We demonstrate that a spin current flowing through a nanocontact into a uniaxial antiferromagnet with first- and second-order anisotropy can excite a self-localized dynamic magnetic soliton, known as a spin-wave droplet in ferromagnets. The droplet nucleates at a certain threshold current with the frequency of the Néel vector precession laying below the antiferromagnetic resonance. The frequency exhibits nonlinear behavior with the increase in the applied current. At the high value of the applied torque, the soliton mode transforms, and the oscillator emits spin waves propagating in the antiferromagnetic layer.
Antiferromagnetic materials (AFMs) have unique properties advantageous for future spintronic applications, including the absence of stray fields, intrinsic high-frequency dynamics, high spin wave (SW) velocities, and abundance in nature.1,2 Utilizing their terahertz (THz) spin dynamics due to strong exchange interaction can bring about solid-state THz nano-devices and, hence, close the THz gap.3,4 One of the most promising candidates of such devices is the AFM-based spin-Hall and spin-transfer torque nano-oscillators (SH/ST-NOs), which can operate as THz sources and detectors.5–17 Furthermore, these devices offer great potential for on-chip THz neuromorphic applications.16–20 Several attempts have been made to understand current-driven spin dynamics in single SH/ST-NOs,5–17 which can further be coupled by propagating THz Slonczewski spin waves,21 similar to ferromagnetic counterparts.22–26
Employing localized spin dynamics is of crucial importance for the operation of spintronics devices.27–29 Contrary to ferromagnets (FMs), where a combination of demagnetization, crystal anisotropy, and external field can form the localizing potential for magnons, the localization of spin dynamics in AFMs is challenging. It can be achieved by exciting self-localized AFM spin textures, such as domain walls, Bloch lines, and skyrmions;28,30,31 however, pure dynamical localized excitations can substantially enrich the scopes of AFM devices. Such dynamical states in the form of AFM solitons were predicted theoretically a long time ago32,33 in the case of zero damping. However, their practical realization was unresolved due to the lack of any excitation method. In contrast, FM dynamic solitons, such as droplets, are experimentally demonstrated in SH/ST-NOs.34–39
In this Letter, we study the excitation of dissipative AFM droplet solitons in a nanocontact (NC)-based SH/ST-NO. We use micromagnetic simulations to investigate the stability and properties of the excited AFM droplets as a function of applied current, magnetic anisotropies, and NC radius. In particular, we compare droplet structures for different NC radii and evaluate their influence on the output signal. Our choice of material is Ru- and Rh-doped hematite ( -Fe2O3), which has been identified as a promising candidate for potential experimental realization.40 While our results show the possibility of a droplet excitation in pure hematite by a large enough NC, doping-enhanced anisotropy substantially increases the stability region of the droplet as a function of the applied current, achievable frequency, range, and power efficiency.
Schematic illustration of an AFM ST oscillator. The nanocontact, which acts as a spin current source, is placed on top of a thin AFM layer with uniaxial anisotropy. The black arrow shows the easy-axis orientation, and the white arrow indicates the direction of the spin current polarization. In the upper left corner, a sketch of the Néel vector precession shows the spin structure across the excited droplet.
Schematic illustration of an AFM ST oscillator. The nanocontact, which acts as a spin current source, is placed on top of a thin AFM layer with uniaxial anisotropy. The black arrow shows the easy-axis orientation, and the white arrow indicates the direction of the spin current polarization. In the upper left corner, a sketch of the Néel vector precession shows the spin structure across the excited droplet.
Equation (1) is reported to describe magnetic anisotropies in hematite in easy axis phase.42 These anisotropy terms can be tuned by doping elements43 in hematite. For example, it is shown that Ru and Rh (Al and Ga) doping increase (decrease) both and .43 Particularly, Ru doping substantially increases AFMR frequency (or in other words ), which reaches 500 GHz already at 2% of a doping level. Since there is a lack of and separate measurements with doping, we assume the constant ratio between them below. The exchange field can also be reduced by any doping in hematite.44,45 These properties make hematite an ideal candidate for realizing AFM droplets in experiments.
The above-mentioned approach is highlighted in Fig. 2, where and droplet profiles are shown at different currents. At the low value of applied current, for all possible and the droplet is absent. However, at a certain threshold , a solution appears with a finite value of , which, in turn, corresponds to a droplet frequency . At higher currents, the condition has two solutions; however, the left one is unstable against an increase in droplet amplitude. While a soliton expands with a current, its frequency gradually decreases toward the limit value .
(a) Dependence on for different values of the applied current. The condition selects the (b) profile of the soliton. Profiles with are shown by the dashed lines. Here, GHz, GHz, nm, and nm. The direction of the current increase is highlighted by the arrow.
(a) Dependence on for different values of the applied current. The condition selects the (b) profile of the soliton. Profiles with are shown by the dashed lines. Here, GHz, GHz, nm, and nm. The direction of the current increase is highlighted by the arrow.
In contrast to ferromagnetic (FM) droplets,47 where , Fig. 2 shows the maximum value of , which follows directly from the above equation with given boundary conditions. The analytical derivation of the maximum droplet amplitude is given in the supplementary material, but it is useful to discuss this difference in terms of FM and AFM dynamics under the action of spin torque. In FMs, spin current induces negative damping when the polarization is anti-parallel to the magnetization but increases damping in a parallel configuration. Hence, the reversed magnetization in the core of the excited droplet is stabilized in this position, , by a spin torque. In contrast, in uniaxial AFM, both polarities of a spin torque induce negative damping, but for the modes with opposite circular polarization (see pp. 26–27 in Ref. 48). This leads to a stable rotation of the Néel vector with , also known as proliferation.5 Thus, in both FMs and AFMs, the spin torque helps to stabilize the core of the droplet, but with different values of .
To investigate the dynamics of an AFM droplet, we carried out micromagnetic simulations using MuMax3 solver49 for a system illustrated in Fig. 1. The structure is composed of an AFM film measuring nm2 with a thickness of 7 nm. At the center of the device, a circular NC having a radius of is placed, supplying a spin current polarized along the easy axis of the AFM. The AFM material properties are set to correspond to -Fe2O3,40,42,50–54 with sublattice saturation magnetization kA/m, exchange stiffness pJ/m, and exchange field T. The AFM unit cell of the size nm3, corresponding to hematite,55 was simulated by two MuMax3 cells with the opposite orientations of magnetization. The intrinsic damping of hematite is rather low (see, e.g., Ref. 53 with reported ), but to account for the damping enhancement due to the spin pumping, we set it to . Similar to the static spin-flop with magnetic anisotropy given by Eq. (1), the excitation induced by the spin current exhibits hysteresis behavior. Consequently, our simulations commence at a higher current, followed by a gradual decrease to the operational value. Hence, the threshold current refers to the minimal operational current that maintains excitations. The primary parameters include the frequency under the NC center and the amplitude defined by the deflection angle at the same point. To ensure that the excitation corresponds to the droplet mode, we check for the condition for frequency to be lower than AFM resonance , where propagating magnons are absent in bulk AFM.
First, we analyzed the case of anisotropy values corresponding to undoped hematite given by and kJ/m3, which gives the characteristic length nm for the maximum speed of km/s. The results of excitation by NC with nm are shown in Figs. 3(a) and 3(b). Notably, a gap between the frequency of the excitation and AFMR appears at the threshold A/m2, similar to what is observed for the droplets in ferromagnetic oscillators. This gap results from an amplitude threshold for the droplet excitation discussed above. As the current increases to A/m2, the area under the NC starts to emit propagating spin waves instead of the localized droplet. The frequency and applied current range within which the droplet persists is relatively narrow. Furthermore, if the NC radius is reduced to the characteristic size of , the initiation of the localized droplet ceases at any current value, yielding only propagating spin waves.
The dependences of (a) the frequency and (b) the deflection angle on the applied current density for the NC with nm and characteristic length nm ( kJ/m3); red dots are extracted from micromagnetic simulations, while black—calculated by a model Eq. (2). The dependences on the anisotropy constants of (c) the excitation frequency at the threshold and (d) the “phase diagram” of the excitation type as a function of the applied current density. The red-filled region in (c) shows the theoretical limits of the droplet frequency in the non-dissipative limit, indicated by the horizontal dashed lines in (a). The blue bar in (a) highlights the range of currents at which the droplet is observed.
The dependences of (a) the frequency and (b) the deflection angle on the applied current density for the NC with nm and characteristic length nm ( kJ/m3); red dots are extracted from micromagnetic simulations, while black—calculated by a model Eq. (2). The dependences on the anisotropy constants of (c) the excitation frequency at the threshold and (d) the “phase diagram” of the excitation type as a function of the applied current density. The red-filled region in (c) shows the theoretical limits of the droplet frequency in the non-dissipative limit, indicated by the horizontal dashed lines in (a). The blue bar in (a) highlights the range of currents at which the droplet is observed.
In order to excite droplets at smaller NC radii, the droplet's characteristic size should be reduced. To achieve this, we investigated the impact of increasing anisotropy values and , keeping their ratio constant at . Ru and Rh doping would help reduce the droplet characteristic size by increasing the anisotropies and decreasing the exchange field, hence, c. The results of simulations, with a NC radius of 43 nm, are depicted in Figs. 3(c) and 3(d). For anisotropy values up to kJ/m3, the excitation frequency at the threshold exceeds the AFMR, and the localization is not forming. At around kJ/m3, the localized droplet region is emerging within a narrow range of applied current, but with the subsequent increase in anisotropy, the region of a droplet excitation extends across a broader range of both frequency and current due to the reduction in the characteristic length .
Now, we advance to analyze the dependences on NC radius . For this, we set the anisotropy constants as kJ/m3 and kJ/m3, thereby determining the characteristic size to equal 17.3 nm. Figure 4 shows the diagram of excitation types depending on the ratio and the applied current density. A scenario solely characterized by the excitation of propagating spin waves is observed for small NC radii, . Pure droplets can be excited at NC radii larger than the characteristic size, yet the region is limited by transforming at to a peculiar object, where droplet excites propagating spin waves. Interestingly, the dependence of the lower threshold on the NC radius corresponds to the one of Slonczewski mode,56 which is described in Ref. 21 [Eq. (3)] for the AFM case. Thus, the maintenance threshold, shown by the solid black line in Fig. 4(c), is calculated using Eq. (3) in Ref. 21 with the substitution , while the nucleation threshold can be described using and .
The dependences of (a) the frequency and (b) the deflection angle on the applied current density for the NC with a radius of (red) 73 nm and (blue) 31 nm. (c) The “phase diagram” of the excitation type as a function of the NC radius and applied current density; vertical dashed lines indicate selected radii used in (a) and (b). The solid black line is calculated using Eq. (3) from Ref. 21. Characteristic length nm.
The dependences of (a) the frequency and (b) the deflection angle on the applied current density for the NC with a radius of (red) 73 nm and (blue) 31 nm. (c) The “phase diagram” of the excitation type as a function of the NC radius and applied current density; vertical dashed lines indicate selected radii used in (a) and (b). The solid black line is calculated using Eq. (3) from Ref. 21. Characteristic length nm.
The profile of the droplet, as well as the frequency, also depend on the ratio, and we examine droplet features for two NC radii: and . For a small NC radius, for example, , the frequency range is narrow, akin to the above-mentioned observations made with smaller anisotropy (see Fig. 3). However, the broader current span allows observing a theoretical prediction that suggests a frequency lowering with increased energy influx and, consequently, increased soliton amplitude . Contrarily, when dealing with a larger NC radius of , the droplet exhibits a considerably broader frequency range, implying improved tunability. However, this case is characterized by an almost precession angle across the current range. This characteristic reduces the output torque , as the ac component of spin pumping is maximized at and diminishes to zero during the proliferation phase5,21 with precession angle. Nevertheless, the total signal could be non-zero since the entire droplet structure under the NC should be considered.
For a detailed analysis of the output signals corresponding to chosen NC radii, we computed the dependence of the average spin accumulation on the applied current densities (see Fig. 5). The total spin accumulation V is the sum of the output torques generated by the precession of the Néel vectors within the NC region S, . We further determined the amplitude of the alternating component using a Fourier transform and normalized it relative to the NC areas to facilitate efficiency comparison. Hence, for a small NC radius , the absence of planar rotation regions within the droplet profile leads to a higher spin accumulation density, although the NC with exhibits a higher total spin accumulation due to its larger interfacial area.
Spin accumulation density, integrated over the NC region, as a function of applied current density for NCs with a radius of (red) 73 nm and (blue) 31 nm. The insets show the cross section of the output torque distribution and the droplet profile.
Spin accumulation density, integrated over the NC region, as a function of applied current density for NCs with a radius of (red) 73 nm and (blue) 31 nm. The insets show the cross section of the output torque distribution and the droplet profile.
While the profiles of a polar angle extracted from micromagnetic simulations are in good agreement with a model of Eq. (2) [see Fig. 4(b)], the behavior of the polar angle is substantially different. Similar distinctions were noticed also for droplets in ferromagnets.41 Contrary to the idealized model of droplet in the AFM, where in-plane angle is constant in space, our simulations suggest that in the dissipative soliton depends strongly on the radial coordinate r (see Fig. 5). This distinction is caused by the assumption in the analytical model that non-conservative terms, i.e., Gilbert damping and STT, are small. The dependence can be described as the presence of highly nonlinear spin waves (SWs) with a frequency equal to precession frequency , confined within the central region of the soliton and propagating from the center of the droplet toward its edge. A qualitative explanation is, in this central area and it can be treated as a local region of a spin-flop state, for which the magnon spectrum has a gapless branch. Thus, for , these SWs are localized inside the soliton, and they can propagate outside it only at . The wavevector of the SW increases with applied current, which gives rise to the frequency of the droplet (see Fig. 4), in contrast to the analytical prediction. These SWs can also apply pressure to the transitional areas of the droplet, pushing them beyond the NC region57 and reducing the ac output.
In conclusion, we demonstrate, both theoretically and through micromagnetic simulations, that AFM dissipative droplet soliton can be excited by applying a spin current through a NC in an extended AFM film featuring first- and second-order uniaxial anisotropy. To stabilize the AFM droplets, the NC radius must be larger than the characteristic size of the soliton. In this case, for a given anisotropy, the droplet mode is stable above a threshold current and below a switching current, at which point the excited AFM droplet transforms and starts to emit propagating waves. Thus, contrary to ferromagnets, the AFM droplet can be used as an effective emitter of high-frequency magnons propagating with high velocity, which is hard to implement by other methods.58 We show the presence of optimal values of the NC radius for maximizing output spin accumulation density and frequency tunability. We also observe the excitation of nonlinear spin waves inside the droplet, which differs from theoretical predictions under non-dissipative conditions. Based on our results, we suggest Ru and Rh-doped hematite ( -Fe2O3) as a perfect material ground for the experimental realization of AFM droplet mode.
SUPPLEMENTARY MATERIAL
See the supplementary material for the details of the AFM model, analytical calculation of the amplitude range of the AFM droplet, and the threshold current, shown by a black line in Fig. 4.
This project was partly funded by the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (Grant No. TOPSPIN No 835068) and the Swedish Research Council Framework Grant No. 2016-05980. M.H. thanks Swiss National Science Foundation (SNSF) for the financial support via Grant No. 177550 and SNSF Postdoc.Mobility Grant No. 222234.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Roman V. Ovcharov: Conceptualization (equal); Data curation (lead); Formal analysis (equal); Investigation (lead); Methodology (equal); Writing – original draft (equal). Mohammad Hamdi: Conceptualization (equal); Formal analysis (equal); Investigation (supporting); Writing – original draft (equal). Boris A. Ivanov: Conceptualization (equal); Methodology (equal); Writing – review & editing (equal). Johan Åkerman: Funding acquisition (lead); Project administration (equal); Supervision (equal); Writing – review & editing (equal). Roman S. Khymyn: Conceptualization (equal); Formal analysis (equal); Methodology (equal); Project administration (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.