Re: Package categories

[Michael Banck <mbanck@debian.org>]

On Mon, Jul 28, 2008 at 07:20:54PM -0430, Muammar El Khatib wrote:
> Are you sure about this? I mean as far as I have studied, DFT is not considered
> as Ab Initio (some people say yes, others no). Maybe calculation quality is as
> good as Ab Initio (DFT takes care about electronic correlation) but DFT starts
> from the premise that the electronic density is taken as the basic quantity. So
> the electronic density is the function taken for the functional (in this case
> the energy) while Ab Initio (like Hartree - Fock for instance) are based on the
> many-electron wavefunctions. On the other hand DFT uses parameters derived from
> _empirical data_, or from more complex calculations and in that moment is when
> DFT does not follow the "calculations from first principles" (where no empirical
> data is used).

While you certainly have valid points concerning the classification of
DFT vs. other methods from a scientific POV, what matters at hand is the
classifaction for the user who wants to use Debian to do stuff.

And at this level, DFT (at least talking from a computational chemistry
point of view here) is identical to other ab-inito methods like MP2 or
CCSD.  Besides, how many empirical parameters you have depends on the
particular exchange-correlation functional you use, in theory I think
one can construct those functionals without using empirical parameters
(see e.g. the PBE functional, Phys. Rev. Lett., 77, 3865).  The very
popular B3LYP functional indeed has a couple of them, but stil in this
case you are doing an all-electron calculation if you choose so.

The devide here is between ab-initio and semi-empirical - where the
latter has a whole different dimension of empirical parameters it uses,
and this is the useful destinction for users.


cheers,

Michael