[DFTB-Plus-User] issues with diamond, free atom energy and DOS

Benjamin Hourahine benjamin.hourahine at strath.ac.uk
Thu Apr 2 00:56:08 CEST 2009

I can clarify a few of these questions:

>1. I was trying to calculate optimize geometry of diamond structure.
>Using the mio-0-1 set for the .skf files. The bulk modulus value what I got is 1.5x10^2 (while the experimental value
>is 4.42x10^2. We varied the lattice parameter manually in order to fit it to the Equation of State. Please find one of the

I would strongly suggest using the pbc set, mio is not optimized for diamond. Also in this case, SCC would not have
any effect on the energy as the non-shell resolved SCC Hamiltonian is being used (the Hubbard U values for the s and p
shells are the same anyway).

>2. In order to calculate the energy of a free C (2s^2, 2p^2) atom, I set up the spin polarisation as below
>are there any trick or flag that sets the populations /fillings properly for a free C atom i.e.,
>SPIN 1Atom Sh.   l   m       Population
>     1   1   0   0       1.00000000
>     1   2   1  -1       1.00000000
>     1   2   1   0       1.00000000
>     1   2   1   1       0.00000000

There seems to be a missing electron somewhere.

The absolute DFTB atom energy is ofset with respect to DFT as the double counting contribution is missing
for the isolated atom, but energy differences are fine (depending on the parameter validity, you may find the cohesive
energy is poor though). You can analytically calculate the non-SCC atom energy, its just the sum of the atomic
eigenvalues times their occupation, similarly the SCC part can be worked out from the Hubbard-U terms and the charge
fluctuations (see any of the review papers for this contribution). Actually, in the context of Kohn-Sham DFT, while
you can force this occupation for the isolated atom,  this isn't a ground state for the Hamiltonian; so, to be pedantic, doesn't
give a meaningful energy (and breaks various rotational symmetries, at least in a collinear picture).

Hope this helps


More information about the DFTB-Plus-User mailing list