[Wien] Mini NEWT and/or PORT - what precision can one expect?

[Peter Blaha]

> > I thought this would be straight-forward by changing to "NEWT 0.1" or
> > "PORT 0.1" in case.inM. However, in NEWT I end up with the
> > "stp.le.tolx", although there is plenty of space between the MT's, and
> > in PORT I run into the "curvature condition failed".
> 
> Probably too obvious, but did you check whether your forces are absolutely
> converged within 0.1 mRy/au? (wrt basis set and k-mesh) Otherwise you get
> 'random' values (at that accuracy level), and any minimizer will find no
> conclusion which can possibly lead to error messages somehow.

And even more important: Did you convege the scf cycle to such high
precisions (in fact, even to higher precision!) ?
run_lapw -fc 1   means that forces may have errors in that range. E.g. when
calculating phonons I need forces converged to 0.05 to 0.01 mRy/bohr.
Of course this takes some more iterations, but without that you cannot
relax positions to such a precision.

                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------