[Pw_forum] K-point convergence help

[Nicola Marzari]

On 25/05/2013 19:04, Ben Palmer wrote:
> Hi everyone,
>
> I made a post some time ago asking for advice on k-point convergence,
> and I had a lot of helpful replies.  I just wanted to ask for a bit more
> help.  I've been converging for Ruthenium and a few other elements.
> They seem to converge at a smearing width of 0.05, then the 18x18x18
> breaks away from the rest at higher smearing widths.  Would this suggest
> 0.05 is the sweet spot, or would it mean I'd have to continue to higher
> numbers of k-points?
>

Dear Ben,


my gut feeling is that at large smearing widths all samplings should
give the same results - so it's a bit odd that the 18x18x18 is off
from the others. Maybe you do not have enough bands at high smearings
so that the occupation of the highest states is significantly larger 
than zero.

In any case, anything interesting is in the region between
0.005 and 0.05 - you need to decide what is the accuracy on the force
you are willing to tolerate, and choose sampling and smearing appropriately.

			nicola




----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL