Current-Induced Domain Wall Motion
4.6 Domain Wall Manipulation in CoFeB/Pt
Earlier results on Co/Cu/CoFeB multilayers report significantly smaller critical current densities [LCM+07]. But the results are difficult to compare since critical fields and currents tend to be smaller in multilayer structures. No random nucle-ation and annihilnucle-ation of DWs is found for current densities up to the breakdown density where structural damage sets in, in contrast to permalloy structures. This is attributed to the higher Curie temperature in CoFeB compared to permalloy.
The results of this section are published in the Journal of Applied Physics [HKB+08a].
4.6 Domain Wall Manipulation in CoFeB/Pt
So far, the motion of a DW by current has been investigated in-plane magne-tized materials like permalloy. Recently, the attention has shifted to out-of-plane magnetized metallic materials with narrow DWs, where spin torque was shown to be more efficient [RLK+05, FRK07, BKW+08]. From a fundamental point of view, such materials allow for the study of the influence of the hotly debated non-adiabatic spin torque on the DW dynamics that is expected to be higher for narrow DWs, due to higher magnetization gradients [TK04].
Experiment
To study the influence of current pulses, 100µm long and 2µm wide wires are de-fined by e-beam lithography and a subsequent lift-off process. The magnetic mate-rial used in this study is composed of a Pt(3 nm)/Co60Fe20B20(0.6 nm)/Pt(2 nm) multilayer that is known to be magnetized out-of-plane and is deposited by sput-tering.
Fig. 4.16(a) shows a SEM image of the investigated structure. The width of the samples is kept large to allow for magnetization configurations with a spin structure that can vary across the wire. A large pad is patterned on one side of the wire to nucleate reverse domains. So far, giant magnetoresistance or the extraordinary Hall effect were mostly used to characterize the influence of the current on the magnetization of out-of-plane magnetized metallic wires [RLK+05, FRK07, BKW+08], but these measurements become hard to interpret as soon as complicated spin structures occur, in contrast to direct imaging using X-PEEM.
The experimental setup is similar to the ones explained before in detail in Chapter 3. The coil integrated into the sample holder, that was used before, generates an in-plane field. It is now replaced by an out-of-plane coil.
84 Current-Induced Domain Wall Motion
Results and Discussion
The magnetic properties are investigated prior to the measurement using a MOKE setup. The hysteresis loop in Fig. 4.16(b) shows a low coercive field of around 1 mT.
This proves that the magnetization is oriented out-of-plane with a remanence of 1 and indicates a reversal with very low pinning. Comparison to conventional Pt/Co(0.6 nm)/Pt with a coercivity of 35 mT makes the low coercivity of the multilayer with CoFeB very conspicuous. Such a low pinning and nucleation field in CoFeB based multilayer may arise from the lower pinning at grain boundaries due to the amorphous nature of CoFeB.
To study the influence of the current injection on the magnetization, a DW is first generated in the wire by preparing a monodomain state with a strong out-of-plane magnetic field pulse and then applying a small magnetic field in the opposite direction. Fig. 4.17(a) shows a magnetic image of a DW in the 2µm wide wire. The magnetic contrast is surprising strong, even though only 0.6 nm magnetic material is present in the sample and the XMCD contrast for out-of-plane materials is about 3.5 times weaker than for in-plane magnetizations (see section 3.1.4).
Starting from this initial configuration, 25µs long current pulses with increas-ing amplitude are injected, beginnincreas-ing at a current density of about1010A/m2. No change in the magnetization structure is observed up to a current density of about 1012A/m2 where the DW vanishes and a new bidomain structure with a DW in the center parallel to the wire is created [Fig. 4.17(b)]. A subsequent current in-jection with an opposite polarity [Fig. 4.17(c)] leads to the equivalent bidomain structure with reversed magnetization directions. By reversing again the current
Figure 4.16: (a) SEM image of a 2µm wide CoFeB/PT wire. (b) MOKE hysteresis loop measured at room temperature with the magnetic field applied out-of-plane. (From [BHR+09])
4.6 Domain Wall Manipulation in CoFeB/Pt 85
Figure 4.17: (a) X-PEEM image of the wire containing a DW. (b) After the injection of a current pulse (1·1012A/m2for 25µs with the current direction indicated by the arrow), the original DW structure completely disappears and a long DW parallel to the wire is created. When the current direction is reversed, the magnetization in the domain also reverses [(c) and (d)]. (From [BHR+09])
polarity, the magnetization in the wire can be switched back [Fig. 4.17(d)]. The magnetization direction in the bidomain structure can thus be switched back and forth by current pulses with opposite polarity. The final state is independent on the initial configuration and only depends on the direction of the injected current.
Interestingly, a bidomain structure with a DW parallel to the wire could also be created from a monodomain state by the sole effect of an out-of-plane magnetic field pulse. This indicates that the bidomain state is energetically close to the monodomain state and that it is favored by the reduction of the stray field energy.
The dependence of the direction of the magnetization in the wire on the cur-rent polarity is clearly consistent with the effect of the Oersted field, that points in opposite directions on the different sides of the wire. To further understand these results, the two dimensional spatial distribution of the Oersted field wire is calcu-lated by solving analytically the Biot-Savart law [Jac98]. The spatial distribution of the Oersted field in the y-z plane for a current flowing in the x-direction with a current density of 1012A/m2 is plotted in Fig. 4.18(a) and the variation of its out-of-plane component (y-direction) in the CoFeB layer is shown in Fig. 4.18(b).
For the calculation a homogeneous current distribution inside the wire is assumed.
This is justified by the much larger thickness of the Pt layers compared to the CoFeB layer, thus the Pt layers contribute strongest to the Oersted field.
As expected, the Oersted field is antisymmetric with respect to the wire center and increases rapidly as one approaches the wire edges with a maximum value of about 8 mT at the edges. This field is strong enough to nucleate a reverse domain on the edge of the wire and switch to the bidomain structure with a DW parallel to the wire. This configuration is clearly favored by the magnetic stray field energy and the symmetry of the Oersted field. The direction of the magnetization in the domains can then be switched back and forth by the Oersted field whose symmetry fits with the one of the bidomain structure. It should be pointed out that the Oersted field needed to nucleate a reversed domain on the
86 Current-Induced Domain Wall Motion
Figure 4.18: (a) Cross-section of a wire with the calculated distribution of the Oersted field. The current (1012A/m2) flows homogeneously in the wire in the -z direction. The wire dimensions are 5.6 nm ×2µm (y- and z-axes are plotted at different scales). The out-of-plane component of the Oersted field Hy is plotted in color. (b)Hy as a function of the lateral position z in the wire. (From [BHR+09])
wire edges is high compared to the 1 mT coercivity measured by Kerr rotation in the continuous film on a macroscopic sample. A possible reason stems from the fact that in the continuous film the defects with the lowest coercivity will initiate the switching by nucleating a reverse domain. In the structured elements, there is a lower probability of the presence of such nucleation sites with low coercivity at the edges of the wire where the field is strongest [Fer02]. As the bidomain state becomes energetically more unfavorable for narrower wires, it is expected that the observed behavior is superseded by spin torque effects for narrower wires. So the observations yield an upper limit for wire dimensions that can be used for current-induced DW motion studies in soft magnetic out-of-plane magnetized materials.
Nevertheless, the observed behavior could be useful in itself, with reproducible switching between two distinct magnetization configurations that does not rely on the spin torque effect.
In the past, theoretical proposals have been put forward to switch elements reversibly by using the pure Oersted field effect [GJ04]. In particular, ring geome-tries were proposed, where the concentric field of a current flowing perpendicular to a multilayer stack was shown to switch the ring [BPBZ01]. More sophisticated geometries based on rings have been used to switch between different magnetic states [CCL+06]. The combined effect of Oersted fields and Joule heating was also used to selectively move DWs [IKH+08]. The ability to control the domain structure of the wire and to switch magnetization back and forth between two well defined magnetic states using the Oersted field opens a further interesting
4.7 Conclusion 87
way to manipulate the magnetization that could be an alternative to spin-torque induced switching in micrometer size structures. Although the actual goal of ob-serving current-induced DW motion in CoFeB/Pt multilayer was not achieved, a new method is found to reliable control the magnetization in magnetic wires.
These results on the reversible current-induced switching of magnetic out-of-plane wires are published in the Journal of Applied Physics [BHR+09].
4.7 Conclusion
In this chapter, the results on the current-induced DW motion and DW transfor-mations were presented. In permalloy both, vortex and transverse walls are found to be displaced by the injected current with average velocities of about 1 m/s for 25µs long current pulses. When injecting 3 ns short current pulses, DW motion with two orders of magnitude higher velocity is observed and the DW motion seems to be more reliable. This difference is attributed to the steep current rise time that creates an additional force on the DW, which is absent for the current pulses with long rise times.
By a systematic study of the critical current density in permalloy, an inverse scaling with the wire width is found. A mechanism based on periodic DW trans-formations is discussed that is able to explain this scaling. These periodic DW transformations that occur above the Walker threshold current density are experi-mentally observed and the spin torque is identified as its origin. The occurrence of these DW transformations has some implications on the non-adiabaticityβ, since if proves that β does not equal the damping constant α, which was suggested by some theoretical papers. A detailed study of the DW velocity as a function of the current density reveals that the DW velocity has a maximum at the Walker threshold. This can only be explained by theory for β > α.
The experiments on Ho doped permalloy show that the slope of the DW veloc-ity vs. current densveloc-ity does not change with increasing damping. This shows that β and α scale similarly with the Ho doping concentration. Also the average DW velocities below the Walker breakdown are independent on the damping constant.
For field-induced DW motion, the DW velocity is found to depend strongly on the damping. This proved how different the underlying mechanisms of field- and current-driven DW motion depend on the material parameters.
CoFeB in contrast to permalloy shows no reproducible current-induced DW motion. Joule-heating induced DW transformations from transverse walls to the
88 Current-Induced Domain Wall Motion
energetically more favorable vortex walls are observed, but structural damage sets in before the DWs start to move reliably.
In soft out-of-plane magnetized multilayer Pt/CoFeB/Pt the influence of the spin torque is found to be dominated by the Oersted-field effect. The Oersted field generates a DW that is parallel to the wire and no spin torque induced DW motion is observed. However, the Oersted-induced switching of the magnetization presents a novel way to control the magnetization inside a wire independent on the spin torque.