Exchange bias can be understood qualitatively by a simple model. In an external field and above the Néel temperature (T_N), antiferromagnetic spins located at the interface with the ferromagnet are aligned, like the ferromagnet, with the field. Once cooled below T_N, these interface spins thereafter keep their orientation and appear "pinned" because they are tightly locked to the spin lattice in the bulk of the antiferromagnet, which is not sensitive to external fields. Consequently these pinned spins produce a constant magnetic field at the interface that causes the hysteresis loop of the ferromagnet to shift. However, this intuitive picture overestimates the magnitude of the loop shift by orders of magnitude.

Exchange bias. The hysteresis loop of a ferromagnet in contact with an antiferromagnet shifts after cooling in an external field to a temperature below the Néel temperature (field-cooling), so that one magnetization direction becomes preferred.

When a Minority
Rules

Sometimes a small minority can have a big voice. Consider the influence of a few atoms on the exchange-bias phenomenon in hard drives in today's desk- or laptop computers. The ability to rapidly and accurately "read" the stored data is due to an ultrathin sandwich of magnetic and nonmagnetic materials in a read head that sits closely over the rapidly whirling disk. Exchange bias provides a magnetic reference, so that the read head can sense binary bits of information as the direction of magnetization in a small area of the disk changes between the two that are allowed.

How exchange bias works has been a mystery for 50 years. Various models are based on the idea that magnetic atoms are like tiny spinning tops that generate minute magnetic fields. If the spin directions are aligned, the collective action of many atoms can give rise to a measurable magnetic field, as in permanent magnets. Such a picture applied to the interface region between the layers of the sandwich can explain exchange bias, but a simple model results in an effect that is many factors of 10 too large. Ohldag et al. have used an x-ray spectroscopy technique that is sensitive to the interface to show that only about 5 percent of the spins there contribute to exchange bias, thereby bringing the model into line with reality.

Numerous efforts have been made to explain the difference, but the problem has been the missing knowledge about the actual spin structure at the interface, which need not be the same as in the bulk. The goal of the collaboration was to detect pinned interfacial spins and use the information to find a direct correlation between their number and the size of the loop shift. The unique ability of x-ray magnetic circular dichroism (XMCD) to probe magnetic properties of individual elements in thin-film samples was the key to success.

Hysteresis loops of cobalt (top) and manganese (bottom) in an exchange-biased layer of ferromagnetic cobalt on antiferromagnetic iridium/manganese. The loops for both elements shift, with the direction (red and blue) of the shift depending on the sample geometry. In addition, there is a vertical shift between the manganese loops. Both shifts are due to a small fraction of pinned spins at the interface.

For this purpose, the researchers used elliptically polarizing undulator (EPU) Beamline 4.0.2 at the ALS to measure hysteresis loops of interfacial spins in samples comprising thin layers of exchange-biased ferromagnetic cobalt or cobalt–iron on thicker layers of antiferromagnetic iridium–manganese, platinum–manganese, or nickel oxide, all on silicon substrates. In each case, they measured hysteresis loops in two sample orientations that gave loop shifts parallel and anti-parallel to the positive field direction, respectively. By tuning the photon energy to the manganese or nickel absorption edges, they detected the extremely small XMCD signal arising only from the interfacial spins in the antiferromagnet (there is no signal from the antiferromagnet itself). These hysteresis loops exhibit a vertical shift in addition to the horizontal shift. Like the preferred magnetization direction, the vertical shift reverses if the sample geometry is reversed.

Left, in an ideal, single-crystal antiferromagnet (green), all spins at the smooth interface (red arrows) are pinned and contribute in the same way to the internal field acting on the top ferromagnetic layer (blue).
Right, in a real sample with a rough interface (horizontal gray) and grain boundaries (vertical gray), most spins at the interface (white arrows) are not pinned and follow the magnetization of the ferromagnet. The small number of pinned spins (red arrows) causes exchange bias.

Two important conclusions follow from the interface hysteresis loops. First of all, most of the interfacial spins couple strongly to the ferromagnet on top and follow its magnetization, thereby exhibiting the same hysteretic behavior. Second, only a small fraction (5%) of interfacial spins is actually pinned, and these cause the horizontal hysteresis loop shifts. In addition, the vertical shift is caused by these pinned interfacial spins, which were aligned by the cooling field and remain pointing in a fixed direction. The very small number of pinned spins observed explains why it was not possible to detect these spins with conventional magnetometry techniques. Furthermore, it shows that the simple model for exchange bias is correct in principle, if one assumes "mixed coupling" across an imperfect antiferromagnetic/ferromagnetic interface instead of ideal interfaces. A next-generation photoemission electron microscope (PEEM-3) coming to the ALS will clarify the origin of the pinning.

Research conducted by H. Ohldag (Stanford Synchrotron Radiation Laboratory, ALS, and Universität Düsseldorf); A. Scholl, E. Arenholz, and A.T. Young (ALS); F. Nolting (Swiss Light Source); S. Maat and M. Carey (Hitachi Global Storage Technologies); and J. Stöhr (Stanford Synchrotron Radiation Laboratory).

Research funding: U.S. Department of Energy, Office of Basic Energy Sciences (BES). Operation of the ALS is supported by BES.

Publication about this research: H. Ohldag, A. Scholl, F. Nolting, E. Arenholz, S. Maat, A.T. Young, M. Carey, and J. Stöhr, "Correlation between exchange bias and pinned interfacial spins," Phys. Rev. Lett. 91, 017203 (2003).

ALSNews Vol. 240, April 28, 2004