Current-Controlled Magnetic Domain-Wall Nanowire Shift Register

See allHide authors and affiliations

Science  11 Apr 2008:
Vol. 320, Issue 5873, pp. 209-211
DOI: 10.1126/science.1154587


The controlled motion of a series of domain walls along magnetic nanowires using spin-polarized current pulses is the essential ingredient of the proposed magnetic racetrack memory, a new class of potential non-volatile storage-class memories. Using permalloy nanowires, we achieved the successive creation, motion, and detection of domain walls by using sequences of properly timed, nanosecond-long, spin-polarized current pulses. The cycle time for the writing and shifting of the domain walls was a few tens of nanoseconds. Our results illustrate the basic concept of a magnetic shift register that relies on the phenomenon of spin-momentum transfer to move series of closely spaced domain walls.

More than three decades ago, the concept of storing information in movable domain walls (DWs) was introduced, and the following years saw intense interest in “magnetic bubble memories” (1). The magnetic bubbles were often arranged in the form of shift registers and manipulated with fixed and alternating magnetic fields (2). However, this required on-chip field generators at the same size scale as the individual magnetic bits, adding considerable complexity and cost to the device and making scaling to smaller dimensions very difficult (3). These obstacles can be overcome by taking advantage of the interaction of spin-polarized current with magnetization in the DWs, which results in a spin-transfer torque on the DW, causing it to move (49). This effect has been observed in a number of magnetic materials (1013), but predominantly in permalloy (Ni81Fe19) nanowires (1417). The use of spin-momentum transfer considerably simplifies the memory device because the current is passed directly across the DW without the need for any additional field generators.

In a permalloy nanowire in which the magnetization lies along the nanowire, adjacent DWs alternate between head-to-head (HH) and tail-to-tail (TT) configurations (Fig. 1, B and C). Under the application of a uniform magnetic field, these DWs will move in opposite directions, leading to their potential annihilation. In contrast, spin-polarized current can, in principle, move a series of neighboring DWs in the same direction. Moreover, spin-momentum transfer becomes more efficient the smaller the size of the magnetic element is (18). We describe here the basic working principle of a current-controlled magnetic DW shift register, the fundamental building block of magnetic racetrack memory (19).

Fig. 1.

(A) SEM image of the permalloy nanowire and the electrical contact lines A and B. PG, pulse generator. (B and C) Probability of TT (B) and HH (C) DWs remaining in section A-B of the nanowire as a function of the injection pulse length. Illustrations of the magnetization configurations of TT (white square) and HH (black square) walls are shown. Black arrows represent the magnetization direction within each domain. (D and E) The ejection time needed to move the injected DW [(D), TT; (E), HH] out of section A-B toward A or B using a second positive or negative voltage pulse, respectively, from PG-3, as a function of the injection pulse length.

In a scanning electron microscopy (SEM) image of a typical permalloy nanowire (Fig. 1A), three pulse generators (PG-1, PG-2, and PG-3) are connected to electrical contact lines A and B, spaced 6 μm apart (20) (SOM text A). The resistance of the nanowire was measured to determine whether a DW was present in the region between line A and B (section A-B). The nanowire resistance also provides information about the structure of the DW (21); that is, whether it is a transverse or a vortex DW (22). In these DWs, the magnetization largely lies in the plane of the nanowire. In the transverse case the magnetization rotates about an axis perpendicular to the length of the nanowire, whereas in the vortex case, the magnetization rotates about a tiny core in which a small net moment points into or out of the plane of the nanowire (Fig. 2 D).

Fig. 2.

(A) SEM image of the shift register device and electrical measurement setup used here and subsequently. (B) Variation of the nanowire dc resistance (R) when a pulse sequence from PG-1 and PG-2 is repeated to successively inject and move TT and HH DWs. (C and D) Resistance variations for the 15th and 16th iterations shown in (B). Dotted lines correspond to transverse, vortex, or no DW present in section A-B. Simulated images of the transverse and vortex DWs are shown to the right; arrows indicate the magnetization direction. The corresponding pulse sequences are shown in (E and F) and (G and H), respectively.

A voltage pulse from one of the pulse generators (PG-1 or PG-2) was used to inject a current into line A as well as into the nanowire. The current that passes through line A generates a highly localized magnetic field that, when sufficiently large and when the magnetization of the nanowire is properly aligned, creates a DW in section A-B. The current that flows into the nanowire drives the DW along the nanowire between A and B in a direction determined by the current flow direction. The nanowire was first magnetized along the –(+) X direction with an external magnetic field H = –(+) 300 Oe, which was then set as close as possible to zero (|H|< ∼0.3 Oe, unless otherwise noted). A HH (or TT) DW was subsequently created in section A-B when a negative voltage pulse was injected from PG-2 (PG-1). Throughout this study, the amplitude of the voltage pulse from PG-1 and PG-2 was fixed at –3.2 V.

Previous experiments showed that the DW created in section A-B moves along the nanowire and exits from line B only when the pulse length exceeds a threshold value τP. This pulse length is determined by the DW velocity, which depended on the current density flowing along the nanowire, and H (17). Only vortex and not transverse DWs could be moved with current under the experimental conditions used (8, 16, 17). For example, the probability of finding a vortex wall in section A-B is plotted as a function of the pulse length (Fig. 1, B and C) for TT and HH walls, respectively, for a current density of ∼ –2.0 × 108 A/cm2. For both walls, τP = ∼35 ns.

When the injection pulse length is shorter than τP, we hypothesized that the DW would move along the nanowire a distance in proportion to the pulse length. This was verified by using an ejection voltage pulse to determine the time needed to move the DW either backward to A or forward to B as a function of the injection pulse length. The ejection pulse was applied from PG-3 to line B with an amplitude of ∼ ±1.6 V, which corresponds to a current of ∼ ∓2×108 A/cm2 in the nanowire; that is, nearly equal to that of the injection pulse current density (fig. S1). Positive voltage applied to line B provides spin-transfer torque, which pushes the DW along the +X direction, whereas negative voltage pushes it in the –X direction.

The pulse length required to eject the DW from the nanowire is plotted as a function of the injection pulse length for TT and HH walls, respectively (Fig. 1, D and E). The length of the ejection pulse, for positive pulses, decreases as the injection pulse length is increased, which shows that when a longer injection pulse is used, the DW travels further from line A, therefore requiring a shorter ejection pulse to complete its motion toward line B. In contrast, when the ejection pulse is of negative polarity, its length increases in proportion to the injection pulse length, indicating that the time needed to move the DW back to line A scales with the injection pulse length. Therefore, the position of the DW along the nanowire can be controlled by varying the injection pulse length.

Up to this point, a large magnetic field (∼±300 Oe) had been applied to reset the magnetic state of the nanowire before each injection of a DW. We developed a method by which we can reset the magnetic state of the nanowire by moving DWs with current, from one end of the nanowire to the other, without using any magnetic field. This requires the successive injection and motion of HH and TT walls. After setting the initial state of the nanowire by magnetizing it along +X with an external field, a TT wall is created in section A-B, using a 10-ns-long pulse applied from PG-1, followed by a 70-ns-long pulse from the same pulse generator to move the TT wall out of this section (Fig. 2F). After this operation, the magnetic state of the nanowire has been reversed. The same operation using PG-2 (Fig. 2E) subsequently generates and moves an HH wall through section A-B, again reversing the magnetization, therefore returning it to its original configuration. Each of these DW operations results in stepwise changes in the nanowire resistance (Fig. 2C) (23). Figure 2B shows the evolution of the nanowire resistance as the pulse sequence was repeated many times. In succession, TT and then HH walls are repeatedly created and moved, which shows that the magnetic state of the nanowire can be reset using current pulses alone.

However, when a transverse wall, which can readily be identified from its resistance value (21), was created in section A-B, the reset operation was disrupted (Fig. 2B, 16th iteration, and Fig. 2, D, G, and H). The injected TT wall remains in section A-B, even after the long pulse is applied so that the injection of the following HH wall annihilates it. These results indicate that transverse walls do not move at the current densities used here.

We next show that two adjacent vortex DWs can be moved together using the same current pulse. Because our measurement technique is limited to probing only the number of DWs stored in the nanowire, the following procedure was used to show that a pair of TT and HH walls move together (see fig. S2 for the detailed pulse sequence). First, a voltage pulse was injected from PG-1 to create a TT wall in section A-B. A voltage pulse from PG-2 was then injected to create a HH wall. The pulse length was made short enough so that the existing TT wall was not ejected from section A-B. After the injection of the two DWs, a voltage pulse from PG-2 was applied to eject only the TT wall by using a proper length. A sufficiently long voltage pulse from PG-2 was then applied to ensure ejection of the HH wall. The expected motion and the corresponding number of DWs in section A-B (Fig. 3B) is consistent with the measured variation in the nanowire resistance (Fig. 3A). The number of DWs changed from the initial state, zero, to 1, 2, 1, and 0, successively, which indicates that the two DWs were indeed moved together (see fig. S3 for the case of the current-induced motion of three neighboring DWs). The same variation in the nanowire resistance is observed when the pulse sequence is repeated many times (Fig. 3C).

Fig. 3.

(A) Variation of the nanowire dc resistance and (B) illustration of the corresponding injection and motion of TT and HH DWs (see fig. S2 for the pulse sequence used). Black squares and white squares in (B) represent HH and TT DWs, respectively. Black arrows represent the magnetization direction within each domain. Blue and red arrows represent the electron flow directions. (C) Variation of the nanowire resistance when the pulse sequence is repeated. (D) DW velocity and (E) the minimum separation distance between a first injected TT DW and a second injected HH DW required to avoid their annihilation as a function of the external magnetic field. The solid line shows the estimated separation distance (fig. S4).

There are occasional errors due, for example, to the injection of a transverse rather than a vortex DW or to the annihilation of the first DW when the second DW was injected (see fig. S2 for details of these probabilities). The latter becomes more probable as the separation distance between the DWs is reduced. This distance can be estimated by measuring the dependence of the probability of annihilation on the first pulse length (SOM text E). However, this measurement turned out to be very sensitive to small deviations of the magnetic field from zero, because the DW velocity, estimated from (section A-B length = 6 μm)/τP, strongly depends on the field (Fig. 3D). The dependence of the minimum separation distance on the magnetic field is plotted in Fig. 3E. Near zero field, the distance is ∼ 2.5 μm. We attribute this large distance to the localized magnetic field created around line A when the pulse is applied as well as to the magnetostatic fields from the magnetic charge on the DWs themselves. The solid line shows the estimated threshold distance when both of these fields are taken into account and it is assumed that a DW will move when subjected to a field that exceeds its propagation field along the nanowire. The dependence of the threshold distance on H can be explained using this model (fig. S4). If the localized magnetic field from the contact line could be eliminated, then the minimum separation distance at zero field was estimated to be ∼ 0.8 μm in the nanowire used here. Local pinning centers can be used to provide stable positions for the DWs, thereby allowing them to be spaced more closely, as well as improving the reliability of their motion (19, 24).

Using the concepts demonstrated above, we realized the construction of a DW shift register memory, namely a three-bit unidirectional serial-in, serial-out memory in which the data are coded as the magnetization direction of individual domains (Fig. 4A and fig. S7). A left-pointing domain represents a 0 and a right-pointing domain represents a 1. The data was written into the leftmost domain in section A-B, was shifted by 2 domains, and then read from the state of the rightmost domain. The magnetization configuration was inferred from the resistance. For example, when a data sequence of 010111 was written into the register, the resistance and the corresponding magnetization configuration evolved as shown in Fig. 4, B and C, respectively (see fig. S5 for the detailed pulse sequence). In Fig. 4B, the shaded regions correspond to a shift operation and the light regions to a write operation. The input sequence is accurately transferred to the output after two write/shift operations. In this example, the cycle time to write and shift one bit is ∼30 ns, which was determined by the write time (here, ∼3 to 4 ns) and the time required to shift the series of DWs by one domain length. The latter time was determined by the velocity of the DWs, which here was fast, ∼150 m/s (Fig. 3D).

Fig. 4.

Demonstration of a three-bit unidirectional magnetic DW shift register. (A) Data are encoded by the magnetization direction of three domains in the nanowire. (B) Nanowire resistance variation when a pulse sequence is used to write and shift along the register the sequence 010111 two times in succession. The light and dark regions indicate writing and shifting operations, respectively. The table shows the corresponding evolution of the states of the three bits during these operations. The highlighted digits show how the input bit sequence is transferred to the output after two write/shift operations. (C) Schematic illustration of the shift-register operation. Black squares and white squares represent HH and TT DWs, respectively. Black arrows represent the magnetization direction within each domain. Blue and red arrows represent the electron flow direction.

The motion of a series of DWs at high speed using nanosecond current pulses not only proves the viability of a shift-register memory but also presages the possibility of current-controlled DW–based logic devices.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article