1) what kind of convergence you haven’t reached yet? a) convergence of 
structural optimisation (atomic positions), or b) convergence of unit cell 
optimisation (e.g. if you use calculation=‘vc-relax’), or c) self-consistency 
cycle

2) if by improving the sampling of the Brillouin zone (2x2x2, 3x3x3, ….) 
nothing happens, maybe it can be that the parameter that prevents you system 
from reaching convergence is not the k-point grid!!! Which pseudo potentials do 
you use, and with which cutoff? For organic crystals, van der Waals 
interactions might be important as well, have you already tried to switch them 
on? Are you sure that the input positions are right, for example by visualising 
your input file with XCrysden (such a practice is advisable BEFORE any, maybe 
computationally expensive, run is started!)

3) the “acceptable" size of the k-point grid depends on the system under 
investigation and on the property you are interested in (for a given grid, 
there are properties that might be converged, and others that might be not!). 

4) concerning grids that look like n x m x l instead of n x n x n , usually if 
for example axes a and b are in the ratio a/b = u, than n and m have the 
inverse ratio (n/m ~ 1/u). For example: a = b = c, use 4 x 4 x 4 ; a = b, c = 
2a, use 4 x 4 x 2 (THIS ARE JUST EXAMPLES, THAT DO NOT PROVIDE CONVERGENCE 
CRITERIA!).

GIOVANNI



> On 04 Dec 2015, at 14:23, Ridwan Agbaoye <[email protected]> wrote:
> 
> 
> 
> On Thursday, December 3, 2015 11:41 AM, Ridwan Agbaoye 
> <[email protected]> wrote:
> 
> 
> I am doing structural optimization for a monoclinic latice structured organic 
> compound with 13 atoms, 4 types of atom (C,H,O and S) and  , i need expertise 
> advice on how my automatic kpoint grid required for optimization will look 
> like, i tried 2 2 2 1 1 1, 3 3 3 1 1 1, 4 4 4 1 1 1... 9 9 9 1 1 1 but the 
> result havent reached convergence yet even with the very very high 
> computation time it require 
> 
> Recently i learnt i am not suppose to use the uniform kpoint grid but the 
> ones that looked like 6 4 6 1 1 1 , 6 1 1 1 1 1, 8 4 8 1 1 1 etc
> 
> please help me with the best automatic kpoint grid required for my system and 
> how to determine the kind of kpoint grid to use for subsequent system 
> calculation 
> 
> Thank you very much
> 
> 
> Agbaoye Ridwan Olamide
> M.Sc. (in view)
> Federal University of Agriculture, Abeokuta
> 
> 
> _______________________________________________
> Pw_forum mailing list
> [email protected]
> http://pwscf.org/mailman/listinfo/pw_forum



_______________________________________________
Pw_forum mailing list
[email protected]
http://pwscf.org/mailman/listinfo/pw_forum

Reply via email to