Domain Wall Motion in CoFeB

For current-induced DW motion, one of the key problems is the required high current density, which results in Joule heating. This can induce random DW dis-placements and even the nucleation and annihilation of DWs [YNT⁺05, JKB⁺07].

To overcome this problem, different approaches have been proposed. The critical

4.5 Domain Wall Motion in CoFeB 79

current density depends on the wire geometry and wider wires lead to a reduction of the required current density (see also section 4.3.1). If the critical current density is governed by intrinsic pinning [TK04], the transverse anisotropy and the damping constant can be engineered to reduce the critical current density [LCM⁺07]. In case of extrinsic pinning at defects, such as edge irregularities and grain boundaries, the critical current density was found to scale with the critical fields for field-induced DW displacement Hcrit[LXV⁺99]. Smaller coercive fields than found in the usu-ally used permalloy have been measured for CoFeB and recently very low critical current densities have been observed for current-induced DW motion in multilay-ers made of Co/Cu/CoFeB [LCM⁺07]. CoFeB that is grown without annealing is known to be amorphous and thus exhibits no grain boundaries where pinning could occur.

Experiment

The setup described in section 3.2.2 is used to inject current pulses into the struc-ture and to apply magnetic fields inside the PEEM. A topographic SEM image of a set of zig-zag wires is presented in Fig. 4.13(a). The wires are fabricated by electron beam lithography (see chapter 2) and sputtering from a CoFeB tar-get (Co₆₆Fe₂₂B₁₂) without subsequent annealing, preserving the amorphous state.

The thickness of the CoFeB is 20 nm with a 2 nm Ru capping layer and the follow-ing wire widths are used: 220, 400, 750 and 1500 nm. One of the drawbacks of usfollow-ing CoFeB compared to 3d-metals or their alloys is the higher resistivity [JYC⁺06]. In the CoFeB wires the resistivity is about a factor 2.5 higher (170µΩcm) than for comparable permalloy wires (70µΩcm). This means that for identical geometries a voltage that is 2.5 times higher than for permalloy has to be applied to obtain

Figure 4.13: (a) SEM image of CoFeB zig-zag wires (20 nm thick, 1500 nm wide). The Au-pads are indicated by the hatched areas at the wire ends. (b) X-PEEM image showing the corresponding magnetization configuration after applying an external magnetic field along the direction indicated by the white arrow and relaxing the field back to zero.

Different shades of gray indicate the direction of the magnetic spins (see gray scale bar) and black arrows are used to visualize them. (c) High resolution image of a transverse wall at one of the kinks. (From [HKB⁺08a])

80 Current-Induced Domain Wall Motion

the same current density and the Joule heating is therefore increased by the same factor.

The magnetization of the sample is initialized by a strong in-plane magnetic field in the direction indicated by the white arrow in Fig. 4.13(b). The resulting magnetization configuration with the magnetization pointing in opposite directions in adjacent branches of the wires and transverse head-to-head or tail-to-tail DWs at the kinks is shown in Fig. 4.13 (b), which was taken after the capping layer was sputtered off. A high resolution image of a transverse head-to-head DW in a 1500 nm wide and 20 nm thick wire is shown in Fig. 4.13 (c).

Results and Discussion

For permalloy wires with the same dimensions of 1500 nm width and 20 nm thick-ness, vortex walls are observed after initialization [LBB⁺06b], which means that compared to permalloy the phase boundary between transverse walls and vortex walls is shifted to larger thicknesses and widths. This is explained by the signifi-cantly smaller saturation magnetization of CoFeB (0.75T [YTFS04]) compared to that of permalloy (>1T). This lowers the stray field energy and makes transverse walls energetically more favorable. As expected, such transverse walls are also observed for all the other smaller wires. The DW widths are similar to DWs in permalloy, corroborating observations that for such geometries and low anisotropy materials the wall widths are governed primarily by the geometry [BSK⁺07].

Figure 4.14: X-PEEM images demonstrating CIDM in CoFeB. (a) Initial transverse wall in the 1500 nm wide and 20 nm thick wire. (b) The DW is displaced in the direction of the current flow (indicated by the arrow) after a pulse injection (j=1·10¹²A/m²for 25µs) and transforms to a vortex wall identifiable by the characteristic dark-bright contrast. (c) and (d) show a similar event for the 750 nm wide wire (j=2·10¹²A/m²). (From [HKB⁺08a])

4.5 Domain Wall Motion in CoFeB 81

In Fig. 4.14(a) a 1500 nm wide CoFeB wire is shown before the current injec-tion, with a transverse wall located at the kink. The current density is gradually increased in steps of 10¹¹A/m² until a change in the magnetization configuration is observed. No changes are found up to a current density of1·10¹²A/m². At this value the DW transforms into a VW and is slightly displaces in the direction of the electron flow as seen in Fig. 4.14(b). In 750 nm wide wires the wall spin structure also transforms from transverse to vortex walls and DW motion is observed at a current density of about2·10¹²A/m² [see Fig. 4.14(c) and (d)]. For the narrower wires no changes up to a breakdown current density of around 5·10¹²A/m² is observed.

At current pulses above2·10¹²A/m², structural damage starts to set in. This is seen in Fig. 4.15 for the 400 nm wide wires. At the interface between the Au pad and the CoFeB wire, the wires are damaged, which is also indicated in an increase of the sample resistance by 50%, but the DWs do not move. Compared to the permalloy wires, a further difference in the behavior of the CoFeB wires is also that no nucleation or annihilation of DWs is observed.

To further study the origin of the high critical current density, MOKE mea-surements are carried out to study the field-induced DW motion⁸. For the 1500nm wide CoFeB wires the depinning field is 16.7±0.3 G whereas for permalloy with

8The MOKE measurements were done by T. A. Moore and P. Möhrke

Figure 4.15: The images in the top row show the topographic contrast for the 400 nm wide wires, and in the lower row the corresponding magnetic images are presented. After applying a vertical magnetic field DWs are formed at the kinks (a) and (d). After the current pulse injection (j=2·10¹²A/m²) the DWs are still located at the kinks (e), but the topographic image (b) reveals the structural damage of the wires, especially of the lower one (highlighted by an arrow). Images (c) and (f) show the strong damage at the interface between of the Au pad and the CoFeB wire where part of the Au evaporated, but the DW in (f) is still unchanged located at the kink. (From [HKB⁺08a])

82 Current-Induced Domain Wall Motion

the same wire geometry the depinning field is only 13.4±0.6 G. The coercitiv-ity (nucleation field) for CoFeB on the other hand is about 30% smaller than in permalloy.

When current is injected, different effects can occur. If the initial DW type observed is a metastable state, that is separated from the lower energy wall type by an energy barrier, the Joule heating will allow this energy barrier to be over-come and the wall can transform to the lower energy wall type [LBB⁺06b]. For sufficiently high current densities the wall will then be moved in the electron flow direction due to the spin torque effect. The observed permanent DW transforma-tions indicate that the energetically lower lying wall type for the wide wires (750 and 1500 nm) is a vortex wall [JKB⁺07, LBB⁺06b] and for wire widths smaller than 750 nm the transverse wall constitutes the lower energy spin structure.

No changes up to high current density (>1·10¹²A/m²) were observed, which is comparable to the critical current density in permalloy (see section 4.3.1). This fact and the high depinning field for CoFeB support the interpretation, that DW depinning either field- or current-induced is dominated by extrinsic pinning at edge irregularities, whereas nucleation of domains is primarily governed by intrinsic pinning.

The critical current density increases from 1·10¹²A/m² for the 1500 nm to about2·10¹²A/m²for the 750 nm lines and to above5·10¹²A/m² for the narrower lines. A similar scaling behavior is observed in permalloy wires (see section 4.3.1).

The fact that in CoFeB no nucleation of DWs is observed at high current densities can be explained by the high Curie temperature of CoFeB (>1300K [NVS⁺00]) that is more than 450 K higher than for permalloy. This means that even for the highest current densities, where already structural damage of the wire occurs, the sample temperature stays below the Curie temperature.

In conclusion, transverse DWs are initially observed for wires up to 1500 nm in width in agreement with the low saturation magnetization present in this material.

High critical current densities of j_c > 1·10¹²A/m² are necessary to induce any magnetization changes in the 1500 nm wide wires. This critical current density increases as the wire width is reduced in agreement with results in permalloy (see section 4.3.1) and explains why no wall motion for wire widths <750 nm is ob-served. The results imply that the DW pinning is primarily due to edge roughness that seems to be significantly higher in CoFeB than in permalloy and not due to pinning at grain boundaries, which should be absent in CoFeB.