Convergence problems in XAS calculations

[marci]

Hi Johannes,

my first guess is that by exciting the core states, at some configurations 
along your trajectory, the homo/lumo gap gets quite small, which causes the 
problems in convergence.
In this case the solution would be to add additional unoccupied states also 
for the ground state calculation, and use the smearing and mixing options 
for the XAS optimization.

regards

marcella

On Monday, May 27, 2013 9:16:29 AM UTC+2, Johannes Niskanen wrote:
>
> Dear all,
>
> I am doing XAS calculations with the half hole TP approximation for a 
> molecule in water. I am using a Dunning's basis for the active site and 
> MOLOPT basis set for the rest of the atoms, to avoid linear dependencies in 
> the basis set.
>
> However some 10% of my calculations (snapshots from MD) do not converge. 
> As I am reading the manual, I can find a couple of options in the XAS 
> calculation and very many in the SCF section. However I seem to be unable 
> to converge the calculations with anything I have tried. 
>
> What kind of options/method do you suggest to use in XAS calculations for 
> the best possible convergence? Could one approach the calculation first 
> with the default method and then with the pressumably more costly one for 
> the snapshots for which SCF does not converge?
>
> Best regards,
> Johannes Niskanen
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20130529/47a3f241/attachment.htm>