Choosing the right system size - supercell

[Matt W]

Hi,

guess this fell through the net because it is so general a question that it 
is impossible to answer in a way that is usefull/but correct in all cases:

The actual answer is that, at the moment without k-point sampling, your 
cell needs to be large enough to converge the property you are interested 
in - in all directions:i.e  a circular answer that is not helpful.

Crudely if you have at least a cell 1 nm in each direction, you'll probably 
get something qualitatively reasonable, but the actual size needed will 
depend on the system and property...

If you can say what band-gap, dielectric constant and properties you want 
to calculate, then you might get a better hint. But still you should be 
happy that your results are well enough converged for your purposes, which 
means testing yourself, not asking a mailing list. 

Cheers,

Matt


On Monday, September 8, 2014 1:58:26 PM UTC+1, Tobias Kraemer wrote:
>
> Dear all, 
>
>
> I have a question regarding the correct choice of system size. I guess 
> this is also aiming at a more general understanding 
> on how periodic DFT works. When is it important to expand the single unit 
> cell to a supercell?? In my particular case I am 
> studying a molecular crystal containing an organometallic fragment (a 
> rhodium complex) embedded in cavities spanned by
> the counterion (the BArF4- borate anion). There are essentially 4 
> asymmetric units in the unit cell (4 cation/anion pairs)
> In another, but related, system there are only two of these asymmetric 
> units involved. So how are the interactions between
> fragments at the boundary of the unit cell described? Since I am setting 
> up the calculations as PERIODIC XYZ, am I getting
> away with just using the unit cell? Are all the interactions due to the 
> presence of the *neighboring *cells in each direction
> automatically taken into account? In particular, here in this system each 
> cation in the crystal is surrounded by 6 counterions 
> (which form kind of an octahedral cavity), but of course in the unit cell 
> each cation has only half (=3) anions in its vicinity.
> I suspect that I might have to use a supercell to describe the system 
> properly, but what would be a good size?
> In the other case mentioned, there are only 2 asymmetric units present in 
> the unit cell, and I wonder if I would introduce 
> some sort of "size-error" into the calculations, which might make a 
> comparison between the two different crystal systems difficult.
> I have attached the xyz for one system for reference. 
>
> A second question, is it generally better to fully relax the complete cell 
> or is a geometry optimization suffcient.
>
>
> Thanks for your help, much appreciated
>
>
> Tobias 
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141001/db99f40f/attachment.htm>