EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Field‐free programmable bipolar magnetic heterostructures for neuromorphic computing

Yaping He, Gaofei Wang, Jinfeng Zhai, Wentao Huang, Zhenyu Zhao, Zheng Zhang, Jiabin Shen, Pan He, Zengguang Cheng, Peng Zhou

InfoMat · 2026-05 · 8 (5)

DOI: 10.1002/inf2.70130 ↗

Kurz gefasst

Fachartikel in der Zeitschrift InfoMat, online erschienen am 09.03.2026. Verfasst von Yaping He und 9 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, Fudan University.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Neuromorphic computing systems inspired by biological principles have garnered significant attention for their superior energy efficiency and computational capabilities, with memristors serving as fundamental building blocks. However, realizing biologically plausible signal processing—particularly in applications like biological vision—requires multi‐state bidirectional (excitatory/inhibitory) responses to emulate synaptic modulation, a functionality unattainable with conventional unidirectional resistive memristors. Here, we demonstrate a field‐free spin–orbit torque magnetic memristor based on Fe 3 GeTe 2 /WTe 2 heterostructures, operating within the 110–150 K temperature range, that overcomes this limitation. The thickness‐dependent ferromagnetic properties of Fe 3 GeTe 2 enable multi‐state Hall resistance modulation, while WTe 2 's localized spin injection precisely controls magnetic domain reversals in Fe 3 GeTe 2 —without an external magnetic field—ensuring the compatibility with CMOS platform. This synergistic approach achieves eight distinct resistance states (3‐bit precision) spanning both polarities. We demonstrate the technology's neuromorphic potential through two benchmarks: high‐fidelity image feature extraction (matching software‐implemented algorithms) and competitive neural network clustering, both achieved with minimal hardware overhead. Our work provides a device‐level solution for implementing signed weight updates in neuromorphic hardware, opening new avenues for energy‐efficient computing systems capable of complex signal processing. image

Förderung

  • National Key Research and Development Program of China
  • National Natural Science Foundation of China
  • Fudan University

Zitieren

APA

He, Y., Wang, G., Zhai, J., Huang, W., Zhao, Z., Zhang, Z., Shen, J., He, P., Cheng, Z., & Zhou, P. (2026). Field‐free programmable bipolar magnetic heterostructures for neuromorphic computing. InfoMat, 8(5). https://doi.org/10.1002/inf2.70130

Vancouver

He Y, Wang G, Zhai J, Huang W, Zhao Z, Zhang Z, et al. Field‐free programmable bipolar magnetic heterostructures for neuromorphic computing. InfoMat. 2026;8(5). doi:10.1002/inf2.70130

BibTeX
@article{He2026,
  author = {He, Yaping and Wang, Gaofei and Zhai, Jinfeng and Huang, Wentao and Zhao, Zhenyu and Zhang, Zheng and Shen, Jiabin and He, Pan and Cheng, Zengguang and Zhou, Peng},
  title = {Field‐free programmable bipolar magnetic heterostructures for neuromorphic computing},
  journal = {InfoMat},
  year = {2026},
  volume = {8},
  number = {5},
  doi = {10.1002/inf2.70130}
}
Kostenlos starten →