Time: 4:30 p.m.
Place: UHW 11/1102
Andrew Horsfield, Imperial College London, London, United Kingdom
Achieving both spin and charge self-consistency in magnetic tight binding using conventional techniques (such as linear and Pulay mixing) can be either unstable or very slow.
By formulating the problem as a search for stationary points of a Harris-Foulkes functional, extended to include spin, a more efficient scheme based on a Newton-Raphson procedure can be developed. However, there are fundamental problems with achieving convergence with respect to spin, and these will be discussed.
A simplification of the algorithm to produce a preconditioner for Pulay mixing will be presented.