muMAG Workshop Report
muMAG MICROMAGNETICS WORKSHOP
Tuesday, April 18, 1995
San Antonio Mariott Rivercenter, TX.
The workshop was held in conjunction with the IEEE Magnetics Society
INTERMAG Conference. The meeting room was filled to capacity, with an
estimated 85-100 people attending. A sign-up sheet to indicate
interest in further participation in muMAG activities and for copies
of viewgraphs from the presentations was offered at the end of the
session, with 30 responses including 14 from 10 U.S. companies.
Presentations
Bob McMichael gave an introduction stating the purpose of the workshop, the
formation of muMAG with invitations to register with the NIST Center for
Theoretical and Computational Materials Science (CTCMS) www home page,
and initial steps towards formation of working groups to address fundamental
problems in micromagnetics. Four presentations followed:
Stan Charap (Carnegie-Mellon University) introduced the distributed center
concept, the NIST CTCMS, and its functions and the inclusion of muMAG in this
program.
Randy Victora (Kodak) emphasized the need for accurate materials
characterization, focusing on moment density, anisotropy, exchange stiffness,
and damping parameters. Several comments underscored this point.
Jimmy Zhu (U. of Minn.) presented modeling of soft and hard materials, using
permalloy and recording media as examples, emphasizing the different energies
and structures that must be taken into account for accurate modeling of these
two types of materials.
Ferenc Vajada (GWU) & Jim Blue (NIST) presented computational issues that
affect results even when material parameters are correct. In particular, stiff
convergence, the necessity for iterative techniques and the existence of
multivalued solutions require specialized code at present.
Points brought up in discussion
Ed Della Torre (GWU), moderator
- Cliff Krowne (NRL) spoke on the availability of a formalism for creating
"modules" of code which are portable to avoid rewriting for new
machines. These modules should carry proprietary rights.
- Several people, including Tom Koehler (IBM Almaden) and Neal Bertram
(Center for Magnetic Recording Research, UCSD) underscored Randy
Victora's point that accurate experimental results are necessary.
- Tom Koehler, (IBM Almaden) pointed out that only modest computational
resources are needed for many micromagnetic calculations. The
computationally intensive reputation of micromagnetics is a
misconception. Also noted is that stable structures, e.g. good MR
heads, take less time to compute than noisy heads.
- Bernie Argyle (IBM Yorktown) spoke about the usefulness of having an
experimental check for modeling results. A suggested example was
motion of vortices in rectangular permalloy films. Ed Della Torre (GWU)
pointed out that the problem of verifiability is not solved by a single
experimental check; many checks are needed.
- Tom Koehler (IBM) talked about a scheme used in the eddy current
modeling community called "team problems' which act as standard
problems to test calculational techniques, and suggested that this might
serve as a model for a standard problem project in micromagnetics.
- A communication gap between modelers and experimentalists was
identified. Frustration was voiced that experimentalists don't turn to
modeling until after the experiment has reached an impossible
complexity. One suggested solution was to have a simple
micromagnetic model program so that experimentalists could get
acquainted with the requirements of modeling with a minimum
investment of time.
- Electronic device modeling is nearly "plug & chug" with SPICE and other
public domain packages. No such plug & chug code is currently
available for magnetic devices. While micromagnetics has special
requirements, a more generally useful code should be possible.
- The sophistication of a model is limited by time; what's "doable"
changes with the years, so a useful public code will be shooting at a
moving target. Currently, there is no code available which is "safe for
the Non-Skeptic". High accuracy which exceeds the accuracy needed
for mechanical finite-element modeling is needed to predict data error
rates in real heads.
The attendees showed enthusiastic involvement through out the workshop, and
an informal show of hands indicated that at least 2/3 would find a public simple
code useful. Almost an equal number thought standard problems with
experimental verification would be useful.
Post Meeting Conversations
- Don't let sensor modeling be forgotten or buried by recording
- J. Fidler notes that Alex Huber has offered code publicly for use as a
benchmark.
The action plan resulting from this workshop is available.
Site Directory
µMAG Apr '95 Workshop Report / NIST CTCMS /
rmcmichael@nist.gov