Too huge total energy with ghost atoms

[Seungchul Kim]

Dear developers,

To test PAOs for Si self diffusion simulations, I compared energies of two different Si-interstitial in Si crystal.
Ghost atoms were used to remove BSSE.

No problems with DZP, DZDP and so on. But problems happened when the 4f orbitals were included (so, we may call it DZPP').

There is one ghost atom, total number of Si atoms is 17.

Total energy without a ghost atom is -4227.189761 eV.
Total energy with a ghost atom is -7130.553808 eV.
So, the difference is too large.

Ekin is negative with this ghost atom.

Please check the attached file.

Thank you very much.

SIESTA 4.0 and SIESTA 4.1-b2
Intel composer XE 2013_sp1.3.174 (compiler and MKL library for scalapack, blacs, lapack and blas)
openmpi-1.8.3

compile options:

FFLAGS=-w -O3 -mp -static-intel
FPPFLAGS= -DMPI -DFC_HAVE_FLUSH -DFC_HAVE_ABORT
LDFLAGS=-Vaxlib
FFLAGS_DEBUG= -g
FFLAGS_CHECKS= -g -O0 -degug full -traceback -C

[Nick Papior (nickpapior)]

Could you try and download this:

https://bazaar.launchpad.net/~siesta-maint/siesta/rel-4.0/tarball/559

and see if the large energy difference still exists?

I can't reproduce it now.

[Nick Papior (nickpapior)]

This was fixed in 4.0r546.