PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
February 2, 2026Advanced Functional Materials0 citations

Areal Trap‐Driven Analog Resistive Switching in InP/ZnSe/ZnS Quantum Dot Memristors for Neuromorphic Computing

View Full Paper
YKYeon Jun KimGBGeun Woo BaekGPGanghyun Park

Key Points

  • To investigate the analog resistive switching capabilities of InP/ZnSe/ZnS quantum dot memristors for neuromorphic applications.
  • Developed a multilayer ZnO/QD/ZnO architecture to enhance charge trapping.
  • Utilized ZnO interlayers to improve electron transport and suppress recombination.
  • Controlled oxidation of the quantum dot surface to create interfacial trap sites.
  • Thinned the quantum dot layer to prevent trap accumulation.
  • Achieved robust trap-controlled analog resistive switching in the device.
  • Reproduced synaptic behaviors similar to biological systems, including potentiation and depression.
  • Demonstrated 92% accuracy on multilayer perceptron and 87.3% on convolutional neural network tasks using CIFAR-10 dataset.

Abstract

ABSTRACT Memristors enable parallel computation in next‐generation memory systems. Among the materials investigated, colloidal quantum dots (QDs) are attractive owing to their size‐tunable electronic structure, solution processability, and versatile surface chemistry. Nevertheless, most QD‐based memristors exhibit digital resistive switching (D‐RS) with abrupt binary transitions, limiting their neuromorphic potential. Here, an InP/ZnSe/ZnS QD‐based memristor exhibiting analog resistive switching (A‐RS) is demonstrated through a multilayer ZnO/QD/ZnO architecture engineered to promote uniform areal charge trapping. ZnO nanoparticle interlayers enhance electron transport, block hole injection to suppress recombination, and induce interfacial trap sites via controlled oxidation of the QD surface. Additional thinning of the QD layer mitigates vertical trap accumulation, ensuring stable and spatially distributed conduction. The resulting device exhibits robust trap‐controlled A‐RS and reproduces synaptic behaviors, including excitatory/inhibitory postsynaptic currents and long‐term potentiation/depression. Benchmarking with hardware‐level neural network simulations demonstrates excellent performance, achieving recognition accuracies of 92% in the multilayer perceptron model and 87.3% in the convolutional neural network trained on the CIFAR‐10 dataset, both approaching the performance of an ideal memristor. These results establish a viable pathway for exploiting the intrinsic properties of colloidal QDs in high‐performance memristors, demonstrating clear advantages over conventional D‐RS counterparts and underscoring their potential for neuromorphic compute‐in‐memory architectures.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Kim et al. (2026) studied this question.

synapsesocial.com/papers/6980ffa4c1c9540dea81239ehttps://doi.org/10.1002/adfm.202532187
Ask AI
Helpful
Bookmark
Share
View Full Paper