This section will summarize the conclusions based on the discussion of the previous sections.
Most of the experimental observation can be understood.
A Pulses of equal polarity change the magnetization by a small amount, figure 6.2 It was observed that some non-deterministic magnetization reversal is visible when applying con-secutive pulses of the same polarity. This can be explained by the grain structure of IrMn. Small grains can be thermally unstable. This can cause random switching of the Co domains.
B Gradual switching, figure 6.3
The gradual switching can be explained partially by current shunting through the Hall cross. As the current density varies, different grains switch at different critical current densities. The dif-ference in exchange bias magnitude and direction among different grains can also induce gradual switching behavior. It influences the critical switching current of the Co below the IrMn.
C Incomplete switching, figure 6.4
Incomplete switching in the center of the hall cross is explained by current shunting. The center of the sample can remains unswitched as the critical current density is not reached.
D Exchange bias discrepancy (Hcomp6= HEB), figure 6.4
The difference in exchange bias is hard to explain as the actual exchange bias in the Hall cross is not exactly known and only determined by measuring Hcomp. A possible explanation for this reduced Hcompcan be due to structural effects when nanostructuring of the sample. A large in-crease in temperature during the measurement is also observed, but only for such a short time that it should not have an effect on the exchange bias. Still, an instantaneous effect of the tem-perature increase on the exchange bias cannot be ruled out. The grain distribution can also play a role in the effective exchange bias, as the exchange bias is bound to crystal axes of the sample and varies between grains. Also, the field-cooling temperature might not be large enough to set all grains in the correct direction.
E Hcomp=2 mT in perpendicular to exchange bias direction, figure 6.5
Anisotropy gradients caused by asymmetries during growth can create a small effective internal field, comparable to the magnitude of Hcompin the perpendicular direction. Misalignment during field cooling can also cause a part of the grains to have an exchange bias along the perpendicular direction large enough to cause a part of the sample to switch deterministically.
F Switch at low current densities for parallel direction, figure 6.6
The grain structure of IrMn can result in grains with a large exchange bias. If such a grain has it exchange bias parallel to the current direction (thus has the correct symmetry for deterministic switching) this can cause deterministic switching at low current densities. As the field cooling sets most of the exchange bias in this direction, this effect is less likely to occur in the perpen-dicular direction. Hence, such a switch at low current density is not visible in the perpenperpen-dicular direction.
Conclusion & Outlook
This chapter summarizes the main conclusions of this work, divided into sections corresponding to chapter 5 and 6. Each section also includes a brief outlook for experiments that can be done in the near future. At the end of this chapter a more long term outlook on the future development of MRAM is given.
7.1 Creating an orthogonal exchange bias
A successful orthogonal exchange bias has been created by using a Pt/Co/IrMn trilayer. An IP exchange bias of 50 mT and an OOP coercivity of 29 mT was found from hysteresis curves.
Inserting a Pt dusting layer between the Co and IrMn greatly increased the quality of this interface and resulted in an increase coercivity (44 mT) and slight increase in exchange bias (59 mT).
Thickness dependence
The different layers of the sample have been optimized layer by layer. The seed and buffer layer thickness did not significantly influence the properties of the sample, although a minimum seed layer thickness of 0.5 nm Ta appeared to be necessary for the growth quality of subsequent layers. A critical amount of Pt (1.3 nm) was necessary to induce an OOP anisotropy in the Co layer. Increasing the Pt thickness resulted in an increased exchange bias and coercivity, which is probably a result of enhanced interfaces.
The Co thickness appeared to be critical to the emergence of orthogonal exchange bias. Only for a small Co thickness range, OOP anisotropy was found. The ranges are found to be 0.4 nm
< tCo < 1.6 nm and 0.6 nm < tCo < 1.6 nm for the sample with and without dusting layer, respectively. Below the critical thickness, a reduced surface anisotropy was found, most likely caused by intermixing with the adjacent IrMn layer. Above 1.6 nm, the shape anisotropy becomes dominant, creating an IP easy-axis.
For the IrMn dependence it was found that low thicknesses (<4 nm) resulted in coherent rotation of the IrMn with the Co, significantly enhancing the coercivity at the cost of the exchange bias.
At 6 nm, the maximum exchange bias was found.
89
Different exchange bias configurations
The exchange bias magnitude was investigated for different field cooling directions and a varying cobalt thickness. The main result is a reduced exchange coupling for the orthogonal exchange bias configuration compared to other configurations. A possible cause can be the spin structure of IrMn which can have different crystalline anisotropies along different crystalline axes. The interface between the IrMn and the Co can also play a significant role in the measured exchange bias.
Temperature effects
It was found that thermal effects can reduce the exchange bias on short and long timescales.
For longer periods of time (a couple of weeks) the exchange bias was significantly reduced at room temperature. Heating for minutes can reduce the exchange bias at a temperature of 400 K. The results can be explained by a thermal fluctuation model taking into account the inde-pendent switching of grains in the polycrystalline structure of IrMn. A very strong dependence on temperature is found for the characteristic stable time τ of a grain, explaning the observed results.
Outlook
There are a few interesting things to investigate in the near future. The magnetic properties of the sample appeared to be significantly influenced by the interface between the Co and IrMn.
It would be interesting to further investigate this interface. By cross-sectional TEM imaging for example, one could determine the intermixing between the layers and the roughness of the in-terface in various samples. This might also reveal how a Pt dusting layer influences the inin-terface.
An open question remains why exactly the exchange bias is lower in the orthogonal configuration.
A more systematic study to determine the IrMn and Co spin structure could help to understand the difference between the configurations. X-ray diffraction could for example determine the dif-ferent crystal structures and see if there is any difference depending on the field cooling direction.
First attempts in this direction, however, proved unsuccessful.
Finally, it would be interesting to try different materials to possibly increase the thermal stabil-ity of the system, as it was seen that this can depend significantly on the grain structure and anisotropy of the materials. For example, PtMn has a higher blocking and Neèl temperature than IrMn [67].