Exchange bias is more than a curious phenomenon. It plays a key role in magnetic-device technology, such as the giant magnetoresistance (GMR)-effect read heads in high-density magnetic data-storage systems (hard disks) already on the market and magnetic random access memory chips under development for low-power, nonvolatile computer memory. But exactly how exchange bias works is not understood. The spin orientation on each side of the antiferromagnet/ferromagnet interface is one of the missing pieces of information. In previous experiments, investigators achieved a major advance by demonstrating that the alignment of the ferromagnetic spins in a cobalt overlayer is in fact correlated with the spin orientation in a LaFeO₃ antiferromagnetic layer. However, owing to the complex crystallography of the LaFeO₃, they could not verify a full three-dimensional correlation between ferromagnetic and antiferromagnetic spins.

Antiferromagnetic domains on NiO(001) in an area 12 µm across. The colored arrows indicate the projections of the antiferromagnetic axes in the surface plane for four types of domains. Domains with identical in-plane projections (e.g., those marked with red and blue arrows) can be distinguished by examining their orientation out of the surface plane, as illustrated in the sketch at the bottom for the area in the dashed box. The green line represents a domain wall where the spins are in-plane.

Spinning Electrons Do the Twist

In quantum mechanics, electrons act like tiny spinning tops. Since the electrons are electrically charged, if enough of them spin in the same direction around the same axis they create a large magnetic field, and voilà, the material they are in is magnetized. Antiferromagnets are more curious in that the magnetically important electrons on neighboring atoms spin around the same axis but in opposite directions, so there is no net field. Nonetheless, manufacturers of data-storage systems (computer hard disks) and other advanced magnetic devices are keenly interested in what happens and why when one places a thin magnetic layer on top of an antiferromagnet. Contrary to what one might think, the antiferromagnet pins the magnet so that it can't change direction to align itself with an external applied field, a still poorly understood effect with the technical name exchange bias. This refusal to budge provides a reference direction that is essential to the function of modern magnetic devices. In the work reported here, Ohldag et al. have used x-ray microscopy techniques to graphically show how the axes of electron spins in an antiferromagnet twist after a magnetic layer is deposited. Their finding will force those trying to explain exchange bias to consider the actual arrangement of electron spins at the boundary between antiferromagnetic and magnetic layers rather than assuming the spin orientation is the same as in the interior.

In the new work, the collaborators used nickel oxide single crystals oriented to have a (100) surface. Nickel oxide single crystals have been well characterized in the literature and exhibit large antiferromagnetic domains that the PEEM with a spatial resolution of 50 nm for magnetic structures can easily image using the technique of x-ray magnetic linear dichroism (XMLD). Image contrast arises because the relative orientation of the polarization of the x-ray beam and the magnetic axis in the antiferromagnetic domains (changeable by rotating the crystal) determines the absorption. To image the cobalt ferromagnetic layer, the group used the now traditional x-ray magnetic circular dichroism (XMCD) with circularly polarized light.

The first set of XMLD measurements, made on bare nickel oxide, revealed a complex domain pattern related to that previously known for bulk single crystals, but with some differences. In the bulk, the domains are defined by the (111) crystallographic planes in which the spins lie and the [211] directions in which they are aligned. Analysis of the PEEM images yielded the same [211] magnetic axes but in a different arrangement. In addition, the PEEM data showed that some of the boundaries between the domains (domain walls) had decreased magnetic symmetry. Noting that the PEEM is sensitive to material only a few nanometers below the surface, the investigators concluded that the surface domain structure deviated from that of the bulk.

Antiferromagnetic (left) and ferromagnetic (right) domains after deposition of eight monolayers of cobalt. The antiferromagnetic axes have rotated into the surface plane so that only two types of domains can now be distinguished. On top of each antiferromagnetic domain, two ferromagnetic domains can be formed with their magnetization in either of two directions parallel to the antiferromagnetic axis underneath. The sketch shows the NiO spins for the area in the dashed box, which is overlaid by two cobalt domains (light and dark areas).

When a ferromagnetic cobalt layer eight monolayers thick was deposited on the nickel oxide, the story changed dramatically. The spins in the nickel oxide reoriented themselves in such a way that only domains with walls in (100) crystallographic planes remained. Moreover, the spins in the domains assumed [110] directions parallel to the interface. XMCD measurements showed that the magnetization in the cobalt domains was aligned, domain by domain, parallel to the magnetic axes of the nickel oxide domains. Heating the sample to above the antiferromagnetic transition (Néel) temperature destroyed this correlation. This graphic demonstration of spin reorientation near the nickel oxide surface means that the exchange-bias mechanism is not based on the bulk spin structure of the antiferromagnet.

Research conducted by H. Ohldag (ALS, Stanford Synchrotron Radiation Laboratory, and Universität Düsseldorf); A. Scholl and S. Anders (ALS); F. Nolting (ALS and Stanford Synchrotron Radiation Laboratory); F.U. Hillebrecht (Max Planck Institute for Microstructure Physics, Halle); 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, S. Anders, F.U. Hillebrecht, and J. Stöhr, "Spin reorientation at the antiferromagnetic NiO(001) surface in response to an adjacent ferromagnet," Phys. Rev. Lett. 86, 2878 (2001).