Free energy from CP2K calculations

[Ashley Dickson]

Hi all, 

I'm currently working on fitting a machine learned potential using DFT data from CP2K, however I am running into some confusion regarding the energies of the simulations. I am using GAP_fit (https://libatoms.github.io/GAP/gap_fit.html), which mentions the importance of fitting the potential to the free energy, in order that the forces are consistent. Am I understanding correctly that the Total energy printed by CP2K is equivalent to the free energy, and includes the electronic entropic energy? 

For instance, the isolated atom energies are required by GAP_fit to fit the binding energies of the potential, so I have performed a SPE calculation on oxygen to give the following output:

Overlap energy of the core charge distribution:               0.00000000000000
  Self energy of the core charge distribution:                -41.52620707899887
  Core Hamiltonian energy:                                     11.40342642720691
  Hartree energy:                                              17.59013700141358
  Exchange-correlation energy:                                 -3.24820705673776
  Electronic entropic energy:                                  -0.02418867401250
  Fermi energy:                                                -0.32715630556588

  Total energy:                                               -15.80503938112864

Would I take -15.80503938112864 as the isolated atom energy here?

Thanks in advance,

Ashley

[Patrick J. Taylor]

Hi Ashley,

Yes, the "Total energy:" includes the electronic entropy and can therefore be considered a free energy in the sense that you are looking for (see also towards the end of this tutorial).

For reference, you should also see a line beginning "ENERGY| Total FORCE_EVAL (QS) ..." elsewhere in your output file, which is the energy extrapolated to sigma = 0 (i.e. the "total energy" that one is usually interested in).

Thanks,

Patrick