Modelling the density response of materials for density mixing in DFT

SpeakerEd Higgins
Time4pm on 12th November 2016

Over the past few decades, density functional theory has become a popular tool for performing accurate quantum mechanical simulations of materials in a reasonable amount of time. This is often done using the self-consistent field (SCF) method in which a set of equations, first proposed by Kohn and Sham[1], are iterated until convergence. While this method has had many successes, it is not guaranteed to be stable. In particular large, anisotropic and/or magnetic systems can often be unstable using current methods.

In order to improve the stability of the SCF method, the input and output charge densities from each iteration are mixed to find an improved input density for the next iteration. The ideal mixing of these can be shown to use the dielectric response matrix. This is usually done using an approximate model, first proposed by Kerker[2]. Similarly, spin densities are mixed using the same model dielectric.

This work looks at the effect of an improved dielectric response model applied to charge density mixing, and proposes a new magnetic response model for spin density mixing. As a result, the number of SCF iterations required is reduced across a wide range of systems, with negligible computational overhead. This enables faster and more robust simulation of such materials.


  1. Kohn W. and Sham L.J. "Self-Consistent Equations Including Exchange and Correlation Effects", Phys. Rev. 140 (1965)
  2. Kerker G.P. "Efficient iteration scheme for self-consistent pseudopotential calculations", Phys. Rev. B 23 (1981)