Re: [Phonopy-users] negative frequencies, cp2k interface

[Kien N. C. <han...@gm...>]

Hi Kadioglu,

Do you see the minimum (or minima) of the negative frequency  at gamma
point or other q-point? If it occurs at other q-points you need to
determine the size of modulated-structures based on that q-point.  The
principle is simple there, just determine the smallest supercell
commensurate with the q-point. Of course you may find several solutions of
supercell sizes, but they should be equivalent based on the symmetry of
your structures. After you have it, use the modulation tag and produce
several (may be many) structures with different aptitudes of displacements
to obtain modulated (supercell) structures. We have to do that because we
want to move atoms far enough in some cases so that it will not be moved
back to the original 'local' energy minimum of the perfect structure during
the energy minimization. Next step, you need to relax the modulated
structures without any symmetry constraints. If the energy of any relaxed
structure is lower than that of your perfect structure, the imaginary
frequency you obtained is real rather than the numerical noise. Of course,
it does not actually mean your structure is not stable at all, it indicates
that your structure is not stable at low temperature within the limit of
the harmonic regime. I'm not using cp2k, but these can be applied for any
DFT calculators.

Instead of using the modulations tags, I saw people run MD or perform
random displacements of atoms in the supercell, but I think it is more
expensive and impractical if you have s large supercell.

On Wed, Jun 24, 2020 at 2:44 PM Kadioglu Özge <ozg...@ep...> wrote:

> Hi,
>
> Thanks a lot for your reply Togo.
>
> Looking at the magnitude of the frequencies, I can say that they are not
> due to numerical noise from cp2k forces. I visualized the phonons in my
> system and saw that the negative frequencies are due to the vibrations of
> the benzene rings.
>
> I am looking at the MODULATION tag right now, I understood how to create
> the tag however I am confused how to choose the q-points to use in the tag.
> When I use the full k-point path to create the band.yaml file, I have 612
> q-points. For each q-position that i specify in the tag, there is one
> MPOSCAR file with the displacements.
>
> Since I am interested in calculating heat capacities with harmonic
> approximation, I think I should be focusing on gamma point. Please correct
> me if i am wrong.
>
> For each negative frequency (band index), I used the modulation tag (at 0,
> 0, 0) and produced structures with the atoms displaced in accordance with
> the corresponding eigenvectors. How am i going to proceed after this point?
> Now I have MPOSCAR files as much as the number of negative frequencies.
>
> If i had one big negative frequency, I would optimize the new structure
> that I got from modulation tag, but now with 8 MPOSCAR files I got a bit
> confused about how to proceed. My question might be more related to the
> theory rather than the implementation of phonopy but I might be misusing
> the modulation tag as well.
>
> I would be very happy if someone who solved that problem before can help
> me.
>
> Thanks in advance!
> Best,
> Özge
>
>
>
>
>
> ---
> Özge Kadioglu
> Doctoral Assistant
> Laboratory of Molecular Simulation (LSMO)
> École Polytechnique Fédérale de Lausanne
>
>
> ------------------------------
> *From:* Atsushi Togo <atz...@gm...>
> *Sent:* Friday, March 13, 2020 5:36:21 AM
> *To:* phonopy-users
> *Subject:* Re: [Phonopy-users] negative frequencies, cp2k interface
>
> Hi,
>
> This is the reply for the question about eigenvector corresponding to
> the imaginary frequency.
> A collective displacement of atoms may be written as
> https://phonopy.github.io/phonopy/formulation.html#mean-square-displacement
> In this equation, what you have to take into account to displace atoms
> are, 1/sqrt(omega), eigenvectors, and phase factor exp(iqr). The
> amplitude depends on phonon state, but the current purpose is to break
> symmetry, so if your code can detect the broken symmetry due to the
> collective displacement, it's OK.
> You may be able to use this feature of phonopy,
>
> https://phonopy.github.io/phonopy/setting-tags.html#create-modulated-structure
> .
> Honestly I don't like this very much, but in some cases, this is
> useful. You have to use this with understanding what you are doing.
> In addition, the imaginary frequency may come from numerical noise of
> cp2k forces. In this case, you may not obtain a good result that you
> expect.
>
> Togo
>
> On Tue, Mar 10, 2020 at 1:04 AM Kadioglu Özge <ozg...@ep...>
> wrote:
> >
> > Hello,
> >
> > I am working on phonon properties of MOFs. There are some challenging
> structures that I constantly have negative frequencies for. I managed to
> decrease the number of negative frequencies by pushing the geometry
> optimization more and using K-points while optimizing on cp2k. However, I
> still had negative frequencies, then I did cell-optimization but afterwards
> the number of negative frequencies increased again.
> >
> > I checked the previous email threads, seems like I should check the
> eigenvectors corresponding to the negative frequencies. I did that by using
> --eigvecs tag and got the eigenvectors written in mesh.yaml file. But i do
> not know how to move on after this point. How do we move the normal modes
> with negative frequencies before trying a new optimization? If you have any
> other suggestion for me to check I would be very glad.
> >
> > Also, with the version 2.4.2, it takes much longer for the displacements
> to be created. It was only couple of seconds with the previous versions of
> phonopy, for a system consisting of 200-300 atoms, now it is 1-2 minutes to
> create cp2k input files with the displacements. May i ask why did it change
> significantly?
> >
> > PS: I am using phonon --cp2k interface.
> >
> > Thanks in advance.
> >
> >
> >
> > ---
> > Özge Kadioglu
> >
> > Doctoral Assistant
> > Laboratory of Molecular Simulation (LSMO)
> > École Polytechnique Fédérale de Lausanne
> >
> >
> > _______________________________________________
> > Phonopy-users mailing list
> > Pho...@li...
> > https://lists.sourceforge.net/lists/listinfo/phonopy-users
>
>
>
> --
> Atsushi Togo
>
>
> _______________________________________________
> Phonopy-users mailing list
> Pho...@li...
> https://lists.sourceforge.net/lists/listinfo/phonopy-users
> _______________________________________________
> Phonopy-users mailing list
> Pho...@li...
> https://lists.sourceforge.net/lists/listinfo/phonopy-users
>


-- 
------------------
Nguyen Kien

electronic