New Algorithms for the Fermi-Löwdin Orbital Self-Interaction Correction Calculations.
ORAL
Abstract
Self--interaction error (SIE) is in most approximate exchange-correlation functionals, and removing SIE is important for improving the performance of the Kohn-Sham density-functional theory (KS-DFT) when applied to systems of chemical and physical interest. The Fermi-Löwdin Orbital Self-Interaction Correction (FLO-SIC) methodology was recently introduced as a unitarily invariant reformulation of the Perdew-Zunger SIC scheme to remove unphysical SIE from DFT.
We propose new algorithms that aim to speed up and extend the applicability of this methodology. The "two-step" algorithm was designed to reduce the number of times that orbital-dependent potentials need to be calculated in the self-consistency cycle, addressing one computational bottleneck of SIC calculations. We also introduced unified Hamiltonian formalism in FLOSIC as a way to solve the system of Schrödinger-like equations. An advantage of the unified Hamiltonian approach is that it can replace a cumbersome "Jacobi sweep" method with a well-optimized diagonalization routine. It also is a convenient formalism for allowing unoccupied states to "see" an SIC potential. We present atomic and molecular applications that show the performance of FLOSIC with these new algorithms.
We propose new algorithms that aim to speed up and extend the applicability of this methodology. The "two-step" algorithm was designed to reduce the number of times that orbital-dependent potentials need to be calculated in the self-consistency cycle, addressing one computational bottleneck of SIC calculations. We also introduced unified Hamiltonian formalism in FLOSIC as a way to solve the system of Schrödinger-like equations. An advantage of the unified Hamiltonian approach is that it can replace a cumbersome "Jacobi sweep" method with a well-optimized diagonalization routine. It also is a convenient formalism for allowing unoccupied states to "see" an SIC potential. We present atomic and molecular applications that show the performance of FLOSIC with these new algorithms.
–
Presenters
-
Kamal Sharkas
Central MIchigan University, Central Michigan Univ
Authors
-
Kamal Sharkas
Central MIchigan University, Central Michigan Univ
-
Juan E Peralta
Central MIchigan University, Central Michigan Univ
-
Koblar Jackson
Central Michigan University, Physics, Central Michigan University, Central MIchigan University, Central Michigan Univ