Computational Studies of Magnetic Materials

Author: James W. Davenport (Brookhaven National Laboratory, Computational Science Center)

This work was performed in collaboration with M. McGuigan, Kab-Seok Kang, J. Glimm, and D. Keyes. Work at Brookhaven National Laboratory supported by the US Department of Energy under contract DE-AC02-98CH10886.

Abstract

Advanced magnetic materials are critical components for electronic data storage, for energy efficiency in power distribution systems, electric motors, and novel refrigerators, and for possible spintronic devices. Yet there is no fundamental theory which describes all of the important length scales from atomic, to nanometer, to macroscopic devices. High end computing promises to change this, enabling simulations ranging from the quantum mechanics of moment formation through the structure of magnetic domains, to relevant macroscopic properties. In this talk I will describe simulations of magnetic systems which include the basics of moment formation at zero temperature as well as the temperature dependence of nanoscale magnetic particles. Calculations of the magnetic moments of metals and alloys can be done from first principles using density functional theory (DFT). These require the self consistent solution of coupled Schrodinger and Poisson equations. These are normally performed using finite basis sets and are generally limited to 100's of atoms and to zero temperature. To model larger systems, the various magnetic states can be fitted to a Heisenberg model for which the first principles result provides the interatomic coupling. Examples of the temperature dependence of nanoscale Heisenberg clusters ranging up to 1 million atoms will be given. Finally I describe some of the features of modern petascale computers such as BlueGene which enable simulations of this type to be performed on samples of unprecedented size.