• Materials Science
• Solid State Physics
• Neurotechnology

Background:

• Silicon p-n junctions are fundamental electronic components.
• Abrupt p-n junctions are a standard semiconductor model.
• Memristive devices and neuromorphic computing are advancing electronics.

Purpose of the Study:

• To create a silicon p-n junction with memristive properties.
• To investigate synaptic behaviors for reservoir computing (RC).
• To demonstrate the device's potential for neuromorphic applications.

Main Methods:

• Fabrication of a silicon p-n junction with an oxide interfacial layer.
• Characterization of resistive switching and rectification properties.
• Evaluation of synaptic behaviors like paired-pulse facilitation and spike-timing-dependent plasticity.

Main Results:

• The device demonstrated high rectification ratio (~5000) and endurance (4.5 × 10^6 cycles) without filament formation.
• Reproducible synaptic behaviors, including short- and long-term memory transitions, were observed.
• Achieved 86.9% accuracy in handwritten digit classification using reservoir computing with 4-bit pulse stimulation.

Conclusions:

• An interface-engineered all-silicon device integrates memristive functions with diode physics.
• The device shows potential for wafer-scale neuromorphic computing platforms.
• Charge-trapping dynamics underpin the observed synaptic functionalities.