Spin-wave reservoir chips with short-term memory for high-speed estimation of external magnetic fields

被引:1
作者
Nagase, Sho [1 ]
Nezu, Shoki [1 ]
Sekiguchi, Koji [2 ,3 ,4 ]
机构
[1] Yokohama Natl Univ, Grad Sch Engn Sci, Tokiwadai 79-5, Yokohama 2408501, Japan
[2] Yokohama Natl Univ, Inst Adv Sci, Tokiwadai 79-5, Yokohama 2408501, Japan
[3] Yokohama Natl Univ, Inst Multidisciplinary Sci, Tokiwadai 79-5, Yokohama 2408501, Japan
[4] Yokohama Natl Univ, Fac Engn, Tokiwadai 79-5, Yokohama 2408501, Japan
来源
PHYSICAL REVIEW APPLIED | 2024年 / 22卷 / 02期
基金
日本学术振兴会;
关键词
COMPUTATION;
D O I
10.1103/PhysRevApplied.22.024072
中图分类号
O59 [应用物理学];
学科分类号
摘要
The experimental realization of a spin-wave reservoir chip employing ferromagnetic permalloy thin films is presented. The novel device facilitates the interference of three spherical wave-excited surface mode spin waves within a rectangular waveguide via strategically positioned slits, enabling the detection of electrical signals from surface mode spin waves across all four observation antennas. Through the experiments conducted, it is confirmed that the device functions as a one-input, four-output reservoir capable of estimating external magnetic fields. Notably, the results demonstrate the device's capacity to retain memory up to one step prior in short-term memory tasks, while confirming the effectiveness of spinwave interference induced by Huygens slits in enhancing nonlinearity, as observed in parity-check tasks. Furthermore, the inclusion of additional detection antennas contributes to improved learning accuracy, highlighting the significant progress achieved by the spin-wave reservoir chip. These findings underscore substantial progress toward practical implementation, with promising avenues for further development and refinement, showing its remarkable ability to process signals at high speeds, even with 0.8-ns pulse sequences.
引用
收藏
页数:10
相关论文
共 44 条
  • [1] New results on recurrent network training: Unifying the algorithms and accelerating convergence
    Atiya, AF
    Parlos, AG
    [J]. IEEE TRANSACTIONS ON NEURAL NETWORKS, 2000, 11 (03): : 697 - 709
  • [2] Real-time computation at the edge of chaos in recurrent neural networks
    Bertschinger, N
    Natschläger, T
    [J]. NEURAL COMPUTATION, 2004, 16 (07) : 1413 - 1436
  • [3] Machine learning based on reservoir computing with time-delayed optoelectronic and photonic systems
    Chembo, Yanne K.
    [J]. CHAOS, 2020, 30 (01)
  • [4] All-optical reservoir computer based on saturation of absorption
    Dejonckheere, Antoine
    Duport, Francois
    Smerieri, Anteo
    Fang, Li
    Oudar, Jean-Louis
    Haelterman, Marc
    Massar, Serge
    [J]. OPTICS EXPRESS, 2014, 22 (09): : 10868 - 10881
  • [5] Mode-dependent magnonic noise
    Furukawa, Ryo
    Nezu, Shoki
    Eguchi, Takuro
    Sekiguchi, Koji
    [J]. NPG ASIA MATERIALS, 2024, 16 (01)
  • [6] Macromagnetic Simulation for Reservoir Computing Utilizing Spin Dynamics in Magnetic Tunnel Junctions
    Furuta, Taishi
    Fujii, Keisuke
    Nakajima, Kohei
    Tsunegi, Sumito
    Kubota, Hitoshi
    Suzuki, Yoshishige
    Miwa, Shinji
    [J]. PHYSICAL REVIEW APPLIED, 2018, 10 (03):
  • [7] Deep reservoir computing: A critical experimental analysis
    Gallicchio, Claudio
    Micheli, Alessio
    Pedrelli, Luca
    [J]. NEUROCOMPUTING, 2017, 268 : 87 - 99
  • [8] Gurevich A., 1996, Magnetization Oscillations and Waves
  • [9] Reservoir Computing Beyond Memory-Nonlinearity Trade-off
    Inubushi, Masanobu
    Yoshimura, Kazuyuki
    [J]. SCIENTIFIC REPORTS, 2017, 7
  • [10] Jaeger H, 2001, 14834 GMD GERM NAT R, V148, P34