Specifying VDW_A1, A2 and S8 when combining functionals with XC

[julien_steffen]

Dear all,

I'm attempting to use the SRP48 functional in VASP 6.5.0 and I'm encountering issues when starting the calculation.
As the functional is not directly implemented in VASP afaik, I'm combining the PBE and RPBE functionals in a 52:48 ratio using the XC and XC_C tags in the INCAR.

Starting the calculation shows the error message that the specified functional is not known and to define VDW_A1, VDW_A2 and VDW_S8 in the INCAR. I've attached the results of the calculation.
Starting a calculation when using either XC = PE or RP, the calculation runs without any issues.

I'd like to know whether there's somehow another way to use the SRP48 functional in VASP that I haven't found yet? Or can I can get the required parameters (VDW_A1, A2 and S8)
by calculating their weighted average from the parameters of PBE and RPBE given in the OUTCARs when I perform a calculation with only the respective functionals?

The calculation works if I supply any one of the asked parameters and it's stated in the OUTCAR that the remaining parameters will be set to the PBE default parameters.
Will just using the PBE parameters to run the calculation with the SRP48 functional drastically change the results?

Best regards

[pedro_melo]

Dear julien_steffen,

Do you know if the version of the functional in question is implemented in Libxc? You can check the list on their website, https://libxc.gitlab.io/functionals/, and if you know the corresponding Libxc tag you can use it and VASP's input tags, https://www.vasp.at/wiki/index.php/LIBXC1, to tell VASP which functional you want to use.

For this to work you will also need to have VASP computed with Libxc support.

Let me know if this helps. Kind regards,
Pedro

[julien_steffen]

Hello pedro,

unfortunately it's not directly implemented in the libxc library.
Also, from my understanding, LIBXC1 and LIBXC2 can only be used to select an exchance and correlation functional.
When selecting an exchange-correlation functional only LIBXC1 can be used.
So considering that the SRP48 functional is not part of the libxc library, the LIBXC tags won't work.

Best regards

[pedro_melo]

Dear julien_steffen,

I spoke with a colleague and mentioned that

XC = PE RP
XC_C = 0.52 0.48

is the correct way to specify the GGA functional SRP48.

The issue with the VDW_A1, VDW_A2, and VDW_S8 tags comes the fact that you are trying to use the D3(BJ) dispersion for the van der Waals interaction, which needs those parameters. However, we are not sure that these have been optimised for this functional in particular. Taking the average of the VDW_A1, VDW_A2, and VDW_S8 parameters for PBE and RPBE (e.g. 0.48* VDW_A1_RPBE + 0.52* VDW_A1_PBE) is probably OK, but we cannot guarantee that it is accurate.

If you have access to the source code, you can find the values for these parameters in the file subdftd3.F in between the lines 2715 and 2726.

Kind regards,
Pedro