[CP2K:1143] Calculations on isolated molecules

[Teodoro Laino]

Ilya,

the question depends of course on the poisson_solver (PS) you decide  
to use. Specifically:

-) Martyna-Tuckermann (MT) PS is fully translational invariant.. This  
means that the molecule can
     be everywhere.
-) Wavelet (W) requires that the molecule is centered in the middle  
of the box (use SUBSYS%TOPOLOGY%CENTER_COORDINATES
     to achieve this purpose)
-) Multipole, is translational invariant

Which one is the best? As usual there's not a real best one..
Keep in mind that MT requires a simulation box at least double the  
size of your system while W requires a box just slightly larger than the
size of the system (i.e. the density must be negligible at the border  
of the simulation box, usually 3-4 Angstrom per side should be  
enough.. but
again this depends on your basis set.. element.. simulation system).

While W could be more convenient because of the smaller cell size, MT  
is undoubtedly more robust during an MD.. In fact is quite likely that
your system will start moving during an MD and if it's reaching the  
border of the simulation box you will need to stop and restart the  
calculation
(quite annoying though)..

Multipole is somehow something that lies in between.. requires that  
density is small at the box boundaries but is translational  
invariant. The main
problem of the multipole is that it is based on a fit.. therefore you  
may have numerical problems and not fully an analytical cancellation  
of the
periodic interactions..

Now.. the choice as usual is yours!
Teo


On 1 Jul 2008, at 12:02, ilya wrote:

>
>
> Hello!
>
> I wanted to clarify where should I place the molecule in the supercell
> to make single molecule calculations. For example if I specify the  
> CELL
> %ABC 10.0 10.0 10.0, should I place the molecule at the origin or in
> 5.0 5.0 5.0? I also should set CELL%PERIODIC NONE and POISSON%PERIODIC
> NONE. What value of POISSON%POISSON_SOLVER is most suited for
> calculations on isolated molecules (the default is PERIODIC)?
>
> Thanks a lot.
> Ilya.
> >