[Wien] Basic question regarding mixing parameter & force optimization

[Laurence Marks]

Some mixer and optimization comments, to clarify.

99.999% of cases where there are convergence problems are due to:

1) Inappropriate physics. A bad model does not have to have a fixed-point
solution, that is there may be no converged solution. (Nothing says there
has to be.)

2) Inappropriate numerics. For instance, a functional that jumps around
very noisily may not be smooth. I know this can happen for 4f electrons
using -eece.

3) Discontinuity at the solution. There is such a discontinuity (electronic
phase transition) for systems such as isolated Fe atoms without spin, T=0.

4) User error. I recently ran a case with 16 symmetry elements which only
should have had 8. While it came close, it would not converge.

Sometimes the pseudo-forces are quite bad with large density residues, so
MSR1a can be slow. There can also be issues if the electron-phonon coupling
is large.

PORT is sometimes better, particularly with repetitive calculations such as
lattice parameter optimizations. However, it is very sensitive to
inconsistent forces and energies.

In my opinion NEWT & NEWT1 should be removed. My opinion.

Always look first at how the model is setup.

--
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu
"Research is to see what everybody else has seen, and to think what
nobody else has thought" Albert Szent-Györgyi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20211006/f57413a7/attachment.htm>