CdS-GeSe Heterostructure-Based Bimodal Memristor with Tunable Synaptic Plasticity and Mixed-Signal Switching for Neuromorphic Hardware

Abstract

In this work, the development and optoelectrical characterization of an Ag/CdS-GeSe/FTO memristor device designed for optically triggered neuromorphic computing applications. The device demonstrates an analog-to-digital transition, exhibiting analog behavior below 0.8 V and digital behavior above 0.8 V (in the pristine cell) after electroforming, the device shows digital switching within +/- 0.5 V, which enables multifunctional operation suitable for both synaptic learning and memory storage. The fabricated memristor shows a high memory window, ensuring stable and reliable data retention with excellent reproducibility over multiple cycles. The device is investigated under both electrical and optical stimuli, mimicking synaptic functions such as potentiation, depression, paired-pulse facilitation (PPF), and paired-pulse depression (PPD) index. It exhibits significant nonlinearity, which is crucial for pattern recognition tasks in neuromorphic systems. In addition to electrical stimulation, optical pulses (405 nm) are applied to the device, enhancing its conductance states and enabling optical modulation of synaptic behavior. Additionally, Fashion-MNIST and digit MNIST datasets are used to evaluate the device's performance in image recognition tasks, demonstrating its application in artificial neural networks (ANNs) for pattern recognition with 75% and 95% accuracy, respectively, in optoelectrical mode, relatively higher than in electrical mode. The findings demonstrate the promise of the Ag/CdS-GeSe/FTO memristor in neuromorphic computing, with potential applications in machine learning, memory storage, and optical modulation.