Large difference in lattice constant between CP2K and VASP

[ASSIDUO Network]

Hi there, I am trying to calculate the lattice constant of bulk Au using CP2K (cell optimization is fine). I have included my input file below.

With CP2K, I get a lattice constant around 4.6 Å, while with VASP I get a value of 4.17 Å. The literature lattice constant for bulk Au should be around 4.08 Å. Why do I get such a massive difference in value between CP2K and VASP/literature?

What would you advise I do with my calculation to improve my results? The largest basis set available for Au (with CP2K) is TZV2P. My supervisor wants me to use the GTH-PBE potential, so I can't change that.

Thank you for your help in advance :)

[fabia...@gmail.com]

Hi,

you are running the optimization without a k-point grid, so you only consider the gamma point. To get a more reasonable result add

    &KPOINTS
      SCHEME MONKHORST-PACK 20 20 20
    &END KPOINTS

to the &DFT section. Adjust the number of kpoints to your needs. When using k-points you should probably change METHOD DIRECT_P_MIXING to METHOD BOYDEN_MIXING.

A few more hints:

with DIAGONALIZATION you dont need the outer_scf section and ADDED_MOS 100 seems excessive.

@SET SCF_CONV 1E-5 is rather inaccurat, I'd recomment decreasing that to 1e-6 or 1e-7 even.

Is there a specific reason to decrease @SET GRIDS 3 from the default? That should not be needed.

When performing simulations with a variable cell size it's recommended to include a &CELL_REF section. Set these lattice vectors longer than the maximum value that you expect during the simulation (for your real cell).

Cheers,

Fabian

[Krack Matthias (PSI)]

Hi

 

for a bulk metal like Au you need to employ smearing and k point sampling.

 

Matthias

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/73e8aa56-4cdb-4c15-8b15-4b672db17cdcn%40googlegroups.com.

[ASSIDUO Network]

Thank you very much for your response. I will definitely look into k-point grids and apply your recommendations. See if that changes things.

Most of the flags/values I used were from my supervisor's (other) student, who recommended these to me. I'm still very new to CP2K, so I am still learning what to use/not to use.