Nanoscale Materials Science  
Forces, surfaces and magnetism
Molecular surface science
Nanostructured coatings
ToF-Sims
SUL
Surface technology
Tribology
Magnetic thin films and devices

Exchange coupling in antferromagnetic/ferromagnetic multilayers

The balance between exchange, anisotropy, and dipolar energies usually determines the domain structure in ferromagnetic (FM) films. However, coupling a ferromagnetic film to an antiferromagnetic (AF) layer can significantly alter the hysteresis processes and domain structure due to the additional FM-AF interfacial exchange [1]. Exchange-biasing results from field cooling such a sample through the Néel temperature (TN) of the AF. The FM-AF interaction provides an effective bias field that gives rise to a horizontal shift in the hysteresis loop. Interaction with AF layers can also establish or increase the anisotropy of the FM layer and often enhances the coercive fields. The effect was discovered in 1956 by Meiklejohn and Bean [2] when studying Co particles embedded in their native antiferromagnetic oxide (CoO). Since then, exchange bias was observed in many different systems containing FM/AF interfaces, such as small particles, inhomogeneous materials, FM films on AF single crystals and thin films. As such, exchange biasing has become an important tool for controlling domain formation in magnetic devices [3]. While routinely exploited, the microscopic origin of exchange bias is still open to debate. It is generally believed that the bias is related to defects in the antiferromagnetic order at the interface to the ferromagnet or in the volume of the antiferromagnet itself, that leads to uncompensated spins inside the AF layers or at the AF interfaces [4].

Recently, we could show that high resolution quantitative MFM in high magnetic fields is a promising tool to efficiently address this topic [5]. The domain reversal of the ferromagnet within the hysteresis loop as well as the stray field from the uncompensated antiferromagnetic spins in a field larger than the saturation field of the ferromagnet. A quantitative analysis of the MFM images made it even possible to quantify the density of uncompensated spins.

MFM in the hysteresis loop

In the experiment at hand, a multilayer of CoPt interspaced with CoO was imaged with MFM. First a large series of MFM images were taken within the loop field and the domain formation as well as annihilation could be studied in detail. Figures 1 and 2 two selected MFM images of the positive and negative branch of the hysteresis loop. Prior to imaging, the sample was cooled down to 8K without an applied field in order to achieve an equal distribution of perpendicular (up and down) domains. Upon ramping up the field, the domains standing against the field direction (white domains) vanish until at about 800mT, the ferromagnet is saturated. The remaining MFM contrast seen is due to the magnetic stray field of the interfacial antiferromagnetic uncompensated spins.

Fig. 1: MFM images of the positive branch of the hysteresis loop.

The last image of figure 2 shows the magnetic multilayer in an applied magnetic field of 7T. It is obvious, that even in 7T field, the antiferromagnetic uncompensated spins are not aligned in field direction.

Fig. 2: MFM images of the negative branch of the hysteresis loop.

The reason why the same domain species (white domains) also annihilate on the negative branch of the hysteresis loop has its origin in the magnetization direction of the tip. The coercivity of the tip magnetization is very weak, therefore, the field needed to reverse the magnetization into the new field direction is lower than 100mT. As a consequence, all domains imaged in white on the positive branch of the hysteresis loop are imaged black on the negative side of the loop and vice versa (cf also figure 3 to the left). From the two images a) and c) it can be observed, that the ferromagnetic domains can be completely recovered, even after the external magnetic field has been ramped up to 7T.<-/p>

Fig. 3: Perfect recovery of the ferromagnetic domains after saturation; problem of cantilever magnetization reversal in opposite fields.

A close inspection of the contrast within the blue boxes (figure 1 a) and e)) shows that the MFM contrast of the ferromagnetic domains and the pattern imprinted in the antiferromagnetic interfacial spins is antiphase. This indicates and antiparallel alignment of  the direction of the uncompensated antiferromagnetic moments and the magnetization direction of the ferromagnetic domains. While parallel alignment of the ferromagnet and the antiferromagnet has been reported several times [6,7], a so called „antiferromagnetic coupling“ across the af/fm interface has not been reported yet. This suggests that the coupling is mediated by a superexchange interaction via the oxygen atoms and implied that the interfacial layer of the antiferromagnet is oxygen terminated.

Fig. 4: Antiparallel alignment of ferromagnetic and antiferromagnetic moments of the MFM contrast.
Quantitative evaluation of the MFM images

To further elaborate on the above conclusions, the MFM data was quantitatively analyzed. First the cantilever was calibrated by a method developed by van Schendel et al.

Fig. 5: Quantitative analysis of the MFM images.

 

 

References

[ 1] A. E. Berkowitz and K. Takano, JMMM 200, p552 (1999).
[ 2] Meiklejohn, Bean PR 102, p1413 (1956)
[ 3] J. C. S. Kools, IEEE Trans. Magn. 32, p3165 (1996).
[ 4] K.Takano et al., PRL. 79, p1130 (1997).
[ 5] P. Kappenberger et al., PRL 91, p267202 (2003).
[ 6] H. Ohldag et al., PRL 91, p017203 (2003).
[ 7] F.T. Parker et al., PRB 61, pR866 (2000).
[ 8] P. J. A. van Schendel et al., JAP 88, p435 (2000).
[ 9] Inside a [111] Co2+ ion layer the magnetic moments are ferromagnetically coupled and aligned with the [117] direction. Therefore, if the magnetic spins are aligned along the [117] direction instead of directly along the [111] surface normal the measured magnet moment will be reduced by cos(43.3°), the angle between the [111] and [117] directions. In neglecting this angle, an error of 28% is introduced into the magnetization. However, omitting the cos term makes the calculations comparable to results from other groups.
[ 10] S. Maat et al., PRL 87, p087202 (2001).

zurück  Links  Artikel versenden  Druckansicht