EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

All‐Electrical Control of Spin Synapses for Neuromorphic Computing: Bridging Multi‐State Memory with Quantization for Efficient Neural Networks

Tzu-Chuan Hsin iD, Chun-Yi Lin iD, Po-Chuan Wang iD, Chun Yang, Chi-Feng Pai iD

Advanced Science · 2025-06 · 12 (22)

DOI: 10.1002/advs.202417735 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Advanced Science, online erschienen am 26.04.2025. Verfasst von Tzu-Chuan Hsin und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Materials Science and Engineering, National Taiwan University, Taipei, 10617, Taiwan. Die Arbeit wurde bisher 9-mal zitiert (Crossref) und nennt 77 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Science and Technology Council, Center of Atomic Initiative for New Materials (AI-Mat), Semiconductor Fabrication Lab of the Consortia of Key Technologies.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The development of energy‐efficient, brain‐inspired neuromorphic computing demands advanced memory devices capable of mimicking synaptic behavior to achieve high accuracy and adaptability. In this study, three types of all‐electrically controlled, field‐free spin synapse devices designed with unique spintronic structures presented: the Néel orange‐peel effect, interlayer Dzyaloshinskii‐Moriya interaction (i‐DMI), and tilted anisotropy. To systematically evaluate their neuromorphic potential, a benchmarking framework is introduced that characterizes cycle‐to‐cycle (CTC) variation, a critical factor for reliable synaptic weight updates. Among these designs, the tilted anisotropy device achieves an 11‐state memory with minimal CTC variation (2%), making it particularly suited for complex synaptic emulation. Through comprehensive benchmarking, this multi‐state device in convolutional neural networks (CNNs) using post‐training quantization is implemented. Results indicate that per‐channel quantization, particularly with the min‐max and mean squared error (MSE) observers, enhances classification accuracy on the CIFAR‐10 dataset, achieving up to 81.51% and 81.12% in ResNet‐18—values that closely approach the baseline accuracy. This evaluation underscores the potential of field‐free spintronic synapses in neuromorphic architectures, offering an area‐efficient solution that integrates multi‐state functionality with robust switching performance. The findings highlight the promise of these devices in advancing neuromorphic computing, contributing to energy‐efficient, high‐performance systems inspired by neural processes.

Schlagwörter

Stichwörter der Autor:innen

field‐free switchingneural networkneuromorphic computingperpendicular magnetic anisotropyspin‐orbit torques

Autor:innen und Institutionen

  • Department of Materials Science and Engineering, National Taiwan University, Taipei, 10617, Taiwan.

ORCID-Profile: Tzu-Chuan Hsin, Chun-Yi Lin, Po-Chuan Wang, Chi-Feng Pai

Förderung

  • National Science and Technology Council
  • Center of Atomic Initiative for New Materials (AI-Mat)
  • Semiconductor Fabrication Lab of the Consortia of Key Technologies
  • Nano-Electro-Mechanical-System Research Center

Zitieren

APA

Hsin, T. C., Lin, C. Y., Wang, P. C., Yang, C., & Pai, C. F. (2025). All‐Electrical Control of Spin Synapses for Neuromorphic Computing: Bridging Multi‐State Memory with Quantization for Efficient Neural Networks. Advanced Science, 12(22). https://doi.org/10.1002/advs.202417735

Vancouver

Hsin TC, Lin CY, Wang PC, Yang C, Pai CF. All‐Electrical Control of Spin Synapses for Neuromorphic Computing: Bridging Multi‐State Memory with Quantization for Efficient Neural Networks. Advanced Science. 2025;12(22). doi:10.1002/advs.202417735

BibTeX
@article{Hsin2025,
  author = {Hsin, Tzu-Chuan and Lin, Chun-Yi and Wang, Po-Chuan and Yang, Chun and Pai, Chi-Feng},
  title = {All‐Electrical Control of Spin Synapses for Neuromorphic Computing: Bridging Multi‐State Memory with Quantization for Efficient Neural Networks},
  journal = {Advanced Science},
  year = {2025},
  volume = {12},
  number = {22},
  doi = {10.1002/advs.202417735}
}
Kostenlos starten →