5.5 Stability of the orthogonal exchange bias
6.1.3 Magnetization reversal under applied magnetic field
The same current sweep as in figure 6.3 is now conducted under various applied external fields.
The external field direction is along the effective exchange bias field as well as the current direc-tion. The results are shown in figure 6.4.
The first observation is the enhanced switching for positive magnetic fields. Apparently, this field helps breaking the symmetry of the system and reduces the critical switching current at some parts of the sample. The effective symmetry breaking field is a combination of the exchange bias field and the externally applied field. This enhancement appears to saturate for fields above 20 mT. This indicates a full reversal of the magnetization.
a.
c.
b.
-5 mT
∆(RAHE)
Figure 6.4: Current-induced magnetization reversal under various IP fields. For an applied field of −5 mT, the magnetization reversal is non-deterministic, indicating a compensation of the internal exchange bias field. a: For Hext> −5 mT, the same behavior as in the field-free case is observed. Increasing the field increases ∆RAHE. b:
Hext< −5 mT cause a reversed magnetization change. c: The change in RAH Eplotted versus the applied field. The arrow in figure a indicates ∆RAHE. At high applied fields, saturation of ∆RAHE indicates full magnetization reversal. An exchange bias compensation field Hcomp=−5 mT is indicated by the dashed line.
Perhaps the most interesting behavior is seen at an external field of exactly -5 mT (shown both in figure 6.4a and 6.4b). Although at current densities above Jpulse=0.5 · 1012A/m2 changes in RAHE are visible, the resulting RAHE is close to zero. As RAHE corresponds to the magnetization magnitude (see equation (4.2)), this indicates an even distribution between up and down do-mains in the F-layer. Apparently, the switching is fully non-deterministic at this specific applied field, which we call the compensation field Hcomp. One could interpret this as a compensation of the effective exchange bias field by the applied external field. A consequence of this inter-pretation is that the effective exchange bias field should be equal to 5 mT, while the measured exchange bias in a sheet of the same material is 59 mT (see section 5.2). This discrepancy is extensively discussed when interpreting the results in section 6.2.
Increasing the magnitude of the negative magnetic field reverses the switching behavior, as seen in figure 6.4b. The curve at −10 mT is identical to the one at 0 mT except being reversed. Hence, the behavior appears to be symmetric around Hcomp. This symmetry is even more prevalent when the change in magnetization ∆(RAHE)in a single current sweep is plotted versus the applied field.
This in shown in figure 6.4c. The curve is close to symmetric around Hcomp=−5 mT. Saturation is found at Hext= Hcomp±15 mT. The curve is very similar to the switching phase diagrams reported by for example Hao et al., with an offset of Hext=−5 mT [57]. A small difference in saturation of RAHE is found comparing the negative and positive fields. The cause of this difference is not known, but can be due to small deviations in magnetic field strength for negative or positive fields.
6.1.4 Comparison of Hcomp in different exchange bias configurations
To investigate if the external field compensates the internal exchange bias field as observed in the previous section, the same experiment is conducted on samples with different exchange bias directions. The resulting ∆(RAHE)versus applied field data is shown in figure 6.5. Three differ-ent configurations are explored. Figure 6.5a shows the result of the previous section, just as a comparison. In this case the applied field, current and field cooling direction are all parallel to each other and resulted in Hcomp=−5 mT.
5 mT 0 mT 2 mT
HEB HEB
Ipulse
Ipulse
Ipulse
a. b. c.
Hext Hext Hext
HEB
Figure 6.5: ∆RAHE for different exchange bias configurations. a: IP exchange bias along the current direction results in Hcomp=−5 mT for fields applied along the cur-rent direction. This corresponds to the effective exchange bias field. b: Determining
∆RAHEfor an as deposited sample results in Hcomp =0 mT. c: If the exchange bias direction is perpendicular to the applied field an current, Hcomp is reduced to 2 mT.
Figure 6.5b shows the results of a sample in which the exchange bias is still in its as-deposited state (refer to chapter 5 for more details about different exchange bias configurations). The external field and the current are applied in the same direction. Hcomp =0 mT is found in this case, indicating the absence of an internal magnetic field sufficiently strong enough to induce deterministic switching.
In figure 6.5c the sample is identical to figure 6.5a but the IP exchange bias direction is now ori-ented perpendicular to the current and the applied field. Hcomp≈2 mT is found. The compensa-tion field is reduced, although is has not become zero. Whether this is due to some misalignment of the field-cooling direction with the applied field or this is caused by other effects, is hard to tell. It will be explored further in section 6.2.
From these experiments it can be concluded that the exchange bias direction is an important parameter for the emergence of deterministic switching without external field applied. In the next section, the exchange bias parallel and perpendicular to the current direction are investigated further.
6.1.5 Switching phase diagrams
To precisely determine the influence of different applied currents and fields on the switching of the sample, a switching phase diagram is made and is presented in figure 6.6.
Parallel to exchange bias Perpendicular to exchange bias
HEB
Ipulse
Ipulse
Hext
Hext
HEB
a. b.
Figure 6.6: For a range of pulse current densities and applied fields, ∆RAHE after the current pulse is determined. It is normalized on the biggest recorded change of RAHE. The graph shows interpolated results from a grid of 25x20 measured points.
a: The current and field are applied along the exchange bias direction. This results in switching without external field for current densities > 7.5 · 1011 A/m2. The red arrow indicates an interesting threshold at which a part of the magnetization switches even for low currents. b: The current and field are applied perpendicular to the exchange bias. Only for Hext>5 mT a substantial switch of the magnetization is observed.
Before each current pulse, the magnetization of the sample is first saturated in the −~z direction by applying the maximum field and current (corresponding to 10 V and a field of -30 mT). The more ideal case of saturating with a magnetic field in the z-direction proved to be impossible as the coercivity of the sample exceeded the maximum field that can be applied. After saturation, RAHEis measured. Subsequently, a current pulse J is applied while simultaneously activating a magnetic field along the exchange bias direction. RAHE is now measured again. The resulting change
∆(RAHE)in Hall resistance is normalized to the largest recorded ∆(RAHE)and plotted verses the applied field and applied current pulse, shown in figure 6.6. The case in which the exchange bias is parallel or perpendicular to the field and current is shown, corresponding to figure 6.5a and 6.5c. The figure shows the interpolated results extracted from a grid of 25 (resolution of J) by 20 (resolution of Ht e x t) measured points.
For the parallel case (figure 6.6a), the magnetization switches when no field (or even a small negative field) is applied. This confirms our previous measurements. An increase in either the field or current increases the switching probability, as is expected from spin Hall driven switching experiments [36]. In figure 6.6b, the same sample is used, but the pulses and field are applied perpendicular to the exchange bias. The graph is shifted up by 5 mT, compared to the parallel case. This corresponds to the previously found Hcomp. At 0 mT, only a maximum switch of 50 % is found, indicating non-deterministic magnetization reversal in the perpendicular case. This direct comparison of both switching diagrams suggest that the effective exchange bias field corresponds to this compensation field, as the exchange bias direction is the only difference between the measurements in figure 6.6a and 6.6b.
In addition to the aforementioned remarks, in the parallel case, non-zero magnetization reversal at current densities exceeding J = 5.5 · 1011 A/m2 is observed (indicated by the red arrow).
Apparently, even for low currents, a part of the magnetization switches. This effect is not observed in the perpendicular case. It will be discussed in section 6.2.