Numerical methods and scalable algorithms for large-scale real-space time dependent density functional theory calculations (Numerical Analysis)
Abstract
This proposal will investigate and develop fast computational methods which may mitigate some of the computational complexity of time dependent density functional theory (TDDFT). The approach will consist of several elements of computational mathematics, adapted and applied to this problem. These include: (i) a local real-space formulation of the equations via reformulating the extended interactions using using ad joint potentials; (ii) finite-element/piecewise polynomial discretization with basis adaption informed by error analysis; (iii) adaptive coarse-graining of the temporal evolution of the wave functions informed by adaptive techniques of numerical analysis; (iv) development of subspace projection techniques that will significantly improve the efficiency and computational complexity of governing equation calculations; (v) a scalable implementation of the algorithms on massively parallel computing platforms. The use of a real-space formulation and finite-element basis is complex and not commonly attempted, but presents several key advantages.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 12, 2017
- Source ID
- W911NF1510158
Entities
People
- Vikram Gavini
Organizations
- Army Contracting Command
- United States Army
- University of Michigan