COMPUTING CLUSTER FOR THE DEVELOPMENT OF FAST AND SCALABLE ALGORITHMS FOR LARGE SCALE REAL-SPACE ELECTRONIC STRUCTURE CALCULATIONS
Abstract
We propose to acquire a cluster comprising of 672 core Intel Xeon processors with 6TB RAM and Infiniband interconnections to support the development of fast, error-controllable, and scalable algorithms for large-scale real-space electronic structure calculations to understand and accurately predict a wide range of materials properties and phenomena relevant to the Army Research Office and the Department of Defense. The DoD-funded research areas to be supported by the acquisition include the development of fast and scalable algorithms for time dependent density functional theory calculations, and the development of computational methods for fast electronic structure calculations using tensor-structured techniques. The proposed acquisition will also facilitate the development of computational methods for real-space all-electron electronic structure calculations, as well as the development of a computational framework for large-scale materials simulations with quantum accuracy, which will tackle two significant approximations in electronic structure calculations, i.e., the pseudopotential approximation and approximations in modeling electron-electron correlations. Finally, the proposed acquisition will assist in technology transfer by allowing codes to be tested on large-scale and applied problems of direct interest to potential ARO and DoD users. Graduate and postdoctoral students in mechanical engineering, materials science and engineering, and applied mathematics will be trained in the development and use of the above computational algorithms for materials simulations, and on the use of distributed memory machines for research in computational materials physics.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 14, 2019
- Source ID
- W911NF1810242
Entities
People
- Vikram Gavini
Organizations
- Army Contracting Command
- United States Army
- University of Michigan