Controlling the creation/annihilation and distribution of magnetic skyrmions by manipulating an externally applied voltage

Magnetic skyrmions are vortex-like spin textures that are stabilized in solids due to topological protection.^1,2 Owing to the applicability of skyrmions as novel information carriers to memory,³ logic,⁴ and neuromorphic devices,⁵ extensive research is conducted on skyrmions. Development of techniques that aim to write, manipulate, and read information utilizing skyrmions is essential to realize a skyrmion computer. Spin polarized electron current^6,7 and voltage-controlled magnetic anisotropy^8–11 effect can be useful to write information using a skyrmion. Additionally, Lorentz tunneling electron microscopy (TEM)¹² and magneto-optical Kerr effect (MOKE) microscopy¹³ or electrical method utilizing topological Hall effect^14,15 or magneto-resistance^16,17 can be utilized to read information using a skyrmion. Methods utilizing an external magnetic field,^18,19 electric current,^20,21 thermal diffusion,^22,23 and strain²⁴ are promising to manipulate a skyrmion functioning as an information carrier. Previous studies have proposed many methods to control the motion of skyrmions; however, methods utilizing an electric current require abundant energy (electric power) and those involving heat and strain face the issues pertaining to an efficient implementation and operational speed. Conversely, the voltage method can be implemented using conventional magnetic tunnel junction technology, and it can be applied to develop skyrmion devices that are more energy efficient and function at high operational speeds. Ma et al. created skyrmions using this method and insisted that the skyrmions were in motion.¹¹ However, this motion was in a racetrack with sloped thickness, and they have not demonstrated any control over the motion of skyrmions using voltage effects within a region of uniform film thickness or magnetic anisotropy. Random motion resulting from thermal diffusion of skyrmions has the potential to perform stochastic calculations^25,26 and realize the development of a Brownian computer.^27,28 We have previously reported the design of a circuit based on the difference in the magnetic anisotropy of skyrmions in Brownian motion²⁹ and also successfully demonstrated information transfer exploiting the Brownian motion and repulsive magnetic dipole interaction of skyrmions.³⁰ To build an unconventional computer based on the Brownian motion of skyrmions, it is essential to develop a technique aimed at enabling input of the information that needs to be transferred. However, no such method has been realized. There are several ways to express information by a skyrmion, including assigning the presence or absence of a skyrmion to a bit 0 and 1 or defining the position of a skyrmion²³ or an arrangement of skyrmions³⁰ as 0 and 1 s. These methods need to externally control the creation and annihilation of the skyrmion or its position. Furthermore, to maintain energy conservation in devices using Brownian motion, it is important to input information by manipulating the skyrmions with low energy. Thus, in this study, we investigated the effect of applying voltage in the direction perpendicular to the Ta/Co–Fe–B/Ta/MgO film stack plane for modulating the magnetic properties and status of the skyrmions. Particularly, we achieved voltage-controlled creation and annihilation of skyrmions in addition to manipulating them based on the repulsive magnetic dipole interactions between the voltage-created skyrmion and that outside the electrode.

To investigate the influence of voltage on the creation of skyrmions at the Ta/Co–Fe–B/Ta/MgO junction, skyrmions confined in a circuit, which is based on the difference in its magnetic anisotropy, were observed using a MOKE microscope. The circuits were partially covered by a semi-transparent electrode to apply voltage. The overall device configuration is shown in Fig. 1. First, the Ta(5)/Co–Fe–B(1.22)/Ta(0.2)/MgO(1.5)/SiO₂(3.0) (described in nm) structures were deposited on a thermally oxidized silicon substrate at room temperature by magnetron-sputtering.^22,23,31 In this sample configuration, Pt and Co–Fe–B were not in direct contact, which prevented any electric state change. The stacks were annealed at 200 °C in vacuum to enhance perpendicular magnetic anisotropy. A 0.2 nm thick Pt film was deposited to form a circuit of skyrmions through the liftoff process. This partially deposited Pt enhances perpendicular magneto-anisotropy in Co–Fe–B, and the difference in magneto-anisotropy acts as a cell or path formed by skyrmions. By selecting an appropriate temperature value, skyrmions only appear in the region comprising of the additionally deposited Pt. Although partial Pt deposition may induce strain to Co–Fe–B, it has been demonstrated that strain only affects the magnetic anisotropy.³² A 50 nm thick SiO₂ layer was deposited as the gate-insulator. Moreover, a 100 nm thick Ru bottom electrode was deposited after milling the film stack to the bottom Ta layer using the Ar ion milling instrument equipped with an end point detector (10IBE, Hakuto). Furthermore, 2 and 100 nm of Ru were utilized as the top electrode, as shown in Fig. 1(b). A 2 nm thick semi-transparent Ru electrode (hereafter called as thin top Ru electrode) partially covered the skyrmion cell formed by Pt. After the deposition of the thin top Ru electrode, 100 nm of Ru (hereafter called as thick top Ru electrode) was deposited on top of the thin top Ru electrode. MOKE microscopic observation of the magnetic domain structure and skyrmions under externally applied voltage was achieved through the thin top Ru electrode. By contacting the DC probe between the top and bottom of the 100 nm Ru electrode, an external voltage was applied to the device and the behavior of the magnetic domain structure and skyrmions were video recorded using the MOKE microscope.

Figures 2(a)–2(c) show the MOKE microscope images observed under an applied voltage of +7, 0, and −7 V to the top electrode at 48.0 °C, respectively. With a decrease in the applied voltage, the width of the magnetic domain was observed to become broader. At 0 V, the width of the magnetic domain walls was different at the area with the thin top Ru electrode compared with that without the Ru electrode. This is because the magnetic anisotropy along the direction perpendicular to the Co–Fe–B junction is enhanced by the presence of the thin top Ru electrode. The similarity in the period of the magnetic domain when applying +7 V voltage confirmed the equality of perpendicular magnetic anisotropy between areas with and without Ru. Contrarily, the perpendicular magnetic anisotropy was enhanced at an applied voltage of 0 V and even greater at −7 V. Thus, we conclude that the voltage has a significant influence because the change in the magnetic structure occurred only under the thin top Ru electrode, whereas it remained unaltered in the area without the electrode. This voltage-induced change in perpendicular magnetic anisotropy can be attributed to the modulation of the interfacial magnetic anisotropy caused by the hybridization of the d orbitals of the magnetic elements in Co–Fe–B and the p orbitals of the oxygen in MgO due to the voltage-induced change in Fermi level.^33,34

Figure 3 shows MOKE microscopic images of a device under an applied voltage in the presence of an external magnetic field of ∼2 Oe (see video S1 of the supplementary material). With a decrease in voltage from positive to negative, the density of skyrmions decreases. This change in the skyrmion density can also be attributed to the fluctuations in the perpendicular magnetic anisotropy similar to that observed in the domain wall experiment. The charge doping caused by applying a voltage perpendicular to the film surface can modulate the perpendicular magnetic anisotropy of the junction to create or annihilate skyrmions.³⁵ 

Next, we investigated whether the position of skyrmions outside the thin top Ru electrode could be controlled by creating voltage-induced skyrmions. As shown in Fig. 4(a), a device equipped with the thin top Ru electrode was fabricated. Figure 4(b) shows a MOKE microscopic image of the device in which a skyrmion appears in the area without the thin top Ru electrode on applying an external magnetic field to the device. In this experiment, we focused on the two skyrmions trapped in the cell right to the thin top Ru electrode [indicated by the white square in Fig. 4(b)]. Brownian motion of the skyrmions was observed at an applied voltage of +5 and −5 V for 25 s each, and their respective positions were investigated. Figures 4(c) and 4(d) show histograms of the positions occupied by the skyrmions, with x = 0 at the right end of the thin top Ru electrode. Figures 4(c) and 4(d) show that at least nine pinning sites (at 1.3, 1.6, 3.8, 5.0, 6.5, 8.2, 9.1, 10.1, and 10.6 μm from the electrode) are frequently occupied by the two skyrmions, that is, skyrmion A and B, respectively. The frequency with which each pinning site is occupied depends on the polarity of the applied voltage. Focusing on skyrmion A in Fig. 4(c), the pinning sites located at 1.3 and 1.6 μm from the electrode were most frequently occupied when a voltage of +5 V was applied. Conversely, when a voltage of −5 V was applied, the frequency of occupying the two pinning sites close to the electrode decreased and the frequency of occupying the pinning site located approximately 4–5 μm from the electrode increased. Owing to its magnetic dipole interaction with skyrmion A, skyrmion B occupied the pinning site further from the electrode more often compared to skyrmion A, whereas the pinning sites at 1.3 and 1.6 μm were not occupied. Pinning sites located at a distance of approximately 5–10 μm were relatively uniformly occupied when a voltage of +5 V was applied, whereas at −5 V, pinning sites at 9 and 10 μm from the electrode were more frequently occupied. This can be explained based on the magnetic dipole interaction that pushed skyrmion B to occupy the pinning sites at 9–10 μm farther from the electrode, as skyrmion A occupied the pinning sites at 5.0 and 6.5 μm. When the voltage was switched from positive to negative, the averaged position of the skyrmions was evaluated and was observed to change from 2.45 to 4.01 μm for skyrmion A and from 8.58 to 9.57 μm for skyrmion B. The change is greater for skyrmion A, which was closer to the electrode. This difference in distance can be attributed to the fact that at an applied voltage of −5 V, the magnetic anisotropy under the thin top Ru electrode changed and the skyrmion phase stabilized to form a skyrmion, which pushed the skyrmion in the area without an electrode through magnetic dipole interaction. The difference in annihilation or creation of skyrmion by negative voltage in the experiments of Figs. 3 and 4 is attributed the difference in temperatures of the experiments. The stability of the skyrmion phase depends on the temperature,^36,37 and the conditions for the appearance of the skyrmion phase in the two experiments differ due to the difference in temperature. In the experiments of Figs. 3 and 4, by application of negative voltage, magnetic phases should be changed from skyrmion phase to perpendicular magnetization phase and from in-plane magnetization phase to skyrmion phase, respectively, by enhancement of perpendicular magnetic anisotropy (see S4 of the supplementary material). Consequently, by subtracting the MOKE microscopic image obtained during the application of −5 V from that obtained at +5 V, Fig. 4(e) shows that the orientation of the magnetic moment under the thin top Ru electrode is along the same direction as the surrounding skyrmions. This result indicates that the position of a voltage-created skyrmion, outside the electrode, exhibiting Brownian motion can be changed owing to the magnetic dipole interactions. A comparable system³⁰ has proven that the magnetic dipole interaction between skyrmions can reach a maximum of roughly 3 μm.

In this study, we fabricated a device in which magnetic anisotropy can be controlled by applying a voltage in the direction perpendicular to the film plane with Co–Fe–B as the magnetic layer. The MOKE microscopic images verified the creation and annihilation of skyrmions under an applied voltage. It was demonstrated that magnetic dipole interactions with a voltage-created skyrmion can change the position of the skyrmion in Brownian motion outside the electrode. Our proposed method can, thus, input information to a stochastic computer with an extremely low-power consumption using the skyrmion in Brownian motion.

See the supplementary material for the following: (S1) and (S2) video version of Figs. 3 and 4(b), respectively; (S3) validity of experiments on control of distributions of skyrmions by voltage-created skyrmion; (S4) effect of voltage and temperature on the creation and annihilation of skyrmions; and (S5) change in saturation magnetization by Pt capping and voltage.

This work was supported by JSPS KAKENHI (S) (Grant No. JP20H05666) and JST CREST (Grant No. JPMJCR20C1), Japan.