bug in cp_fm_invert ?

flo fsch... at pci.uzh.ch
Fri Nov 26 10:14:20 CET 2010


Hi Hanning,

commit at least the fix for the permutations. I will try to have a
look into the determinants soon.
Just in case you are going to use the method I have some remarks. From
discussions with other people, it turned out, that the method has
problems with strongly coupled systems. Even constraining the electron
density to given charges, does not perfectly prevent it from
delocalizing a bit. As the method seems to be very sensitive to such
tiny changes, you will need very well converged wavefunctions.
Futhermore it might be good to use hybrid functionals, as they
localize the density a bit more and might be easier to constrain to a
diabatic state.

I will let you know as soon as I fixed the cp_fm_invert.

Flo


More information about the CP2K-user mailing list