[Pw_forum] Questions regarding EXX implementation

Tyler Grassman tgrassma at ucsd.edu
Wed Oct 3 21:00:17 CEST 2007


Hello all.  I have some questions regarding the hybrid functionals
implementation in PWscf.  First of all, how "experimental" is it, exactly?
The "DO NOT USE IT FOR ANY PRODUCTION RUN" warning is definitely strong, but
I've been doing some of my own testing with it (PBE0 for Ge, bulk and slab,
so far), and the results look pretty good, definitely in-line with other
EXX-based results in the literature; the calculated lattice paramter is
quite good, and there is a band gap (slightly overestimated, but according
to the literature, this is normal for Ge).  Is there anything in particular
about the implementation that is suspect and should be specifically watched
out for?

Also, does the implementation still only support NC-PPs?  The last mention
of this I was able to find in the forums is from 2005, and the README file
in the EXX_example also says this, so I wondering if this was still the
case, or if perhaps this had changed in a more recent (perhaps CVS) version.
The calculations are indeed extremely slow compared to "normal" SCF runs;
fortunatunately, the R-and-G parallelization seems to be quite efficient,
even for the EXX stuff, so the scaling is good (at least up to 32 CPUs),
which helps.  But, with US-PPs, the lower cutoff energy would certainly help
a good deal, as well.

If anyone has some experience with the hybrid functional implementation here
in PWscf and has found any sort of parameters that really help improve the
efficiency of these jobs, I'd love to hear about it.  Thank you very much.

Regards,

Tyler Grassman
Dept. of Chemistry/Biochemistry, Dept. of Materials Science & Engineering
University of California, San Diego




More information about the users mailing list