Event‐Driven Neuroplasticity and Spiking Modulation in a Photoelectric Neuristor Configured by Threshold Switching Memristor and Optoelectronic Transistor

• Published in: Advanced functional materials Vol. 35; no. 2
• Main Authors: Chen, Kuan‐Ting, Lin, Pei‐Lin, Huang, Ya‐Chi, Chen, Shuai‐Ming, Liao, Zih‐Siao, Chen, Jen‐Sue
• Format: Journal Article
• Language: English
• Published: Hoboken Wiley Subscription Services, Inc 01.01.2025
• Subjects: artificial neuron; associative learning; Electrical stimuli; Gallium; Indium gallium zinc oxide; leaky integrate‐and‐fire; Memristors; Neural networks; Neuromorphic computing; Neuroplasticity; Optical properties; Optoelectronics; optosynapse; Photoelectricity; Spiking; spiking neural network; Transistors
• DOI: 10.1002/adfm.202412452
• ISSN: 1616-301X

Integrating and implementing spiking neurons and synapse into neuromorphic hardware aligned with spiking neural networks (SNNs) offer significant promise for energy‐efficient operation and decision making. In this work, a stacked artificial synapse and spiking neuron utilizing an indium gallium zinc oxide (IGZO) optosynaptic transistor paired with a vanadium‐based volatile threshold switching memristor are constructed. This compact neuristor encompasses multiple functionalities including the conversion of optical impulses into electrical signals, modifiable post‐synaptic current‐enhanced features, and the implementation of leaky integrate‐and‐fire (LIF) spiking generation behavior, showcasing the capability of information delivery in SNNs. The spiking activity within the proposed configuration can be effectively modulated through the interplay of optical and electrical stimuli. Additionally, the excitatory and inhibitory properties manifested by the spiking behavior underscore the gate‐tunable neuron excitability. Notably, the capacity for accommodating hybrid inputs operation makes achievement of spike‐based associative learning by reviving the Pavlov's dog experiment in the proposed device. Moreover, this research unveils the synaptic weight‐governed spiking activity, demonstrating the sophisticated input–output characteristics of spiking behavior. The stacked memristor and transistor assembly can advance the neuromorphic technologies and lay the foundation for the realization of physical SNNs. A stacked memristor and transistor assembly is engineered to integrate optical‐to‐electrical signal conversion, adjustable current potentiation, and controlled spiking activity. This device enables spiking behavior modulation through optical and electrical inputs, facilitating associative learning via persistent photoconductivity and threshold switching. This study also demonstrates synaptic weight‐governed spiking activity using the leaky integrate‐and‐fire principle, highlighting its applicability for neuromorphic computing.