[CPMD-list] Trying to understand screened exchange B3LYP for PBC solids

[Axel Kohlmeyer]

On Thu, 15 Mar 2007 driver at mps.ohio-state.edu wrote:

KD> Hi Axel,

hi kevin,


KD> Thanks for your initial response.
KD> 
KD> I haven't changed the code in any way. The B3LYP calculation gives a
KD> quite different total energy than a plain BLYP calculation. From what
KD> I've seen, the code at least gives the impression that it is doing
KD> Hartree Fock exchange for k-points. And, the B3LYP energy difference I
KD> get matches an all-electron B3LYP calculation I've done previously in
KD> another code.

interesting.

KD> It is puzzeling though, that in hfx.F there is the line of code
KD>       IF(TKPNT) CALL STOPGM('HFX','K-POINTS NOT IMPLEMENTED')
KD> But, the code doesn't stop, which could indicate several things I suppose.

for starters, that the HFX routine is not called from the corresponding
'driver'. having all functionally implemented consistently across all
various 'drivers' is a big challenge. many people implement their new
features only for the 'driver' they need or are familiar with, which is
understandable, but somewhat problematic for 'only user'-users.

[...]

KD> So is the code actually doing Hartree Fock exchange for k-points? And,
KD> if it's not actually running screened B3LYP, then what is it running?
KD> Any suggestions on how to find the answers to these questions?

that would require some deeper digging into the sourcecode. right now,
i'm not in the condition to look into it (just survived a two week 
HPC workshop with >130 participants alltogether). i'm curious as well, 
however the k-point support has been lagging behind the gamma-point 
branch quite a bit over the last 5 or so years. i suspect, it will
not be doable without 'getting your hands dirty'. please get back to
me personally in about a week, when i'm back to normal life, and i'll
help you figure out what is going on. would you mind putting together 
a few input/test/reference file, so i don't have to come up with them
myself? thanks.

KD> I know a lot of people that would be excited to run the code if it is
KD> actually doing hybrid functionals with k-points.

i don't doubt that. in return, i also know a number rather overloaded 
people that would be glad, if somebody would look into fixing it for
them. ;-)

cheers,
    axel.

KD> 
KD> Thanks for the feedback!
KD> 
KD> Kevin Driver
KD> The Ohio State University, Physics Dept.
KD> driver at mps.ohio-state.edu
KD> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.