[CP2K:6828] SCF convergence in NEB calculation

[Tobias Kraemer]

Dear Juerg,


thanks for your answer. Whilst I am using an SCF convergence of 10e-6 as 
you suggest, I am not using the conjugate gradient OT minimizer but DIIS 
instead. Perhaps something to try out, I will give it a go. What I was 
wondering about is, if there is a possibility to control these sort of 
parameters for individual replicas.
The electronic structure will change quite a bit as one traverses along the 
profile between starting and last point, hence I wouldn't expect too much 
of a problem with the first point, but more towards the transition state. 
Anyways, I guess it is more reasonable to control the SCF globally for all 
points.

Another question I have in this context, is there a rule of thumb for how 
many intermediate structures for NEB should be provided? At the moment I am 
using a starting structure, a guess for the TS plus a final structure. With 
8 replicas defined, the algorithm should interpolate between these. I was 
successful to localise
one transition state with this already, however, for other transition 
states things are more complicated. Also, in cases I observe rather strange 
energy profiles, in which the first (sometimes also the last) point lies 
higher in energy then the remaining points (see attachment). This is 
somewhat surprising, since the first point corresponds to an optimised 
structure (I am assuming that the first replica correlates with my first 
starting structure). Inspection of the movie.xyz generated from the 
individual BAND files reassures me that I am looking at a reasonable 
trajectory. Is is worth waiting for this calculation to converge to some 
stationary point, or should I come up with a better setup in terms of 
structures?    

Thanks for the advice


Tobias
    
   

  



On Wednesday, July 29, 2015 at 8:33:36 AM UTC+1, jgh wrote:
>
> Hi Tobias 
>
> I would start using a very conservative setting for the SCF 
> optimizer. For example 
>
>     &SCF 
>       MAX_SCF 20 
>       SCF_GUESS ATOMIC 
>       EPS_SCF  1.e-06 
>       &OT ON 
>         MINIMIZER CG 
>         ENERGY_GAP 0.002 
>         LINESEARCH 3PNT 
>         PRECONDITIONER FULL_ALL 
>       &END OT 
>       &OUTER_SCF 
>         MAX_SCF 20 
>         EPS_SCF  1.e-06 
>       &END OUTER_SCF 
>     &END SCF 
>
> But maybe you already do this? 
>
> Juerg 
>
> -------------------------------------------------------------- 
> Juerg Hutter                         Phone : ++41 44 635 4491 
> Institut für Chemie C                FAX   : ++41 44 635 6838 
> Universität Zürich                   E-mail: hut... at chem.uzh.ch 
> <javascript:> 
> Winterthurerstrasse 190 
> CH-8057 Zürich, Switzerland 
> --------------------------------------------------------------- 
>
> -----cp... at googlegroups.com <javascript:> wrote: -----To: cp2k <
> cp... at googlegroups.com <javascript:>> 
> From: Tobias Kraemer <161br... at gmail.com <javascript:>> 
> Sent by: cp... at googlegroups.com <javascript:> 
> Date: 07/28/2015 12:44PM 
> Subject: [CP2K:6828] SCF convergence in NEB calculation 
>
> Hi all, 
>
>
> I am running CI-NEB calculations in order to locate transition states 
> along a reaction profile involving organometallic species in the solid 
> state. More precisely, this is a hydrogenation reaction of double bonds, 
> i.e. addition of H2 to an alkene. This should be a straightforward task. 
> Anyway, for some cases it seems the SCF encounters problems to converge, or 
> converges to the wrong electronic state. In particular, those replicas are 
> affected which lead to the  transition 
>  state. I can see that the electronic structure may be a bit tricky, 
> since in a TS notable rearrangement of the electrons takes place. It goes 
> without saying, that having a wrong electronic structure in the TS leads to 
> unreasonably high barriers. I should add that these type of calculations 
> haven't imposed any problems so far using other codes. What causes some 
> additional puzzlement is the fact that for the same TS search slightly 
> different input  geometries lead to very different behaviour of the SCF. In 
> one case there is reasonable convergence to a decent electronic structure, 
> in the other case the SCF goes completely havoc.  
>
> So my question really is, how can I control the SCF convergence for a 
> particular replica? Can this be done in the main input file? 
>
> your comments are appreciated 
>
>
> Tobi 
>
>
>
>
>
> -- 
>
> You received this message because you are subscribed to the Google Groups 
> "cp2k" group. 
>
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to cp2k+... at googlegroups.com <javascript:>. 
>
> To post to this group, send email to cp... at googlegroups.com <javascript:>. 
>
>
> Visit this group at http://groups.google.com/group/cp2k. 
>
> For more options, visit https://groups.google.com/d/optout. 
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150729/06d3c75e/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: output.png
Type: image/png
Size: 2653 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150729/06d3c75e/attachment.png>