OOMMF software requirements

We use three pieces of software to perform micromagnetic studies with the finite difference method. Each of these packages is an extension on other widely-available applications (see figures 2.15 and E.1).

Figure 2.15: The simplified simulation process. The left-hand side of the chart represents the input files for the simulation packages for OOMMF (finite difference method; the micromagnetic information format file contains material, simulation and geometric parameters) or magpar (hybrid finite element/boundary element method, material; simulation and geometric (mesh) parameters as individual files). The results from each of these packages are transformed into unified output formats (right-hand side) for analysis (xmgrace (Turner, 1995)) and visualisation (VTK)

The first piece of software is a proprietary program, mifmaker, which we developed to create simulation environments. Ordinarily a significant amount of manual effort is required to generate a simulation, as the problem must be directly defined in a Tcl-based format which OOMMF can recognise. There is no method built-in to OOMMF which allows this process to be automated for three-dimensional problems. The mifmaker program is a command-line application which can accept a series of geometric, material and simulation parameters and generate a valid OOMMF problem description.

Using mifmaker it is straightforward to generate batches of valid simulation problems which can subsequently be solved. This is ideal for performing size-dependence studies and generating phase diagrams. The operation of mifmaker depends on Python (Hetland, 2002, Lutz and Ascher, 2003, van Rossum and Drake, 2001).

After the appropriate magnetic problem has been generated by mifmaker, this is sent to OOMMF -- the Object Oriented MicroMagnetic Framework -- developed by the National Institute of Standards and Technology. OOMMF can then solve the micromagnetic problem which we have presented to it. OOMMF is heavily dependent on Tcl/Tk (Ball, 1999, Flynt, 1999, Raines and Tranter, 1999, Smith, 2000, Welch, 1999).

Figure 2.16 shows how the requirements for system memory in OOMMF scale in a cubic system (i.e. the length of the $x$, $y$ and $z$ sides of the cube are the same length and have the same discrete cell size) as a function of the number of cells.

Figure 2.16: The memory requirements of OOMMF as a function of the number of discrete simulation cells per edge for a three-dimensional geometry