Density matrix purification: computation of large eigenspaces in O(n)

Emanuel Rubensson
Department of Theoretical Chemistry
KTH
Stockholm


Abstract:

In electronic structure calculations such as Hartree-Fock or Kohn-Sham Density Functional calculations, an important operation is the computation of the density matrix for a given potential; the density matrix is used to represent the electron density. The traditional approach to compute the density matrix uses diagonalization of the potential matrix (the Fock/Kohn-Sham matrix). The computational cost of this operation increases cubically with system size. In recent years, methods for which the needed computational effort increases only linearly with system size have become popular. In this seminar I will talk about one of these methods: density matrix purification. I will focus on how errors due to computational approximations can be controlled.

See the article in J. Chem. Phys. 128, 074106 (2008).