[Wien] NON-ZERO forces for zinc-blend GaAs

[Laurence Marks]

On 3/12/07, Oleg Rubel <oleg.rubel at physik.uni-marburg.de> wrote:
> Thank you Stefaan,
> thank you Laurence,
>
>
> > But once more, a force of 0.0x is zero in practice.
>
> Oh yes, but for orthorhombic (primitive) structures forces are exactly
> zero for ALL SCF iterations. Interesting: the first SCF cycle right after
> dstart does give zero forces!?

Almost certainly because the forces are zero by symmetry in this case.
They are presumably not zero by symmetry.

>
> > Also, -ec 0.00005 is not a very high convergence; -ec 0.00001 is
> > better and you should also use -cc
>
> What would a "good" limit for -cc? May I set several convergence criteria
> simultaneously, e.g. 'run_lapw -ec 0.00001 -fc 0.001 -cc ?????' ?

I generally use something like -ec 0.00001 -fc 1 -cc 0.0001

>
> I experienced problem with convergency when I go up to 20-40 atoms per
> supercell: the more atoms the more difficult to converge. But, for some
> reasons, it does not happen for 'P'-lattice. Is it a general property of
> 'R'-lattices?

Maybe a property of some of the transformations within the code. Make
the PW scaling small (0.01), this helps for reasons that I only
partially understand. Leave the CLM scaling alone (it's redundant) and
use -in1new once you have partial convergence, e.g.
run_lapw -ec 0.05 -cc 0.05
rm case.*bro*
run_lapw -ec 0.00001 -fc 1 -cc 0.0001 -in1new 1

>
>
> Thank you once again.
>
> Yours,
>
> Oleg Rubel
>
>
> On Mon, 12 Mar 2007, Stefaan Cottenier wrote:
>
> >
> > >I exchanged TOT by FOR in all *.in2* files
> > >and run 'run_lapw -ec 0.00005' again. Here is the result:
> > >
> > >marc-hn:~/wien_work/GaAs111APB/4atoms/noAPB> grep :FOR *.scf
> > >:FOR001:   1.ATOM      0.072      0.000      0.000     -0.072
> > >
> > >The problem is not gone
> > >
> > >
> > Only now I notice that you did use -fc, as in your previous post you see
> > the value for e.g. :FOR001 changing in the last iteration (thought all
> > these lines were for one structure file, sorry):
> >
> > >:FOR001:   1.ATOM      0.699      0.000      0.000      0.699
> > >
> > >:FOR001:   1.ATOM      0.603      0.000      0.000      0.603
> > >
> > >:FOR001:   1.ATOM      0.021      0.000      0.000      0.021
> > >
> > Two things about that:
> >
> > 1) Such a small force (this is mRy/au, not meV/Ang !) is essentially zero.
> >
> > 2) The difference betwee the final-minus-1 and final-minus-2 iterations
> > is at the 0.1 level. That means that alos your final force is converged
> > only upon that level (your new 0.072 value when you add one iteration
> > confirms that). At that level of convergence, even the exact 0.000 lies
> > within the numerical spread.
> >
> > 3) If you converge further, you will probably still find a force at the
> > 0.0x level (or perhaps a bit smaller). Increasing basis set size (up to
> > even 9.0) would be needed to bring that one closer to zero. But once
> > more, a force of 0.0x is zero in practice.
> >
> > Stefaan
> >
> >
> > Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
> >
> > _______________________________________________
> > Wien mailing list
> > Wien at zeus.theochem.tuwien.ac.at
> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> >
> >
>
>
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>


-- 
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60208, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
Web: www.numis.northwestern.edu
EMM2007 http://ns.crys.ras.ru/EMMM07/