[Pw_forum] Convergence test on big model

[Ari P Seitsonen]

Dear Yangchuan Li,

   In principle the values of cut-off are fully transferable, for the same 
kind of pseudo potential: As the highest Fourier components are required 
due to more localised (ie fastly varying) electronic wave functions close 
to the nuclei and there the changes going from one chemical environment to 
a different one are small, in most cases the cut-off tested in one system 
are indeed transferable (a reminder: using the same pseudo potential, so 
the recommended values for the cut-off energies are not strictly element- 
but even pseudo potential-specific).

   About the k points, rather the bigger the system, less k points: That is 
"hidden" in the word "reciprocal" space. Please have a look in the 
discussions on the back-folding of k points etc. The bottom line is, to 
get the same kind of accuracy in systems of different sizes the distance 
between the k points should be more or less the same (as they are used to 
approximate an integral). And if the cell in real space gets bigger, the 
Brillouin zone gets smaller: Thus less k points needed if the distance 
between the points is kept the same.

   And the density depends on the kind of system: In general one needs more 
points in metallic systems (in order to integrate the states close to the 
Fermi level accurately) than in semi-conducting or insulating systems. But 
if in the latter there are some more exotic bands somewhere needing higher 
integration sampling in the reciprocal space, more k points are also 
needed there. So shortly, no, one cannot give a general "preferred" number 
of k points; also because that depends on the quantity one is looking at 
(the structure converges faster than for example the phonon modes etc).

     Greetings from Sleepy Paris,

        apsi

-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
   Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/
     Ecole Normale Supérieure (ENS), Département de Chimie, Paris
     Mobile (F) : +33 789 37 24 25    (CH) : +41 79 71 90 935


On Thu, 18 Feb 2016, Yangchuan Li wrote:

> Hi everyone,
> 
> I want to ask whether a converged Ecut and smearing width can be used in another model. I have a big model (functionalized carbon nanotube) and it's time-consuming to do scf
> calculation. I notice Stefano mentioned that Ecut is not strongly depended on particular configuration. (In topic Technique for converging Ecut and K-points?)
> 
> So can I used small model like diamond, nitrogen gas to do convergence test and then use the preferred parameter in my big model? Why I can or can't do that?
> 
> Also, if the dimension of the system is big. Is it necessary to select a very dense k-mesh? Is there a preferred k-point density in reciprocal space?
> 
> Thank you for your time. Any comment will be appreciated.
> 
> Yangchuan Li
> UT Austin | ME
> 
> 
> 
>