Meaning of negative ENCUTGW value in OUTCAR

[ashwin_r]

Hello,

I am trying to compare the outputs of TDHF and TIMEEV runs (with VASP 6.6.1). The VASP Wiki indicates that, to bring the Casida results into closer agreement with time-propagation, one should increase ENCUTGW beyond its default value and set ANTIRES=2 in the Casida TDDFT calculation.

As an example, for a TDHF run that uses ENCUT=450 eV, the OUTCAR shows ENCUTGW=300 eV, and ENCUTGWSOFT=240 eV (exactly as per the defaults noted on the Wiki).

On the other hand, for the TIMEEV run, the output is a bit more cryptic: for ENCUT = 450 eV, the OUTCAR shows ENCUTGW=-1.6 and ENCUTGWSOFT=-1.6 -- it is not immediately obvious what the actual values are and, consequently, I am not able to port these defaults to the Casida run for consistency.

Any clarification is most appreciated!

Ashwin

[pedro_melo]

Dear Ashwin,

ENCUTGW and ENCUTGWSOFT are set to -1.6 by construction in the TIMEEV run. Since these are not specified in the INCAR, VASP internally initialises them to -2 (meaning "not set by the user"), and later some safeguards multiply them by 0.8, hence the -1.6 value you observe.

The TIMEEV algorithm does not use ENCUTGW nor ENCUTGWSOFT. The Hartree and exchange parts of the Hamiltonian in TIMEEV are built using the same Hartree-Fock routines that are used for ground-state/hybrid functional calculations. The cutoffs are controlled by ENCUT/PRECFOCK/ENCUTFOCK. So there is no equivalent to ENCUTGW or ENCUTGWSOFT that you can use in the INCAR in the TDHF run.

Essentially you are performing a convergence run. You need to increase the value of ENCUTGW until the spectrum stops changing. The advice here is to perform this on a job where all the other parameters are not converged, (NBANDSO, NBANDSV, KPOINTS, etc). This will make your jobs faster and require fewer resources.

Hope this helps. Kind regards,
Pedro

[ashwin_r]

Thanks for the clarification, Pedro.

How come the OUTCAR and standard output display this warning for TIMEEV: "The default for ENCUTGWSOFT is different in this version of VASP. If you wish to obtain identical results as using pre vasp.6.3 please manually set ENCUTGWSOFT =-2 in the INCAR file."? Based on what you noted, I assume that this can be ignored, correct?

Ashwin