The ABINIT Users Mailing List ( CLOSED )

[Dr MJ Verstraete <mjv500@york.ac.uk>]

You can also try the wavelet poisson solver with 0 boundary conditions (to 
reduce your cell size and make sure you have no interaction with 
neighbors).

or (and? don't know if they work together but they should) you can try 
PAW.

I would also expect that the results would converge when you crank up the 
ecut, and even that they would converge fairly smoothly, but with d metals 
and vacuum states, you can get to certain plateaux of convergence then the 
PW resolve the next level of detail in localized states or whatever and 
your convergence continues... In the end it should work. Get more RAM? or 
do band parallelism on several nodes of a cluster. If your system is tough 
there is no shame in using a parallel job for 2 atoms.

To save a bit of memory, use a FCC primitive unit cell. The volume and PW 
are less and the minimum distance between images is the same (12 NN images 
instead of 6 for a cubic box).

Finally, for your Erxn, are you sure the atomic energies are converged? 
Otherwise you may not be going to the right limit.

"So many tricks, so little time..."

Matthieu

================================================================
Dr. Matthieu Verstraete                 mailto:mjv500@york.ac.uk
Dept. of Physics, University of York,     tel: +44 1904 43 22 08
Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14