[gpaw-users] : RuntimeError: Ghost!

[Ask Hjorth Larsen]

Hi Yuelin

On Mon, 15 Nov 2010, Yuelin Li wrote:

> Ask,
>
> thanks. You are suggesting that one has to re-generate the pseud-potential 
> for Ca? I think each element is unique in this sense how can parameters for 
> other elements helpful? And what elements do you have in mind?
>
> Best,
>
> Yuelin

I'm not saying anything very specific.  If a particular PAW setup is 
troublesome, it may be worth generating a different one.  To get an idea 
about which parameters to use, it makes sense to look at different atoms 
with some of the same properties (e.g. same group) - that will tell you 
whether e.g. the filtering parameter should be 0.1-ish or 0.4-ish or 
1e6-ish.

But another thing you can try is to use the HGH pseudopotentials (although 
that'll make the calculation more expensive as it may require a finer 
grid-spacing, e.g. h=0.12).  It doesn't necessarily make the calculation 
converge, either.

Did you try increasing the Fermi temperature to something rather high?

In conclusion,
   1) There is no entirely systematic way to make something converge
   2) There is no single optimal way to generate a setup

I hope you can find something that works.

Regards
Ask