encut convergence in ACFDT

[kdoblhoff]

Dear Vasp community,
I have a potentially stupid question: I am performing an ACFDT (finite temp, low scaling algorithm) calculation of a molecule on a metal slab. In converging the plane wave cutoff energy, I find a "jumping" behavior (see orange line in the attached figure), while the DFT calculation converges smoothly when increasing the plane wave cutoff (blue line). The jumping behavior is not related to the ENCUTGWSOFT and the jumps appear both in the exchange and in the correlation energy. Where do these jumps come from?
Thanks for explaining,
Katharina

[merzuk.kaltak]

Dear Katharina,

Since you observe the same for the exchange, this might be related to PRECFOCK.
The used setting is written in the OUTCAR file.
If you do not set this value in the INCAR, vasp reverts to its default, which for ALGO=RPAR is fast.
I recommend PRECFOCK=normal or even accurate.

[kdoblhoff]

Dear Merzuk,
Argh! You were right of course! The PRECFOCK solves the issue (see plot).
Thank you!

[kdoblhoff]

One more comment maybe for other users: This is currently a mistake in the documentation. At the moment, the documentation states:

PRECFOCK
PRECFOCK = Normal | Accurate | Fast | Medium | Single | Low
Default: PRECFOCK = Normal

As noted by merzuk.kaltak, this is not the default behavior for RPAR.
Best regards,
Katharina

[merzuk.kaltak]

Thanks for letting me know. The wiki is changed accordingly now.