Specifying both GGA and meta-GGA tags in INCAR

[aaron_kaplan2]

I've found that specifying both GGA and meta-GGA tags can change total energies by an order of magnitude (this was not intentional specifying both). An example calculation is attached for Al2O3 (sapphire). The directory ISMEAR_-5 is a simple static using only METAGGA = R2SCAN, and ISMEAR_-5_gga uses both METAGGA = R2SCAN and GGA = PS.

One can see that the total energies differ by an order of magnitude, but the forces in vasprun.xml are completely identical.

Can you confirm that changing GGA with METAGGA specified induces only a constant energy shift, so that total energies computed with the same combination of (GGA, METAGGA) are internally consistent?

[fabien_tran1]

Hi,

I could reproduce this change in the total energy when "GGA = PS" is also in INCAR. This is not very nice, but this should be essentially harmless because this change in the total energy comes from atomic quantities. That means that it is a shift that depends only on the number and type of atoms in your system and not on the geometry. The calculation of the cohesive energy is also not affected (I have checked it). Note that there is no such shift when the GGA is the same as in the POTCAR, i.e. PBE. Also note that since vasp.6.4.0 a calculation stops if GGA and METAGGA are both specified in INCAR.

[aaron_kaplan2]

Thanks, Fabien.

I'm finding that a small set of calculations (~5% of 1,100 calculations) run with the same METAGGA tag but different GGA tag give appreciably different forces. The majority give exactly the same forces, or differ within an acceptable tolerance.

While I'm not sure this is important, letting you know in case this is an issue with how VASP reconstructs the core charge density / potential when using a functional different from the one specified in LEXCH.

[fabien_tran1]

Thank you for the information. Could you please provide the files of such a case that belongs to these 5%?