µMAG Standard Problem Strategy

Introduction

The strategy we used in collecting results for standard problem #1 involved a promise to keeping the identities of the contributers private. This strategy was adopted for a couple of reasons:

to reduce the likelihood of ``proof by reputation'', and
to increase participation by providing a shielded critique.

At our workshop in January, a show-of-hands poll indicated that standard problem solutions would be more likely if the results were to be published as regular papers. As we make the transition to published standard problem results, there are a number of issues that we should address.

New Strategy

It will be important, as we discuss new standard problems, to remember to keep the standard problem simple. For each new standard problem that is proposed, there are an infinite number of variations that can be thought up. For example, fields can be applied in any direction, and in different sequences to calculate hysteresis loops, torque magnetometer loops, ferromagnetic resonance spectra, AC susceptibility, etc. as a function of material parameters, and geometries, etc. These calculations may have scientific value and they may be suitable for publication in archival journals. The standard problem, however, should be limited in scope. If the standard problem is simple, initial papers may be published which deal solely with a given standard problem. Later publications dealing with variations may include standard problem results in a separate section as a starting point. Eventually, however, I envision a situation where standard problem results come to be regarded as reliable, and new computational publications will mention a comparison against standard problem results in much the same way as a good experimental paper would mention instrumental calibration. In order for this to happen, the standard problem must be simple enough that it doesn't take a lot of time away from more investigative work.

Data Comparison

One very important requirement of a good standard problem is that a precise comparison of results should be possible. For standard problem #1, we accomplished this by requiring data be submitted in a particular format, but such a requirement is clearly not possible for journal publications.

Proposal

We would still like to provide this web site as a collection point for standard problem results. So, as incentive for providing data, we will offer some free publicity on the µMAG web site in the form of

tables of your data,
summary plots created from the tables, if appropriate,
a link to your preprint server, and
a full journal citation.

We want to encourage published solutions of each problem from a number of groups, and having data tables available on the web will allow subsequent publications to easily find, incorporate and cite your work!

Publication

It is important to be able to compare results, but the reason for doing the comparisons is to evaluate a variety of micromagnetic techniques, and to learn how to do micromagnetic computations reliably and, when possible, quickly. When you are preparing a standard problem paper, please remember that colleagues will want to know about all of the aspects of your calculation, sooner or later. Please include descriptions of:

your formulation of the problem (FEM, BEM, finite difference, etc.),
how you determined that the disretization was fine enough,
your initial conditions,
how you relaxed the system,
how you decided that the solution had converged,
how you stepped the applied field,
how you handled magnetostatic effects,
how you handled exchange and anisotropy effects,
and anything else that is special about your technique.

Without this information, your solution will have much less value.

Site Directory

µMAG organization / NIST CTCMS / rmcmichael@nist.gov
10-FEB-1998