Optically Tunable Electrical Oscillations in Oxide-Based Memristors for Neuromorphic Computing.

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Hardware-based neural networks require efficient sensory input integration.
• Volatile threshold switching materials offer potential for neuromorphic computing.

Purpose of the Study:

• To investigate direct optical control of oscillatory neurons using V₃O₅.
• To explore the underlying mechanisms of optical tuning in V₃O₅ devices.
• To demonstrate the application of these devices in neuromorphic computing.

Main Methods:

• Fabrication and characterization of V₃O₅ based oscillatory neuron devices.
• Temperature-dependent electrical measurements, conductive atomic force microscopy (C-AFM), and in situ thermal imaging.
• Lumped element modeling to understand device physics.
• Demonstration of in-sensor reservoir computing and optical encoding for spiking neural networks (SNNs).

Main Results:

• Direct optical control of V₃O₅ oscillatory neurons achieved.
• Electroforming-free operation with optically tunable switching parameters.
• V₃O₅ exhibits significant photoconductive and bolometric characteristics, increasing conductivity.
• Identified V₃O₅ as a novel bolometric material with a high TCR of -4.6% K^-1 at 423 K.
• Successful demonstration of in-sensor reservoir computing and optical SNN encoding.

Conclusions:

• V₃O₅ is a promising material for optically controlled neuromorphic hardware.
• The interplay of photoconductive and bolometric effects governs device behavior and oscillation dynamics.
• These devices offer a pathway towards energy-efficient and compact in-sensor computing architectures.