Abstract
Oak Ridge National Laboratory has embarked on several computational i r m d Challenges, which require the close cooperation of physicists, mathematicians, and computer scientists. One of these projects is the determination of the material properties of alloys from first principles and, in particular, the electronic structure of high-temperature superconductors. The physical basis for high Tc superconductivity is not well understood. The design of materials with higher critical temperatures and the ability to carry higher current densities can be greatly facilitated by the modeling and detailed study of the electronic structure of existing superconductors. While the present focus of the project is on superconductivity, the approach is general enough to permit study of other properties of metallic alloys such as strength and magnetic properties. This paper describes the progress to date on this project. We include a description of a self-consistent KKR-CPA method, parallelization of the model, and the incorporation of a dynamic load balancing scheme into the algorithm. We also describe the development and performance of a consolidated KKR-CPA code capable of running on CRAYs, workstations, and several parallel computers without source code modification. Performance of this code on the Intel iPSC/860 is also compared to a CMY 2, CRAY YMP, and several workstations. The code runs at over 1.6 Gflops on a 128 processor iPSC/860. Finally, some density of state calculations of two perovskite superconductors are given.
Original language | English |
---|---|
Title of host publication | Applications |
Editors | David W. Walker, Quentin F. Stout |
Publisher | Institute of Electrical and Electronics Engineers Inc. |
Pages | 504-512 |
Number of pages | 9 |
ISBN (Electronic) | 0818621133, 9780818621130 |
DOIs | |
State | Published - 1990 |
Event | 5th Distributed Memory Computing Conference, DMCC 1990 - Charleston, United States Duration: Apr 8 1990 → Apr 12 1990 |
Publication series
Name | Proceedings of the 5th Distributed Memory Computing Conference, DMCC 1990 |
---|---|
Volume | 1 |
Conference
Conference | 5th Distributed Memory Computing Conference, DMCC 1990 |
---|---|
Country/Territory | United States |
City | Charleston |
Period | 04/8/90 → 04/12/90 |
Funding
*This research was supported by the Applied Mathematical Sciences Research Program, Office of Energy Research, U.S. Department of Energy, under contract DEAC05-840R21400 with Martin Marietta Energy Systems, Inc.