[CPMD-list] slow convergence with ODIIS

[Christian Tuma]

Hello,

these wavefunction convergence problems with Vanderbilt-PPs seem to be
quite common. Chapter 11.1 of the manual mentions them, there have been
postings in the mailing-list, and Axel Kohlmeyer just gave some hints
again. However, my own experience is that even when the geometry of the
system is well optimised, one still should not believe that the wavefunction
always converges properly when using these keywords:

INITIALIZE WAVEFUNCTION RANDOM
PCG MINIMIZE
TIMESTEP
 20

I was running calculations on zeolite systems (using Vanderbilt-PPs for
hydrogen, oxygen, and carbon) and encountered situations where the
final result of a wavefunction optimisation using these keywords depends
on the chosen TIMESTEP value, e.g. (using the BP86 functional):

 TIMESTEP     result
 3            -2613.5011 Hartree
 5            -2613.4979
 8            -2613.5011
 9            -2613.4979
 10           -2613.5011
 11           -2613.4951
 12           -2613.4979
 15           -2613.4979
 20           (no convergence, oscillation)

This corresponds to relative energy differences of several kJ/mol, and
e.g. for some questions of chemical interest this cannot be neglected.

However, testing several values for TIMESTEP is no guarantee to find the
"correct" wavefunction and so the convergence problem still persists.
Axel, you suggested to put some initial ATOMIC CHARGES. I would like
to try this and to see whether this improves the situation, but, I
have no clear idea which values to choose. When doing a properties
calculation for atomic charges I get a set of charges derived from
real space integration and another one from the electrostatic potential,
the absolute values from the latter being about one order of magnitude bigger
than the ones from real space integration. Which of these two sets should I
use as an orientation for the ATOMIC CHARGES to follow your suggestion?

Does anybody have another new idea how to make life easier when trying
to converge the wavefunction using Vanderbilt-PPs?

cheers,
Christian.

PS: I agree to Atte, PBE calculations seem to converge easier than e.g.
BP86 calculations, at least when using normconserving PPs.

-- 
Christian Tuma             Humboldt-Universitaet Berlin
ct at chemie.hu-berlin.de     Arbeitsgruppe Quantenchemie (Prof. Sauer)
phone: +49-30-20937140     Brook-Taylor-Str. 2, 12489 Berlin, GERMANY
fax:   +49-30-20937136     http://www.chemie.hu-berlin.de/ag_sauer