The ABINIT Users Mailing List ( CLOSED )

[Matthieu Verstraete <mjv500@york.ac.uk>]

Hello Fer,

for each qpoint you need an explicit WFQ file containing the k+q eigenstates 
and eigenvalues. With your 4x4x4 grid, the qpoint was a differnce of 2 kpoints, 
so you could use the original GS file as a WFQ file: ABINIT tries to do the 
juggling of the kpoints from the GS file if no getwfq is specified.

With 6x6x6, 1/4 0 0 is no longer a difference of 2 kpoints in the GS grid. You 
must do the non scf calculation at k+(1/4 0 0) explicitly, as you do for DS2 
(where it is actually useless since q=0,0,0 is always a difference of two 
kpoints). Actually, Umesh Wagmare told us at the abinit school this summer that 
it is not very good practice at all to use qpoints not contained in the kpoint 
mesh, and that systematic errors can be introduced.

At a guess, I would say you'll need a lot more kpoints than 6x6x6 to converge, 
so you might as well use a multiple of 4 (to contain all the qpoints you use) 
and increase it further to 8x8x8, and probably even 12x12x12 or 16x16x16.

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