Nanomagnetism Beyond the PetaFLOP/s Barrier

Using a hybrid code combining statistical mechanics with first principles calculations, we were able to be one of the first two groups to sustain a performance of more than one PetaFLOP/s in a scientific calculation. "The Search for Stable Storage". The code exploits multiple levels of parallelism and shows superb scaling behavior and performance on massively parallel computers, such as the new Cary XT5 Jaguar system at the National Center for Computational Sciences at ORNL.

The code (gWL-LSMS) will allow the investigation of the thermal behavior of material properties using first principles density functional theory (DFT) calculations without the need to resort to model parameters. While DFT has proven to be a useful tool for the study of the ground state of many materials, for finite temperatures the situation is less favorable, as it is often necessary to rely on models with parameters either fitted to first principles or experimental results. This is especially unsatisfactory in inhomogeneous systems, nano-particles, or other systems where the model parameters should vary significantly from one site to another. To overcome this problem we have combined our first principles locally-selfconsistent multiple scattering code (LSMS) with the classical thermodynamics Wang-Landau (WL) Monte-Carlo method. Given the success of LSMS in dealing with magnetic systems, gWL-LSMS will be used to investigate the magnetic properties of nano-particles at finite temperatures, which is of significant importance for understanding the stability of information stored on magnetic recording devices.

This work is performed by Markus Eisenbach (NCCS), Don Nicholson (CSM), and Thomas Schulthess (ETH Zürich and CSCS).