inconsistency VASP 6.1.2 vs. VASP 6.2 and 6.3 in output of RPA energies

[kdoblhoff]

Dear VASP developers,
Is the following change made on purpose?: In VASP 6.1.2, I got output similar to the following for an ACFDT calculation:

Code: Select all

  HF-free energy TOTEN  =        -4.53841188 eV
  exchange ACFDT corr.  =        -0.05397473  see jH, gK, PRB 81, 115126
 HF+correlation energy  =        -9.40117008
 HF+E_corr(extrapolated)=        -9.47664187
 

In VASP 6.2 and 6.3 the output has changed to the following:

Code: Select all

HF-free energy      FHF    =        -4.53841188 eV
  HF+RPA corr. energy TOTEN  =        -9.34719538 eV
  HF+E_corr(extrapolated)    =        -9.42266718 eV
  ACFDT corr.    EXHF_ACFDT  =        -0.05397473 eV (see jH, gK, PRB 81, 115126)

Importantly, the correction for the partial occupations in the XC energy are added in for the value called "HF+E_corr(extrapolated)" in VASP 6.1.2, while they are not added in for the same value for VASP 6.2 and 6.3 (according to the documentation this was also the case for VASP 5, but I did not check).

Personally, I found this change very confusing, as I tended to simply grep the line with "HF+E_corr(extrapolated)" - the meaning of which suddenly changed between VASP 6.1 and 6.2. What was the rational behind the change? And could maybe a note be added to the documentation?

Thank you and best regards,
Katharina

[merzuk.kaltak]

Dear Katharina,

The derivation of the EXHF_ACFDT term for the zero-temperature formalism of RPA was discussed in PRB 81 (115126).
Prior to vasp.6.2 this term -- related to the partial occupancies -- was also added to the total energy at T>0.
However, to the best of our knowledge, one can not rigorously derive this term within the Matsubara formalism.
This is the main reason why we decided to remove the term from the total energy as of version 6.2.
Obviously, this energy contribution is still written to the OUTCAR file, and hence,
you may add it back manually to the total energy, if you wish to do that.

Just recall that the forces are the derivative of the total finite temperature energy, without this term!
We also noticed that the derivative of this term with respect to the ionic positions is very difficult or impossible to compute.

[kdoblhoff]

Dear Merzuk,
Thank you for your reply! That makes sense!