The Magnetics and Information Science Center - MISC

Magnetic Information Technology at
Washington University in St. Louis

Introduction

Washington University has been active in magnetic information technologies since 1986 when Professor Marcel Muller gained the support of the School of Engineering to create a research program in this area. Later that year, with funding from the National Science Foundation, in collaboration with Hewlett-Packard Laboratories in Palo Alto, research was begun. In 1988 Professor Ronald Indeck joined Muller after a year long fellowship in Japan. This created a strong program in both experimental and theoretical research. Professor Joseph O'Sullivan, an expert in information theory and signal processing, began a research collaboration in 1991. The Magnetics and Information Science Center (MISC) was established in 1992 as an interdisciplinary theoretical and experimental center focusing on fundamental recording physics and information science. Complementing the Center are personnel in coding theory (Rimoldi), electronics (Morley), applied physics and integrated optics (Krchnavek), materials science (Kelton, Sastry), and chemistry (Buhro).

Principal Projects

The magnetics effort comprises four principal projects:

Computational analysis/modeling, especially massively parallel processing
Magnetic thin film transducers and thermal analysis including multi-gap/multi-sensor heads, synthetic aperturing, and servoing
Medium noise studies, especially magnetic microstructure
Physically-based information science for magnetic recording
Novel storage systems

1. Computational analysis/modeling

Washington University has several numerical analytical tools applicable to magnetic information storage. These include magnetic field analysis, thermal analysis, stress analysis, and micromagnetic computation. Uniquely available to researchers at MISC are several massively parallel computers including DEC MasPar and NCUBE machines. The structures of the different machines have applicability to various problems of interest to the Center as described below in connection with the applications.

We have developed tools for computational micromagnetics comprising particularly computationally tractable algorithms. The modeling is easily carried out on readily available workstations. The tools yield results in agreement with experiment and with other more elaborate algorithms. This computational ease offers us more sophisticated analysis of various recording phenomena. We recently developed analysis allowing for a random tiling of the medium (Voronoi tessellation). This development is more physically accurate than the previous regular tiling and removes some of the nagging problems associated with the former modeling. Modeling storage media and transducer-storage medium interaction is ongoing. We expect to continue obtaining results in good agreement with experimental findings, and with predictive value, without the computational demands of the successful earlier models.

2. Magnetic thin film transducers and thermal analysis

Novel magnetic recording transducers are of profound interest to the Center, generating many new ideas. We have modeled and selectively fabricated and tested novel transducer designs such as high gradient write heads, magnetoresistive gradiometers, and side field profiles. Unconventional devices incorporating multiple sensing elements have been a major direction for our research. These include application to high spatial resolution, parallel data transfer and servoing.

Heat generation is becoming increasingly problematic as devices become physically smaller: heating limits signal output and affects sensitivity through stress. Among thermal studies of transducers we have looked at heat conduction through the air bearing in rigid disk systems.

We have investigated device connectivity and devised a stacked lead structure with no crossfeed. Recently we have embarked on a systems level interconnect study and we are investigating options for optical signal input and output connections to the transducers. This project profits from our expertise in polymeric interconnects.

3. Medium noise studies

The medium noise problem is recognized as crucial in the achievement of the highest storage density and has received much attention recently. The experimental results of our noise studies have been obtained from our specially designed and assembled precision tester; our system provides fine three-dimensional spatial control, which means that we can determine voltage waveforms for precise media positions and do not rely on bulk averaging as is commonly done with a spectrum analyzer. Besides the sophisticated magnetic recording testers our laboratory facilities include state of the art sputter deposition and conventional thermal evaporation, magnetic characteristic measurement, state of the art SEM and TEM facilities and a new magnetic force microscope.

Our work on medium noise includes:

track edge noise: its character, limitations on track width, use as a narrow source for head characterization, and use as a continuously monitored track following servo;
dc media and transition noise: characterizing their random and deterministic components and analyzing transverse correlation lengths in media;
media modeling and signal processing: with an aim toward system design including adaptive noise reduction and communications channel analysis.

Increases in areal storage density require reductions both of bit length and of track width. As tracks are made narrower, their edge regions do not scale with the track width and become an increasingly large fraction of the storage medium. Irregularities of the track edge assume greater importance as noise sources, and interactions between the edges, both of the same track, and with adjacent tracks may be limiting factors of track narrowing. In our studies we are investigating the size and distribution of track edge fluctuations; the chirality of track edges and their mutual interaction: at what distance are track edge walls uncorrelated; when, if at all, do they become correlated; can we maintain uncorrelated walls even at very high track densities.

The track edge research is one aspect of the experimental and theoretical study of the relation between the physical and magnetic microstructures of magnetic information storage media and transducers and their function in recording systems, with a view to the eventual application of the results to new methods of increasing storage density and access rate. Another aspect of the work explores the magnetic microstructures of several macromagnetically well characterized storage media in various states of magnetization. Magnetic medium noise differs from other types of random processes in that the randomness built into the medium structure itself - the size, shape, orientation, anisotropy, magnetoelastic properties etc. of the grains or particles - are permanently established when the medium is fabricated. This part of the randomness, and the errors that it introduces into the record, could in principle be reduced or eliminated by signal processing techniques. These measurements show that a large portion of the noise, especially of the conventionally noisier media, is repeatable and potentially capable of reduction. A prerequisite for any such work is detailed knowledge of the magnetic microstructure and its relation to the recording process.

A new series of experiments has been initiated to study the interaction between microstructural features and the reproducibility of written transitions. We have investigated transition jitter, its spatial and track width dependence. As indicated from our other work on deterministic medium noise, we observe significant spatial dependence of transition jitter: a transition written repeatedly in the same place on the medium's surface has a determinable shift from its intended position. This shift varies from spot to spot on the medium. As the trackwidth decreases, the amount of jitter increases. This is predicted from theory utilizing random, fixed magnetic microstructure. Fortunately, these determinable shifts can, in principle, be corrected.

These medium studies are continuing and have implications for further work. We are also carrying out corroborative work on the physical microstructure using TEM and Lorentz microscopy, SEM and MFM.

4. Physically-based information science for magnetic recording

An initiative that widens the scope of the research is the development of a physically based information science of magnetic recording. Instead of the conventional design approach to magnetic recording systems, which involves statistically characterizing the distribution of the signal response of the recording channel, we are connecting results from the investigations of the magnetic microstructure with magnetic recording models to determine the storage capacity of the magnetic system. This is a novel approach that is potentially applicable to other information storage systems.

A first task has been to develop the methods and analyses for determining the information capacity of the recording medium using the medium's magnetic character. (This involves modeling of the medium as a set of random magnetic regions, which are allowed to interact with one another, and characterizing the possible stable states.) We have developed an analysis of capacity bounds that enables us to help in the design of recording media. Media with differing materials parameters may be equivalent from the viewpoint of storage capacity bounds, offering material property choices that may be preferable for ease of fabrication, durability, or cost.

Using the results of our research on magnetic media that have shown that the magnetic microstructure is quite repeatable (deterministic), we are working toward real time adaptive signal processing that corrects for these non-random fluctuations. Our work has demonstrated algorithms that yield as much as 5dB improvement in SNR. Additional work remains to create better algorithms and corresponding devices that make these concepts practical. Another application of these phenomena has been in medium authentication. The deterministic material randomness, through the development of Magneprint , is used to detect authenticity and is available for such applications as bank cards and checks.

We expect new ways to process and store information to emerge through this novel and unconstrained approach. This may include processing information in a non-binary fashion (m-ary or analog) and writing and reading of information that need not be tied to circular, longitudinal, or angled tracks and may not be written or read with conventional transducers.

5. Novel storage systems

Our newest and perhaps most exciting system we are developing is an associative searching of massive databases. We have applied our knowledge of the recording process and extensive prior work on various signal processing to massive databases. Today, databases hold seemingly boundless information such as that from millions of web pages, thousands of employee records, and DNA genome sequencing. Searching through billions of bytes of information requires building massive reverse databases (which is difficult unless the database is static and has a small dictionary) or systematically searching the database by shifting blocks into and out of main memory while the CPU acts on it (using valuable computing and data transfer resources, i.e., compute cycles and bus bandwidth). In principle, we utilize the speed internal to a storage system bypassing the need to access the data bus, and employ readily available fast, inexpensive data processing electronics directly in line with the storage system read channel. With this technique, using inexpensive storage devices, orders of magnitude improvement in searching performance have been approximated. We are currently designing prototype systems that use these techniques.

Summary

At Washington University we have established, besides the research, a program in education as a key ingredient of our work. A magnetic recording course, a materials properties course, a course on sensors and microactuators have been developed, and new courses are being developed and will be taught in upcoming academic years. Interaction with local industry through seminars and with professional societies is increasing. Indeck, Krchnavek, Muller and O'Sullivan have been continuously active at the local and international levels of the IEEE. Our work has created a principal nucleus in the exciting field of information storage technology for the central United State.

Last modified 25 June 2004