PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
December 8, 2025Applied Physics Letters3 citations

Electro-optically co-modulated ZnS synaptic memristor for neuromorphic recognition systems

View Full Paper
HLHailong LiHSHao SunXZXiang Zhang

Key Points

  • Classification accuracy in handwritten digit recognition reached 88.25%, with implications for practical applications.
  • Key metrics include an On/Off ratio of about 26 and a retention time exceeding 104 seconds, showcasing advanced performance.
  • The devices exhibit reliable resistive switching with significant dependencies on electrical stimulation and optical signals.
  • Findings support the potential for developing enhanced neuromorphic computing systems using ZnS-based memristors.

Abstract

Memristor-based neuromorphic computing offers a revolutionary strategy to address the limitations of traditional computing architectures. Developing synaptic memristors co-modulated by electrical and optical signals is crucial for realizing neural networks with high-efficiency parallel processing and in-memory computing, yet it remains a significant challenge. Herein, wide-bandgap zinc sulfide (ZnS) is introduced to design Ag/ZnS/FTO optoelectronic synaptic memristors. The devices verify reliable resistive switching (RS) behavior, primarily attributed to being dominated by sulfur vacancies (VS), with a narrow Set/Reset distribution (variation 0.04/0.03 V), an On/Off ratio of ∼26, and a retention time exceeding 104 s. Under electrical, especially near-infrared light (808 and 980 nm) stimulation, these memristors accurately mimic diverse synaptic plasticity functions, including excitatory post-synaptic current, short-term/long-term memory, long-term potentiation/depression, paired-pulse facilitation/depression, spike-timing-dependent plasticity, spike-voltage-dependent plasticity, spike-dependent dynamic plasticity, spike-rate-dependent plasticity, and Ebbinghaus learning–forgetting behaviors. Notably, applying to handwritten digit recognition on the MNIST dataset, the system achieves an 88.25% classification accuracy, demonstrating its potential for practical neuromorphic applications. These findings open an avenue for the development of sulfide-based optoelectronic synaptic devices and advanced neuromorphic computing systems.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Li et al. (2025) studied this question.

synapsesocial.com/papers/69401f0f2d562116f28fa1d7https://doi.org/10.1063/5.0289512
Ask AI
Helpful
Bookmark
Share
View Full Paper