[Pw_forum] SCF convergence - proposal

[Nicola Marzari]

> 
> Anyway: your suggestion is most welcome. An algorithm that
> converges no matter what would be highly desirable.

Well, just to make the point - we developed one years ago
(ensemble-DFT, Marzari, Vanderbilt, and Payne, PRL 79 1337 (1997)).

This is a truly variational formulation of the problem on all
degrees of freedom (wavefunction, orbitals, and subspace rotations),
so that the total (free) energy is bound to get closer to
self-consistency at every time step.

We have used it happily everafter (in our own code, that is
a Castep derivative), and a few other groups have implemented
it (Francois Gygi in Jeep, Sohrab Ismail-Beigi in DFT++). Another
implementation is in PARATEC, and one is in some stage of completion
(I believe) for the Accelrys commercial copy of Castep.

The CP code linked to PWSCF/Democritos will soon have it, albeit
with Car-Parrinello evolution on the orbitals (this complicates
matters, and decreases efficiency). Putting such an algorithm in
PWSCF is not out of the question, but requires someone doing
dedicated work.

Also, for systems where charge-density mixing converges smoothly,
getting to selfconsistency from scratch tends to be faster.

The strengths of ensemble-DFT are in the ab-initio molecular
dynamics, and in its formal guarantee to converge "no matter
what"...

All the best,

			nicola

---------------------------------------------------------------------
Prof Nicola Marzari   Department of Materials Science and Engineering
13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
tel 617.4522758  fax 617.2586534  marzari at mit.edu  http://nnn.mit.edu