講演情報
[15p-P07-33]積層反強磁性体における非線形ダイナミクスとリザーバー計算
〇新海 拓夢1、飯浜 賢志1、林 兼輔1、森山 貴広1 (1.名古屋大工)
キーワード：
スピン波リザバー計算、積層反強磁性体
Artificial intelligence (AI) using neural networks has become an indispensable part in modern society. However, these AIs use a large number of conventional CPUs and GPUs, leading to huge energy consumption. In this respect, physical implementation of neural networks such as physical reservoir computing (PRC) has attracted much attention. The PRC based on non-local magnetization dynamics, i.e. spin -wave, is promising towards high-performance reservoir computing at nano-scale with low energy consumption [1]. Synthetic antiferromagnets (SAF), where ferromagnetic layers coupled antiferromagnetically separated by a nonmagnetic layer, have complicated magnetization dynamics and rich physics, which are essential properties for PRC. In this study, we investigated nonlinear and complicated non-local magnetization dynamics in SAF based on micromagnetic simulation.
Mumax3 was used to perform micromagnetic simulations. Figure 1(a) shows a schematic illustration of the setup used in this study. Physical nodes were arranged in a lattice pattern on the SAF and a current-induced spin-transfer torque was applied on the top ferromagnetic layer to excite spin -waves. Figure 1(b) shows the heatmap of maximum signals of fast Fourier transform (FFT) spectra of magnetization dynamics plotted against the out-of-plane magnetic field and the current density. Spin-torque auto-oscillation in SAF was observed at the red region. We will discuss the nonlinear current-induced magnetization dynamics in SAF and their performance for PRC.
Figure 1 (a) Schematic illustration of the setup for current-induced magnetization dynamics in SAF, (b) Heat map of the FFT intensity of magnetization dynamics against the out-of-plane magnetic field and the current density.
Reference: [1] Iihama et al., npj Spintronics 2, 5 (2024).
Mumax3 was used to perform micromagnetic simulations. Figure 1(a) shows a schematic illustration of the setup used in this study. Physical nodes were arranged in a lattice pattern on the SAF and a current-induced spin-transfer torque was applied on the top ferromagnetic layer to excite spin -waves. Figure 1(b) shows the heatmap of maximum signals of fast Fourier transform (FFT) spectra of magnetization dynamics plotted against the out-of-plane magnetic field and the current density. Spin-torque auto-oscillation in SAF was observed at the red region. We will discuss the nonlinear current-induced magnetization dynamics in SAF and their performance for PRC.
Figure 1 (a) Schematic illustration of the setup for current-induced magnetization dynamics in SAF, (b) Heat map of the FFT intensity of magnetization dynamics against the out-of-plane magnetic field and the current density.
Reference: [1] Iihama et al., npj Spintronics 2, 5 (2024).