Magnetic bubblecade memory based on chiral domain walls

Unidirectional motion of magnetic domain walls is the key concept underlying next-generation domain-wall-mediated memory and logic devices. Such motion has been achieved either by injecting large electric currents into nanowires or by employing domain-wall tension induced by sophisticated structural modulation. Herein, we demonstrate a new scheme without any current injection or structural modulation. This scheme utilizes the recently discovered chiral domain walls, which exhibit asymmetry in their speed with respect to magnetic fields. Because of this asymmetry, an alternating magnetic field results in the coherent motion of the domain walls in one direction. Such coherent unidirectional motion is achieved even for an array of magnetic bubble domains, enabling the design of a new device prototype—magnetic bubblecade memory—with two-dimensional data-storage capability.


Sample preparation
For this study, metallic ferromagnetic Ta/Pt/Co/Pt films were deposited on Si substrates with 100-nm-thick SiO 2 layer by means of the dc-magnetron sputtering. The thicknesses of the Ta and Co layers are fixed to 5.0 and 0.3 nm, respectively, and the thicknesses of the upper and lower Pt layers are adjusted from 1.0 to 3.0 nm to tune the magnetic properties 31 . To enhance the sharpness of the layer interfaces, the films were deposited with a small deposition rate (0.25 Å /sec) through adjustment of the Ar sputtering pressure (~2 mTorr) and power (~10 W).
All the films exhibit clear circular domain expansion with weak pinning strength. The results in Figs. 2 and 3 were obtained from 5.0-nm Ta/2.5-nm Pt/0.3-nm Co/3.0-nm Pt film (Sample A) that shows the fastest bubble speed under the present experimental condition, possibly due to the weak coercive field (7.1 mT). The results in Fig. 4 were obtained from 5.0-nm Ta/2.5nm Pt/0.3-nm Co/1.0-nm Pt film (Sample B) that allows regular bubble-array writing with small irregularities due to the relatively large coercive field (16.2 mT). The DMI-induced magnetic field DMI was measured to be 40 and 22 mT for Samples A and B, respectively, by analyzing the asymmetric DW motion 25,26 . The DMI constant is then estimated to be about 0.3 and 0.1 mJ/m 2 for Samples A and B, respectively, by use of the saturation magnetization (1.3×10 6 A/m) measured by a vibrating sample magnetometer and the typical DW width (5 nm).

Experimental setup and procedure
The magnetic domain images were observed by use of a magneto-optical Kerr effect (MOKE) microscope equipped with a charge-coupled device (CCD) camera on the focal plane 10 . To apply the magnetic field onto the films, two electromagnets and two small coils are attached to the sample stage. One of the electromagnets is used to apply the in-plane magnetic field bias up to 200 mT. The smallest coil (~1 mm in radius) is used to apply the out-of-plane magnetic field pulses up to 68 mT with a fast rising time (< 1 s). The combination of the inplane electromagnet and the smallest coil was used to obtain the results shown in Fig. 2. The other coil (~2 mm in radius) is designed to apply the alternating sinusoidal magnetic field with adjustable tilting angle, which was used to obtain the results shown in Fig. 3. For field uniformity over the wide range (> 2 mm) of the film, two electromagnets were used to apply the in-plane and out-of-plane magnetic fields to obtain the results shown in Fig. 4.

Speed of the bubblecade motion
By adopting the Taylor expansion with respect to , the DW speed ∥ at the rightmost point of the bubble domain can be written as where . Since ∥ is an odd function with respect to , From the relation = [ ∥ ( , ) + ∥ (− , − )]/2, the speed of the bubble motion can be thus expressed as The experimental observation (Fig. 2d) indicates that it is good enough to approximate Eq.
within the present experimental range of , by confirming that the higher-order terms are negligible compared to the linear term.
According to Ref. 25, the DW energy density DW is given by a function of as where 0 is the DW energy of the Bloch configuration,  is the DW width, S is the saturation magnetization, and DMI is the DMI-induced effective magnetic field. Here, D (≡ 4 D S ⁄ ) is the DW anisotropy field that is required to rotate ̂D W from the Bloch configuration to the Néel configuration, where D is the DW anisotropy constant. Based on the assumption that the dependence of ∥ on is solely attributed to the variation of DW due to , one finds the relation which is then written as where 1 ≡ 4 DW (0) sgn( DMI ).
where the DW chirality DMI is defined by DMI bubble inside the DW at the rightmost point of the bubble domain and the tilting angle of the magnetic field is defined by atan( / ).
We confirmed the dependence on sgn( ) by a repeated experiment with opposite sign of (not shown). Such dependence on sgn( ) can be also verified even in the present experimental results, by rotating the observation coordinate by 180 degree with respect to the z axis. On the other hand, to confirm the dependence on DMI , we repeated the experiment by use of Pt/Co/MgO films that are known to have the left-handed chirality S2 , opposite to the right-handed chirality in the Pt/Co/Pt films 25,26 . Figure S1 summarizes the results from the Pt/Co/MgO films. The results truly show that the direction of the bubble motion in the Pt/Co/MgO films is opposite to that of the Pt/Co/Pt films (Fig. 2), verifying Eq. (S12).

Thermomagnetic writing of bubble domains
To create bubble domains, the thermomagnetic writing scheme S1,S3 is adopted. In this scheme, the magnetization of the film is first saturated by applying an out-of-plane magnetic field pulse (-30 mT, 1 s). A laser beam (60 mW) is then focused on a small spot (~1 m in diameter) of the film, causing reduction of the coercive field inside the spot by increasing the temperature. At this instant, a reversed magnetic field pulse (8 mT, 12 ms) is applied. Since the strength of the reversed magnetic field is adjusted to be slightly larger than the reduced coercive field inside the spot, the magnetization reversal occurs only in the area of the spot.
Consequently, a bubble-shaped reversed domain is created. By repeating this procedure with motorized sample stage, arbitrary pattern of bubble domain array can be recorded. Due to the time delay for multiple-bubble recording with stage translation, the clock pulses were interleaved in the present demonstration of the magnetic bubblecade memory operation shown in Fig. 4f. The spin-transfer torque scheme with multiple nanopillar structures will possibly provide a parallel writing capability required for real-time device operation.