[QE-developers] OpenMix

[Laurence Marks]

I would defend the perhaps bold statement that the mixer in Wien2k is state
of the art. It has sophisticated implicit and explicit trust region
controls (not all details published yet), so needs essentially no user
input. It is also fast, and the algorithm which does simultaneous density
and atomic optimization is powerful.

That said, it is also complicated. Rebuilding it from scratch for another
DFT code would be a significant task.

In addition, maintaining it in Wien2k would not be trivial if I stopped
work. While I am not retiring tomorrow, I also won’t keep working into my
old age.

What to do?

One thought is to create OpenMix. This would be an open source (e.g.
GitHub) code which includes the trust region controls of the mixer in
Wien2k with the existing MSEC, MSR1 algorithms as well as others such as
DIIS. Build a core with hooks so any and every DFT code can directly use
the algorithms via either reverse or forward communication.

I am pondering this, and would be interested to know if you think it would
be a good idea and would be supportive.

-- 
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu
Corrosion in 4D: www.numis.northwestern.edu/MURI
Co-Editor, Acta Cryst A
"Research is to see what everybody else has seen, and to think what nobody
else has thought"
Albert Szent-Gyorgi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20190920/f63052e1/attachment.html>