Roscoe Giles

Demagnetizing field computation for dynamic simulation of the magnetization reversal process

The magnetic-field distribution for a thin magnetic film is computed using the fast Fourier transform technique. The method is quite general and accommodates any two-dimensional magnetization distribution. It allows the computation of fields both inside the film (demagnetizing fields) and outside (stray fields and leakage). >

Distributed data and immersive collaboration

illiSnails. Conformal projections to Euclidean space of Blaine Lawsons ruled minimal surfaces in 3D spherical space.

Simulation of the magnetization-reversal dynamics on the Connection Machine

Large two‐dimensional lattices of interacting magnetic dipoles representing thin films of amorphous rare‐earth–transition‐metal alloys are studied on the Connection Machine. The accuracy of the results is established by comparing the simulated dynamics of isolated Bloch walls with theoretical predictions. The effects of random axis anisotropy on the process of magnetization reversal are then investigated, and the nucleation and growth of reverse‐magnetized domains are found to be the mechanism of reversal. The obtained hysteresis loops have very high squareness, and sample coercivities are well within the range of experimental values.

Coercivity of domain‐wall motion in thin films of amorphous rare‐earth–transition‐metal alloys

Computer simulations of a two‐dimensional lattice of magnetic dipoles are performed on the Connection Machine. The lattice is a discrete model for thin films of amorphous rare‐earth–transition‐metal alloys with application to erasable optical data‐storage systems. Simulated dipoles follow the dynamic equation of Landau, Lifshitz, and Gilbert under the influence of an effective magnetic field arising from local anisotropy, near‐neighbor exchange, classical dipole‐dipole interactions, and externally applied fields. By introducing several types of defects and inhomogeneities in the lattice, we show that the motion of domain walls can be hampered in various ways and to varying degrees.

Three‐dimensional micromagnetic simulations on the connection machine

We are using the connection machine to perform three‐dimensional micromagnetic simulations of films with interesting topological structures such as vertical Bloch lines. A critical element of the computations is the calculation of the demagnetizing field. We have extended the Fourier method of Mansuripur and Giles to three‐dimensional films and implemented it on the connection machine. We discuss the algorithm and its parallel implementation.

Possible sources of coercivity in thin films of amorphous rare earth‐transition metal alloys

Computer simulations of a two‐dimensional lattice of magnetic dipoles are performed on the Connection Machine. The lattice is a discrete model for thin films of amorphous rare earth‐transition metal alloys, which have application as the storage media in erasable optical data storage systems. In these simulations the dipoles follow the dynamic equation of Landau–Lifshitz–Gilbert under the influence of an effective field arising from local anisotropy, near‐neighbor exchange, classical dipole–dipole interactions, and an externally applied field. The effect of random axis anisotropy on the coercive field is studied and it is found that the fields required for the nucleation of reverse‐magnetized domains are generally higher than those observed in the experiments. Various ‘‘defects’’ are then introduced in the magnetic state of the lattice and the values of coercivity corresponding to different types, sizes, and strengths of these ‘‘defects’’ are computed. It was found, for instance, that voids have insignificant effects on the value of the coercive field, but that reverse‐magnetized seeds of nucleation, formed and stabilized in areas with large local anisotropy, can substantially reduce the coercivity.

Molecular dynamics simulation of liquids on the Connection Machine

A molecular dynamics algorithm for performing large‐scale simulations on the Connection Machine, a massively parallel supercomputer, is discussed. The algorithm uses a cell data structure to obtain the near neighbors of each particle as the fluid evolves. The main features of the data structure are that one processor per particle is used to integrate the equations of motion of each particle and one processor per cell is assigned to compute the interparticle forces. The results for Lennard‐Jones fluids of 15 000 to 500 000 particles indicate that the algorithm scales linearly with the number of particles.

Parallel micromagnetic simulations using Fourier methods on a regular hexagonal lattice

Computer simulations of the microscopic magnetic dynamics of thin films provides a means for developing theoretical understanding of the behavior of magnetic recording materials and a way of relating material parameters to magnetic behavior. Such simulations have become possible because of the availability of high-performance supercomputers (such as the Connection Machine) and by improvements in algorithms for evaluating magnetic interactions. The proper implementation of the Fourier technique on a two-dimensional hexagonal lattice is described. A naive transcription of the method as used for a rectangular lattice leads to serious aliasing problems for short wavelengths. The implementation described reduces these effects. >

Coercivity mechanisms in magneto-optical recording media

The sources for the nucleation coercivity and wall motion coercivity in magneto‐optical recording thin films are investigated based on the Connection Machine simulations. It was assumed that the thin films consist of nanoscale patches which are magnetic structure other than columnar structure or crystal grains. The postulated inhomogeneities have not been observed directly, but we believe that magnetic structure must exist in view of the fact that the media suitable for practical applications do not have columnar structure and crystal grains, yet they exhibit coercivity phenomena which can only be caused by nanoscale inhomogeneities. The nucleation and wall motion processes were simulated in the patchy films with different types of inhomogeneities and with the average patch size ranging from 60 to 200 A. The simulation results show that the nucleation coercivity depends sensitively on the patch size characterizing the fluctuation of the anisotropy constant, but weakly on the exchange at the patch borders and the fluctuation of easy‐axis orientation. To account for the observed nucleation coercivity in magneto‐optical thin films the average patch size should be on the order of 100 A. In contrast, the wall motion coercivity is mostly caused by the fluctuation in the exchange stiffness constant and the patch‐to‐patch variations of the easy‐axis orientation. It was found that the domain wall, when encountering a high anisotropy region, first encircles it and then reverses its magnetization due to the domain wall force. This suggests a scenario other than thermal fluctuation by which a wall can go over a local energy barrier.

A numerical investigation of domain wall and horizontal Bloch line motion in thin films with perpendicular anisotropy

The dynamic structure of a magnetic domain wall in a thin film with large uniaxial anisotropy perpendicular to the plane of the material has been determined by numerical integration of the Landau-Lifshitz-Gilbert equation. The geometry corresponds to a domain wall whose structure is uniform along the wall, but can vary normal to the wall and normal to the plane of the material. For external fields in the range of a few Oersted, a constant mobility is observed. In this region, the wall remains flat and horizontal Bloch lines (HBL) are not formed. At larger fields, HBLs are formed that propagate through the wall and punch through at the opposite surface from which they formed. During punch through, the region of the wall undergoing a quick 180 degrees rotation is seen to move backward compared with the rest of the wall. This causes the average wall position to be stationary during punch through. At larger fields, the structure of the domain wall is significantly more complex and multiple HBLs are found in the domain wall. >