FOR IMMEDIATE RELEASE CONTACT: December 2, 1998 Michael Schneider Pittsburgh Supercomputing Center 412-268-4960 firstname.lastname@example.org
PITTSBURGH "This method is a key for understanding magnetism at the theoretical level," says Pittsburgh Supercomputing Center (PSC) scientist Yang Wang, "and it has application in many areas, including read-head technology for disk storage media, and bulk magnets used in power generation and the sensors in cars." Wang is referring to the locally self-consistent multiple scattering (LSMS) method, software for simulating magnetism that earlier this month became the first research application to break the teraflop barrier in supercomputing.
On Nov. 9, running on a 1,480 processor CRAY T3E system, LSMS achieved sustained performance of 1.02 Teraflops (trillions of calculations per second). For their work on this project, a team of scientists from Oak Ridge National Lab, the National Energy Research Scientific Computing Center, University of Bristol (UK) and PSC won the 1998 Gordon Bell Prize, given for best achievement in high-performance computing.
The team leader, Malcolm Stocks of Oak Ridge, notes that Wang's contribution was critical to the teraflop feat: "He took the code to Pittsburgh and made it run on the T3E."
From 1993 to 1996, Wang worked with Stocks as a post-doctoral fellow at Oak Ridge where they developed LSMS on Oak Ridge's Intel Paragon system. When Wang came to Pittsburgh in 1996, he worked together with Oak Ridge scientist Bill Shelton to translate LSMS to run on PSC's CRAY T3D, predecessor to the T3E. The greater raw computing power of the T3E's processors ultimately made it possible to crack the teraflop barrier.
LSMS, say Stocks and Wang, provides a method to realistically simulate the microscopic, atomic-level details of magnetism. Although magnetic materials are ubiquitous in consumer electronics and computing hardware, there are significant gaps in understanding how magnetism works. "If you could even slightly increase the maximum energy product of a permanent magnet," says Stocks, "it would have enormous effect on technology."
To solve many important practical problems, it's necessary to simulate magnetic metals with "unit cells" the atomic size of the system from hundreds to thousands of atoms. This is beyond the range of most existing computational methods. Like other methods, LSMS relies on density-functional theory (for which physicist Walter Kohn won the 1998 Nobel Prize in chemistry) to describe the interactions between electrons and atoms in magnetic materials, but unlike most other methods it uses algorithms that permit very large systems to be addressed on today's supercomputers.
The Nov. 9 teraflop calculation simulated a 1,458 atom unit cell of iron. "In size this is not much smaller than nano-particles being made today," says Stocks. LSMS exploits the power of massively parallel systems like the T3E by assigning each atom of the unit cell to a separate computer processor.
Wang is currently working to refine LSMS by eliminating an approximation built into the method. In future calculations, Stocks, Wang and collaborators plan to address a number of problems in magnetic multi-layers and magnetic nano-particles.
This research is sponsored by the U.S. Department of Energy's Office of Basic Energy Sciences, Division of Materials Science and DOE's Office of Computational and Technology Research, Mathematics, Information and Computational Sciences.
The Pittsburgh Supercomputing Center is a joint effort of Carnegie Mellon University and the University of Pittsburgh together with Westinghouse Electric Company. It was established in 1986 and is supported by several federal agencies, the Commonwealth of Pennsylvania and private industry.