[CP2K-user] Cell Optimization for Copper.

[Travis]

Hi,

Firstly, your initial cell parameters should be the same as the structure 
you downloaded. You don't need vacuum to optimize the bulk. If you are 
trying to mimic having a larger grid, set a larger REF_CELL instead.

For cell optimization you need a large supercell or you need kpoints,

1) MULTIPLE_UNIT_CELL X Y Z in both &CELL and &TOPOLOGY sections
2) KPOINTS as below

&DFT
 ...
 &KPOINTS
  FULL_GRID 
<https://manual.cp2k.org/cp2k-6_1-branch/CP2K_INPUT/FORCE_EVAL/DFT/KPOINTS.html#list_FULL_GRID> 
T
  SCHEME 
<https://manual.cp2k.org/cp2k-6_1-branch/CP2K_INPUT/FORCE_EVAL/DFT/KPOINTS.html#list_SCHEME> 
MONKHORST-PACK X Y Z
 &END KPOINTS
 ...
&END DFT

EPS_DEFAULT 1.0E-8 is usually set smaller as well (default is 1e-10). The 
kpoint route should converge within just a handful of optimization steps, 
supposing your stopping criteria are comparable to VASP. Then generate your 
slab.


-T


On Wednesday, August 14, 2019 at 8:25:15 PM UTC-4, Hasan Al-Mahayni wrote:
>
> Hello all,
>
> I am fairly new to cp2k so I believe this issue is easily resolvable. My 
> end goal is calculating adsorption energy of different molecules on copper 
> surfaces (slab). However, the first step I need to do is optimize the cell 
> parameters of copper(cell size). I know from materials project site that it 
> is 3.62126 Angstroms. That is the configuration I start with and is 
> attached by the name of "copper.xyz". The problem I am having is that 
> when I run cell optimization on this (input file is "copper.inp"), I get 
> the wrong results as the cell size shrinks (I get 3.4A). I think this is 
> because of the vacuum between the copper cell and the simulation cell. My 
> question is, what is the right way to set up the input file and especially 
> the &CELL part. I started off by making the A B C cell size 10 A bigger 
> than 3.62A, but that gives me the wrong result. If the A B C cell size is 
> too small, I get an error saying geometry is wrong. Also, maybe GEO_OPT 
> would be a more appropriate run type... I tried making supercell 2x2x2, 
> 3x3x3, and 4x4x4 while keeping ABC cell size 10 A bigger every time, and 
> the result gets closer to 3.62 but this method is not ideal as the system 
> gets bigger and bigger and the calculation is more computationally 
> expensive. Any ideas?
>
> Thanks,
>
> Hasan.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190814/264a9c0f/attachment.htm>