Dear Kyohn,
I finally got a chance to look into your problem, and I can reproduce
the behavior you describe. Without SO everything worked fine. With SO,
the projection clearly went awry: The spreads were too large (~16 Ų vs
~4 Ų in the non-SO case), the centers were off, and the Wannier
bandstructure was bad also. Is that what you saw?
I then tried the example on a different machine, because I had a
different Wien2k version there -- I thought maybe something about SO had
changed and broken the w2w procedure. There, I really did get good
results, but it turns out that the key difference was the compiler and
LAPACK library used for wannier90.
The problem happened when I used ‘ifort’ together with the ATLAS
library. I tried various combinations of ifort/gfortran and
MKL/ATLAS/netlib, but only ifort+ATLAS seems to have this problem.
Can you please try if a similar strategy works in your case?
Elias
_______________________________________________
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
