betaFeedback

WorksPhysics and Astronomy

Unraveling the dynamics of magnetization in topological insulator-ferromagnet heterostructures via spin-orbit torque

Abstract

Spin-orbit coupling stands as a pivotal determinant in the realm of condensed matter physics. In recent, its profound influence on spin dynamics opens up a captivating arena with promising applications. Notably, the topological insulator-ferromagnet heterostructure has been recognized for inducing spin dynamics through applied current, driven by spin-orbit torque. Building upon recent observations revealing spin flip signals within this heterostructure, our study elucidates the conditions governing spin flips by studying the magnetization dynamics. We establish that the interplay between spin-anisotropy and spin-orbit torque plays a crucial role in shaping the physics of magnetization dynamics within the heterostructure. Furthermore, we categorize various modes of magnetization dynamics, constructing a comprehensive phase diagram across distinct energy scales, damping constants, and applied frequencies. This research not only offers insights into controlling spin direction but also charts a new pathway to the practical application of spin-orbit coupled systems.

Introduction

Spin-orbit coupling (SOC), acknowledged as one of the fundamental interactions in materials, [1] reveals a fertile landscape within condensed matter physics. A standout illustration of its impact is evident in the burgeoning field of spintronics, where SOC-driven spin dynamics unveils a promising horizon for practical applications. Accordingly, recent attention has been drawn to topological insulator-ferromagnet (TI-FM) heterostructures, as evinced by notable reports. [2, 3, 4, 5, 6, 7] The surge in interest stems from their remarkable efficiency in catalyzing spin dynamics, primarily attributed to the spin-orbit torque (SOT) induced by an applied current. [8, 9]

These heterostructures exhibit intriguing phenomena that closely follow its intricate dynamics of spins. For instance, an emergent inductance from the motive force in spiral magnets [10, 11, 12] was extended to the TI-FM heterostructures. [13, 14, 15] Nonreciprocal transport phenomena associated with the SOC was explored. [16, 17] The significant increase of Curie temperature was reported as well. [18, 19, 20, 21] Notably, it has been acknowledged that the SOT at the TI-FM interface is robust enough to flip magnetization and induce the sign change of Hall Effect. [22, 23, 24, 25] [26]

Such properties hold promise for applications in spintronics devices, particularly in utilizing topological SOTs for magnetic memories. 27–32

Motivated by such intriguing phenomena due to the SOC and SOT, this study explores magnetization dynamics in TI-FM heterostructures with an applied current, aiming to unveil the conditions for magnetization flip. Considering both spin-anisotropy and SOT, we establish the equilibrium state of magnetization and develop a model describing its dynamics under direct current (DC) or alternating current (AC). In the absence of damping, we identify oscillating, faltering, and flipping modes for DC, with the latest inducing magnetization. tization flip. The choice between modes is determined by the relative strengths of SOT to spin-anisotropy. With the introduction of damping, the occurrence of spin flip hinges on the duration elapsed until the crossover from flipping to oscillating modes takes place.

Magnetization dynamics of TI-FM heterostructure

Figure 1. Magnetization dynamics of TI-FM heterostructure. (a) The schematics of the TI-FM heterostructure. (b) The equilibrium state of magnetization in the different relative strengths of SOT x=σ/Kx = \sigma/K.

For AC, on the other hand, we discern three final states—adiabatic, resonating, and chaotic. We classify five distinct modes in the adiabatic state at low frequencies, which is derived from DC modes. By providing the phase diagrams of adiabatic modes, we illustrate that the modes are determined by SOT, spin-anisotropy, driving frequency, and initial driving phase. Lastly, we explore their transition to resonating and chaotic states at higher frequencies in the viewpoint of the Fourier transform, revealing that the periodic array of peaks gives rise to the chaotic state. By overhauling the dynamics, we provide insights into the complex dynamics of magnetization in TI-FM heterostructures.

Results

Model and its equilibrium

The physical configuration is illustrated in Fig. 1(a), where a current flows along the xx-axis, the itinerant electron spin aligns with the yy-axis, and the ferromagnet’s magnetization lie along the +z+z-axis at t=0t=0. The polar and azimuthal angles with respect to the yy-axis are denoted as θ\theta and ϕ\phi. We represent the magnetization as m⃗=m(sin⁡θsin⁡ϕ,cos⁡θ,sin⁡θcos⁡ϕ)\vec{m}=m(\sin\theta\sin\phi,\cos\theta,\sin\theta\cos\phi) and the itinerant electron spin as σ⃗=σ(t)(0,1,0)\vec{\sigma}=\sigma(t)(0,1,0). For simplicity, we exclusively focus on the dynamics of the single m⃗\vec{m}, and ignore that of σ⃗\vec{\sigma}. This is justified by the stronger spin-momentum locking of the TI surface state compared to anisotropy. [3, 15, 35, 36] [33, 34] For instance, Ni has the easy-axis anisotropy energy about 2.7 μeV2.7\ \mu\mathrm{eV}/atom in the bulk[37] and 4.0 meV4.0\ \mathrm{meV}/atom in the monolayer,[38] while the spin-orbit coupling of Bi2Se3\mathrm{Bi}_{2}\mathrm{Se}_{3} is estimated as ∼1.2 eV\sim1.2\ \mathrm{eV}/atom. Accordingly, we set our unit of energy to be 1.0 μeV∼2.4 GHz1.0\ \mu\mathrm{eV}\sim2.4\ \mathrm{GHz} and the unit of time to be (2.4 GHz)−1≈0.42 ns(2.4\ \mathrm{GHz})^{-1}\approx0.42\ \mathrm{ns}. The spin flip is then denoted as the switching of spin direction between +z+z and −z-z half space.

Two crucial potential energies emerge: VA=−K2mz2V_A=-\frac{K}{2}m_z^2 (K>0K>0), representing magnetic anisotropy in the ferromagnet, and VS=−γm⃗⋅σ⃗V_S=-\gamma\vec{m}\cdot\vec{\sigma} signifying the Rashba-field-like[39, 40] (or domain-damping-like)[41] SOT. The equilibrium state of magnetization is obtained, showcased in Fig. 1(b) for fixed σ⃗\vec{\sigma}. Normalizing m=γ=1m=\gamma=1, key energy scales become the anisotropy energy KK and SOT σ\sigma. The dimensionless parameter x=σKx=\frac{\sigma}{K} is defined. As xx increases from 00, the equilibrium states initially align along ±z\pm z-axis and gradually rotate towards the yy-axis. In terms of angles, sin⁡θ=x\sin\theta=x, ϕ=±π/2\phi=\pm\pi/2. Before reaching the yy-axis, the equilibrium states are twofold degenerate. For x≥1x\ge1, the degenerate equilibrium states converge at the yy-axis.

To describe magnetization dynamics out of equilibrium, the Lagrangian density is considered:

L=LB−VA−VS,LB=m(1−cos⁡θ)ϕ˙.(1)\mathcal{L}=\mathcal{L}_B-V_A-V_S,\qquad\mathcal{L}_B=m(1-\cos\theta)\dot{\phi}. \tag*{(1)}

Note that LB\mathcal{L}_B is the Berry phase term, making polar and azimuthal angles conjugate. [42, 43]

Thus, as magnetization rotates from the zz-axis to its equilibrium state, deviations from the yzyz-plane induce complex motion. Introducing Gilbert damping through the Rayleigh dissipation function, [44] R=η2∂m⃗∂t⋅∂m⃗∂t\mathcal{R} = \frac{\eta}{2}\frac{\partial\vec{m}}{\partial t}\cdot\frac{\partial\vec{m}}{\partial t}, with the dimensionless damping constant η\eta, yields the equations of motion:

(1+η2)θ˙=−K2sin⁡θsin⁡2ϕ+ηK2sin⁡2θcos⁡2ϕ−ησ(t)sin⁡θ, and(2)(1+\eta^2)\dot{\theta}=-\frac{K}{2}\sin\theta\sin2\phi+\frac{\eta K}{2}\sin2\theta\cos^2\phi-\eta\sigma(t)\sin\theta,\ \text{and} \tag*{(2)}
(1+η2)ϕ˙=−Kcos⁡θcos⁡2ϕ+σ(t)−ηK2sin⁡2ϕ.(3)(1+\eta^2)\dot{\phi}=-K\cos\theta\cos^2\phi+\sigma(t)-\frac{\eta K}{2}\sin2\phi. \tag*{(3)}

The initial conditions are set at θ=π/2\theta=\pi/2 and ϕ=0\phi=0. This system of equations governs physics for both DC and AC scenarios. The numerical computation is performed by Mathematica.

It should be noted that the field-like SOT σ(t)\sigma(t) precesses the magnetization around yy-axis while the damping-like SOT ησ(t)\eta\sigma(t) cants the magnetization toward yy-axis. This is consistent with the Landau-Lifshitz-Gilbert equation. It is noteworthy that the relation between the damping constant η\eta and the Gilbert damping constant α\alpha is α=ηγ\alpha=\eta\gamma, where γ\gamma is the gyromagnetic ratio. [44] We already set γ=1\gamma=1, resulting in α=η\alpha=\eta. As a consequence, the field-like SOT is responsible for the spin flip, whereas the damping-like SOT deters it.

Magnetization dynamics under DC

In the exploration of magnetization dynamics under DC, we initiate with a straightforward scenario where α=0\alpha=0 and σ(t)=0\sigma(t)=0 for t<0t<0 and σ(t)=σ\sigma(t)=\sigma for t≥0t\geq0. Lacking damping, we anticipate permanent magnetization precession, described by the simplified equations:

θ˙=−Ksin⁡θsin⁡ϕcos⁡ϕ,ϕ˙=−Kcos⁡θcos⁡2ϕ+σ.(4)\dot{\theta}=-K\sin\theta\sin\phi\cos\phi,\quad\dot{\phi}=-K\cos\theta\cos^2\phi+\sigma. \tag*{(4)}

These equations are intricate to solve analytically, but insight can be gained by examining two limiting cases, x≪1x\ll1 and x≫1x\gg1 with α=0\alpha=0. In the first limit, we find oscillatory behavior:

θ=π2+x(cos⁡(Kt)−1),ϕ=xsin⁡(Kt).(5)\theta= \frac{\pi}{2} + x(\cos(Kt)-1),\quad\phi= x\sin(Kt). \tag*{(5)}

Both θ\theta and ϕ\phi oscillate with amplitude xx and frequency KK, termed the oscillating mode. In the second limit:

θ=π2+14x(cos⁡(2σt)−1),ϕ=σt.(6)\theta= \frac{\pi}{2} + \frac{1}{4x}(\cos(2\sigma t)-1),\quad\phi= \sigma t. \tag*{(6)}

Here, θ\theta oscillates with amplitude 1/(4x)1/(4x) and frequency 2σ2\sigma, while ϕ\phi monotonically increases. This implies the spin precession about the yy-axis and a spin flip occurring at period of π/σ\pi/\sigma, termed the flipping mode. Comparing the frequency of θ\theta dynamics in each mode, we speculate that the transition from the oscillating to the flipping modes might occur at x=1/2x = 1/2 where K=2σK = 2\sigma.

The guess can be verified by numerical computations of frequency in units of KK for σ(t)=σ\sigma(t) = \sigma for t≥0t \ge0 and α=0\alpha= 0 in Fig. 2(a). The frequency variation exhibits ωDC≈K=1\omega_{DC} \approx K = 1 for x≪1x \ll1 and ωDC≈2σ\omega_{DC} \approx2\sigma for x≫1x \gg1. The singularity at x=1/2x = 1/2 indicates the expected transition. Only at this transition point, the spins are faltering between zz axis and xyxy-plane in period, so we call this faltering mode. For DC, we focus on the oscillating and flipping modes, as the faltering mode requires a fine tuning.

When damping α\alpha is introduced under DC, the magnetization eventually attains an equilibrium state. For x≥1x \ge1, the equilibrium state aligns with yy-axis, so the spin flip is absent. However, since two equilibrium states exists for x<1x < 1, the selection between them occurs, which can flip the spin. The selection is determined by both xx and α\alpha. Figure 2(b) displays two examples of dynamics leading to an even-flipped (x=0.6,α=0.05)(x = 0.6, \alpha= 0.05) and odd-flipped (x=0.6,α=0.1)(x = 0.6, \alpha= 0.1) final state, respectively. The magnetization in the even-flipped state stays in +z+z half space, while that in the odd-flipped state stays in −z-z half space. In both examples, owing to the damping effect, the crossover from the flipping to the oscillating mode occurs after a specific time τc\tau_c. In Fig. 2(b), for the former, after τc≈10.8\tau_c \approx10.8 ns, while for the latter, τc≈4.2\tau_c \approx4.2 ns. Empirically, we find that τc∝e4.53xα−1\tau_c \propto e^{4.53x\alpha^{-1}} for small α\alpha. [See Supporting Information (SI).] These show that the variation in τc\tau_c caused by the interplay of xx and α\alpha serves as a determining factor for the final state.

Image not extracted

Figure 2. The magnetization dynamics under DC. (a) Computed frequencies of oscillating (x<0.5x < 0.5), faltering (x=0.5x = 0.5), and flipping modes (x>0.5x > 0.5) under DC without damping in varying x=σ/Kx = \sigma/K. (b) Dynamics of magnetization at x=0.6x = 0.6, α=0.05\alpha= 0.05 (lower) and x=0.6x = 0.6, α=0.1\alpha= 0.1 (upper) in time. Chromatically, the red indicates +z+z, the green indicates in-plane, and the blue indicates −z-z directions of magnetization. The direction inward the paper is the yy direction. (c) A 2D phase diagram of the final states in varying xx and α\alpha. The dark blue indicates the even-flipped final state, and the yellow indicates the odd-flipped final state.

We further delve into the relation of the final state with xx and α\alpha by a 2D phase diagram in Fig. 2(c). When x<1/2x < 1/2, exclusively the even-flipped final state manifests, whereas for x≥1/2x \ge1/2, both odd-flipped and even-flipped states emerge, forming a fan-like configuration. It should be noted that the fan-like configuration is not clear in α<0.01\alpha< 0.01 due to the resolution. The exclusive appearance of even-flipped state for x<1/2x < 1/2 is due to the absence of spin flip in the oscillating mode. However, in the case of x≥1/2x \ge1/2, the spin flip can occur multiple times before the crossover from the flipping to the oscillating modes takes place. If the crossover time exceeds half of the flipping mode period, the spin flips to −z-z half space. With a longer crossover time surpassing a full flipping mode period, the spin flips twice, returning to +z+z half space. By extending the crossover time further, the spin flips repeatedly. Even (odd) numbers of spin flips lead to an even-flipped (odd-flipped) final state, giving rise to the distinctive fan-like feature in x≥1/2x \ge1/2. This pattern persists until x=1x = 1, where the equilibrium state converges to yy-axis.

Magnetization dynamics under AC

Based on above results, we here explore the magnetization dynamics under AC, which is given by σ(t)=σsin⁡(ωACt+δ)\sigma(t) = \sigma\sin(\omega_{\mathrm{AC}}t + \delta) for t≥0t \ge0 and σ(t)=0\sigma(t) = 0 for t<0t < 0. ωAC\omega_{\mathrm{AC}} is the driving frequency, and δ\delta is the initial phase of AC. The initial state of the magnetization is again aligned with +z+z-axis. Owing to the damping α\alpha, the relative strength of SOT xx, and driving frequency ωAC\omega_{\mathrm{AC}}, the initial state overcomes the irregular dynamics and transits to the final states after some time. We identify three distinct final states in Fig. 3(a): adiabatic, resonating, and chaotic states. Both adiabatic and resonating states are the steady states coming after the decay of initial state. At each time tt, σ(t)\sigma(t) determines an equilbrium state as shown in Fig. 1(b). The adiabatic state denotes that the magnetization mostly adheres to the equilibrium state at each time. The resonating state denotes that the magnetization dynamics is periodic but detach from the equilibrium state at each time. On the other hand, the chaotic state is evolved from the initial state, retaining its irregularity, in which the magnetization never has a periodic motion.

We primarily observe the adiabatic states for low-frequency or strong damping regime. We observe five distinct modes in the adiabatic states: I) an even-flipped oscillating mode, II) an odd-flipped oscillating mode, III) an even-flipped faltering mode, IV) an odd-flipped faltering mode, and V) a periodically flipping mode. These modes are depicted in Fig. 3(b).

The even-flipped (odd-flipped) oscillating mode or Mode I (II) is the oscillation of magnetization within +z+z (−z-z) half space. The even-flipped (odd-flipped) faltering mode or Mode III (IV) is the repeated faltering of magnetization between xyxy-plane and +z+z (−z-z) half space. The periodically flipping mode or Mode V is the repeated magnetization flip. Notably, Modes I and II (or III and IV) are almost identical, but differ in the number of spin flips before decaying to the adiabatic state. Only Mode V shows the continuous spin flip in time in its adiabatic state. As one can expect from their names, the adiabatic modes originate from DC modes. Modes I and II come from oscillating mode, Modes III and IV come from the faltering mode, and Mode V comes from the flipping mode.

The modes are chosen by the interplay of xx, ωAC\omega_{\mathrm{AC}}, δ\delta, and α\alpha. Both xx and ωAC\omega_{\mathrm{AC}} underscore their importance after decaying to adiabatic states while δ\delta and α\alpha play a pivotal role during the decay process. Primary investigation is performed by phase diagrams for the modes in xx and ωAC\omega_{\mathrm{AC}} at α=0.03\alpha= 0.03 with different δ\delta presented in Figs. 3(c,d). For x<1x < 1, Modes I and II appear, while for x≥1x \ge1, Modes III, IV, and V manifest. This transition occurs around x∼1x \sim1 since x=1x = 1 is where two equilibrium states meet at the yy-axis. Specifically, for x<1x < 1, as the equilibrium position does not reach yy-axis at any time, the magnetization oscillates within either +z+z or −z-z half space. For x≥1x \ge1, however, as the equilibrium reaches yy-axis, the magnetization either falters on the xyxy-plane or flips its position between +z+z and −z-z half spaces periodically.

ωAC\omega_{\mathrm{AC}} also plays a role in determining the modes. Primarily, ωAC\omega_{\mathrm{AC}} chooses the faltering and flipping modes when x≥1x \ge1, since it determines the time duration τy\tau_y that the equilibrium state at each time stays at the yy-axis. One can obtain the duration by finding the maximum τy\tau_y satisfying σ(t)≥1\sigma(t) \ge1 in t∈[t0,t0+τy]t \in[t_0, t_0 + \tau_y]. In the case of sinusoidal σ(t)\sigma(t), this can be expressed as

τy=1ωAC(π−2ζ),\tau_y = \frac{1}{\omega_{\mathrm{AC}}}(\pi- 2\zeta),

where ζ=arcsin⁡(1/x)∈(0,π/2]\zeta= \arcsin(1/x) \in(0,\pi/2]. For τy\tau_y, as the equilibrium state is at the yy-axis, the magnetization modulates near the yy-axis. After τy\tau_y, the equilibrium position is divided again and deviates away from the yy-axis. Then, depending on its modulated position, the magnetization chooses one of the equilibrium state, which leads to either faltering or

Image not extracted

Figure 3. The modes in the adiabatic state under AC. (a) The classification and comparison of final states. The final states transit with the driving frequency of AC, from adiabatic to resonating, and to chaotic state in sequence. While adiabatic and resonating states have periodicity, chaotic state does not. While adiabatic state adheres to the equilibrium state at each tt, others do not. (b) Schematics of distinct AC adiabatic modes after decay. (c–d) 2D phase diagrams of adiabatic modes in xx and ωAC\omega_{\mathrm{AC}} with α=0.03\alpha= 0.03, at (c) δ=0\delta= 0 and (d) δ=π/2\delta= \pi/2. (e–f) 2D phase diagrams of adiabatic modes in δ\delta and ωAC\omega_{\mathrm{AC}} at (e) x=0.6x = 0.6 and (f) x=0.8x = 0.8. The violet denotes Mode I, the red denotes Mode II, the blue denotes Mode III, the yellow denotes Mode IV, and the green denotes Mode V.

flipping modes. Additionally, the window of xx for every mode at higher ωAC\omega_{\mathrm{AC}} is opened up wider than that at lower ωAC\omega_{\mathrm{AC}}, showing the fan-like feature in Figs. 3(c,d). Unlike the faltering mode under DC, the window of Modes III and IV, originating from the faltering mode, expands to the finite range of xx. Lastly, it is noteworthy that the phase of ωAC=0\omega_{\mathrm{AC}} = 0 in Figure 3(d) is the same as the phase in α=0.03\alpha= 0.03 line of Fig. 2(c). This substantiates that adiabatic modes can be derived from DC modes.

On the other hand, δ\delta acts as a switch to turn on the spin flip during the decay process to adiabatic states. Specifically, comparing Figs. 3(b) to (c), Modes II and IV barely appear when δ=0\delta= 0, which means that the spin flip is turned off near δ=0\delta= 0. This happens because the average of σ(t)\sigma(t) during the decay time is small, so the spin cannot flip before decaying the adiabatic state. This can be supported by Figs. 3(e-f). In Fig. 3(e), we present phase diagrams for the modes in δ\delta and ωAC\omega_{\mathrm{AC}} at x=0.6x = 0.6. The transition from Modes I to II occurs once during the increase of δ\delta. This happens because the spin flip is turned on by finite δ\delta. Increasing xx, the spin flips more times just as in DC case, so the repeated transition between Modes I and II can also be observed in Fig. 3(f). We should note that the increase of α\alpha reduces the decay time. Above arguments holds also for the triangular wave instead of sinusoidal wave. [See SI.]

We move on to high-ωAC\omega_{\mathrm{AC}} and low-α\alpha regime to investigate ferromagnetic magnetoresonance (FMR). One could expect that the resonance frequency is closely related to the frequencies ωDC\omega_{\mathrm{DC}} in Fig. 2(a). In fact, when ωAC\omega_{\mathrm{AC}} approaches ωDC\omega_{\mathrm{DC}}, the oscillating amplitude becomes larger, and eventually the adiabatic state changes to the resonating state and to the chaotic state sequentially. By checking the final state after t∼1t \sim1 ms, we present 2D phase diagrams in ωAC\omega_{\mathrm{AC}} and α\alpha at x=0.3x = 0.3 and 0.70.7 in Fig. 4(a). Both diagrams show similar behavior. It begins with the adiabatic state at low frequencies, confronts the transition to the resonating state, and reaches the chaotic state at a certain range of ωAC\omega_{\mathrm{AC}}. For x=0.7x = 0.7, the window of chaotic states is opened up widely as SOT becomes large.

We further illuminate the transition of states by Fourier transform. The transition is illustrated in Fig. 4(b), where the change in ∣ϕ(ω)∣|\phi(\omega)| by ωAC\omega_{\mathrm{AC}} is schematically presented. Here, ϕ(ω)\phi(\omega) corresponds to the Fourier transform of azimuthal angle ϕ(t)\phi(t) from t=0t=0 to t=0.21t=0.21 ms (5×1055 \times10^{5} unit times). Figure 4(c) showcases representative examples of Fourier transforms at x=0.3x=0.3 and α=0\alpha=0. For adiabatic and resonating states, four distinct peaks are observed in the top panels: i) a main peak at ω1\omega_1 derived from ωDC\omega_{\mathrm{DC}}, ii) another main peak driven by AC at ω2=ωAC\omega_2=\omega_{\mathrm{AC}}, iii) induced peaks from i) and ii) at ω3=ω1+n(ω1−ωAC)\omega_3=\omega_1+n(\omega_1-\omega_{\mathrm{AC}}), and iv) subpeaks at ω4=∣ω1,2,3±2nωAC∣\omega_4=|\omega_{1,2,3}\pm2n\omega_{\mathrm{AC}}| (n∈Nn \in\mathbb{N}). Notably, as the time interval is elongated from 0.21 ms, Peaks i and iii decay while Peaks ii and iv gain intensity. [See SI.] This means that Peaks i and iii are related to the decay process while Peaks ii and iv are related to the stable state after decaying. Increasing ωAC\omega_{\mathrm{AC}} to the higher frequency, the change only occurs to the distance between peaks, as the frequency of Peak i decreases and that of Peak ii increases. Thus, adiabatic and resonating states are indistinguishable solely by Fourier transform. This is substantiated by the phase diagrams in Fig. 4(a), where the transition from adiabatic and resonating states is not distinctly delineated. This behavior is consistent for all modes in

Figure 4. The transition from adiabatic to resonating and chaotic states under AC. (a) 2D phase diagrams of the final states in ωAC\omega_{\mathrm{AC}} and α\alpha. The adiabatic state is in dark blue, the resonating state is in green, and the chaotic state is in dark red. The top panel is at x=0.3x=0.3 and the bottom panel is at x=0.7x=0.7. (b) The schematics of change in Fourier transform in for each state. (c) The corresponding examples to each state of (b) at x=0.3x=0.3 and α=0\alpha=0.

Fig. 3(b). Near the chaotic state, the peaks form array in a period of ωAC−ω1\omega_{\mathrm{AC}}-\omega_1 as shown in the middle panels. As ωAC\omega_{\mathrm{AC}} increases more, chaos sets in, destroying all peaks and being ϕ(ω)∝1/ω\phi(\omega) \propto1/\omega as shown in the bottom panels. This behavior does not change although the range of time is elongated.

Discussion

We address here about the typical values of parameters in the realistic systems. The typical value of ferromagnetic anisotropy is 1−10 μeV/atom1-10\,\mu\mathrm{eV}/\mathrm{atom}26, that of the current density in the experiments is ∼107 A⋅cm−2\sim10^{7}\,\mathrm{A}\cdot\mathrm{cm}^{-2}, and that of the itinerant spin polarization by Rashba-Edelstein Effect is estimated about 10−4ℏ10^{-4}\hbar per unit cell.[45] The typical value of Gilbert damping constant is ∼10−3−10−2\sim10^{-3}-10^{-2}[46], and that of the unit of AC frequency is estimated about 1−10 GHz1-10\,\mathrm{GHz}.

So far, we discuss the magnetization dynamics at the interface of TI-FM heterostructure. Under DC, we observe the oscillating and flipping modes without damping, which is determined by the relative strength of anisotropy and SOT. With damping, the number of spin flips is determined by the duration of the crossover from flipping to the oscillating mode. Under AC, we observe five distinct modes in the low frequency regime and their evolution to the resonating and chaotic states in the high frequency regime. Although we mainly discuss the TI-FM heterostructure due to its high efficiency, our work can be applied to the general systems with strong Rashba spin-orbit coupling. This is because the assumption underlying in our work is only that the current carries finite spin due to the Rashba-Edelstein Effect. Our work offers insights into the spin control by spin-orbit coupling, underscoring the practical aspects of the world of spintronics.

Acknowledgement

We thank Wataru Koshibae for the fruitful discussions. This work was supported by JST, CREST Grant Number JPMJCR1874, Japan.

Supporting Information Available

This material is available free of charge via the Internet at https://pubs.acs.org/

  • The dependence of crossover time from flipping to oscillating mode on xx and α\alpha, the dependence of decay time to adiabatic modes on α\alpha, the comparison between triangular and sinusoidal waves, and the time evolution of peaks in Fourier transform.

Supporting information for “Unraveling the dynamics of magnetization in topological insulator-ferromagnet heterostructures via spin-orbit torque”

Taekoo Oh; Naoto Nagaosa

The dependence of crossover time of DC in xx and α\alpha

Image not extracted

Figure 1 (PDF p. 16). τc\tau_c in xx and α\alpha. (a) τc\tau_c in xx with various α\alpha. (b) τc\tau_c in α\alpha with various xx.

Under DC, the system shows flipping, faltering, and oscillating modes without damping.

When damping is introduced, the crossover from flipping or faltering mode to oscillating mode occurs. We acquire the crossover time τc\tau_c empirically. Figure 1(a) shows the relation of xx and τc\tau_c, while Figure 1(b) exhibits the relation of α\alpha and τc\tau_c. For small α\alpha, one could note that the empirical relation of τc\tau_c to xx and α\alpha is τc∝e4.53xαα−1\tau_c \propto e^{4.53x\alpha}\alpha^{-1}.

The dependence of decay time of AC in α\alpha

Under AC and damping, the system decays into an resonating or an adiabatic state or evolves into a chaotic state. The decay time is defined as the duration of time reaching to resonating or adiabatic states from the initial state. We empirically observe the decay time at different xx, ωAC\omega_{\mathrm{AC}} and α\alpha in Fig. 2 One could note that the decay time is proportional to α−1\alpha^{-1}.

Image not extracted

Figure 2 (PDF p. 17). The decay time in α\alpha. The decay time in α\alpha with different xx and ωAC\omega_{\mathrm{AC}}.

The comparison of triangular and sinusoidal waves

We mainly discuss the magnetization dynamics under sinusoidal waves of AC. One could replace the sinusoidal waves with triangular waves. The triangular wave is given by

σ(t)=∣mod⁡[2ωACσπ(t−t0),4σ]−2σ∣−σ.(7)\sigma(t) = \left|\operatorname{mod}\left[\frac{2\omega_{\mathrm{AC}}\sigma}{\pi}(t-t_0),4\sigma\right]-2\sigma\right|-\sigma. \tag*{(7)}

The amplitude of this function is σ\sigma, and the frequency of this function is ωAC\omega_{\mathrm{AC}}. At t0=π/(2ωAC)t_0 = \pi/(2\omega_{\mathrm{AC}}), the function becomes sine-like, while at t0=0t_0 = 0, the function becomes cosine-like, as shown in Fig. 3(a).

We set ωAC=2π/240≈0.026\omega_{\mathrm{AC}} = 2\pi/240 \approx0.026, α=0.03\alpha= 0.03, and compare the adiabatic modes for sinusoidal and triangular waves in Fig. 3(b). In the manuscript, we describe that Modes II and IV does not appear under sine waves. The argument is also consistent with triangular waves, since the phase diagram under sine-like triangular waves does not show Modes II and IV as well. The difference between sinusoidal and triangular waves is that the window for each mode widens up for triangular waves.

Magnetization dynamics under triangular waves

Figure 3 (PDF p. 18). The magnetization dynamics under triangular waves. (a) (upper panel) a sine-like triangular wave, (lower panel) a cosine-like triangular wave. (b) The adiabatic mode diagrams at ωAC=2π/240\omega_{\mathrm{AC}} = 2\pi/240 and α=0.03\alpha= 0.03 for sine (first), sine-like triangular (second), cosine (third), and cosine-like triangular (fourth) waves. The violet denotes Mode I, the red denotes Mode II, the blue denotes Mode III, the yellow denotes Mode IV, and the green denotes Mode V.

The time evolution of peaks in Fourier transform

In our manuscript, we describe four types of peaks observed in the Fourier transform of ϕ(t)\phi(t). These include: i) the main peak originating from the DC modes denoted as ω1\omega_1, ii) a peak corresponding to the AC driven frequency ω2=ωAC\omega_2 = \omega_{\mathrm{AC}}, iii) induced peaks resulting from the frequency difference between i) and ii), given by ω3=ω1+n(ω1−ω2)\omega_3 = \omega_1 + n(\omega_1 - \omega_2), and iv) subpeaks represented by ω4=∣ω1,2,3±2nωAC∣\omega_4 = \lvert\omega_{1,2,3} \pm2n\omega_{\mathrm{AC}}\rvert.

Peaks i) and iii) are associated with the decaying process, whereas peaks ii) and iv) represent the stable state after decay. Figure 4 illustrates this phenomenon. In Fig. 4(a), the Fourier transform is performed from t=0t = 0 to 1.26 μs1.26\ \mu\mathrm{s}, while in Fig. 4(b), it is performed from t=0.12474t = 0.12474 to 0.126 ms0.126\ \mathrm{ms}. As time progresses, Peaks i and iii diminish due to damping, while Peaks ii and iv remain robust owing to the driven frequency.

Image not extracted

Figure 4 (PDF p. 19). The time evolution of peaks in Fourier transform of ϕ(t)\phi(t) (a) The Fourier transform of ϕ(t)\phi(t) in t=0t = 0 to 1.26 μs1.26\,\mu\mathrm{s}. (b) The Fourier transform of ϕ(t)\phi(t) near t=0.12474t = 0.12474 to 0.126 ms0.126\,\mathrm{ms}.

TOC Graphic

Schematic of a ferromagnet on a topological insulator

Figure 1 (PDF p. 26).

References

  1. [1]Witczak-Krempa, W.; Chen, G.; Kim, Y. B.; Balents, L. Correlated quantum phenomena in the strong spin-orbit regime. Annu. Rev. Condens. Matter Phys. 2014, 5, 57–82.
  2. [2]Hasan, M. Z.; Kane, C. L. Colloquium: topological insulators. Reviews of modern physics 2010, 82, 3045.
  3. [3]Qi, X.-L.; Zhang, S.-C. Topological insulators and superconductors. Reviews of Modern Physics 2011, 83, 1057.DOI
  4. [4]Pi, U. H.; Won Kim, K.; Bae, J. Y.; Lee, S. C.; Cho, Y. J.; Kim, K. S.; Seo, S. Tilting of the spin orientation induced by Rashba effect in ferromagnetic metal layer. Applied Physics Letters 2010, 97.DOI
  5. [5]Liu, L.; Moriyama, T.; Ralph, D.; Buhrman, R. Spin-torque ferromagnetic resonance induced by the spin Hall effect. Physical review letters 2011, 106, 036601.pubmed.ncbi.nlm.nih.gov/21405285
  6. [6]Kim, J.; Sinha, J.; Hayashi, M.; Yamanouchi, M.; Fukami, S.; Suzuki, T.; Mitani, S.; Ohno, H. Layer thickness dependence of the current-induced effective field vector in Ta— CoFeB— MgO. Nature materials 2013, 12, 240–245.DOI
  7. [7]Che, X.; Pan, Q.; Vareskic, B.; Zou, J.; Pan, L.; Zhang, P.; Yin, G.; Wu, H.; Shao, Q.; Deng, P., et al. Strongly surface state carrier-dependent spin–orbit torque in magnetic topological insulators. Advanced Materials 2020, 32, 1907661.DOI
  8. [8]Brataas, A.; Kent, A. D.; Ohno, H. Current-induced torques in magnetic materials. Nature materials 2012, 11, 372–381.DOI
  9. [9]Shao, Q.; Li, P.; Liu, L.; Yang, H.; Fukami, S.; Razavi, A.; Wu, H.; Wang, K.; Freimuth, F.; Mokrousov, Y., et al. Roadmap of spin–orbit torques. IEEE transactions on magnetics 2021, 57, 1–39.
  10. [10]Nagaosa, N. Emergent inductor by spiral magnets. Japanese Journal of Applied Physics 2019, 58, 120909.DOI
  11. [11]Yokouchi, T.; Kagawa, F.; Hirschberger, M.; Otani, Y.; Nagaosa, N.; Tokura, Y. Emergent electromagnetic induction in a helical-spin magnet. Nature 2020, 586, 232–236.DOI
  12. [12]Kitoari, A.; Kanazawa, N.; Yokouchi, T.; Kagawa, F.; Nagaosa, N.; Tokura, Y. Emergent electromagnetic induction beyond room temperature. Proceedings of the National Academy of Sciences 2021, 118, e2105422118.
  13. [13]Ieda, J.; Yamane, Y. Intrinsic and extrinsic tunability of Rashba spin-orbit coupled emergent inductors. Physical Review B 2021, 103, L100402.DOI
  14. [14]Yamane, Y.; Fukami, S.; Ieda, J. Theory of Emergent Inductance with Spin-Orbit Coupling Effects. Physical review letters 2022, 128, 147201.DOI
  15. [15]Araki, Y.; Ieda, J. Emergence of inductance and capacitance from topological electromagnetism. Journal of the Physical Society of Japan 2023, 92, 074705.DOI
  16. [16]Yasuda, K.; Tsukazaki, A.; Yoshimi, R.; Takahashi, K.; Kawasaki, M.; Tokura, Y. Large unidirectional magnetoresistance in a magnetic topological insulator. Physical review letters 2016, 117, 127202.DOI
  17. [17]Yasuda, K.; Tsukazaki, A.; Yoshimi, R.; Kondou, K.; Takahashi, K.; Otani, Y.; Kawasaki, M.; Tokura, Y. Current-nonlinear Hall effect and spin-orbit torque magnetization switching in a magnetic topological insulator. Physical review letters 2017, 119, 137204.DOI
  18. [18]Alegria, L.; Ji, H.; Yao, N.; Clarke, J.; Cava, R. J.; Petta, J. R. Large anomalous Hall effect in ferromagnetic insulator-topological insulator heterostructures. Applied Physics Letters 2014, 105.DOI
  19. [19]Katmis, F.; Lauter, V.; Nogueira, F. S.; Assaf, B. A.; Jamer, M. E.; Wei, P.; Satpati, B.; Freeland, J. W.; Eremin, I.; Heiman, D., et al. A high-temperature ferromagnetic topological insulating phase by proximity coupling. Nature 2016, 533, 513–516.DOI
  20. [20]Wang, H.; Liu, Y.; Wu, P.; Hou, W.; Jiang, Y.; Li, X.; Pandey, C.; Chen, D.; Yang, Q.; Wang, H., et al. Above room-temperature ferromagnetism in wafer-scale two-dimensional van der Waals Fe3GeTe2 tailored by a topological insulator. ACS nano 2020, 14, 10045–10053.DOI
  21. [21]Ou, Y.; Mirzhalilov, M.; Nemes, N. M.; Martinez, J. L.; Rocci, M.; Akey, A.; Ge, W.; Suri, D.; Wang, Y.; Ambaye, H., et al. Enhanced Ferromagnetism in Monolayer Cr2Te3 via Topological Insulator Coupling. arXiv preprint arXiv:2312.15028 2023,DOI
  22. [22]Fan, Y.; Upadhyaya, P.; Kou, X.; Lang, M.; Takei, S.; Wang, Z.; Tang, J.; He, L.; Chang, L.-T.; Montazeri, M., et al. Magnetization switching through giant spin–orbit torque in a magnetically doped topological insulator heterostructure. Nature materials 2014, 13, 699–704.DOI
  23. [23]Mellnik, A.; Lee, J.; Richardella, A.; Grab, J.; Mintun, P.; Fischer, M. H.; Vaezi, A.; Manchon, A.; Kim, E.-A.; Samarth, N., et al. Spin-transfer torque generated by a topological insulator. Nature 2014, 511, 449–451.
  24. [24]Han, J.; Richardella, A.; Siddiqui, S. A.; Finley, J.; Samarth, N.; Liu, L. Room-temperature spin-orbit torque switching induced by a topological insulator. Physical review letters 2017, 119, 077702.DOI
  25. [25]Wang, Y.; Zhu, D.; Wu, Y.; Yang, Y.; Yu, J.; Ramaswamy, R.; Mishra, R.; Shi, S.; Elyasi, M.; Teo, K.-L., et al. Room temperature magnetization switching in topological insulator-ferromagnet heterostructures by spin-orbit torques. Nature communications 2017, 8, 1364.DOI
  26. [35]Cullity, B. D.; Graham, C. D. Introduction to magnetic materials; John Wiley & Sons, 2011.
  27. [36]Xu, Y.-K.; Li, H.; He, B.-G.; Cheng, Z.-P.; Zhang, W.-B. Electronic structure and magnetic anisotropy of single-layer rare-earth oxybromide. ACS omega 2020, 5, 14194–14201.DOI
  28. [37]Fritsche, L.; Nöffke, J.; Eckardt, H. A relativistic treatment of interacting spin-aligned electron systems: application to ferromagnetic iron, nickel and palladium metal. Journal of Physics F: Metal Physics 1987, 17, 943.DOI
  29. [38]Gay, J.; Richter, R. Spin anisotropy of ferromagnetic films. Physical review letters 1986, 56, 2728.pubmed.ncbi.nlm.nih.gov/10033075
  30. [39]Liu, L.; Lee, O.; Gudmundsen, T.; Ralph, D.; Buhrman, R. Current-induced switching of perpendicularly magnetized magnetic layers using spin torque from the spin Hall effect. Physical review letters 2012, 109, 096602.pubmed.ncbi.nlm.nih.gov/23002867
  31. [40]Manchon, A.; Železný, J.; Miron, I. M.; Jungwirth, T.; Sinova, J.; Thiaville, A.; Garello, K.; Gambardella, P. Current-induced spin-orbit torques in ferromagnetic and antiferromagnetic systems. Reviews of Modern Physics 2019, 91, 035004.DOI
  32. [41]Kim, B. S. Skyrmions and Hall transport. Journal of Physics: Condensed Matter 2019, 31, 383001.
  33. [42]Auerbach, A. Interacting electrons and quantum magnetism; Springer Science & Business Media, 2012.
  34. [43]Tatara, G.; Fukuyama, H. Phasons and excitations in skyrmion lattice. Journal of the Physical Society of Japan 2014, 83, 104711.DOI
  35. [44]Gilbert, T. L. A phenomenological theory of damping in ferromagnetic materials. IEEE transactions on magnetics 2004, 40, 3443–3449.DOI
  36. [45]Chen, W. Edelstein and inverse Edelstein effects caused by the pristine surface states of topological insulators. Journal of Physics: Condensed Matter 2019, 32, 035809.DOI
  37. [46]Barati, E.; Cinal, M.; Edwards, D.; Umerski, A. Calculation of Gilbert damping in ferromagnetic films. EPJ Web of Conferences. 2013; p 18003.DOI

Paper details

vlnc:W_2ses7qxw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents