Quickstep: SCF Energy increasing after few outer-SCF cycles

[Sharma SRK Chaitanya Yamijala]

Dear Flo, Matt and CP2K users,

I have  changed the preconditioner as well as the energy gap as suggested 
by Flo and Matt and now the runs are exiting normally. When I just changed 
the energy gap it has worked for some systems but not for all. Whereas, 
when I used both of them, then it is working for all the systems (till now 
what I have considered). Yet, I haven't checked the results only by 
changing the preconditioner to FULL_SINGLE_INVERSE. Once I get them, I will 
post.

Many thanks to Flo and Matt,
Sincerely,
Sharma.

On Thursday, August 22, 2013 6:12:11 PM UTC+5:30, Florian Schiffmann wrote:
>
> Hi, 
>
> this problem has indeed to do with the preconditioner. The statement on 
> the website that a 0.001 value for the energy gap should improve 
> convergence is only partially true. The problem is that it is used as a 
> scaling factor, which for well behaved system (wide band gap usually) is 
> not a problem, but can lead to more random results (problems in inversion 
> due to almost linear dependencies) for small band gap systems and thus mess 
> up your electronic structure (see electronic density on regular grids). 
> Changing the value for the energy gap to 0.1 should thus help as well. As 
> always check as well your plane wave convergence, especially for BLYP 
> larger values might be required.
>
> Flo
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20130828/e475c578/attachment.htm>