[Pw_forum] something about Pt.pw91-n-van.UPF

[Nicola Marzari]

Gabriele Sclauzero wrote:

> In order to get a better estimate of the isolated atom energy, I suggest to do a spin 
> polarized calculation (i.e. nspin=2) either using a very small smearing (and check 
> converge with respect to degauss), or fixing occupations (specify occupations='from_input' 
> in the electrons namelist and add the input card OCCUPATIONS after the namelists).
> You can also fix the total magnetization (=2.0) or, equivalently, the spin multiplicity 
> (=3)instead of occupations.



A note of caution - if the atom ends up with fractional occupations
on equivalent p or d levels, the code will print out in the
total energy of the system also the entropic contribution (i.e.
it will write E-TS), where S is not zero. (That's why I actually
like to call the energy written by PWSCF total free energy, rather than
total energy). Looking at the total free energy is the right thing
to do in a metal, but not in an atom. Of course, even if S
is large (because occupations are not 0 or 1), in the limit of T
going to zero -TS goes to zero, but it would be more appropriate to
actually look at E (taking the total (free) energy that PWSCF writes,
i.e. E-TS, and subtracting to that -TS, that the code also writes,
in order to get E).

Checking that as a function of smearing [(E-TS)-(-TS)] is
constant (this is the difference between the two quantities
written by PWSCF in output), while the total (free) energy
E-TS is linear, would be a good  consistency check.


			nicola

---------------------------------------------------------------------
Prof Nicola Marzari   Department of Materials Science and Engineering
13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu