Room-Temperature Skyrmion Shift Device for Memory Application


Room-Temperature Skyrmion Shift Device for Memory Application...

0 downloads 74 Views 3MB Size

Subscriber access provided by University of Newcastle, Australia

Communication

Room-temperature skyrmion shift device for memory application Guoqiang Yu, Pramey Upadhyaya, Qiming Shao, Hao Wu, Gen Yin, Xiang Li, Congli He, Wanjun Jiang, Xiufeng Han, Pedram Khalili Amiri, and Kang L. Wang Nano Lett., Just Accepted Manuscript • DOI: 10.1021/acs.nanolett.6b04010 • Publication Date (Web): 14 Dec 2016 Downloaded from http://pubs.acs.org on December 17, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Nano Letters is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Room-temperature skyrmion shift device for memory application Guoqiang Yu1*, Pramey Upadhyaya1, Qiming Shao1, Hao Wu2, Gen Yin1, Xiang Li1, Congli He, Wanjun Jiang3,4, Xiufeng Han2, Pedram Khalili Amiri1* and Kang L. Wang1*

1

Department of Electrical Engineering, University of California, Los Angeles, California 90095, United States 2

Beijing National Laboratory for Condensed Matter Physics, Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China 3

State Key Laboratory of Low-Dimensional Quantum Physics and Department of Physics, Tsinghua University, Beijing 100084, China 4

Collaborative Innovation Center of Quantum Matter, Beijing 100084, China

Email address: [email protected], [email protected], [email protected]

1 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract Magnetic skyrmions are intensively explored for potential applications in ultralow-energy data storage and computing. To create practical skyrmionic memory devices, it is necessary to electrically create and manipulate these topologically-protected information carriers in thin films, thus realizing both writing and addressing functions. Although roomtemperature skyrmions have been previously observed, fully electrically controllable skyrmionic memory devices, integrating both of these functions, have not been developed to date. Here, we demonstrate a room-temperature skyrmion shift memory device, where individual skyrmions are controllably generated and shifted using current-induced spinorbit torques. Particularly, it is shown that one can select the device operation mode in between: (i) writing new single skyrmions, or (ii) shifting existing skyrmions, by controlling the magnitude and duration of current pulses. Thus, we electrically realize both writing and addressing of a stream of skyrmions in the device. This prototype demonstration brings skyrmions closer to real-world computing applications.

KEYWORDS: Skyrmion, shift memory device, room temperature, thin films, spin-orbit torque.

2 ACS Paragon Plus Environment

Page 2 of 35

Page 3 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Magnetic skyrmions are swirling spin textures with a spherical spin topology1, 2. Their topological protection property results in the efficient manipulation with low current densities, which thus makes skyrmions attractive as information carriers in next-generation memory devices3-8. For example, in the theoretically proposed skyrmion shift (SKS) memory devices4-8, a train of individual skyrmions is used as information carriers and shifted by currents for encoding information. The SKS memory device is new and different from the earlier domain wall-based memories9, in which the skyrmions can be manipulated at lower energy cost compared with using domain walls, simultaneously offering higher densities by storing each bit of information in a smaller area. However, most studies to date on skyrmion memory devices have been theoretical and/or lacked a clear device demonstration4-8. Here we experimentally demonstrate a functional SKS memory device at room temperature, thus bringing skyrmion-based devices significantly closer towards practical applications.

Magnetic skyrmions were firstly observed in bulk B20 compounds3,

10-12

, where a collective

skyrmion lattice phase can be stabilized by a chiral Dzyaloshinskii-Moriya interaction (DMI)13, 14

. This kind of chiral DMI is an asymmetric exchange interaction intrinsically induced by the

broken inversion symmetry of the bulk crystal structure. Alternatively, chiral DMI can also be extrinsically introduced into ultra-thin magnetic films by interfacing with heavy nonmagnetic elements with strong spin-orbit coupling (SOC)

15-18

. This type of materials system supports

Néel-type skyrmions due to the nature of the interfacial DMI vector

19-26

, which is sensitive to

the quality of the interface, the thickness of the chosen SOC underlayer and the related SOC strength. This layered structure thus provides rich possibilities and engineering parameters for device developments. Furthermore, the fabrication of these thin film structures is compatible 3 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

with the existing semiconductor and magnetic recording manufacturing techniques. The SKS memory devices proposed in this work are thus based on the Néel-type skyrmions in magnetic thin films. In order to realize a memory device, one requires controllable generation, manipulation and detection of individual skyrmions for data write, transmission (addressing) and readout operations. In this work, we demonstrate the first two of these requirements at room temperature, as shown in Fig. 1. There are two constriction regions at the contact edges, which are designed for generating magnetic skyrmions. For practical memory applications, the readout can be done by using a rather standard technique, tunnel magnetoresistance (TMR) in the indicated magnetic tunnel junction (MTJ) on the end of the device, as shown in Fig. 1. With an out-of-plane pinned layer (PL) provided as a reference, the presence or absence of a skyrmion in the free layer (FL) could be detected in the form of electrical resistance. Once a skyrmion reaches the junction region, it can be detected via a standard MTJ readout. However, for detailed studies of the skyrmion dynamics, we instead use magneto-optical Kerr microscopy for readout in this experiment. One key advance presented here is the controlled generation of individual skyrmions in the SKS memory, the first step for realizing a practical device. Although it has previously been demonstrated that single skyrmions can be created in thin films, e.g. by using local spinpolarized currents from a scanning tunneling microscope at low temperature20, the roomtemperature controllable generation of individual skyrmions is required for practical applications. This control over the number and position of individually created skyrmions is a key ingredient for practical memory devices, which is addressed in this work. Our SKS memory device is a step beyond the recent work of Jiang et al. In that work, a chiral stripe domain can be dynamically transformed into ensembles of skyrmion bubbles22, by using 4 ACS Paragon Plus Environment

Page 4 of 35

Page 5 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

spatially divergent current-induced spin-orbit torque (SOT)27,

28

. The skyrmion number

generation in a controlled manner was not achieved. Here, we solve this problem by demonstrating a precise control of the generation of single skyrmions in a single shot fashion in a simple two-terminal device. One key factor for realizing this is that the generation of a stripe domain can be controlled through spin-orbit torque-induced asymmetric domain nucleation due to the presence of DMI that tilts magnetization at the device boundary. This two-terminal device integrating both skyrmion generation and manipulation functions is more advantageous by simply applying an “in-line” electron current between two terminals. This is also in contrast to the previous demonstration by Jiang et al., which requires four terminals to realize writing and shifting, among which two are used for generating skyrmions and the other two for shifting them29, 30, requiring at least two pulse generators for operating the device. For practical memory application, the realization of “in-line” injection of skyrmions can significantly simplify the memory device structure31.

In the presented memory device, a single skyrmion bubble can be controlled (including creation and manipulation) purely by current pulses. Through controlling the magnitude and duration of the applied current pulses, we show that it is possible to select the device operation mode between either: (i) injecting (i.e. writing) a single skyrmion, or (ii) shifting skyrmions. In addition, the direction of motion is determined by the sign of current. In this manner, we demonstrate a fully electrically operated SKS memory device with a simple structure.

5 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Sample structure: The film structures used to build the device consist of Ta(5 nm)/Co40Fe40B20(t = 1.1, 1.2 and 1.3 nm)/TaOx. The metal layers were deposited on thermally oxidized wafers by d.c. magnetron sputtering at room temperature. The TaOx layer was formed by oxidizing a 1.5 nm thick Ta layer using a radio-frequency O2/Ar plasma. The films were then annealed to enhance their crystallization and improve their perpendicular magnetic anisotropy. Figure 2a shows the high-resolution cross-section transmission electron microscope (TEM) image of the sample with t =1.2 nm, where the sharp interfaces can be identified. The magnetic properties of the films were measured using vibrating sample magnetometer (VSM) measurements at room temperature. For the sample with t =1.2 nm, the out-of-plane and in-plane magnetic hysteresis loops (as shown in Fig. 2b) indicate that the CoFeB layer is perpendicularly magnetized. The saturation magnetization was determined to be Ms = 724 emu/cc. The perpendicular hysteresis loop in the low field regime (see the inset in Fig. 2b) shows a gradual transition with multiple intermediate states during switching, indicating a complex spin-texture evolution during the reversal. By contrast, the t =1.1 nm sample exhibits a higher perpendicular anisotropy (i.e. a larger in-plane saturation field and sharper switching behavior), and the t =1.3 nm sample exhibits an in-plane anisotropy (see details in Figs. S1a and c). The film with t = 1.2 nm was fabricated into devices and studied for the skyrmion creation and manipulation. This is motivated by its intermediate perpendicular anisotropy, which gives rise to a metastable skyrmion bubble state22, as will be demonstrated later. The evolution of magnetic domain configurations in the film under an out-of-plane applied magnetic field was studied using a polar magneto-optical Kerr effect (polar-MOKE) imaging system26, 32, as shown in Fig. 2c. Labyrinthine domains were observed at zero magnetic field, due to the competition among the short-ranged exchange interaction, the long-ranged dipole-dipole

6 ACS Paragon Plus Environment

Page 6 of 35

Page 7 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

interaction, the perpendicular magnetic anisotropy and the interfacial DMI. When an out of plane magnetic field was applied, the magnetic domains favoring the direction of the applied field were expanded, in order to lower the Zeeman energy. With further increase of the magnetic field, the multi-domain state was transformed to a single domain, with all the magnetization pointing along the external magnetic field direction. The evolution of the magnetic domain configuration with the magnetic field is consistent with the magnetization measurements, as shown in the inset in Fig. 2b.

Previous work had demonstrated that a skyrmion phase could be obtained by controlling perpendicular magnetic anisotropy in a perpendicularly magnetized CoFeB layer26. In this work, however, there was no magnetic skyrmion phase immediately observed for any of the three samples (t = 1.1, 1.2 and 1.3 nm) in the magnetic field dependent measurements. This is due to the fact that the required anisotropy for realizing the magnetic skyrmion phase lies within a very narrow window, which falls between the t = 1.2 nm and 1.3 nm samples in our case (See Ref. 26 for a similar study with a finer gradient of the thickness, illustrating this anisotropy window). However, the magnetic skyrmions can still be created as a metastable state in the sample with t = 1.2 nm, which is verified by calculating the static magnetic energy for a single skyrmion using the experimentally extracted DMI parameter (|D| = 0.17 mJ/m2). (See details in Supplementary Section S1).

Generation of an ensemble of skyrmions: Our first experiment is targeted at demonstrating that skyrmions can be injected from one edge of the device, by applying an electric current pulse

7 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

through the constriction region. This is realized through firstly generating a stripe domain at the edge of the device via current-induced SOT, and then translating the stripe domain to skyrmions in the non-uniform geometrical region. (The current-induced SOT effective field and spin Hall angle are quantified using a separate experiment, as discussed in the Supplementary Section S2.) The t =1.2 nm film was patterned into the device with a constriction region on one edge, as shown in Fig. 3a, by standard photolithography and dry etching techniques. When a DC current of 0.23 mA (∼ 1.3 MA/cm2 in the constriction region) is applied in the presence of a small outof-plane magnetic field (H⊥ = −1.7 Oe), skyrmions can be continuously generated in a stream, as shown in Figs. 3b-d. Figures 3b and c show the process during the application of a DC current, while Figure 3d shows the magnetic pattern after the DC current is removed. It is clear that there is a stripe domain present within the constricted region when the DC current is applied.

Fig. 3e illustrates schematically the evolution of the magnetic textures during skyrmion generation. Before applying a current, the magnetization is first saturated to Mz > 0. Once a positive electron current is applied at t = t0, the current-induced SOT starts to generate a stripe domain, which then expands from the edge (t0 < t ≤ t1). The stripe subsequently expands out of the constriction region and transforms into a skyrmion (t1 < t ≤ t3) as the current is continuously applied. Further maintaining the applied current generates extra skyrmions (t = t4). Thus, by controlling the duration of the current pulse, the creation of individual skyrmion can be precisely controlled. One can observe that there are two key steps involved in the skyrmion generation, as illustrated in Fig. 3e:

8 ACS Paragon Plus Environment

Page 8 of 35

Page 9 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

(i)

A stripe domain is firstly formed from the edge of the constriction region and expanded within the constriction region (t0 < t ≤ t1). This step is essential for starting the skyrmion generation process.

(ii)

When a critical current is reached, the stripe domain is transformed into a skyrmion after getting out from the constriction region (t2 < t ≤ t3). This is a result of the spatially diverging current-induced SOT22.

To understand the formation of the stripe domain induced by electric current, a wire without constriction regions at the edges was fabricated for comparison. The experimental results are shown in Figs. 3f-i. When applying an electron current flowing to the left in the presence of an external out-of-plane magnetic field (pointing up), a stripe domain is nucleated on the right side (ground terminal), and then expands along the electron flow direction, as shown in Fig. 3f. No nucleation is observed on the left side. By contrast, when the electron current reverses its direction, i.e. flowing to the right, the nucleation happens at the left edge, as shown in Fig. 3g. The same behavior is observed when the external magnetic field is reversed, as shown in Figs. 3h-i. The stripe domains are always injected and propagated along the electron current direction.

Model: The above observations can be explained by a toy model, considering the local magnetization tilting at the sample edges induced by the DMI33, as illustrated schematically in Fig. 3j. When the magnetization is pointing up, the magnetic moments (blue arrows) tilt clockwise and anticlockwise, respectively, at the left and right edges of the wire, due to the presence of DMI33. The current-induced damping-like effective fields ( ~  × ) point in the directions as shown by the red arrows. Here,  is the magnetization unit vector and is the electron spin direction. As a result, the energy barriers for nucleation at the two different edges 9 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 35

are different. The effective field acting on the local magnetization on the right edge needs to overcome a higher energy barrier, caused by the DMI-induced tilt, in order to switch the local magnetization. Therefore, a positive electron current can more easily cause a nucleation on the left edge, which is consistent with the observed results as shown in Fig. 3g. (Note that it is possible that the nucleation occurs at the right edge, but with a relatively larger energy requirement.) When the external magnetic field is reversed (the lower sketch in Fig. 3j), the effect is reversed as shown in Fig. 3j. The nucleation edge is thus only determined by the current direction and the sign of DMI, rather than the magnetization direction. It is worth noting that our experimental results are consistent with recent simulations, which have also demonstrated that DMI-induced tilting of magnetization can result in an asymmetric nucleation of domains driven by the current-induced SOT34, 35. This asymmetric nucleation is crucial for realizing the creation and manipulation of individual skyrmions in the device with only two terminals.

The second step is to transform the stripe domain into magnetic skyrmions. This is accomplished through a divergent SOT induced by the non-uniform current22, which has been demonstrated before to create skyrmions. We found that the generated skyrmions tend to be deflected downwards for the skyrmions with the inner magnetization pointing up (i.e. Mz > 0), as shown in Figs. 3b-d. By contrast, when the inner magnetization points down (i.e. Mz < 0), the skyrmions were deflected upwards (see Fig. S4). Similar behaviors have been observed in a previous work26, where the skyrmions were moved at an angle with respect to the electron current direction and the deflection angle was dependent on the inner magnetization direction. The observed deflection of the skyrmion motion is similar to the so-called “skyrmion Hall effect” due to the Magnus force 2, 36. The deflection angle (θ = 13 ± 1°) is extracted by tracking the current10 ACS Paragon Plus Environment

Page 11 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

driven skyrmion motion in a device with a uniform width (see Fig. S5 for details). This value is comparable with the theoretical estimation (see Supplementary section S3) and the recently reported values36. Due to the presence of the current inhomogeneity created by the constriction, the skyrmion Hall angle observed here contains contribution from the inhomogeneous topological Magnus force. It is noted that the current-induced Oersted field has a negligible contribution to the deflection angle as shown in Supplementary section S3.

Controlled generation of individual skyrmions: Having demonstrated the “in-line” generation of skyrmion ensembles at the constricted contact, we next turn to describe the controllable generation of skyrmions – one at a time – by using a single electric current pulse. This is achieved by timing the write pulse appropriately to ensure that only a single skyrmion is allowed to break from the generated stripe, after which the current is turned off. This process is then considered as a correct data-write, as shown in Fig. 4a. There is also a non-zero probability that the applied pulse results in the generation of unwanted magnetic configurations – such as no skyrmion, more than one skyrmion, or a stripe. A few such “write errors” observed in our devices are shown in Fig. 4a. The write error rate is expressed as WER = 1-P, where P is the probability of a correct write. The WER needs to be minimized for reliable device operation. As illustrated in Fig. 3e, only an appropriately tailored pulse duration (t2 < t < t4) is able to generate the desired single skyrmion with a low WER.

We studied the probability of generating a single skyrmion as a function of the pulse magnitude and duration, as shown in Fig. 4b. For a given pulse magnitude, for example, 2 V, the write

11 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 35

probability P (defined as the probability of accurate generation of a single skyrmion by one current pulse, measured in 20 trials) exhibits a Gaussian distribution with the pulse duration (see the blue solid line for Gaussian fitting in Fig.4b). When the duration is shorter, no skyrmion generation is the most probable outcome, which we refer to here as Error Type I. On the other hand, when the duration is longer than the median, the errors consist of producing two or three skyrmions, referred to as Error Types II and III, respectively. In addition, Error Type IV is defined as producing a stripe; this, however, is barely observed in our devices.

We next set out to design pulse shapes in order to either: (i) maximize single skyrmion creation probability; or (ii) maximize the probability of shifting existing skyrmions, without activating new skyrmions; which can then be used to perform reliable data write and shifting operations, respectively. One can observe that as the applied pulse magnitude increases, the write probability distributions still follow a Gaussian function, but the center of the distribution is shifted to a smaller pulse width value. The center of the distribution as a function of pulse magnitude is shown in Fig. 4c, exhibiting a monotonic decrease with increasing pulse magnitude. On the other hand, the maximum probability shows a non-monotonic dependence on pulse magnitude, as shown in Fig. 4d. The write probability for a single skyrmion reaches the maximum (around 80%) for 1 V pulse magnitude and gradually decreases for higher pulse magnitudes. In addition, the probability value for the optimized magnitude and duration shows a very weak dependence on the width of the device channel within the studied width region, which provides the possibility to reduce channel width for device applications.

12 ACS Paragon Plus Environment

Page 13 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

We next discuss the origin of the WER, which is an important issue for practical applications. As illustrated in Fig. 3e, an appropriate pulse duration (t2 < t < t4) can in principle generate a skyrmion without errors. A WER ~ 0 is, however, not realized in our experimental observations. We attribute the WER to the stochastic behavior during the magnetic domain nucleation37, 38 and domain wall depinning39-41, induced by the non-uniformity in the film thickness, defects in the film and at the edges, and thermal fluctuations. These stochastic parameters result in the fluctuation of the observed time window for the generation of one single skyrmion, both within a single device and across multiple devices, thus giving rise to the WER. The effect of the stochastic nucleation and motion process is also dependent on the strength of the driving forces41, which may explain the dependence of the maximum P (write probability) on the pulse magnitude. By reducing the defects in the films and at the device edges, a better control over the energy landscape in the device can be expected, thus reducing the WER.

SKS memory device operation: Based on the measured dependence of the probability as a function of pulse duration and magnitude, the proposed SKS memory device can be realized as follows. For writing a data bit (injecting a single skyrmion), a pulse with the magnitude corresponding to the maximum probability for creating a single skyrmion is applied. After creating the single skyrmion, it can be shifted by a pulse with a higher magnitude and a shorter duration, which lowers the probability for injecting a magnetic skyrmion (i.e. corresponding to increasing the Type I Error), hence only moving the existing skyrmions. The experimental demonstration of such a device is shown in Figs. 5a-e. The optical image of device is shown in Fig. 5a. The patterned device shows a labyrinthine magnetic domain structure, similar to those observed in continuous films (see Fig. S7). The system was firstly 13 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 35

saturated by a large enough magnetic field in the positive z-direction. The magnetic field was then gradually reduced to 2.7 Oe, where the net magnetization points up. Subsequently, a pulse voltage of amplitude −2V and of duration 5 ms was applied to inject a magnetic skyrmion into the device channel. It is noted that the optimized magnitude and duration of the pulse for injecting a skyrmion in the SKS memory device are different to those in the device shown in Fig. 4a, due to different device geometries and therefore different current density distribution. The produced skyrmion is seen in black color in the image, representing the inner magnetization pointing down. Next, a series of pulses with an 8 V magnitude and a 10 µs duration were applied, which drive the existing skyrmions, moving them along the electron current direction, without generating additional skyrmions, as shown in Fig. 5b. Hence, each skyrmion is only injected along the electron current direction and moves along the electron current flow. When the electron current reverses its direction, the skyrmion injection happens from the other (left) side of the device, as shown in Fig. 5c. Similarly, the injection of skyrmions with the inner magnetization pointing up can also be realized (see Fig. S8.).

The skyrmion velocity as a function of current density was also studied. Figure 5d shows the average displacement per pulse (based on monitoring the skyrmion over 20 applied pulses) as a function of pulse duration for different current densities. The result is approximately linear, yielding a constant velocity at each fixed current density. This velocity increases with the applied current density (see the blue solid dots in Fig. 5e). Below a voltage of 1 V (the corresponding current density is 0.35 MA/m2), skyrmion motion was not observed. This threshold value is ~10× higher than a previously observed critical current density for skyrmion motion22, which may be related to the increased pinning field induced by the edges. To verify this hypothesis, we also 14 ACS Paragon Plus Environment

Page 15 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

extract the velocity of skyrmions in the electron current direction in the middle of a wide device (shown in Fig. S5), where the edge-skyrmion distance is at least 10 µm (black open squares in Fig. 5e). It is apparent that these skyrmions move faster and have a lower threshold current (close to that measured in Ref. 22) in comparison to the skyrmions moving near the edges. Reducing the influence of the device edges will be important for improving the device performance, which may be realized through optimizing the material and fabrication process to minimize the edge defects.

Significantly, the multi-bit operation was also achieved, where a series of skyrmions were written in the wire and shifted together. Here, the magnetic skyrmion was firstly generated by a write pulse and then moved by a shift pulse across the device with a length of 90 µm. After the generation of the first skyrmion, the second one was generated by a subsequent writing pulse. Using this method, two (see Fig. S9), three (see Fig. S10), and up to seven magnetic skyrmions were controllably created and shifted in the present devices. Figure 6a shows the creation of seven skyrmions through the injection (IWrite) and shift (IShift) processes. There is a ~10% writing error in injecting single skyrmions in the demonstrated device, as shown in Fig. 6b. This approach can be easily scaled to larger numbers of skyrmions if the device length is extended, which however is not within the scope of this paper.

It is worth pointing out that the pulse current induced skyrmion displacement is not identical, resulting in a “displacement error”. This indicates that the motion dynamics is governed by the creep motion that is due to randomly distributed pinning centers within the device and at the

15 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 35

edges. For practical device performance, this problem can be solved by having better control of the channel material, etch process, and the resulting pinning centers, to ensure a more uniform motion of the skyrmions. In addition, a larger current pulse is required to drive the motion dynamics into the steady flow motion regime where equivalent skyrmions displacement is expected.

One important motivation to utilize skyrmions as information carriers in computing elements is to take the advantage of a lower threshold of current density for initiating skyrmion motion, which significantly reduces the energy cost of device operations. Another advantage of skyrmions is their small size, leading to higher data storage density compared to the current available technology. In the present study, the skyrmions are relatively large. The threshold current density to induce skyrmion motion is in the range of few MA/cm2, comparable with that required to induce conventional domain wall motion at a similar velocity42. We speculate that this is because the size of skyrmions in our case is larger than the average distance between defect-induced pinning centers in the films, and hence there is no route to bypass the defects during the skyrmion motion. In the case of single-crystal B20 materials, the pinning center density is much lower and the skyrmion size is much smaller compared to the thin film materials used in the present study. Therefore, skyrmions in B20 materials can bypass the pinning centers during motion, resulting in a much lower driving current compared to the thin-film system used in our device. In principle, optimizing the material system to increase the DMI could reduce skyrmion size23, 43, which may also help to reduce the driving current.

16 ACS Paragon Plus Environment

Page 17 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

In summary, we have demonstrated a prototypical two-terminal SKS memory devices. By tuning the magnitudes and duration of pulse current, individual skyrmions can be generated and shifted controllably. This observation can be attributed to the asymmetric edge nucleation of a chiral stripe domain due to the DMI-induced tilting of edge magnetization, and subsequent transformation from the stripe domain to skyrmions due to spatially diverging spin-orbit torque. The pulse shape corresponding to the minimum write error was used for generating single skyrmions in our devices, while current pulses with higher magnitudes and shorter durations were used for shifting skyrmions without creating new ones. Furthermore, the generation and controlled manipulation of up to seven skyrmions in a single device were also demonstrated. The proposed scheme for individually-controlled skyrmion creation and motion forms the basis for SKS memory devices. Materials and methods The Ta/CoFeB/TaOx stack was fabricated from Ta(5 nm)/Co40Fe40B20(t = 1.1, 1,2 and 1.3 nm)/Ta(1.5 nm) sputtered films. The metal layers were deposited on a thermally oxidized wafer by d.c. magnetron sputtering at room temperature in an AJA international physical vapor deposition system. The TaOx layer was then formed by exposing the sample to a radio-frequency O2/Ar plasma. The films were subsequently annealed at 200 ºC for 30 min. The sample was subsequently patterned into devices by standard photolithography and dry etching techniques. The polar MOKE imaging experiments were carried out using a spatially (360 nm resolution) and temporally (20 ms resolution) resolved polar MOKE microscope. The external magnetic field was generated by a Helmholtz coil driven by a Kepco power supply. The current pulses were applied using an Agilent 33250A generator. All measurements were carried out at room temperature.

17 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 35

Acknowledgements This work acknowledges the support by the Spins and Heat in Nanoscale Electronic Systems (SHINES), an Energy Frontier Research Center funded by the US Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES) under award #SC0012670. This work was supported in part by C-SPIN and FAME, two of six centers of STARnet, a Semiconductor Research Corporation program, sponsored by MARCO and DARPA. This work was also supported by the National Science Foundation (ECCS 1611570) and Nanosystems Engineering Research Center for Translational Applications of Nanoscale Multiferroic Systems (TANMS) Cooperative Agreement Award EEC-1160504. We would like to acknowledge the collaboration of this research with the King Abdul-Aziz City for Science and Technology (KACST) via The Center of Excellence for Green Nanotechnologies (CEGN). Supporting Information Determination of DMI strength and estimation of skyrmion size; determining Spin Hall angle in the Ta/CoFeB/TaOx system; calculation of deflection angle and the contribution of Oersted field gradient Note The authors declare that they have no competing interests.

18 ACS Paragon Plus Environment

Page 19 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

22. 23.

24.

25. 26. 27. 28. 29. 30. 31. 32. 33.

Fert, A.; Cros, V.; Sampaio, J. Nature Nanotechnology 2013, 8, (3), 152-156. Nagaosa, N.; Tokura, Y. Nature Nanotechnology 2013, 8, (12), 899-911. Jonietz, F.; Muhlbauer, S.; Pfleiderer, C.; Neubauer, A.; Munzer, W.; Bauer, A.; Adams, T.; Georgii, R.; Boni, P.; Duine, R. A.; Everschor, K.; Garst, M.; Rosch, A. Science 2010, 330, (6011), 1648-1651. Tomasello, R.; Martinez, E.; Zivieri, R.; Torres, L.; Carpentieri, M.; Finocchio, G. Scientific Reports 2014, 4, 6784. Zhang, X.; Zhao, G.; Fangohr, H.; Liu, J.; Xia, W.; Xia, J.; Morvan, F. Scientific Reports 2015, 5, 7643. Sampaio, J.; Cros, V.; Rohart, S.; Thiaville, A.; Fert, A. Nature Nanotechnology 2013, 8, (11), 839-844. Zhou, Y.; Ezawa, M. Nature Communications 2014, 5, 4652. Kang, W.; Huang, Y.; Zheng, C.; Lv, W.; Lei, N.; Zhang, Y.; Zhang, X.; Zhou, Y.; Zhao, W., Magnetic Skyrmion Racetrack Memory: from Physics to Electronics. arXiv:1507.08001: 2015. Parkin, S.; Hayashi, M.; Thomas, L. Science 2008, 320, (5873), 190-194. Muhlbauer, S.; Binz, B.; Jonietz, F.; Pfleiderer, C.; Rosch, A.; Neubauer, A.; Georgii, R.; Boni, P. Science 2009, 323, (5916), 915-919. Yu, X. Z.; Onose, Y.; Kanazawa, N.; Park, J. H.; Han, J. H.; Matsui, Y.; Nagaosa, N.; Tokura, Y. Nature 2010, 465, (7300), 901-904. Du, H.; DeGrave, J.; Xue, F.; Liang, D.; Ning, W.; Yang, J.; Tian, M.; Zhang, Y.; Jin, S. Nano Letters 2014, 14, (4), 2026-2032. Dzyaloshinsky, I. Soviet Physics Jetp-Ussr 1957, 5, (6), 1259-1272. Moriya, T. Physical Review 1960, 120, (1), 91. Bode, M.; Heide, M.; von Bergmann, K.; Ferriani, P.; Heinze, S.; Bihlmayer, G.; Kubetzka, A.; Pietzsch, O.; Blugel, S.; Wiesendanger, R. Nature 2007, 447, (7141), 190-193. Ferriani, P.; von Bergmann, K.; Vedmedenko, E.; Heinze, S.; Bode, M.; Heide, M.; Bihlmayer, G.; Blugel, S.; Wiesendanger, R. Physical Review Letters 2008, 101, (2), 027201. Heide, M.; Bihlmayer, G.; Blugel, S. Physical Review B 2008, 78, (14), 140403(R). Chen, G.; Zhu, J.; Quesada, A.; Li, J.; N'Diaye, A.; Huo, Y.; Ma, T.; Chen, Y.; Kwon, H.; Won, C.; Qiu, Z.; Schmid, A.; Wu, Y. Physical Review Letters 2013, 110, (17), 77204. Heinze, S.; von Bergmann, K.; Menzel, M.; Brede, J.; Kubetzka, A.; Wiesendanger, R.; Bihlmayer, G.; Blugel, S. Nature Physics 2011, 7, (9), 713-718. Romming, N.; Hanneken, C.; Menzel, M.; Bickel, J. E.; Wolter, B.; von Bergmann, K.; Kubetzka, A.; Wiesendanger, R. Science 2013, 341, (6146), 636-639. Woo, S.; Litzius, K.; Kruger, B.; Im, M.; Caretta, L.; Richter, K.; Mann, M.; Krone, A.; Reeve, R.; Weigand, M.; Agrawal, P.; Lemesh, I.; Mawass, M.; Fischer, P.; Klaui, M.; Beach, G. Nature Materials 2016, 15, (5), 501-506. Jiang, W.; Upadhyaya, P.; Zhang, W.; Yu, G.; Jungfleisch, M.; Fradin, F.; Pearson, J.; Tserkovnyak, Y.; Wang, K.; Heinonen, O.; te Velthuis, S.; Hoffmann, A. Science 2015, 349, (6245), 283-286. Moreau-Luchaire, C.; Moutafis, C.; Reyren, N.; Sampaio, J.; Vaz, C.; Van Horne, N.; Bouzehouane, K.; Garcia, K.; Deranlot, C.; Warnicke, P.; Wohlhuter, P.; George, J.; Weigand, M.; Raabe, J.; Cros, V.; Fert, A. Nature Nanotechnology 2016, 11, (8), 731-731. Boulle, O.; Vogel, J.; Yang, H.; Pizzini, S.; Chaves, D.; Locatelli, A.; Mentes, T.; Sala, A.; Buda-Prejbeanu, L.; Klein, O.; Belmeguenai, M.; Roussigne, Y.; Stashkevich, A.; Cherif, S.; Aballe, L.; Foerster, M.; Chshiev, M.; Auffret, S.; Miron, I.; Gaudin, G. Nature Nanotechnology 2016, 11, (5), 449-454. Chen, G.; Mascaraque, A.; N'Diaye, A.; Schmid, A. Applied Physics Letters 2015, 106, (24), 242404. Yu, G.; Upadhyaya, P.; Li, X.; Li, W.; Kim, S.; Fan, Y.; Wong, K.; Tserkovnyak, Y.; Amiri, P.; Wang, K. Nano Letters 2016, 16, (3), 1981-1988. Miron, I. M.; Garello, K.; Gaudin, G.; Zermatten, P. J.; Costache, M. V.; Auffret, S.; Bandiera, S.; Rodmacq, B.; Schuhl, A.; Gambardella, P. Nature 2011, 476, (7359), 189-193. Liu, L.; Pai, C.; Li, Y.; Tseng, H.; Ralph, D.; Buhrman, R. Science 2012, 336, (6081), 555-558. von Bergmann, K. Science 2015, 349, (6245), 234-235. Jiang, W.; Zhang, W.; Yu, G.; Jungfleisch, M.; Upadhyaya, P.; Somaily, H.; Pearson, J.; Tserkovnyak, Y.; Wang, K.; Heinonen, O.; te Velthuis, S.; Hoffmann, A. Aip Advances 2016, 6, (5), 055602. Phung, T.; Pushp, A.; Thomas, L.; Rettner, C.; Yang, S.; Ryu, K.; Baglin, J.; Hughes, B.; Parkin, S. Nano Letters 2015, 15, (2), 835-841. Yu, G.; Upadhyaya, P.; Wong, K. L.; Jiang, W. J.; Alzate, J. G.; Tang, J. S.; Amiri, P. K.; Wang, K. L. Physical Review B 2014, 89, (10), 104421. Pizzini, S.; Vogel, J.; Rohart, S.; Buda-Prejbeanu, L.; Jue, E.; Boulle, O.; Miron, I.; Safeer, C.; Auffret, S.;

19 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

34. 35. 36. 37. 38. 39. 40. 41. 42. 43.

Page 20 of 35

Gaudin, G.; Thiaville, A. Physical Review Letters 2014, 113, (4), 047203. Mikuszeit, N.; Boulle, O.; Miron, I. M.; Garello, K.; Gambardella, P.; Gaudin, G.; Buda-Prejbeanu, L. D. Physical Review B 2015, 92, (14), 144424. Martinez, E.; Torres, L.; Perez, N.; Hernandez, M.; Raposo, V.; Moretti, S. Scientific Reports 2015, 5, 10156. Jiang, W.; Zhang, X.; Yu, G.; Zhang, W.; Jungfleisch, M. B.; Pearson, J. E.; Heinonen, O.; Wang, K. L.; Zhou, Y.; Hoffmann, A.; Velthuis, S. G. E. t. Nature Physics 2016. Sethi, P.; Murapaka, C.; Lim, G.; Lew, W. Applied Physics Letters 2015, 107, (19), 192401. Im, M.; Fischer, P.; Kim, D.; Lee, K.; Lee, S.; Shin, S. Advanced Materials 2008, 20, (9), 1750-1754. Burrowes, C.; Ravelosona, D.; Chappert, C.; Mangin, S.; Fullerton, E.; Katine, J.; Terris, B. Applied Physics Letters 2008, 93, (17), 172513. Im, M.; Bocklage, L.; Fischer, P.; Meier, G. Physical Review Letters 2009, 102, (14), 147204. Burrowes, C.; Mihai, A.; Ravelosona, D.; Kim, J.; Chappert, C.; Vila, L.; Marty, A.; Samson, Y.; GarciaSanchez, F.; Buda-Prejbeanu, L.; Tudosa, I.; Fullerton, E.; Attane, J. Nature Physics 2010, 6, (1), 17-21. Emori, S.; Bauer, U.; Ahn, S. M.; Martinez, E.; Beach, G. S. D. Nature Materials 2013, 12, (7), 611-616. Hrabec, A.; Porter, N.; Wells, A.; Benitez, M.; Burnell, G.; McVitie, S.; McGrouther, D.; Moore, T.; Marrows, C. Physical Review B 2014, 90, (2), 020402(R).

20 ACS Paragon Plus Environment

Page 21 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Figure 1. Proposed SKS (Skyrmion Shift) memory device. The blue circles represent generated skyrmions in the CoFeB layer. The spin textures in the skyrmions are shown in the sketch in the lower right corner. The blue stripe at the left edge of the device channel represents the stripe domain generated by the current-induced SOT. The regions with orange (blue) color represent Mz > 0 (Mz < 0) of the CoFeB layer. Je represents the direction of a positive electron current. The red and blue arrows on the side show the spin directions of polarized electrons in the heavy metal (HM) due to the spin Hall effect. The writing of individual skyrmions is realized by applying a low amplitude current pulse (IWrite) with a given duration, while the shifting or moving of existing skyrmions is realized by applying current pulses with a shorter duration and a higher magnitude (IShift). Readout may be accomplished using a magnetic tunnel junction (MTJ), with an out-of-plane pinned layer as a reference, to detect the presence (representing “1”) or absence (representing “0”) of a skyrmion from tunnel magnetoresistance (TMR). The PL and FL refer to the pinned layer and free layer in the MTJ, respectively.

21 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 35

Figure 2. Film properties. a, High-resolution cross-section TEM image of the thin film. b, Perpendicular (black) and in-plane (red) M/Ms-H loops of a Ta (5)/CoFeB (1.2)/TaOx (thickness in nm) multilayer. The inset shows the enlarged perpendicular loop, which indicates that the switching is accomplished in a gradual fashion. c, Evolution of magnetic domain patterns with an out-of-plane external magnetic field. The labyrinthine domain at zero field is consistent with the gradual switching shown in (b) inset. The bright (dark) color represents the magnetization pointing up (down).

22 ACS Paragon Plus Environment

Page 23 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Figure 3. Device structure and skyrmion generation. a, Optical microscope image of fabricated devices. b-d, Polar-MOKE images illustrating the process of generating skyrmions driven by a DC current of I = 0.23 mA in the presence of a small out-of-plane magnetic field H⊥ = −1.7 Oe (where the direction of magnetic field is pointing along –z, pointing into the paper). The electron current flows to the right. The figures are obtained from the captured movie, which is attached in the Supplementary Materials as Movie I. e, Sketch of the time sequence for current-driven generation of skyrmions. The parameter t0 represents the time passed before applying a current, while parameter t4 represents the time corresponding to generating two skyrmions. The regions shown in orange and blue colors represent Mz > 0 and Mz < 0. f-i, Current-driven generation of stripe domains in the wire without the constriction regions at the two ends (the direction of magnetic field is pointing up ((f) and (g)) and down ((h) and (i))). The applied current density is 1.03 MA/m2. The stripe is controllably created along the electron current motion direction (red arrows). There is no creation of skyrmion in this case, as the current is uniform without divergence. j, Sideview sketch illustrating the asymmetric nucleation

23 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

of stripe domains driven by a current-induced damping-like SOT. The blue arrows represent the magnetization direction. The magnetic moments at the edges tilt due to the DMI interactioninduced chiral micromagnetic structure. The effective field due to the SOT is  =  ( × ). The effective fields on the tilted magnetic moments at the edges are shown by red arrows. is the electron spin direction as a result of an applied current by spin Hall effect, which is in the form of ~ −  ×  , and Je is along the electron current flow direction.

24 ACS Paragon Plus Environment

Page 24 of 35

Page 25 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Figure 4. Controlled generation of single skyrmions. a, Single magnetic skyrmion generation driven by a single positive electron current pulse (flowing to the right). The generation of a single skyrmion is defined as a correct writing event. The observed write error types include no skyrmion generation (Error I), two skyrmions (Error II), three skyrmions (Error III) and stripe generation (Error IV). An out-of-plane magnetic field H⊥ = −1.5 Oe (along the –z direction) is applied during measurements. b, Probability (i.e. measured probability of generating a single skyrmion) as a function of applied current pulse duration for different pulse magnitudes. The solid lines are the fits to Gaussian function. The device resistance is R = 1.414 kΩ. For 1 V pulse, the corresponding current density in the constriction region is 2.3 MA/cm2. c, The pulse duration corresponding to the fitted peak values as a function of applied pulse magnitude. d, The fitted peak probability value as a function of applied pulse voltage. The inset shows the probability value for the optimized magnitude and duration of the pulse as a function of the width of the device channel.

25 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 26 of 35

Figure 5. Shifting SKS memory device bits along the wire. a, Optical microscope image of the SKS memory device. b, c, Single skyrmion generation and motion driven by a current pulse for negative (b) and positive (c) electron currents, respectively. A large enough magnetic field was firstly applied to saturate the magnetization along the +z direction. The field was then reduced to H⊥ = 2.7 Oe. The single skyrmion was generated by a current pulse with a magnitude of 2 V and a duration of 5 ms. The created skyrmions were then shifted by a series of pulses with a magnitude of 8 V and a duration of 10 µs. The device resistance is R = 4.65 kΩ. During the shifting of skyrmions, no new skyrmion was generated, confirming the selective device operation controlled by the pulse shape. d, Average skyrmion displacement per one pulse as a function of the pulse duration, under different pulse magnitudes. The solid lines are linear fits to the experimental data. e, Average skyrmion velocity as a function of the pulse magnitude in the SKS device (blue solid dots), which is extracted from the fitting of the displacement versus pulse

26 ACS Paragon Plus Environment

Page 27 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

duration, as shown in (d). The black open squares show the component of the skyrmion velocity in the electron current direction in a stripe device with a uniform width (see Fig. S5). The edgeskyrmion distance is at least 10 µm to exclude the influence from device edges.

Figure 6. a, Demonstration of controlled generation of up to seven skyrmions in the SKS memory device by using a series of write pulses (with a magnitude of 2 V and a duration of 5 ms) together with shift pulses (with a magnitude of 8 V and a duration of 10 µs). The images show the whole process of the generation and shifting of the skyrmions for a negative electron current. From the top to the bottom, each figure represents a snapshot of the spin texture during the sequence of operations. The pulse sequence [write (red) and shift (blue)] applied to obtain each snapshot is labeled on the right side. b, Skyrmion injections by current pulses as a function 27 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

of the number of pulses. “1” represents a successful injection of a single skyrmion and “0” represents a failed injection of a single skyrmion. The probability based on 30 trials is 27/30, i.e. 90% for this case.

28 ACS Paragon Plus Environment

Page 28 of 35

Page 29 of 35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

TOC Graphic:

29 ACS Paragon Plus Environment

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

IWrite

Read V

IShift

Page 30 of 35

Write & Shift

PL Barrier FL

Je

ACS Paragon Plus Environment

Page 31 of 35

a

c Pt TaOx CoFeB Ta

5 nm 1

SiO2

25 µm 0.0 Oe

Out-of-plane In-plane

0.7 Oe 1

0 M/Ms

b

M/Ms

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

Nano Letters

0

-1 -10 -5

-1 -500

-250

0

0

5

H (Oe)

250

10

500

H (Oe)

ACS Paragon Plus Environment

1.4 Oe

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

a 20 µm

b

c

d

f

e t0

Page 32 of 35

20 µm

g H

h t1 t2 t3

i j z

H

x

σ

Je H

t4

ࡴୗ୓୘ ൌ ‫ܪ‬ୗ୓୘ ሺ࢓ ൈ ࣌ሻ

ACS Paragon Plus Environment

Page 33 of 35

b Write probability (100%)

Correct

Error I

Error III

Error IV

4.0 V 2.5 V 2.0 V 1.5 V 1.0 V 0.7 V

75 50 25 0

1

10

100

1000

Pulse duration (ms)

c

d 500 400 300 200 100 0 0

1

2

3

4

Voltage (V)

ACS Paragon Plus Environment

100 Probability (%)

Error II

100

Max probability (100%)

a

Optimized duration (ms)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

Nano Letters

75 50

100 50 0

10 20 30 40 50 60

Width (m)

25 0

0

1

2

3

Voltage (V)

4

Nano Letters

a

d 20 µm

b Write

Je < 0

c

Je > 0 Write

Displacement (m)

12

1.04 MA/cm2 1.39 MA/cm2 1.73 MA/cm2 2.08 MA/cm2 2.43 MA/cm2 2.77 MA/cm2 3.12 MA/cm2 3.47 MA/cm2

8 4 0

0

Shift

Shift

200

400

600

800 1000

Pulse duration (s)

e Velocity (m/s)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

Page 34 of 35

with edge influence without edge influence

0.4

0.2

0.0 0

1

2

3

4 2

Current density (MA/cm )

ACS Paragon Plus Environment

Page 35 of 35

a Je

Vw ൈ 1 Vs ൈ 3 Vw ൈ 1 Vw ൈ 1 Vs ൈ 5 Vw ൈ 1 Vs ൈ 9 Vw ൈ 1 Vs ൈ 5 Vw ൈ 1 Vs ൈ 5 Vw ൈ 1

b Injection

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

Nano Letters

1

Pwrite = 27/30 0 0

5

10

15

20

25

Nwrite

ACS Paragon Plus Environment

30