Report submitted by Bob McMichael.
The meeting was held in Ballroom B from 7:30 p.m. to approximately 9:30 p.m. We had some problems with our publicity scheme; unfortunately, a sizeable portion of the people registering for the conference did not receive a flyer about the meeting in their registration packet as planned. Despite this setback, about 100 people came and we extend apologies to those who may have been affected by this problem.
In keeping with the organization of muMAG projects, the agenda came in two parts: one on standard problems, and one on public code.
Tom Koehler made some comments on the effects of the choice of stopping criteria, saying that in determination of coercive field values, the stopping criteria do not necessarily have a large effect. If the iterations stop prematurely on one field step, they would be expected to start up again when the field is incremented. Subsequent computations should then be carried out with smaller field spacing and tighter stopping criteria to approach a value of the coercivity.
Jason Eicke showed results that described remanent states and hysteresis loops that differed depending on the state of the magnetization *before* raising the field to 50 mT to start the loop. This implies that the specified maximum field is not large enough to select a unique remanent state if the calculation is done as specified.
Ed Della Torre led a discussion on the possibilities for new standard problems. In particular, Ed asked the participants to think about standard problems in terms of the goal of standard problems, whether the goal is to understand the physics of the problem, to validate code, or to determine the validity limits of micromagnetic techniques. Ed also proposed that we look for standard problems with analytic solutions (Stoner-Wohlfarth, and domain walls for example).
Neal Bertram described a proposal for a new standard problem which involves only magnetostatic energy, exchange energy, and the applied field. An intrinsic length scale is determined by the exchange stiffness and the magnetization, so that the dimensions of the particle can be normalized to this length scale, delta. The applied field can also be normalized in terms of the magnetization. The suggestion was to determine coercivity of a block of material as a function of width/delta (delta = intrinsic exchange length) with the expectation that when the particle dimensions are less than delta, we should see agreement between models and a divergence of computed results as the particle dimensions grow larger than delta.
Alex Hubert suggested that we may not wish to be tied only to coercivity as a measure of agreement since computation times can be expected to diverge near a critical field, and that a rigorous determination of a critical field could be expected to require extrapolation in terms of both applied field and discretization. As an alternative, he proposed a calculation of the "single domain limit" of a cubic particle with uniaxial anisotropy Ku = 0.1 Kd. The goal is to find the size L of equal energy for the so-called flower state (which one may also call a splayed state or a modified single-domain state) on the one hand, and the vortex or curling state on the other hand. The transition is expected to be found in the neighborhood of L/delta = 8.
This proposal was well-received, and Jacques Miltat gave us an important reminder that such results must always be demonstrated to be independent of the discretization density.
Without much discussion, Bob McMichael called for a show of hands poll on the following, carefully phrased question: ``Do you think your colleagues would be more likely to submit a solution if a) it is to be anonymous, or b) it is to be published as a regular paper?'' The response was clearly in favor of published standard problem results.
John's program is based on magnetostatically interacting single domain elements in a 3D-environment. The elements may have specified magnetization, anisotropy, pinning fields and resistivity. Electrical, current- carrying circuits may be constructed by magnetic and non-magnetic elements. As examples, John showed a simulation of a spin-valve head flying over a perpendicularly magnetized medium, and a 3D NIST logo constructed from single-domain blocks.
Mike Donahue described the Object Oriented MicroMagnetic computing Framework (OOMMF) software which is currently a `pre-release' which can be found at http://math.nist.gov/oommf. along with a postscript version of Mike's viewgraphs from the workshop.
The software is written in modular form with modules communicating over network sockets. Currently available modules include a problem editor and a number of display modules including graphs and vector plots. At the heart of the suite is a solver which currently handles a 2D array of 3D spins.
The user interfaces are written entirely in Tcl/Tk for maximum portability with. TCL/TK is a freely available scripting language that runs on a wide variety of platforms. The computationally intensive portions of the code are written in C++. The software has been successfully compiled for Windows 95, Windows NT, Linux/Alpha, Linux/x86, Sun, and SGI systems.
Using another micromagnetic NIST logo, Don Porter closed the meeting with a demonstration of the OOMMF software and the flexibility provided by the modular design. Use of the software includes specifying the problem in the problem editor window, starting up a solver, possibly on a remote machine, and telling it where to get the problem description, specifying output formats and starting up display widgets to handle the output from the solver.
Although they were not discussed at the workshop, the following tasks appear to me to deserve attention.