[CP2K-user] Large difference in lattice constant between CP2K and VASP

[fa...@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
On Monday, 18 January 2021 at 14:16:16 UTC+1 ASSIDUO Network wrote:

> 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 :)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210118/0732311b/attachment.htm>