The ABINIT Users Mailing List ( CLOSED )

[Steven Homolya <steven.homolya@spme.monash.edu.au>]

Josef W. Zwanziger wrote:

> Dear Colleagues,
>
> I'm a little confused about what ecutsm does. The explanation in the
> help file seems clear enough, but I find the following behavior.
> Starting from an experimentally determined crystal structure, I optimize
> the atom positions and unit cell dimensions with
>
> ionmov 2
> ntime 200
> tolmxf 2.0D-5
> toldff 2.0D-6
>
> optcell 2
> dilatmx 1.1
> ecutsm 0.5
>
> This procedure seems to work fine--the crystal structure changes
> slightly, with the result that the stress tensor calculated for the
> final structure has elements of order 10-8 H/Bohr^3. Then if I take the
> output structure (new acell values, new xred values, rprim didn't change
> of course) and compute the energy with toldff 2.0D-6, but no ecutsm, the
> computed stress tensor rises to nearly 10-5 H/Bohr^3. If I then put back
> in the ecutsm 0.5 line, the stress tensor goes back down. Can someone
> shed some light on this? I already have ecut set rather high (70 H) and
> am using 4x4x4 M-P grid with shiftk 0.5 0.5 0.5. 
>
> Thanks much for your help,
>
> Joe
>
>  

Dear Joe,

Sorry, I don't have an answer for you but I am interested in what the 
explanation might be for your curious results. As no one else replied to 
your post I'm sending this reply.

Do any other physical quantities change when you change ecutsm? In 
particular, do the forces on the atoms change at all? Would you mind 
sending me your input file (with the optimised geometry) and, if needed 
to reproduce yor results, pseudopotential files?

Cheers,

Steve