[CP2K:6369] Re: How to freeze the density in QM/MM calculation ?

[Valérie Vaissier]

Hi Martin,

Thanks a lot for your help. I wanted to get back to you with either "yes it
works" or "no it does not" but the truth is .. I'm not sure.

So I set EPS_SCF to 100 (1 was actually too small to have it "converge"
within one step) and set ALPHA parameter to something very small (I tried 0
but, while parsing the input file ok, it's unhappy when calling a
subroutine where it has to do something%ALPHA - fair enough). And I froze
my QM atoms (I had forgotten that in the input file I attached last time).

It runs and terminates normally. Now I wanted to do few sanity check to
make sure nothing too weird happens. Then I realize that it's actually
quite difficult to find something to track in a QM/MM calculation to check
that the density stays the same. I mean all energies will depend one way or
another on the MM region so ... In the end I printed out the electron
density and I'm not so happy with what it gives:

- I have a .wfn file (let's call it file1.wfn) from which I start a 100
step NVE MD at the QM/MM level (Run 1). Frozen QM atoms and mixing density
with alpha = 10^-20.
    &SCF
      SCF_GUESS RESTART
      MAX_SCF 10
      EPS_SCF 100
      EPS_DIIS 100
      &MIXING
        ALPHA 1e-20
      &END MIXING
    &END SCF

I print a new restart file, file2.wfn for this run. I print out the
electron density cube files for each step
- I start a new 100step QM/MM (Run 2) from file2.wfn for which I also print
out the electron density cube files.

I'm not so sure about that but I guessed that the .wfn file was being
translated into the electron density. So I expected the last electron
density of Run 1 to be identical to the first electron density of Run 2
(through file2.wfn). But that is not the case. This is a different issue I
suppose but I'd like to be confident on the restart scheme too

The the first electron density cube file differs from the last electron
density cube file for each run. Which surprises me with a machine-zero
value for the coefficient. I'm thinking that if it generates deviations
after 100 steps, what would happen after millions ... Then I tried
alpha=10^-30 and there, all electron density cube files are empty. Which is
even more surprising. I thought the mixing scheme was density = alpha*new
density + (1- alpha)* old density. But that's not consistent with these
results ...

puzzling!






On 26 March 2015 at 07:51, M. Brehm <brehmin... at googlemail.com> wrote:

> Hi,
>
> when I was just reading your question, I was thinking about a dirty hack
> which could do what you want:
>
> Set EPS_SCF to something very large (e.g., 1), such that the SCF cycle
> will always converge after the first step. Then, turn on classical mixing,
> and set the ALPHA parameter to 0. This means that a new wave function is
> computed, but only "0%" of it are mixed into the old one, i.e., the initial
> wave function should remain constant.
>
> Maybe this works :-)
>
> Best regards,
> Martin
>
> --
> You received this message because you are subscribed to a topic in the
> Google Groups "cp2k" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/cp2k/cgAdbo-2V8E/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> cp2k+uns... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> 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/20150327/65ce6ebb/attachment.htm>