Data transfer rates of order 100 MB/second in computer data storage will require an improved understanding of high-speed switching in magnetic materials. Researchers at NIST (National Institute for Science and Technology) in Boulder, CO, have recently done experiments (Kabos, Kaka, Russek, and Silva, IEEE Trans. Mag. 36, p. 3050 (2000)) on high-speed magnetization dynamics in materials used in recently introduced MRAM (magnetic read-only memory) that show unexpected results -- under certain circumstances the magnetization can get "stuck" in a certain direction. To be able to write memory reliably, it is important to understand this phenomenon. The UA group has done the first simulations that reproduce this metastability. They have also shown the mathematical equivalence of magnetization switching and the motion of a roller coaster, and developed a visualization method based on this equivalence that makes it much easier to visualize and understand the behavior of the magnetization.
Here is a link to a 13MB quicktime movie. On the left is a top view of a thin film of permalloy, such as might be used in an MRAM device. A 2 x 20 array of cells is shown, using periodic boundary conditions. In each cell is a magnetization vector. The vectors are nearly in the plane of the screen, so (since their lengths are fixed) their tips follow a circular path as they rotate. This path is shown at right, with a roller coaster track above it, with a slightly tilted perspective so the viewer can see the height of the track, which is just the magnetic energy. The equations of motion of the roller coaster are mathematically identical to those of the magnetization vector. Initially the magnetizations (represented by a line sticking up from the track) all point to the left, then a downward magnetic field pulse is applied which makes the track much steeper and makes the magnetizations swing down (toward the viewer, in the roller coaster view). The pulse is turned off after a couple of oscillations (restoring the original flatter track shape) and the system is left in a metastable state, which eventually breaks into domains (one with rightward magnetization, the other left); a small leftward bias field causes the domain walls to drift so one domain disappears and the system is left in its original state. [Caveat: In this simulation we have not included magnetostatic interactions. They are not important as long as the magnetization is nearly uniform, or nearly perpendicular to the domain walls. Toward the end of this simulation this condition is not satisfied, and demag fields will prevent the large-angle rotations seen in this simulation. Instead, we expect "ripple domains" with smaller-angle boundaries.]
Time variation of magnetization angle is shown in this graph:
This work was published as Xuebing Feng, P. B. Visscher, and D. M. Apalkov, "Micromagnetic simulation of thermal ripple in thin films: 'Roller-coaster' visualization", J. Appl. Phys. 93 no. 10, 15 May 2003.
It was presented at the Nov. 02 MMM conference, paper FG12.